| blob | 4901cb1fd0742d7a1b08589d9b8cb0b398f11269 |
1 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2 Chung-Ju Wu <jasonwucj@gmail.com>
4 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5 * config/nds32/nds32-graywolf.md: New file.
6 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
7 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
8 pipeline.
9 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
10 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
11 * config/nds32/nds32.md (pipeline_model): Add graywolf.
12 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
13 * config/nds32/pipelines.md: Include n15 settings.
15 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
16 Chung-Ju Wu <jasonwucj@gmail.com>
18 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
19 * config/nds32/nds32-n13.md: New file.
20 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
21 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
22 pipeline.
23 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
24 * config/nds32/nds32.md (pipeline_model): Add n13.
25 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
26 * config/nds32/pipelines.md: Include n13 settings.
28 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
29 Chung-Ju Wu <jasonwucj@gmail.com>
31 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
32 * config/nds32/nds32-n10.md: New file.
33 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
34 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
35 pipeline.
36 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
37 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
38 * config/nds32/nds32.md (pipeline_model): Add n10.
39 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
40 * config/nds32/pipelines.md: Include n10 settings.
42 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
43 Kito Cheng <kito.cheng@gmail.com>
44 Chung-Ju Wu <jasonwucj@gmail.com>
46 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
47 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
48 Add enum values for DSP extension instructions.
49 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
50 New constraints.
51 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
52 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
53 New code iterators.
54 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
55 * config/nds32/nds32-dspext.md: New file for DSP implementation.
56 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
57 * config/nds32/nds32-intrinsic.md: Likewise.
58 * config/nds32/nds32_intrinsic.h: Likewise.
59 * config/nds32/nds32-md-auxiliary.c: Likewise.
60 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
61 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
62 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
63 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
64 * config/nds32/nds32-protos.h: New declarations for DSP extension.
65 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
66 TYPE_DMAC in switch statement.
67 * config/nds32/nds32.c: New checking and implementation for DSP
68 extension instructions.
69 * config/nds32/nds32.h: Likewise.
70 * config/nds32/nds32.md: Likewise.
71 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
72 * config/nds32/predicates.md: Implement new predicates for DSP
73 extension.
75 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
77 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
78 Reformat alternatives and attributes so it is easier to identify
79 which constraints/attributes go with which instruction.
80 (mov<mode>_hardfloat32, FMOVE64): Likewise.
81 (mov<mode>_softfloat32, FMOVE64): Likewise.
82 (mov<mode>_hardfloat64, FMOVE64): Likewise.
83 (mov<mode>_softfloat64, FMOVE64): Likewise.
85 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
87 * doc/extend.texi (PowerPC Built-in Functions): Rename this
88 subsection.
89 (Basic PowerPC Built-in Functions): The new name of the
90 subsection previously known as "PowerPC Built-in Functions".
91 (Basic PowerPC Built-in Functions Available on all Configurations):
92 New subsubsection.
93 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
94 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
95 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
96 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
98 2018-05-11 Martin Jambor <mjambor@suse.cz>
100 PR ipa/85655
101 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
102 single const.
104 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
106 PR target/85733
107 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
109 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
111 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
112 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
113 (ix86_handle_option): Handle -mwaitpkg.
114 * config.gcc: New header.
115 * config/i386/cpuid.h (bit_WAITPKG): New bit.
116 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
117 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
118 function type.
119 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
120 OPTION_MASK_ISA_WAITPKG.
121 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
122 (ix86_option_override_internal): Add PTA_WAITPKG.
123 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
124 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
125 IX86_BUILTIN_TPAUSE.
126 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
127 __builtin_ia32_umwait and __builtin_ia32_tpause.
128 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
129 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
130 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
131 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
132 UNSPECV_TPAUSE): New.
133 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
134 * config/i386/i386.opt: Add -mwaitpkg.
135 * config/i386/waitpkgintrin.h: New file.
136 * config/i386/x86intrin.h: New header.
137 * doc/invoke.texi: Add -mwaitpkg.
139 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
141 PR target/85606
142 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
143 equivalent.
144 (cortex-m0): Use armv6s-m isa.
145 (cortex-m0plus): Likewise.
146 (cortex-m1): Likewise.
147 (cortex-m0.small-multiply): Likewise.
148 (cortex-m0plus.small-multiply): Likewise.
149 (cortex-m1.small-multiply): Likewise.
151 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
152 Jakub Jelinek <jakub@redhat.com>
154 PR tree-optimization/85692
155 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
156 source permute as well.
158 2018-05-11 Martin Liska <mliska@suse.cz>
160 PR sanitizer/85556
161 * doc/extend.texi: Document LLVM style format for no_sanitize
162 attribute.
164 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
166 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
167 mode_supports_vsx_dform_quad to mode_supports_dq_form.
168 (mode_supports_vsx_dform_quad): Likewise.
169 (mode_supports_vmx_dform): Move these functions to be next to the
170 other mode_supports functions.
171 (mode_supports_dq_form): Likewise.
172 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
173 mode_supports_dq_form.
174 (reg_offset_addressing_ok_p): Likewise.
175 (offsettable_ok_by_alignment): Likewise.
176 (rs6000_legitimate_offset_address_p): Likewise.
177 (legitimate_lo_sum_address_p): Likewise.
178 (rs6000_legitimize_address): Likewise.
179 (rs6000_legitimize_reload_address): Likewise.
180 (rs6000_secondary_reload_inner): Likewise.
181 (rs6000_preferred_reload_class): Likewise.
182 (rs6000_output_move_128bit): Likewise.
184 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
186 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
187 Generate SImode target register for null target.
188 <case IX86_BUILTIN_XGETBV>: Ditto.
189 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
190 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
192 2018-05-10 Carl Love <cel@us.ibm.com>
194 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
195 dcbtt and dcbtstt if operands[2] is 0.
197 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
199 PR target/85693
200 * config/i386/sse.md (usadv64qi): New expander.
202 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
204 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
205 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
206 -maltivec=be support.
207 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
208 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
209 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
210 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
211 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
212 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
213 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
214 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
215 altivec_vsumsws): Adjust.
216 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
217 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
218 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
219 support.
220 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
221 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
222 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
223 (altivec_lve<VI_char>x): Delete expand.
224 (*altivec_lve<VI_char>x_internal): Rename to...
225 (altivec_lve<VI_char>x): ... this.
226 (altivec_lvxl_<mode>): Delete expand.
227 (*altivec_lvxl_<mode>_internal): Rename to ...
228 (altivec_lvxl_<mode>): ... this.
229 (altivec_stvxl_<mode>): Delete expand.
230 (*altivec_stvxl_<mode>_internal): Rename to ...
231 (altivec_stvxl_<mode>): ... this.
232 (altivec_stve<VI_char>x): Delete expand.
233 (*altivec_stve<VI_char>x_internal): Rename to ...
234 (altivec_stve<VI_char>x): ... this.
235 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
236 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
237 reduc_plus_scal_<mode>): Adjust.
238 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
239 comment.
240 (rs6000_cpu_cpp_builtins): Adjust.
241 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
242 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
243 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
244 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
245 -maltivec=be support.
246 (rs6000_split_vec_extract_var): Adjust.
247 (rs6000_split_v4si_init): Adjust.
248 (swap_selector_for_mode): Delete.
249 (altivec_expand_lvx_be, altivec_expand_stvx_be,
250 altivec_expand_stvex_be): Delete.
251 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
252 -maltivec=be support.
253 (rs6000_gimple_fold_builtin): Ditto.
254 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
255 Adjust.
256 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
257 (TARGET_DIRECT_MOVE_64BIT): Adjust.
258 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
259 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
260 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
261 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
262 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
263 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
264 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
265 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
266 anonymous split): Adjust.
267 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
268 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
270 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
272 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
273 when --with-gxx-include-dir is also specified.
274 * configure: Regenerate.
276 2018-05-09 Jim Wilson <jimw@sifive.com>
278 PR target/84797
279 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
280 * config/riscv/t-withmultilib: New.
281 * config/riscv/withmultilib.h: New.
282 * doc/install.texi: Document RISC-V --with-multilib-list support.
284 2018-05-09 Richard Biener <rguenther@suse.de>
286 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
287 vector.
288 (vect_bb_vectorization_profitable_p): Adjust. Compute
289 actual scalar cost using the cost vector and the add_stmt_cost
290 machinery.
292 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
294 PR rtl-optimization/85645
295 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
296 in the REG_CFA_REGISTER note for LR, don't leave it empty.
298 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
300 PR rtl-optimization/85645
301 * shrink-wrap.c (spread_components): Return a boolean saying if
302 anything was changed.
303 (try_shrink_wrapping_separate): Iterate spread_components until
304 nothing changes anymore.
306 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
308 PR rtl-optimization/85645
309 * regrename.c (build_def_use): Also kill the chains that include the
310 destination of a REG_CFA_REGISTER note.
312 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
314 PR rtl-optimization/85645
315 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
316 insn that has a REG_CFA_REGISTER note.
318 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
320 * cfgexpand.c (expand_clobber): New function.
321 (expand_gimple_stmt_1): Use it.
322 * tree-vect-stmts.c (vect_clobber_variable): New function,
323 split out from...
324 (vectorizable_simd_clone_call): ...here.
325 (vectorizable_store): Emit a clobber either side of an
326 IFN_STORE_LANES sequence.
327 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
329 2018-05-09 Tom de Vries <tom@codesourcery.com>
331 PR target/85626
332 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
333 (define_insn "trap_if_false"): Add exit after trap.
335 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
337 PR rtl-optimization/85638
338 * bb-reorder.c: Include common/common-target.h.
339 (create_forwarder_block): New function extracted from...
340 (fix_up_crossing_landing_pad): ...here. Rename into...
341 (dw2_fix_up_crossing_landing_pad): ...this.
342 (sjlj_fix_up_crossing_landing_pad): New function.
343 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
344 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
345 from both partitions and exit the loop after one iteration.
347 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
349 Revert:
350 * doc/extend.texi (PowerPC Built-in Functions): Rename this
351 subsection.
352 (Basic PowerPC Built-in Functions): The new name of the
353 subsection previously known as "PowerPC Built-in Functions".
354 (Basic PowerPC Built-in Functions Available on all Configurations):
355 New subsubsection.
356 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
357 subsubsection.
358 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
359 subsubsection.
360 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
361 subsubsection.
362 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
363 subsubsection.
365 2018-05-08 Jim Wilson <jimw@sifive.com>
367 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
368 (LD_EMUL_SUFFIX): New.
369 (LINK_SPEC): Use it.
371 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
373 * doc/extend.texi (PowerPC Built-in Functions): Rename this
374 subsection.
375 (Basic PowerPC Built-in Functions): The new name of the
376 subsection previously known as "PowerPC Built-in Functions".
377 (Basic PowerPC Built-in Functions Available on all Configurations):
378 New subsubsection.
379 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
380 subsubsection.
381 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
382 subsubsection.
383 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
384 subsubsection.
385 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
386 subsubsection.
388 2018-05-08 Jakub Jelinek <jakub@redhat.com>
390 PR target/85683
391 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
392 after cmpelim optimization.
394 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
396 * config.gcc: Support "goldmont".
397 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
398 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
399 PROCESSOR_GOLDMONT.
400 * config/i386/i386.c (m_GOLDMONT): Define.
401 (processor_target_table): Add "goldmont".
402 (PTA_GOLDMONT): Define.
403 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
404 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
405 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
406 (fold_builtin_cpu): Add "goldmont".
407 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
408 (ix86_option_override_internal): Add "goldmont".
409 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
410 (processor_type): Add PROCESSOR_GOLDMONT.
411 * config/i386/i386.md: Add CPU "glm".
412 * config/i386/glm.md: New file.
413 * config/i386/x86-tune.def: Add m_GOLDMONT.
414 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
416 2018-05-08 Jakub Jelinek <jakub@redhat.com>
418 PR target/85572
419 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
420 E_V4DImode.
421 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
422 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
423 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
425 PR target/85317
426 * config/i386/i386.c (ix86_fold_builtin): Handle
427 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
429 PR target/85480
430 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
431 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
433 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
435 PR target/85658
436 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
437 (check_arch): Likewise.
438 (check_fpu): Return the result rather than printing it.
439 (end arch): Fix operator precedence.
440 (end cpu): Likewise.
441 (END): Print the result from check_fpu.
443 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
444 Alan Hayward <alan.hayward@arm.com>
445 David Sherwood <david.sherwood@arm.com>
447 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
448 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
449 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
450 (*fcmuo<mode>_and): New patterns.
452 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
454 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
455 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
456 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
457 (cmp_op, sve_imm_con): New code attributes.
458 (SVE_COND_INT_CMP, imm_con): Delete.
459 (cmp_op): Remove above unspecs from int attribute.
460 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
461 to...
462 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
463 comparison-specific unspecs.
464 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
465 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
466 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
467 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
468 (*vec_fcm<cmp_op><mode>): Rename to...
469 (*fcm<cmp_op><mode>): ...this and adjust likewise.
470 (*vec_fcmuo<mode>): Rename to...
471 (*fcmuo<mode>): ...this and adjust likewise.
472 (*pred_fcm<cmp_op><mode>): New pattern.
473 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
474 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
475 functions.
476 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
477 and UNORDERED.
478 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
479 (aarch64_emit_sve_predicated_cond): New function.
480 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
481 (aarch64_emit_unspec_cond_or): Replace with...
482 (aarch64_emit_sve_or_conds): ...this new function. Use
483 aarch64_emit_sve_ptrue_op for the individual comparisons and
484 aarch64_emit_binop to OR them together.
485 (aarch64_emit_inverted_unspec_cond): Replace with...
486 (aarch64_emit_sve_inverted_cond): ...this new function. Use
487 aarch64_emit_sve_ptrue_op for the comparison and
488 aarch64_emit_unop to invert the result.
489 (aarch64_expand_sve_vec_cmp_float): Update after the above
490 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
492 2018-05-07 Nathan Sidwell <nathan@acm.org>
494 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
495 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
496 (Backwards Compatibility): Likewise.
498 2018-05-07 Luis Machado <luis.machado@linaro.org>
500 PR bootstrap/85681
501 Revert:
502 2018-05-07 Luis Machado <luis.machado@linaro.org>
504 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
505 <prefetch_dynamic_strides>: New const bool field.
506 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
507 prefetch_dynamic_strides.
508 (exynosm1_prefetch_tune): Likewise.
509 (thunderxt88_prefetch_tune): Likewise.
510 (thunderx_prefetch_tune): Likewise.
511 (thunderx2t99_prefetch_tune): Likewise.
512 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
513 (aarch64_override_options_internal): Update to set
514 PARAM_PREFETCH_DYNAMIC_STRIDES.
515 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
516 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
517 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
518 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
519 prefetch-dynamic-strides setting.
521 2018-05-07 Luis Machado <luis.machado@linaro.org>
523 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
524 <minimum_stride>: New const int field.
525 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
526 minimum_stride field.
527 (exynosm1_prefetch_tune): Likewise.
528 (thunderxt88_prefetch_tune): Likewise.
529 (thunderx_prefetch_tune): Likewise.
530 (thunderx2t99_prefetch_tune): Likewise.
531 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
532 (aarch64_override_options_internal): Update to set
533 PARAM_PREFETCH_MINIMUM_STRIDE.
534 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
535 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
536 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
537 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
538 stride is constant and is below the minimum stride threshold.
540 2018-05-07 Luis Machado <luis.machado@linaro.org>
542 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
543 to 512.
545 2018-05-07 Luis Machado <luis.machado@linaro.org>
547 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
548 <prefetch_dynamic_strides>: New const bool field.
549 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
550 prefetch_dynamic_strides.
551 (exynosm1_prefetch_tune): Likewise.
552 (thunderxt88_prefetch_tune): Likewise.
553 (thunderx_prefetch_tune): Likewise.
554 (thunderx2t99_prefetch_tune): Likewise.
555 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
556 (aarch64_override_options_internal): Update to set
557 PARAM_PREFETCH_DYNAMIC_STRIDES.
558 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
559 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
560 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
561 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
562 prefetch-dynamic-strides setting.
564 2018-05-07 Luis Machado <luis.machado@linaro.org>
566 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
567 <minimum_stride>: New const int field.
568 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
569 minimum_stride field.
570 (exynosm1_prefetch_tune): Likewise.
571 (thunderxt88_prefetch_tune): Likewise.
572 (thunderx_prefetch_tune): Likewise.
573 (thunderx2t99_prefetch_tune): Likewise.
574 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
575 (aarch64_override_options_internal): Update to set
576 PARAM_PREFETCH_MINIMUM_STRIDE.
577 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
578 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
579 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
580 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
581 stride is constant and is below the minimum stride threshold.
583 2018-05-06 Jakub Jelinek <jakub@redhat.com>
585 PR c++/85659
586 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
587 the type is addressable. Don't force op into register if it has
588 BLKmode.
590 2018-05-05 Roland McGrath <mcgrathr@google.com>
592 PR other/77609
593 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
594 any section for which we don't know a specific type it should have,
595 regardless of name. Previously this was done only for the exact
596 names ".init_array", ".fini_array", and ".preinit_array".
597 (default_elf_asm_named_section): Add comment about
598 relationship with default_section_type_flags and SECTION_NOTYPE.
599 (get_section): Don't consider it a type conflict if one side has
600 SECTION_NOTYPE and the other doesn't, as long as neither has the
601 SECTION_BSS et al used in the default_section_type_flags logic.
603 2018-05-05 Tom de Vries <tom@codesourcery.com>
605 PR target/85653
606 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
607 (workaround_barsyncs): New function.
608 (nvptx_reorg): Use workaround_barsyncs.
609 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
610 (define_expand "nvptx_membar_cta"): New define_expand.
611 (define_insn "*nvptx_membar_cta"): New insn.
613 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
615 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
616 To improve optimization opportunities.
617 * builtin-types.def: The new needed builtin types for the above.
619 2018-05-04 Richard Biener <rguenther@suse.de>
621 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
622 * gimple-ssa-store-merging.c
623 (imm_store_chain_info::output_merged_store): Remove redundant create,
624 release split_store vector contents on failure.
625 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
626 scalar stmt vector on cache hit.
628 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
630 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
631 Xilinx FP support.
632 * config.gcc (powerpc-xilinx-eabi*): Remove.
633 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
634 support.
635 (fusion_addis_mem_combo_load): Ditto.
636 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
637 FP support.
638 (rs6000_cpu_cpp_builtins): Ditto.
639 * config/rs6000/rs6000-linux.c
640 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
641 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
642 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
643 support.
644 (rs6000_setup_reg_addr_masks): Ditto.
645 (rs6000_init_hard_regno_mode_ok): Ditto.
646 (rs6000_option_override_internal): Ditto.
647 (legitimate_lo_sum_address_p): Ditto.
648 (rs6000_legitimize_address): Ditto.
649 (rs6000_legitimize_reload_address): Ditto.
650 (rs6000_legitimate_address_p): Ditto.
651 (abi_v4_pass_in_fpr): Ditto.
652 (setup_incoming_varargs): Ditto.
653 (rs6000_gimplify_va_arg): Ditto.
654 (rs6000_split_multireg_move): Ditto.
655 (rs6000_savres_strategy): Ditto.
656 (rs6000_emit_prologue_components): Ditto.
657 (rs6000_emit_epilogue_components): Ditto.
658 (rs6000_emit_prologue): Ditto.
659 (rs6000_emit_epilogue): Ditto.
660 (rs6000_elf_file_end): Ditto.
661 (rs6000_function_value): Ditto.
662 (rs6000_libcall_value): Ditto.
663 * config/rs6000/rs6000.h: Ditto.
664 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
665 (TARGET_MINMAX): ... this. New.
666 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
667 * config/rs6000/rs6000.md: Remove Xilinx FP support.
668 (*movsi_internal1_single): Delete.
669 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
670 mfpu=, mxilinx-fpu): Delete.
671 * config/rs6000/singlefp.h: Delete.
672 * config/rs6000/sysv4.h: Remove Xilinx FP support.
673 * config/rs6000/t-rs6000: Ditto.
674 * config/rs6000/t-xilinx: Delete.
675 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
676 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
677 (VStype_simple): Delete.
678 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
679 * config/rs6000/xfpu.h: Delete.
680 * config/rs6000/xfpu.md: Delete.
681 * config/rs6000/xilinx.h: Delete.
682 * config/rs6000/xilinx.opt: Delete.
683 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
684 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
686 2018-05-04 Tom de Vries <tom@codesourcery.com>
688 PR libgomp/85639
689 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
690 if ignore == 0.
692 2018-05-04 Richard Biener <rguenther@suse.de>
694 PR middle-end/85627
695 * tree-complex.c (update_complex_assignment): We are always in SSA form.
696 (expand_complex_div_wide): Likewise.
697 (expand_complex_operations_1): Likewise.
698 (expand_complex_libcall): Preserve EH info of the original stmt.
699 (tree_lower_complex): Handle removed blocks.
700 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
701 on complex multiplication and division libcall builtins.
703 2018-05-04 Richard Biener <rguenther@suse.de>
705 PR middle-end/85574
706 * fold-const.c (negate_expr_p): Restrict negation of operand
707 zero of a division to when we know that can happen without
708 overflow.
709 (fold_negate_expr_1): Likewise.
711 2018-05-04 Jakub Jelinek <jakub@redhat.com>
713 PR libstdc++/85466
714 * real.h (real_nextafter): Declare.
715 * real.c (real_nextafter): New function.
716 * fold-const-call.c (fold_const_nextafter): New function.
717 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
718 CASE_CFN_NEXTTOWARD.
719 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
720 even when arg1_mode is different from arg0_mode.
722 2018-05-03 Nathan Sidwell <nathan@acm.org>
724 * doc/extend.texi (Deprecated Features): Remove
725 -ffriend-injection.
726 (Backwards Compatibility): Likewise.
727 * doc/invoke.texi (C++ Language Options): Likewise.
728 (C++ Dialect Options): Likewise.
730 2018-05-03 Jakub Jelinek <jakub@redhat.com>
732 PR target/85530
733 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
734 _mm512_mask_mullox_epi64): New intrinsics.
736 2018-05-03 Tom de Vries <tom@codesourcery.com>
738 PR testsuite/85106
739 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
740 dump files): Add offload-tree.
742 2018-05-03 Richard Biener <rguenther@suse.de>
744 PR tree-optimization/85615
745 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
746 to loops not nested in BBs loop father to avoid creating multi-entry
747 loops.
749 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
751 PR tree-optimization/70291
752 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
753 arguments. Change return type to tree. Emit libcall as a new
754 statement rather than replacing existing one when inplace_p is true.
755 (expand_complex_multiplication_components): New function.
756 (expand_complex_multiplication): Expand floating-point complex
757 multiplication using the above.
758 (expand_complex_division): Rename inner_type parameter to type.
759 Update expand_complex_libcall call-site.
760 (expand_complex_operations_1): Update expand_complex_multiplication
761 and expand_complex_division call-sites.
763 2018-05-02 Jakub Jelinek <jakub@redhat.com>
765 PR target/85582
766 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
767 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
768 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
769 the highest significant bit of the shift count mask is clear. In
770 check whether and[sq]i3 is needed verify that all significant bits
771 of the shift count other than the highest are set.
773 2018-05-02 Tom de Vries <tom@codesourcery.com>
775 PR libgomp/82428
776 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
777 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
778 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
779 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
780 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
781 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
782 __builtin_goacc_parlevel_size.
784 2018-05-02 Richard Biener <rguenther@suse.de>
786 PR tree-optimization/85597
787 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
788 do not use split vect_get_vec_defs call but call vect_get_slp_defs
789 directly.
791 2018-05-02 Tom de Vries <tom@codesourcery.com>
793 PR testsuite/85106
794 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
795 dump files): Add ltrans-tree.
797 2018-05-02 Tom de Vries <tom@codesourcery.com>
799 PR testsuite/85106
800 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
801 dump files): Add wpa-ipa.
803 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
805 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
806 powerpc*-*-linux*paired* target.
807 * config/rs6000/750cl.h: Delete.
808 * config/rs6000/paired.h: Delete.
809 * config/rs6000/paired.md: Delete.
810 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
811 float support.
812 * config/rs6000/rs6000-builtin.def: Remove paired float support.
813 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
814 comment. Remove paired float support.
815 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
816 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
817 VECTOR_PAIRED.
818 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
819 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
820 declarations.
821 * config/rs6000/rs6000.c: Remove paired float support.
822 (paired_expand_vector_init, paired_expand_vector_move,
823 paired_emit_vector_compare, paired_emit_vector_cond_expr,
824 (paired_expand_lv_builtin, paired_expand_stv_builtin,
825 paired_expand_builtin, paired_expand_predicate_builtin,
826 paired_init_builtins): Delete.
827 * config/rs6000/rs6000.h: Remove paired float support.
828 * config/rs6000/rs6000.md: Remove paired float support.
829 (move_from_CR_ov_bit): Delete.
830 * config/rs6000/rs6000.opt (mpaired): Delete.
831 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
832 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
834 2018-05-02 Richard Biener <rguenther@suse.de>
836 PR middle-end/85567
837 * gimplify.c (gimplify_save_expr): When in SSA form allow
838 SAVE_EXPRs to compute to SSA vars.
840 2018-05-02 Jakub Jelinek <jakub@redhat.com>
842 PR target/85582
843 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
844 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
845 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
846 clobber operands[2], instead use a new pseudo. Formatting fixes.
848 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
850 PR tree-optimization/85586
851 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
852 exit early for statements in the same group if the accesses are
853 not strided.
855 2018-05-02 Tom de Vries <tom@codesourcery.com>
857 PR lto/85451
858 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
859 error message.
861 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
863 PR tree-optimization/85143
864 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
866 2018-05-01 Tom de Vries <tom@codesourcery.com>
868 PR lto/85451
869 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
870 not found" error message.
872 2018-05-01 Tom de Vries <tom@codesourcery.com>
874 PR other/83786
875 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
876 * vec.c (test_ordered_remove_if): New function.
877 (vec_c_tests): Call test_ordered_remove_if.
878 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
879 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
880 * tree-vect-patterns.c (vect_pattern_recog_1): Use
881 VEC_ORDERED_REMOVE_IF.
883 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
885 PR tree-optimization/82665
886 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
887 pointer subtraction where arguments come from a memchr call.
889 2018-05-01 Jakub Jelinek <jakub@redhat.com>
891 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
892 --push-state --as-needed and --pop-state instead of --as-needed and
893 --no-as-needed if ld supports it.
894 * configure: Regenerated.
896 PR web/85578
897 * doc/install.texi2html: Replace _002d with - and _002a with * in
898 generated html files using sed.
900 2018-04-30 David Malcolm <dmalcolm@redhat.com>
902 PR c++/85523
903 * gcc-rich-location.c (blank_line_before_p): New function.
904 (use_new_line): New function.
905 (gcc_rich_location::add_fixit_insert_formatted): New function.
906 * gcc-rich-location.h
907 (gcc_rich_location::add_fixit_insert_formatted): New function.
909 2018-04-30 David Malcolm <dmalcolm@redhat.com>
911 * selftest.c (assert_streq): Rename "expected" and "actual" to
912 "val1" and "val2". Extend NULL-handling to cover both inputs
913 symmetrically, while still requiring both to be non-NULL for a pass.
914 * selftest.h (assert_streq): Rename "expected" and "actual" to
915 "val1" and "val2".
916 (ASSERT_EQ): Likewise.
917 (ASSERT_EQ_AT): Likewise.
918 (ASSERT_KNOWN_EQ): Likewise.
919 (ASSERT_KNOWN_EQ_AT): Likewise.
920 (ASSERT_NE): Likewise.
921 (ASSERT_MAYBE_NE): Likewise.
922 (ASSERT_MAYBE_NE_AT): Likewise.
923 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
924 the assertion to pass.
925 (ASSERT_STREQ_AT): Likewise.
927 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
929 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
930 interaction with -pie.
932 2018-04-30 David Malcolm <dmalcolm@redhat.com>
934 * selftest.h: Fix alphabetization of per-source-file selftest
935 declarations.
937 2018-04-30 Jason Merrill <jason@redhat.com>
939 PR c++/61982 - dead stores to destroyed objects.
940 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
941 of clobber.
943 2018-04-30 Jason Merrill <jason@redhat.com>
945 * tree.c (build_clobber): New.
946 * tree.h: Declare it.
947 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
949 2018-04-30 David Malcolm <dmalcolm@redhat.com>
951 * diagnostic-show-locus.c (layout::layout): Update for
952 location_get_source_line returning a char_span.
953 (struct char_span): Move to input.h.
954 (struct correction): Update for fields in char_span becoming
955 private.
956 (struct source_line): Update for location_get_source_line
957 returning a char_span.
958 (layout::print_line): Likewise.
959 * edit-context.c (edited_file::print_content): Likewise.
960 (edited_file::print_diff_hunk): Likewise.
961 (edited_file::print_run_of_changed_lines): Likewise.
962 (edited_file::get_num_lines): Likewise.
963 (edited_line::edited_line): Likewise.
964 * final.c (asm_show_source): Likewise.
965 * input.c (location_get_source_line): Convert return type
966 from const char * to char_span, losing the final "line_len"
967 param.
968 (dump_location_info): Update for the above.
969 (get_substring_ranges_for_loc): Likewise. Use a char_span
970 when handling the literal within the line.
971 (test_reading_source_line): Update for location_get_source_line
972 returning a char_span.
973 * input.h (class char_span): Move here from
974 diagnostic-show-locus.c, converting from a struct to a class.
975 Make data members private.
976 (char_span::operator bool): New.
977 (char_span::length): New.
978 (char_span::get_buffer): New.
979 (char_span::operator[]): New.
980 (char_span::subspan): Make const.
981 (char_span::xstrdup): New.
982 (location_get_source_line): Convert return type from const char *
983 to char_span, losing the final "line_size" param.
985 2018-04-30 Jan Hubicka <jh@suse.cz>
987 * lto-wrapper.c (ltrans_priorities): New static var.
988 (cmp_priority): New.
989 (run_gcc): Read priorities and if doing parallel build order
990 the Makefile by them.
992 2018-04-30 David Malcolm <dmalcolm@redhat.com>
994 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
996 2018-04-30 Richard Biener <rguenther@suse.de>
998 * tree-cfg.c (verify_address): Remove base argument, add
999 flag whether to check TREE_ADDRESSABLE and do that.
1000 (verify_expr): Remove.
1001 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1002 (verify_gimple_assign_single): Likewise.
1003 (verify_gimple_switch): Likewise.
1004 (verify_expr_location_1): Dereference tp once. Add (disabled)
1005 piece from verify_expr.
1006 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1008 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1010 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1012 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1014 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1015 (small_data_pattern): Likewise.
1016 (arc_rewrite_small_data): Likewise.
1017 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1018 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1019 (get_symbol_alignment): New function.
1020 (legitimate_small_data_address_p): Likewise.
1021 (legitimate_scaled_address): Update, call
1022 legitimate_small_data_address_p.
1023 (output_sdata): New static variable.
1024 (arc_print_operand): Update how we handle small data operands.
1025 (arc_print_operand_address): Likewise.
1026 (arc_legitimate_address_p): Update, use
1027 legitimate_small_data_address_p.
1028 (arc_rewrite_small_data_p): Remove.
1029 (arc_rewrite_small_data_1): Likewise.
1030 (arc_rewrite_small_data): Likewise.
1031 (small_data_pattern): Likewise.
1032 (compact_sda_memory_operand): Update to use
1033 legitimate_small_data_address_p and get_symbol_alignment.
1034 (prepare_move_operands): Don't rewite sdata pattern.
1035 (prepare_extend_operands): Remove.
1036 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1037 pattern.
1038 (zero_extendqisi2): Likewise.
1039 (zero_extendhisi2): Likewise.
1040 (extendqihi2): Likewise.
1041 (extendqisi2): Likewise.
1042 (extendhisi2): Likewise.
1043 (addsi3): Likewise.
1044 (subsi3): Likewise.
1045 (andsi3): Likewise.
1046 * config/arc/constraints.md (Usd): Change it to memory constraint.
1048 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1050 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1051 as source of std instructions.
1052 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1053 allow 6-bit constants as source for store instructions.
1054 (movdi_insn): Update instruction pattern to allow 6-bit constants
1055 as source for store instructions.
1057 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1059 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1061 2018-04-30 Nathan Sidwell <nathan@acm.org>
1062 Sandra Loosemore <sandra@codesourcery.com>
1064 * dumpfile.c (dump_open): Allow '-' for stdout.
1065 * doc/invoke.texi (Developer Options): Document dump filename
1066 determination early. Document stdin/stdout selection.
1068 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1070 Microblaze Target: PIC data text relative
1072 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1073 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1074 Add declaration.
1075 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1076 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1077 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1078 New addressing mode for data-text relative position indepenedent code.
1079 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1080 'ADDRESS_SYMBOLIC_TXT_REL'.
1081 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1082 (microblaze_legitimate_pic_operand): Exclude function calls from
1083 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1084 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1085 addresses cases.
1086 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1087 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1088 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1089 for 'address + offset'.
1090 (microblaze_expand_prologue): Add new function prologue call for
1091 'r20' assignation.
1092 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1093 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1094 table in case of TARGET_PIC_DATA_TEXT_REL.
1095 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1096 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1097 Add new macros 'UNSPEC_TEXT',
1098 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1099 + exclude function calls from 'UNSPEC_PLT' in case of data text
1100 relative mode.
1101 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1102 new target hook for generating address diff vector tables in case of
1103 flag_pic.
1104 * doc/tm.texi : Regenerate.
1105 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1106 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1107 of addr diff vector generation.
1108 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1109 target hook definition.
1110 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1111 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1112 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1113 option for data text relative.
1115 2018-04-30 Richard Biener <rguenther@suse.de>
1117 * tree-chrec.h (evolution_function_is_constant_p): Remove
1118 redundant check.
1119 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1121 2018-04-30 Richard Biener <rguenther@suse.de>
1123 PR bootstrap/85571
1124 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1126 2018-04-30 Richard Biener <rguenther@suse.de>
1128 PR tree-optimization/28364
1129 PR tree-optimization/85275
1130 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1131 copying first exit test.
1133 2018-04-28 Mark Wielaard <mark@klomp.org>
1135 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1136 dwarf_version >= 5.
1137 (dwarf_AT): Handle DW_AT_addr_base.
1138 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1140 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1142 PR target/84431
1143 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1144 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1145 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1146 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1148 2018-04-28 Richard Biener <rguenther@suse.de>
1150 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1151 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1152 to reflect use. Only add interesting stmts.
1154 2018-04-27 Martin Jambor <mjambor@suse.cz>
1156 PR ipa/85549
1157 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1158 the jump function allows for passing through aggregate values.
1160 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1162 * input.h (in_system_header_at): Convert from macro to inline
1163 function.
1164 (from_macro_expansion_at): Likewise.
1165 (from_macro_definition_at): Likewise.
1167 2018-04-27 Jeff Law <law@redhat.com>
1169 * config.gcc: Mark tile* targets as deprecated/obsolete.
1171 2018-04-27 Richard Biener <rguenther@suse.de>
1173 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1174 fix for ILP32.
1176 2018-04-27 Richard Biener <rguenther@suse.de>
1178 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1180 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1182 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1183 with Yd constraint. Set "preferred_for_speed" attribute from
1184 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1185 with Yd constraint.
1186 (*movdi_internal): Ditto.
1187 (movti_interunit splitters): Remove
1188 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1189 (movdi_interunit splitters): Ditto.
1190 * config/i386/constraints.md (Ye): Remove.
1191 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1193 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1195 PR target/85512
1196 * config/aarch64/constraints.md (Usg): Limit to 31.
1197 (Usj): Limit to 63.
1199 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1201 PR tree-optimization/85529
1202 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1203 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1204 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1205 zero extension or masking of the MSB bit.
1206 (optimize_range_tests): Add FIRST_BB argument, pass it through
1207 to optimize_range_tests_var_bound.
1208 (maybe_optimize_range_tests, reassociate_bb): Adjust
1209 optimize_range_tests callers.
1211 2018-04-26 Richard Biener <rguenther@suse.de>
1212 Jakub Jelinek <jakub@redhat.com>
1214 * cgraph.h (symbol_table): Just declare debug method here.
1215 * symtab.c (symbol_table::debug): Define.
1217 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1219 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1221 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1223 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1224 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1225 (*movdi_internal): Substitute Yi and Yj constraint with x
1226 and Ym and Yn constraint with y constraint. Update "isa"
1227 attribute and set "preferred_for_speed" attribute from
1228 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1229 (*movsi_internal): Ditto.
1230 (*movdf_internal): Ditto.
1231 (*movsf_internal): Ditto.
1232 (*zero_extendsidi2): Ditto.
1233 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1234 (sse2_loadld): Ditto.
1235 (*vec_extract<ssevecmodelower>_0): Ditto.
1236 (*vec_extractv4si_0_zext_sse4): Ditto.
1237 (vec_concatv2di): Ditto.
1238 (*vec_dup<mode>): Ditto.
1239 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1240 * config/i386/constraints.md (Yi): Remove.
1241 (Yj): Remove.
1242 (Ym): Remove.
1243 (Yn): Remove.
1245 2018-04-26 Nathan Sidwell <nathan@acm.org>
1247 * dumpfile.c (dump_open): New.
1248 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1249 (dump_finish): Detect stdio/stderr by value not name.
1251 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1253 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1255 2018-04-26 Tom de Vries <tom@codesourcery.com>
1257 PR target/84952
1258 * config/nvptx/nvptx.c (verify_neutering_jumps)
1259 (verify_neutering_labels): New function
1260 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1262 2018-04-26 Tom de Vries <tom@codesourcery.com>
1264 PR target/84025
1265 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1266 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1267 need neutering.
1269 2018-04-26 Richard Biener <rguenther@suse.de>
1270 Tom de Vries <tom@codesourcery.com>
1272 PR lto/85422
1273 * lto-streamer-out.c (output_function): Fixup loops if required to match
1274 discovery done in the reader.
1276 2018-04-26 Richard Biener <rguenther@suse.de>
1278 PR tree-optimization/85116
1279 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1280 have a loop exit from the single latch predecessor. Remove
1281 case of header with just condition.
1282 (ch_base::copy_headers): Exclude infinite loops from any
1283 processing.
1284 (pass_ch::execute): Record exits.
1286 2018-04-26 Richard Biener <rguenther@suse.de>
1288 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1289 prologue cost vector and pass it to vect_get_load_cost.
1290 (vect_get_peeling_costs_all_drs): Likewise.
1291 (vect_peeling_hash_get_lowest_cost): Likewise.
1292 (vect_enhance_data_refs_alignment): Likewise.
1294 2018-04-26 Richard Biener <rguenther@suse.de>
1296 PR middle-end/85450
1297 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1298 checking of integer<->pointer conversions.
1299 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1300 sign-/zero-extending pointer types.
1301 (expand_omp_for_static_chunk): Likewise.
1303 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1304 Jean Lee <xiaoyur347@gmail.com>
1306 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1307 (TARGET_ASAN_SHADOW_OFFSET): Define.
1308 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1309 true for -fsanitize=address.
1311 2018-04-25 Mark Wielaard <mark@klomp.org>
1313 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1314 shorter ones.
1316 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1318 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1319 than "alu", remove explicit "memory" and "imm_disp" attributes.
1320 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1322 PR middle-end/85414
1323 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1324 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1325 gen_lowpart_no_emit.
1327 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1329 PR target/85473
1330 * config/i386/i386.c (ix86_expand_builtin): Change memory
1331 operand to XI, extend p0 to Pmode.
1332 * config/i386/i386.md: Change unspec volatile and operand
1333 1 mode to XI, change operand 0 mode to P.
1335 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1337 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1338 GET_MODE_MASK before any checking.
1339 (nds32_can_use_bset_p): Likewise.
1340 (nds32_can_use_btgl_p): Likewise.
1342 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1344 * config/nds32/nds32-doubleword.md: New define_split pattern for
1345 illegal register number.
1347 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1349 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1351 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1353 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1355 2018-04-25 Richard Biener <rguenther@suse.de>
1357 * lto-streamer.h (LTO_major_version): Bump to 8.
1359 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1361 * BASE-VER: Set to 9.0.0.
1363 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1365 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1366 in __abskf2 and __powikf2.
1368 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1370 PR target/85512
1371 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1372 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1373 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1374 Use the above on operand 2. Reindent.
1375 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1377 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1379 PR target/85485
1380 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1381 handle OPT_mcet.
1382 * config/i386/i386.opt (mcet): Removed.
1383 * doc/install.texi: Remove -mcet documentation.
1384 * doc/invoke.texi: Likewise.
1386 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1388 PR target/85485
1389 * doc/install.texi: Remove -mcet from bootstrap-cet.
1391 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1393 PR target/85511
1394 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1395 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1396 if TARGET_64BIT.
1398 PR target/85503
1399 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1400 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1401 containing a CONST_VECTOR.
1403 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1405 * doc/install.texi: Update newlib dependency for nvptx.
1407 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1409 PR target/85508
1410 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1411 instead of INTVAL when shifting x left.
1413 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1415 PR tree-optimization/85478
1416 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1417 vect_grouped_store_supported for single element vectors.
1419 2018-04-24 Richard Biener <rguenther@suse.de>
1421 PR target/85491
1422 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1423 load cost increase to the case of non-constant step.
1425 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1427 PR target/84828
1428 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1429 destination if any_malformed_asm.
1431 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1433 PR middle-end/85496
1434 * expr.c (store_field): In the bitfield case, if the value comes from
1435 a function call and is returned in registers by means of a PARALLEL,
1436 do not change the mode of the temporary unless BLKmode and VOIDmode.
1438 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1440 PR rtl-optimization/85423
1441 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1442 dependencies to debug insns when the previous insn is non-debug.
1444 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1446 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1447 enums into a single definition.
1448 (fls): Fix predicates and printing.
1449 (seti): Likewise.
1451 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1453 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1454 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1455 and short u6 immediate.
1456 (check_if_valid_sleep_operand): Remove.
1457 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1459 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1461 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1462 flag_always_save_lp condition.
1463 * config/nds32/nds32.opt (malways-save-lp): New option.
1465 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1467 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1468 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1469 * config/nds32/nds32.h
1470 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1471 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1473 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1475 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1476 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1478 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1479 Chung-Ju Wu <jasonwucj@gmail.com>
1481 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1482 Declare.
1483 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1484 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1486 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1488 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1490 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1492 * config/nds32/nds32-protos.h (nds32_data_alignment,
1493 nds32_local_alignment): Declare.
1494 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1495 nds32_local_alignment): New functions.
1496 (TARGET_CONSTANT_ALIGNMENT): Define.
1497 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1499 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1501 * config/nds32/nds32.c
1502 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1503 (TARGET_MODES_TIEABLE_P): Likewise.
1505 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1507 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1508 level Ofast and Og.
1510 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1511 Chung-Ju Wu <jasonwucj@gmail.com>
1513 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1514 for unaligned access.
1515 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1516 unaligned access.
1517 * config/nds32/nds32-intrinsic.md: Likewise.
1518 * config/nds32/nds32_intrinsic.h: Likewise.
1519 * config/nds32/nds32.h (nds32_builtins): Likewise.
1520 * config/nds32/nds32.opt (munaligned-access): New option.
1521 * config/nds32/nds32.c (nds32_asm_file_start): Display
1522 flag_unaligned_access status.
1524 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1526 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1527 -mno-relax is present.
1528 * config/riscv/linux.h (LINK_SPEC): Ditto.
1530 2018-04-20 Martin Sebor <msebor@redhat.com>
1532 PR c/85365
1533 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1534 for null pointers.
1535 (gimple_fold_builtin_stxcpy_chk): Same.
1536 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1538 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1540 PR target/85456
1541 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1542 __powikf2 when long double is IEEE 128-bit.
1544 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1546 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1547 step to make sure stack always aligned.
1549 2018-04-20 Carl Love <cel@us.ibm.com>
1551 PR target/83402
1552 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1553 size check for arg0.
1555 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1556 Tom de Vries <tom@codesourcery.com>
1558 PR target/85445
1559 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1560 Emit insns for calls too.
1561 (nvptx_find_par): Always look for worker-level predecessor insn.
1562 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1563 calls.
1564 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1565 (nvptx_process_pars): Propagate frames for calls.
1567 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1569 PR target/85469
1570 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1571 Removed.
1572 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1573 (ix86_handle_option): Don't handle OPT_mibt.
1574 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1575 __SHSTK__.
1576 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1577 has_ibt and ibt.
1578 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1579 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1580 (ix86_target_macros): Define __CET__ with flag_cf_protection
1581 for -fcf-protection.
1582 * config/i386/i386.c (isa2_opts): Remove -mibt.
1583 * config/i386/i386.h (TARGET_IBT): Removed.
1584 (TARGET_IBT_P): Likewise.
1585 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1586 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1587 * config/i386/i386.opt (mcet): Update help message.
1588 (mshstk): Likewise.
1589 (mibt): Removed.
1590 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1591 -mcet as an alias for -mshstk.
1593 2018-04-20 Richard Biener <rguenther@suse.de>
1595 PR middle-end/85475
1596 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1597 complexity by forcing a single use of the multiply operand.
1599 2018-04-20 Martin Jambor <mjambor@suse.cz>
1601 ipa/85449
1602 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1603 recursion dependency to only apply to non-clones.
1605 2018-04-20 Martin Jambor <mjambor@suse.cz>
1607 ipa/85447
1608 * ipa-cp.c (create_specialized_node): Check that clones of
1609 self-recursive edges exist during IPA-CP.
1611 2018-04-19 Toon Moene <toon@moene.org>
1613 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1614 by -O3.
1616 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1618 PR tree-optimization/85467
1619 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1620 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1621 VECTOR_CST element to type.
1623 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1625 PR target/85397
1626 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1627 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1628 (builtin_longjmp): Likewise.
1629 (save_stack_nonlocal): New pattern.
1630 (restore_stack_nonlocal): Likewise.
1632 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1634 PR target/85404
1635 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1636 Replace ASM_OUTPUT_LABEL with fprintf.
1638 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1640 PR target/85417
1641 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1642 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1643 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1644 define __IBT__ and __SHSTK__ for -fcf-protection.
1645 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1646 TARGET_IBT.
1647 (ix86_trampoline_init): Likewise.
1648 (x86_output_mi_thunk): Likewise.
1649 (ix86_notrack_prefixed_insn_p): Likewise.
1650 (ix86_option_override_internal): Don't disallow -fcf-protection.
1651 * config/i386/i386.md (rdssp<mode>): Also enable for
1652 -fcf-protection.
1653 (incssp<mode>): Likewise.
1654 (nop_endbr): Likewise.
1655 * config/i386/i386.opt (mcet): Change help message to built-in
1656 functions only.
1657 (mibt): Likewise.
1658 (mshstk): Likewise.
1659 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1660 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1661 enable CET built-in functions.
1663 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1665 * common/config/i386/i386-common.c
1666 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1667 OPTION_MASK_ISA_MOVDIRI_UNSET,
1668 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1669 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1670 * config.gcc (movdirintrin.h): New header.
1671 * config/i386/cpuid.h (bit_MOVDIRI,
1672 bit_MOVDIR64B): New bits.
1673 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1674 and -mmvodir64b.
1675 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1676 (VOID, PVOID, PCVOID)): New function types.
1677 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1678 __builtin_ia32_directstoreu_u64,
1679 __builtin_ia32_movdir64b): New builtins.
1680 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1681 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1682 and -mmovdiri.
1683 (ix86_valid_target_attribute_inner_p): Ditto.
1684 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1685 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1686 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1687 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1688 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1689 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1690 (movdiri<mode>, movdir64b_<mode>): New.
1691 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1692 * config/i386/immintrin.h: Include movdirintrin.h.
1693 * config/i386/movdirintrin.h: New file.
1694 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1696 2018-04-19 Richard Biener <rguenther@suse.de>
1698 PR middle-end/85455
1699 * cfg.c (clear_bb_flags): When loop state says we have
1700 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1702 2018-04-19 Richard Biener <rguenther@suse.de>
1704 PR tree-optimization/84737
1705 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1706 copying restrict info.
1707 (vect_setup_realignment): Use it.
1708 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1709 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1710 the first DR to all generated stores.
1711 (vectorizable_load): Likewise for loads.
1713 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1715 PR tree-optimization/85446
1716 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1717 the integral and pointer types to have the same precision.
1719 * doc/install.texi: Document --disable-cet being the default and
1720 --enable-cet=auto.
1722 2018-04-18 Martin Liska <mliska@suse.cz>
1724 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1725 style.
1727 2018-04-18 Martin Liska <mliska@suse.cz>
1729 Revert
1730 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1732 PR ipa/83983
1733 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1734 arguments if they are comparable.
1736 2018-04-18 Martin Liska <mliska@suse.cz>
1738 Revert
1739 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1741 PR lto/84805
1742 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1743 incomplete types.
1745 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1747 PR target/85388
1748 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1749 ENDBR after calling __morestack.
1751 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1753 PR jit/85384
1754 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1755 by using gcc_base_ver to generate a gcc_driver_version, and use
1756 it when generating GCC_DRIVER_NAME.
1757 * configure: Regenerate.
1759 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1761 PR target/81084
1762 * config.gcc: Obsolete powerpc*-*-*spe*.
1764 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1766 PR debug/84637
1767 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1768 (stabstr_D): Change type of unum from unsigned int to
1769 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1770 type.
1772 2018-04-17 Jim Wilson <jimw@sifive.com>
1774 PR 84856
1775 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1776 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1777 Set arg_pointer_offset after using pretend_args_size.
1779 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1781 PR rtl-optimization/85431
1782 * dse.c (record_store): Ignore zero width stores.
1784 PR sanitizer/85230
1785 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1786 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1787 __builtin_stack_restore rather than after it.
1788 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1789 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1790 argument instead of virtual_dynamic_stack_rtx.
1792 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1794 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1795 New prototype.
1796 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1797 Add note to error message to explain internal mapping of overloaded
1798 built-in function name to non-overloaded built-in function name.
1799 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1800 function.
1802 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1804 PR target/85424
1805 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1806 where the inputs overlap with the output.
1808 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1810 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1811 (=v, v) alternative and explicit "memory" attribute.
1812 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1813 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1814 attributes.
1815 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1816 "sselog1" type instead of "sselog".
1817 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1818 "sselog". Remove explicit "memory" attribute.
1819 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1820 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1821 attributes.
1822 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1823 "sselog1" type instead of "sselog". Remove explicit "memory"
1824 attribute.
1825 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1826 use "sselog1" type instead of "sselog". Remove explicit "memory"
1827 attribute.
1828 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1829 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1830 attributes.
1831 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1832 "sselog1" type instead of "sselog". Remove explicit "memory"
1833 attribute.
1834 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1835 use "sselog1" type instead of "sselog". Remove explicit "memory"
1836 attribute.
1838 PR target/85430
1839 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1841 PR middle-end/85414
1842 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1843 on a SUBREG.
1845 2018-04-17 Martin Jambor <mjambor@suse.cz>
1847 PR ipa/85421
1848 * ipa-cp.c (create_specialized_node): Call
1849 expand_all_artificial_thunks if necessary.
1851 2018-04-17 Martin Liska <mliska@suse.cz>
1853 PR lto/85405
1854 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1855 in message, remote space in between '_G' and '('.
1857 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1859 PR target/85281
1860 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1861 avx512f_vmcmp<mode>3<round_saeonly_name>,
1862 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1863 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1864 avx512f_rndscale<mode><round_saeonly_name>,
1865 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1866 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1867 Use %<iptr>2 instead of %2 for -masm=intel.
1868 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1869 avx512f_vcvttss2usi<round_saeonly_name>,
1870 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1871 -masm=intel.
1872 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1873 avx512f_vcvttsd2usi<round_saeonly_name>,
1874 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1875 Use %q1 instead of %1 for -masm=intel.
1876 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1877 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1878 of %3 for -masm=intel.
1879 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1880 -masm=intel.
1881 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1882 -masm=intel.
1883 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1884 -masm=intel.
1885 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1886 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1887 %g1.
1888 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1889 -masm=intel.
1890 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1891 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1892 %g1 and one with %0 and %1.
1893 (avx512er_vmrcp28<mode><round_saeonly_name>,
1894 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1895 %1 for -masm=intel.
1896 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1897 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1898 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1899 of %0 and %{%4%} for -masm=intel.
1900 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1901 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1902 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1903 order of %0 and %{%5%}%{z%} for -masm=intel.
1905 2018-04-17 Jan Hubicka <jh@suse.cz>
1907 PR lto/85405
1908 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1910 2018-04-17 Martin Liska <mliska@suse.cz>
1912 PR ipa/85329
1913 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1914 for target_clone error message. Make default implementation
1915 clone to be a local declaration.
1916 (separate_attrs): Add new argument and check for an empty
1917 string.
1918 (expand_target_clones): Handle it.
1919 (ipa_target_clone): Make redirection just for target_clones
1920 functions.
1922 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1923 Tom de Vries <tom@codesourcery.com>
1925 PR middle-end/84955
1926 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1927 tiled basic blocks without omp continue statements.
1929 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1931 PR target/83660
1932 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1933 vec_extract expression as having side effects to make sure it gets
1934 a cleanup point.
1936 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1938 PR target/85403
1939 * config/i386/i386.c (get_builtin_code_for_version): Check
1940 error_mark_node.
1942 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1944 PR target/84331
1945 * gcc/config.gcc: Support "skylake".
1946 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1947 PROCESSOR_SKYLAKE.
1948 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1949 (processor_target_table): Add "skylake".
1950 (ix86_option_override_internal): Add "skylake".
1951 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1952 PROCESSOR_CANNONLAKE.
1953 (get_builtin_code_for_version): Fix priority for
1954 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1955 PROCESSOR_SKYLAKE-AVX512.
1956 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1957 (processor_type): Add PROCESSOR_SKYLAKE.
1959 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1960 Jason Merrill <jason@redhat.com>
1962 PR c++/85112
1963 * convert.c (convert_to_integer_1): Use direct recursion for
1964 enumeral types and types with a precision less than the number
1965 of bits in their mode.
1967 2018-04-16 Julia Koval <julia.koval@intel.com>
1969 PR target/84413
1970 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1971 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1973 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1975 PR target/85293
1976 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1977 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1978 and -mno-direct-move.
1980 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1982 PR target/83402
1983 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1984 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1985 Ensure negative shifts result in {0}.
1987 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1989 PR rtl-optimization/79916
1990 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1991 regs (if any) to define how to gnerate SD moves when LRA is in
1992 progress.
1994 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1996 PR rtl-optimization/85393
1997 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1998 * except.c (expand_dw2_landing_pad_for_region): Make static.
1999 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2000 a label and unconditional jump to old_bb, rather than
2001 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2002 basic block.
2004 PR rtl-optimization/85376
2005 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2006 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2007 instead of a specific value.
2009 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2010 Bin Cheng <bin.cheng@arm.com>
2012 PR tree-optimization/82965
2013 PR tree-optimization/83991
2014 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2015 by_profile_only parameter.
2016 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2017 information if the loop was predicted to iterate too many times.
2018 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2020 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2022 PR lto/71991
2023 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2024 always inline.
2026 2018-04-13 Martin Liska <mliska@suse.cz>
2027 Jakub Jelinek <jakub@redhat.com>
2029 PR middle-end/81657
2030 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2031 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2032 * builtins.c (expand_builtin_memory_copy_args): Use
2033 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2034 handle dest_addr == pc_rtx.
2036 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2038 PR target/85291
2039 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2040 asked to not generate direct moves.
2041 (fix_trunc<mode>si2_stfiwx): Similar.
2042 (fix_trunc<mode>si2_internal): Similar.
2044 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2046 PR debug/83157
2047 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2048 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2049 lookup if dest in some wider mode is known to be const0_rtx and
2050 if so, record permanent equivalence for it to be ZERO_EXTEND of
2051 the narrower mode destination.
2053 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2055 * lto-streamer-out.c (output_function): Revert 259346.
2056 * omp-expand.c (expand_oacc_for): Likewise.
2058 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2060 PR rtl-optimization/85354
2061 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2062 * sel-sched.c (sel_global_init): ... here.
2064 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2066 PR target/85238
2067 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2068 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2069 mode for PE-COFF targets.
2070 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2071 (i386_pe_asm_lto_end): Likewise.
2072 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2073 (TARGET_ASM_LTO_END): Likewise.
2074 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2075 (i386_pe_asm_lto_start): New function.
2076 (i386_pe_asm_lto_end): Likewise.
2078 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2079 Richard Biener <rguenther@suse.de>
2081 PR middle-end/84955
2082 * lto-streamer-out.c (output_function): Fix CFG loop state before
2083 streaming out.
2084 * omp-expand.c (expand_oacc_for): Handle calls to internal
2085 functions like regular functions.
2087 2018-04-12 Richard Biener <rguenther@suse.de>
2089 PR lto/85371
2090 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2091 for the early LTO debug to properly generate references to it
2092 during DIE emission. Do not re-use that for the skeleton for
2093 split-dwarf.
2094 (dwarf2out_early_finish): Likewise.
2096 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2098 PR target/85328
2099 * config/i386/sse.md
2100 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2101 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2102 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2103 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2104 and output is a reg, avoid creating invalid lowpart subreg, but
2105 instead split into a 512-bit move. Don't split if not AVX512VL,
2106 input is xmm16+ reg and output is a mem.
2107 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2108 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2109 xmm16+ reg and output is a mem.
2111 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2113 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2114 also for flag_dwarf2_cfi_asm.
2116 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2118 PR rtl-optimization/85342
2119 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2120 a bool scalar var inside of the loop instead. Don't try to update
2121 recog_data.operand after failed apply_change_group.
2123 2018-04-12 Tom de Vries <tom@codesourcery.com>
2125 PR target/85296
2126 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2127 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2128 array with flexible array member as array without given dimension.
2129 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2130 argument for undefined param to true.
2132 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2134 PR target/85321
2135 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2136 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2137 from PowerPC section.
2138 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2139 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2140 help text that is too long.
2141 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2142 help text that is too long.
2143 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2144 help text that is too long.
2146 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2148 * config/alpha/alpha.md (stack_probe_internal): Rename
2149 from "probe_stack". Update all callers.
2151 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2153 PR rtl-optimization/84566
2154 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2155 sched_macro_fuse_insns.
2157 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2159 PR target/84301
2160 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2161 (compute_block_dependences): ... from here.
2163 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2165 PR tree-optimization/85331
2166 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2167 from int to HOST_WIDE_INT.
2169 2018-04-11 Martin Jambor <mjambor@suse.cz>
2171 PR ipa/84149
2172 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2173 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2174 not the same as the source val.
2175 (cgraph_edge_brings_value_p): New parameter.
2176 (gather_edges_for_value): Pass destination value to
2177 cgraph_edge_brings_value_p.
2178 (perhaps_add_new_callers): Likewise.
2179 (get_info_about_necessary_edges): Likewise and exclude values brought
2180 only by self-recursive edges.
2181 (create_specialized_node): Redirect only clones of self-calling edges.
2182 (+self_recursive_pass_through_p): New function.
2183 (find_more_scalar_values_for_callers_subset): Use it.
2184 (find_aggregate_values_for_callers_subset): Likewise.
2185 (known_aggs_to_agg_replacement_list): Removed.
2186 (decide_whether_version_node): Re-calculate known constants for all
2187 remaining context clones.
2189 2018-04-11 Richard Biener <rguenther@suse.de>
2191 PR lto/85339
2192 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2193 from early DWARF output.
2194 (dwarf2out_early_finish): Output line info unconditionally into
2195 early DWARF and add reference to it.
2197 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2199 PR target/85281
2200 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2201 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2202 other than V2DFmode using iptr mode attribute.
2203 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2205 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2207 PR rtl-optimization/84659
2208 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2210 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2212 PR debug/85302
2213 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2214 SIZEP is NULL.
2215 (output_loc_list): Pass address of a dummy size variable even in the
2216 locview handling loop.
2217 (index_location_lists): Add comment on why skip_loc_list_entry can't
2218 call size_of_locs.
2220 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2222 PR target/85261
2223 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2224 into register.
2226 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2228 PR target/85321
2229 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2230 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2231 and -mstring-compare-inline-limit.
2233 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2235 PR target/85287
2236 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2237 for stack clash protection in a register whenever we need it to be in
2238 a register.
2240 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2242 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2243 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2245 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2247 PR target/85321
2248 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2249 the help text.
2250 (mlong-double-): Ditto.
2251 * config/rs6000/sysv4.opt (msdata=): Ditto.
2252 (mtls-size=): Ditto.
2254 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2256 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2257 erroneous entries for
2258 "vector int vec_ldl (int, long int *)", and
2259 "vector unsigned int vec_ldl (int, unsigned long int *)".
2260 Add comments and entries for
2261 "vector bool char vec_ldl (int, bool char *)",
2262 "vector bool short vec_ldl (int, bool short *)",
2263 "vector bool int vec_ldl (int, bool int *)",
2264 "vector bool long long vec_ldl (int, bool long long *)",
2265 "vector pixel vec_ldl (int, pixel *)",
2266 "vector long long vec_ldl (int, long long *)",
2267 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2268 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2269 type tree bool_long_long_type_node and correct definition of
2270 bool_V2DI_type_node to make reference to this new type tree.
2271 (rs6000_mangle_type): Replace erroneous reference to
2272 bool_long_type_node with bool_long_long_type_node.
2273 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2274 comments to emphasize sign distinctions for char and int types and
2275 replace RS6000_BTI_bool_long constant with
2276 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2277 use of RS6000_BTI_pixel.
2278 (bool_long_type_node): Remove this macro definition.
2279 (bool_long_long_type_node): New macro definition
2281 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2283 PR rtl-optimization/85300
2284 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2285 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2286 simplify_unary_operation fails.
2288 2018-04-10 Martin Liska <mliska@suse.cz>
2290 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2291 cgraph_edge and ipa_ref.
2293 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2295 PR target/85177
2296 PR target/85255
2297 * config/i386/sse.md
2298 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2299 computation of the VEC_MERGE selector from mask.
2300 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2301 Fix decoding of the VEC_MERGE selector into mask.
2303 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2305 PR tree-optimization/85286
2306 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2308 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2310 * final.c (final_1): Set insn_last_address as well as
2311 insn_current_address.
2313 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2315 PR target/85173
2316 * explow.c (emit_stack_probe): Call validize_mem on memory location
2317 before passing it to gen_probe_stack. Create address operand and
2318 legitimize it for the probe_stack_address case.
2320 2018-04-09 Jan Hubicka <jh@suse.cz>
2322 PR lto/85078
2323 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2324 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2325 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2326 walk basetypes.
2327 (free_lang_data): Rebuild type inheritance graph.
2329 2018-04-09 Martin Sebor <msebor@redhat.com>
2331 * invoke.texi (-finline-small-functions): Mention other optimization
2332 options.
2333 (-findirect-inlining, -fpartial-inlining): Same.
2334 (-finline-functions-called-once): Same.
2335 (-freorder-blocks-and-partition): Same.
2337 2018-04-09 Jan Hubicka <jh@suse.cz>
2339 PR rtl/84058
2340 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2341 jumps; choose last target that matches the criteria (i.e.
2342 no partition changes for non-crossing jumps).
2343 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2344 support for redirecting crossing jumps to non-crossing.
2346 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2348 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2349 also for naked functions.
2351 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2353 * config/arc/arc.md (add_shift): New pattern.
2354 (add_shift2): Likewise.
2355 (sub_shift): Likewise.
2356 (sub_shift_cmp0_noout): Likewise.
2357 (compare_si_ashiftsi): Likewise.
2358 (xbfu_cmp0_noout): New combine pattern.
2359 (xbfu_cmp0"): Likewise.
2360 (movsi_set_cc_insn): Place the predicable variant first.
2361 (commutative_binary_cmp0_noout): Remove clobber.
2362 (commutative_binary_cmp0): New pattern.
2363 (noncommutative_binary_cmp0): Likewise.
2364 (noncommutative_binary_cmp0_noout): Likewise.
2365 (noncommutative_binary_comparison_result_used): Removed.
2366 (rsub_cmp0): New pattern.
2367 (rsub_cmp0_noout): Likewise.
2368 (extzvsi): Changed, keep only meaningful variants.
2369 (SQH, SEZ): New iterators.
2370 (SQH_postfix): New mode attribute.
2371 (SEZ_prefix): New code attribute.
2372 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2373 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2374 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2375 of numerical value.
2376 (noncommutative_operator): Check the availability of barrel
2377 shifter option.
2379 2018-04-09 Richard Biener <rguenther@suse.de>
2381 PR tree-optimization/85284
2382 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2383 Only use the niter constraining form of simple_iv when the exit
2384 is always executed.
2386 2018-04-09 Tom de Vries <tom@codesourcery.com>
2388 PR target/84041
2389 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2390 (define_expand "*memory_barrier"): New define_expand.
2391 (define_insn "memory_barrier"): New insn.
2393 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2395 PR rtl-optimization/80463
2396 PR rtl-optimization/83972
2397 PR rtl-optimization/83480
2399 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2400 correct producer for the insn.
2401 (tidy_control_flow): Fixup seqnos in case of debug insns.
2403 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2405 PR rtl-optimization/83913
2407 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2408 different sched-times when merging exprs.
2410 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2412 PR rtl-optimization/83962
2414 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2415 tidy_fallthru_edge and tidy_control_flow.
2417 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2419 PR rtl-optimization/83530
2421 * sel-sched.c (force_next_insn): New global variable.
2422 (remove_insn_for_debug): When force_next_insn is true, also leave only
2423 next insn in the ready list.
2424 (sel_sched_region): When the region wasn't scheduled, make another pass
2425 over it with force_next_insn set to 1.
2427 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2429 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2430 into tm_file.
2431 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2432 for interrupt control.
2433 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2434 functions for interrupt control.
2435 * config/nds32/nds32-intrinsic.md: Likewise.
2436 * config/nds32/nds32_intrinsic.h: Likewise.
2437 * config/nds32/nds32.h (nds32_builtins): Likewise.
2439 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2441 * config/nds32/nds32.c (nds32_init_machine_status,
2442 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2443 strict_aligned_p field.
2444 (nds32_expand_to_rtl_hook): New function.
2445 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2446 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2448 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2449 Chung-Ju Wu <jasonwucj@gmail.com>
2451 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2452 * config/nds32/nds32-n7.md: New file.
2453 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2454 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2455 pipeline.
2456 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2457 * config/nds32/nds32.md (pipeline_model): Add n7.
2458 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2459 * config/nds32/pipelines.md: Include n7 settings.
2461 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2462 Chung-Ju Wu <jasonwucj@gmail.com>
2464 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2465 * config/nds32/nds32-e8.md: New file.
2466 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2467 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2468 pipeline.
2469 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2470 * config/nds32/nds32.md (pipeline_model): Add e8.
2471 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2472 * config/nds32/pipelines.md: Include e8 settings.
2474 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2475 Chung-Ju Wu <jasonwucj@gmail.com>
2477 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2478 * config/nds32/nds32-n8.md: New file.
2479 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2480 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2481 pipeline.
2482 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2483 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2484 * config/nds32/nds32.md (pipeline_model): Add n8.
2485 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2486 * config/nds32/pipelines.md: Include n8 settings.
2488 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2489 Chung-Ju Wu <jasonwucj@gmail.com>
2491 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2492 * config/nds32/nds32-n9-2r1w.md: New file.
2493 * config/nds32/nds32-n9-3r2w.md: New file.
2494 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2495 nds32_register_ports): New or modify for cpu n9.
2496 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2497 pipeline.
2498 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2499 * config/nds32/nds32-utils.c: New file.
2500 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2501 TARGET_MUL_SLOW): Define.
2502 * config/nds32/nds32.md (pipeline_model): New attribute.
2503 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2504 New options that support cpu n9.
2505 * config/nds32/pipelines.md: Include n9 settings.
2506 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2508 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2510 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2511 information if necessary.
2512 (output_cond_branch_compare_zero): Likewise.
2513 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2514 (nds32_target_alignment): Refine for alignment.
2515 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2516 (FUNCTION_BOUNDARY): Modify.
2517 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2518 align case.
2519 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2521 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2523 * config/nds32/constants.md (unspec_volatile_element): Add values for
2524 TLB operation and data prefetch.
2525 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2526 functions for TLB operation and data prefetch.
2527 * config/nds32/nds32-intrinsic.md: Likewise.
2528 * config/nds32/nds32_intrinsic.h: Likewise.
2529 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2530 (nds32_print_operand): Likewise.
2531 * config/nds32/nds32.h (nds32_builtins): Likewise.
2533 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2534 Andrew Pinski <pinsika@gcc.gnu.org>
2536 PR middle-end/82976
2537 * match.pd: Use constant_boolean_node of correct type instead of
2538 boolean_true_node or boolean_false_node for simplifying
2539 pointer comparisons to zero.
2541 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2543 PR tree-optimization/80021
2544 * tree.c (verify_type_variant): Make error call in verify_variant_match
2545 translatable and remove final full stop.
2547 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2549 * config/nds32/constants.md (unspec_volatile_element): Add
2550 UNSPEC_VOLATILE_EH_RETURN.
2551 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2552 nds32_output_stack_pop): Support dwarf exception handling process.
2553 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2554 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2555 exception handling process.
2556 (nds32_compute_stack_frame): Likewise.
2557 (nds32_return_addr_rtx): Likewise.
2558 (nds32_initial_elimination_offset): Likewise.
2559 (nds32_expand_prologue): Likewise.
2560 (nds32_expand_epilogue): Likewise.
2561 (nds32_dynamic_chain_address): New function.
2562 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2563 exception handling.
2564 (DYNAMIC_CHAIN_ADDRESS): Define.
2565 (EH_RETURN_DATA_REGNO): Define.
2566 (EH_RETURN_STACKADJ_RTX): Define.
2567 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2568 patterns for dwarf exception handling.
2570 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2572 * config/nds32/nds32.h: Clean up obsolete macros.
2574 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2576 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2577 Add enum values for particular instructions.
2578 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2579 particular intrinsic functions.
2580 * config/nds32/nds32-intrinsic.md: Likewise.
2581 * config/nds32/nds32_intrinsic.h: Likewise.
2582 * config/nds32/nds32.h (nds32_builtins): Likewise.
2583 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2584 (btst, ave): New patterns for particular instructions.
2586 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2588 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2589 Add enum values for atomic load/store and memory sync.
2590 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2591 and memory sync.
2592 * config/nds32/nds32-intrinsic.md: Likewise.
2593 * config/nds32/nds32_intrinsic.h: Likewise.
2594 * config/nds32/nds32.h (nds32_builtins): Likewise.
2596 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2598 PR tree-optimization/85257
2599 * fold-const.c (native_encode_vector): If not all elts could fit
2600 and off is -1, return 0 rather than offset.
2601 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2602 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2603 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2604 adjust buffer in native_interpret_expr call.
2606 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2608 * config/nds32/constants.md (unspec_volatile_element): Add cache
2609 control enum values.
2610 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2611 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2612 * config/nds32/nds32.c (nds32_cctl_names): New.
2613 (nds32_print_operand): Handle cache control register names.
2614 * config/nds32/nds32.h (nds32_builtins): New enum values.
2615 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2616 macros.
2617 * config/nds32/nds32.md (type): Add mmu.
2618 * config/nds32/pipelines.md (simple_insn): Add mmu.
2620 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2622 * config/nds32/nds32.md (type): Remove call.
2623 * config/nds32/pipelines.md (simple_insn): Likewise.
2625 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2627 * config/nds32/constants.md (unspec_volatile_element): Add
2628 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2629 UNSPEC_VOLATILE_FMFCFG.
2630 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2631 description for fmfcfg and fmfcsr.
2632 (bdesc_1arg): Add fmtcsr.
2633 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2634 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2635 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2636 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2637 unspec_fmfcfg): New patterns.
2638 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2639 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2640 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2641 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2642 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2643 __nds32__fmfcfg): Define.
2645 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2647 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2648 intrinsic register names.
2649 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2650 intrinsic register enum values and macros.
2652 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2654 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2655 for load/store addressing form.
2656 (nds32_print_operand_address): Likewise.
2658 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2660 PR target/85196
2661 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2662 based on LABEL_REF. Remove useless assertion.
2663 (pic_address_needs_scratch): Fix formatting.
2664 (sparc_legitimize_pic_address): Minor tweaks.
2665 (sparc_delegitimize_address): Adjust assertion accordingly.
2666 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2667 into symbolic_operand.
2668 (movsi_high_pic_label_ref): Likewise.
2669 (movsi_lo_sum_pic_label_ref): Likewise.
2670 (movdi_pic_label_ref): Likewise.
2671 (movdi_high_pic_label_ref): Likewise.
2672 (movdi_lo_sum_pic_label_ref): Likewise.
2674 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2676 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2677 custom LIB_SPEC setup.
2679 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2680 Kito Cheng <kito.cheng@gmail.com>
2682 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2683 * config/riscv/freebsd.h: New.
2685 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2687 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2688 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2689 file.
2691 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2692 Kito Cheng <kito.cheng@gmail.com>
2694 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2695 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2696 * config/nds32/nds32-protos.h (nds32_output_call,
2697 nds32_output_return): Declare.
2698 * config/nds32/nds32.md: Refine all the call and return patterns.
2700 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2702 PR debug/85252
2703 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2704 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2706 PR rtl-optimization/84872
2707 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2708 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2709 EDGE_CROSSING edge.
2711 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2713 * expr.c (copy_blkmode_to_reg): Revert 254862.
2714 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2716 2018-04-06 Richard Biener <rguenther@suse.de>
2718 PR middle-end/85244
2719 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2720 after seeing a component reference with an adjacent field. Treat
2721 refs to arrays at struct end of external decls similar to
2722 refs to unconstrained commons.
2724 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2726 PR sanitizer/85213
2727 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2728 look through SAVE_EXPRs with non-side-effects argument. Adjust
2729 recursive calls.
2730 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2731 save_p here.
2733 2018-04-06 Richard Biener <rguenther@suse.de>
2735 PR middle-end/85180
2736 * alias.c (find_base_term): New wrapper around find_base_term
2737 unwinding CSELIB_VAL_PTR changes.
2738 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2739 recursion.
2741 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2743 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2744 instructions.
2745 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2746 constant definitions.
2747 ("nop"): lr 0,0 -> nopr r0
2748 ("nop_lr0", "nop_lr1"): New insn definitions.
2750 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2752 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2753 NDS32_V3PUSH_AVAILABLE_P macro.
2755 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2756 Chung-Ju Wu <jasonwucj@gmail.com>
2758 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2759 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2760 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2761 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2762 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2763 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2764 * config/nds32/constraints.md: New constraints and checking for hard
2765 float configuration.
2766 * config/nds32/iterators.md: New mode iterator and attribute for hard
2767 float configuration.
2768 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2769 patterns.
2770 * config/nds32/nds32-fpu.md: New file.
2771 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2772 deal with hard float code generation.
2773 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2774 ARCH_V3S.
2775 (abi_type, float_reg_number): New enum type.
2776 * config/nds32/nds32-predicates.c: New predicates for hard float.
2777 * config/nds32/nds32-protos.h: Declare functions for hard float.
2778 * config/nds32/nds32.c: Implementation for hard float configuration.
2779 * config/nds32/nds32.h: Definitions for hard float configuration.
2780 * config/nds32/nds32.md: Include hard float machine description and
2781 modify patterns for hard float configuration.
2782 * config/nds32/nds32.opt: New options for hard float configuration.
2783 * config/nds32/predicates.md: New predicates for hard float
2784 configuration.
2786 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2788 * common/config/nds32/nds32-common.c
2789 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2791 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2793 PR middle-end/85195
2794 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2795 CONSTRUCTOR_ELT (ctor, ...)->value.
2797 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2799 PR target/85193
2800 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2802 2018-04-05 Tom de Vries <tom@codesourcery.com>
2804 PR target/85204
2805 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2806 cond jump.
2808 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2809 Kito Cheng <kito.cheng@gmail.com>
2811 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2812 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2813 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2814 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2816 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2817 Kito Cheng <kito.cheng@gmail.com>
2819 * config/nds32/constraints.md (Ufe): New memory constraint.
2820 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2821 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2822 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2823 operands.
2824 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2825 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2827 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2829 * config/nds32/nds32.md: Use optimize_size in the condition for
2830 alu-shift instructions.
2832 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2834 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2836 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2838 * config/nds32/nds32.md (negsi2): Refine pattern.
2840 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2841 Chung-Ju Wu <jasonwucj@gmail.com>
2843 * config/nds32/iterators.md (shift_rotate): New code iterator.
2844 (shift): New code attribute.
2845 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2846 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2847 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2848 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2849 bit-wise operations.
2850 (andsi3, *andsi3): Ditto.
2851 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2852 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2853 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2854 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2855 nds32_ior_operand, nds32_xor_operand): New predicates.
2857 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2859 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2860 (addsi3, subsi3): ... this.
2862 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2864 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2866 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2868 * config/nds32/nds32.md: Adjust indention.
2870 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2872 * config/nds32/nds32.md (feature): New attribute.
2874 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2876 * config/nds32/nds32.md (subtype): New attribute.
2878 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2880 PR target/85203
2881 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2882 expansion to perform a bitwise AND of the argument followed by a
2883 boolean negation of the result.
2885 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2887 PR rtl-optimization/84878
2888 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2889 the basic block. Assert the use reference is not artificial and that
2890 it has an associated insn.
2892 2018-04-04 Michael Matz <matz@suse.de>
2894 * builtins.c (compute_objsize): Pass correct operand
2895 to array_at_struct_end_p.
2897 2018-04-04 Richard Biener <rguenther@suse.de>
2899 PR lto/85176
2900 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2901 from contexts for DINFO_LEVEL_TERSE and below.
2903 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2905 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2906 resiter_operand condition.
2907 * config/nds32/nds32.md (*move<mode>): Ditto.
2909 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2910 Monk Chiang <sh.chiang04@gmail.com>
2912 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2914 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2916 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2918 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2919 Kito Cheng <kito.cheng@gmail.com>
2921 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2922 nds32_cond_code_str, output_cond_branch,
2923 output_cond_branch_compare_zero, nds32_expand_cbranch,
2924 nds32_expand_cstore, nds32_expand_movcc,
2925 nds32_output_cbranchsi4_equality_zero,
2926 nds32_output_cbranchsi4_equality_reg,
2927 nds32_output_cbranchsi4_equality_reg_or_const_int,
2928 nds32_output_cbranchsi4_greater_less_zero: New functions.
2929 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2930 nds32_expand_cstore, nds32_expand_movcc,
2931 nds32_output_cbranchsi4_equality_zero,
2932 nds32_output_cbranchsi4_equality_reg,
2933 nds32_output_cbranchsi4_equality_reg_or_const_int,
2934 nds32_output_cbranchsi4_greater_less_zero): Declare.
2935 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2936 nds32_rimm11s_operand): New predicates.
2937 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2938 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2939 patterns.
2941 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2943 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2944 * config/nds32/nds32.md: Ditto.
2945 * config/nds32/pipelines.md: Ditto.
2947 2018-04-04 Richard Biener <rguenther@suse.de>
2949 PR tree-optimization/85168
2950 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2951 propagating abnormals.
2953 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2955 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2957 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2958 Kito Cheng <kito.cheng@gmail.com>
2960 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2961 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2962 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2963 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2964 * config/nds32/nds32.md (sibcall_internal): New.
2965 (sibcall_register): Remove.
2966 (sibcall_immediate): Remove.
2967 (sibcall_value_internal): New.
2968 (sibcall_value_register): Remove.
2969 (sibcall_value_immediate): Remove.
2970 * config/nds32/predicates.md (nds32_general_register_operand): New.
2971 (nds32_call_address_operand): New.
2973 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2975 PR rtl-optimization/85167
2976 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2977 bb_defs if *split_p, instead preinitialize it to NULL.
2979 PR tree-optimization/85156
2980 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2981 evaluating the argument multiple times.
2983 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2985 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2986 than vector.
2987 (_mm_cvtpd_ps): Likewise.
2988 (_mm_cvttpd_epi32): Likewise.
2989 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2990 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2991 vector, pixel, and bool following altivec.h include.
2993 2018-04-03 Martin Sebor <msebor@redhat.com>
2995 * doc/extend.texi (Common Function Attributes): Clarify.
2996 (const attribute): Likewise.
2997 (pure attribute): Likewise.
2999 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3001 PR target/85169
3002 * config/i386/i386.c (ix86_expand_vector_set): Use
3003 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3005 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3007 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3008 instructions when changing rounding bits to preserve precision bits
3009 in the x87 control word.
3011 2018-04-03 Martin Liska <mliska@suse.cz>
3013 PR tree-optimization/82491
3014 * rtl.h (strip_offset_and_add): Replace += suboffset with
3015 poly_uint64 () + suboffset.
3017 2018-03-29 Martin Liska <mliska@suse.cz>
3018 Martin Jambor <mjambor@suse.cz>
3020 PR ipa/84947
3021 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3022 param_type is not an integral or pointer type.
3024 2018-04-03 Richard Biener <rguenther@suse.de>
3026 * sese.h (recompute_all_dominators): Remove.
3028 2018-04-02 Martin Sebor <msebor@redhat.com>
3030 * doc/invoke.texi (-Wrestrict): Fix typos.
3032 2018-04-02 Jim Wilson <jimw@sifive.com>
3034 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3035 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3036 (<optab>di3, <optab>si3_extend): Likewise.
3037 (<optab>si3_mask, <optab>si3_mask_1): New.
3038 (<optab>di3_mask, <optab>di3_mask_1): New.
3039 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3040 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3041 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3043 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3045 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3046 example.
3048 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3050 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3051 (nds32_canonicalize_comparison): New function.
3053 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3054 Kito Cheng <kito.cheng@gmail.com>
3055 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3057 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3058 * config/nds32/constants.md (unspec_volatile_element): Add
3059 UNSPEC_VOLATILE_RELAX_GROUP.
3060 * config/nds32/nds32-relax-opt.c: New file.
3061 * config/nds32/nds32-predicates.c
3062 (nds32_symbol_load_store_p): New function.
3063 * config/nds32/nds32-protos.h
3064 (nds32_symbol_load_store_p): Declare function.
3065 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3066 * config/nds32/nds32.c
3067 (nds32_register_pass): New function to register pass.
3068 (nds32_register_passes): New function to register passes.
3069 * config/nds32/nds32.md (relax_group): New pattern.
3070 * config/nds32/nds32.opt (mrelax-hint): New option.
3071 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3073 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3075 * config/nds32/t-nds32: Modify files dependency.
3077 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3079 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3080 (PROFILE_HOOK): Define its implementation.
3082 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3084 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3085 type and 32-bit size.
3087 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3089 PR middle-end/85090
3090 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3091 (V_128_256): New mode iterator.
3092 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3093 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3094 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3095 of V.
3096 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3097 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3099 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3101 PR target/83315
3102 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3103 NaN inputs correctly.
3105 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3107 PR target/80546
3108 * config/rs6000/vsx.md (??r): New mode attribute.
3109 (*vsx_mov<mode>_64bit): Use it.
3110 (*vsx_mov<mode>_32bit): Likewise.
3112 2018-03-30 Martin Sebor <msebor@redhat.com>
3114 PR tree-optimization/84818
3115 * builtins.c (check_access): Use warning_n.
3117 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3119 PR target/83822
3120 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3121 condition.
3122 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3123 condition.
3125 2018-03-30 Julia Koval <julia.koval@intel.com>
3127 PR target/84413
3128 * x86-tune.def (movx, partial_reg_dependency): Enable for
3129 m_SKYLAKE_AVX512.
3131 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3133 PR inline-asm/84985
3134 * lra-constraints.c (process_alt_operands): Move setting
3135 this_alternative_matches below.
3137 2018-03-29 Martin Liska <mliska@suse.cz>
3139 PR lto/84995.
3140 * doc/invoke.texi: Document how LTO works with debug info.
3141 Describe auto-load support of binutils. Mention 'x86-64'
3142 as valid option value of -march option.
3144 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3146 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3148 PR c/85094
3149 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3150 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3151 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3152 checking.
3154 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3156 PR target/84912
3157 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3158 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3159 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3160 for RS6000_BTM_POWERPC64.
3161 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3162 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3163 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3164 definition.
3165 (DIVDE): Use it.
3166 (DIVDEU): Likewise.
3168 2018-03-28 Carl Love <cel@us.ibm.com>
3170 Revert
3171 2017-09-27 Carl Love <cel@us.ibm.com>
3173 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3174 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3175 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3176 fctiw instruction.
3178 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3180 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3181 instead of __vector bool.
3182 (_mm_max_pu8): Likewise.
3183 (_mm_min_pi16): Likewise.
3185 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3187 PR target/84912
3188 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3189 (DIVWEUO): Likewise.
3190 (DIVDEO): Likewise.
3191 (DIVDEUO): Likewise.
3192 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3193 DIVWEUO and DIVDEUO.
3194 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3195 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3196 (div_extend): Likewise.
3197 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3198 builtin function.
3199 (__builtin_divweuo): Likewise.
3200 (__builtin_divdeo): Likewise.
3201 (__builtin_divdeuo): Likewise.
3203 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3205 PR target/85095
3206 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3207 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3209 PR tree-optimization/82004
3210 * gimple-match-head.c (optimize_pow_to_exp): New function.
3211 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3212 Don't fold to exp if optimize_pow_to_exp is false.
3214 2018-03-28 Martin Liska <mliska@suse.cz>
3216 PR other/84819
3217 * calls.c (initialize_argument_information): Fix trailing space.
3218 * common.opt: Fix typo and provide better explanation for
3219 -fsanitize-coverage option.
3220 * config/i386/i386.opt: Fix typo.
3222 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3223 Martin Liska <mliska@suse.cz>
3225 PR sanitizer/85081
3226 * gimplify.c (asan_poison_variable): Don't do the check for
3227 gimplify_omp_ctxp here.
3228 (gimplify_decl_expr): Do it here.
3229 (gimplify_target_expr): Likewise.
3231 2018-03-28 Martin Liska <mliska@suse.cz>
3233 PR target/84988
3234 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3235 chkp_type_bounds_count if MPX is not enabled.
3237 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3239 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3241 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3243 PR target/84914
3244 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3245 function to create the function decl for complex long double
3246 multiply and divide for -mabi=ieeelongdouble.
3247 (init_float128_ieee): Call it.
3249 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3251 PR target/85044
3252 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3253 -fcf-protection=branch -mibt.
3254 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3256 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3258 PR target/81863
3259 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3261 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3263 PR target/85056
3264 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3265 extern array declarations.
3267 2018-03-27 Richard Biener <rguenther@suse.de>
3269 PR middle-end/84067
3270 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3271 explicit single_use checks.
3273 2018-03-27 Richard Biener <rguenther@suse.de>
3275 PR tree-optimization/85082
3276 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3277 Valueize the VUSE.
3279 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3281 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3282 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3283 Turn on fasynchronous-unwind-tables and funwind-tables.
3285 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3287 PR target/85073
3288 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3289 (*bmi_blsr_<mode>_ccz): Ditto.
3291 2018-03-26 Tom de Vries <tom@codesourcery.com>
3293 PR tree-optimization/85063
3294 * omp-general.c (offloading_function_p): New function. Factor out
3295 of ...
3296 * omp-offload.c (pass_omp_target_link::gate): ... here.
3297 * omp-general.h (offloading_function_p): Declare.
3298 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3299 with attribute omp declare target for offloading functions.
3301 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3303 PR tree-optimization/84005
3304 * tree-data-ref.h (get_base_for_alignment): Declare.
3305 * tree-data-ref.c (get_base_for_alignment_1): New function.
3306 (get_base_for_alignment): Likewise.
3307 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3308 get_base_for_alignment to find a suitable base object, instead
3309 of always using drb->base_address.
3311 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3313 PR inline-asm/85022
3314 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3315 known size by default.
3317 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3319 PR inline-asm/85030
3320 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3321 and non BLKmode operands.
3323 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3325 PR target/85026
3326 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3327 Clean up attributes.
3329 2018-03-23 Richard Biener <rguenther@suse.de>
3331 PR debug/85020
3332 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3333 we are going to emit early debug for LTO.
3335 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3337 PR inline-asm/85034
3338 * function.c (match_asm_constraints_1): Don't optimize if input
3339 doesn't satisfy general_operand predicate for output's mode.
3341 PR inline-asm/85022
3342 * alias.c (write_dependence_p): Don't require for x_canonicalized
3343 non-VOIDmode if x has VOIDmode.
3345 PR sanitizer/85029
3346 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3347 just don't try to optimize it rather than assert it never happens.
3349 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3351 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3352 macro expansions for definition of ST_INTERNAL_<mode> and
3353 LD_INTERNAL_<mode> builtins.
3354 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3355 Remove prototype.
3356 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3357 function.
3358 (altivec_expand_st_builtin): Likewise.
3359 (altivec_expand_builtin): Remove calls to deleted functions.
3360 (rs6000_address_for_altivec): Delete this function.
3361 * config/rs6000/vector.md: Remove expands for
3362 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3364 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3366 PR target/84826
3367 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3368 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3369 re-computing once computed.
3370 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3371 (arm_init_machine_status): Initialize
3372 machine->static_chain_stack_bytes.
3374 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3376 PR target/84760
3377 * doc/extend.texi: Add four new prototypes for vec_ld.
3378 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3379 definitions for more logical presentation.
3380 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3381 entries for V1TI variants of __builtin_altivec_ld builtin.
3382 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3383 handling of V1TI variant of LVX icode pattern.
3384 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3385 (rs6000_gimple_fold_builtin): Likewise.
3386 (altivec_init_builtins): Add code to define
3387 __builtin_altivec_lvx_v1ti function.
3389 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3391 PR inline-asm/84941
3392 * function.c (match_asm_constraints_1): Don't do the optimization
3393 if input isn't a REG, SUBREG, MEM or constant.
3395 2018-03-22 Tom de Vries <tom@codesourcery.com>
3397 PR tree-optimization/84956
3398 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3399 bb_has_abnormal_pred.
3401 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3403 PR sanitizer/85018
3404 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3405 DECL_INITIAL (decl) to decl at the end.
3406 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3407 adjust the comment.
3409 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3411 * doc/extend.texi (__builtin_tgmath): Document when complex
3412 integer types are treated as _Complex _Float64.
3414 2018-03-21 Tom de Vries <tom@codesourcery.com>
3416 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3418 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3420 PR tree-optimization/84960
3421 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3422 if it is ENTRY block, move them into single succ of ENTRY in that case.
3424 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3426 PR tree-optimization/84811
3427 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3428 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3429 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3430 (known_subrange_p): Update accordingly. Cast each value involved
3431 in the size comparison, rather than casting the result of the
3432 subtraction.
3434 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3436 PR tree-optimization/84982
3437 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3438 by flipping the least significant bit rather than all bits from
3439 bitpos to bitpos + bitsize - 1.
3441 2018-03-21 Nathan Sidwell <nathan@acm.org>
3443 * doc/extend.texi (Deprecated Features): Remove mention of
3444 long-deleted deprecations.
3446 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3448 PR jit/84288
3449 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3450 * configure: Regenerate.
3452 2018-03-21 Tom de Vries <tom@codesourcery.com>
3454 PR tree-optimization/83126
3455 * tree-parloops.c (num_phis): New function.
3456 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3458 2018-03-21 Nathan Sidwell <nathan@acm.org>
3460 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3461 mention anon-struct/union members and trailing attributes.
3463 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3465 PR tree-optimization/84969
3466 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3467 builtin memset partitions if they set different rhs values.
3469 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3471 PR rtl-optimization/84989
3472 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3473 VEC_DUPLICATE with scalar result mode.
3475 2018-03-21 Martin Liska <mliska@suse.cz>
3477 PR ipa/84963
3478 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3479 not intended return statement.
3481 2018-03-21 Martin Liska <mliska@suse.cz>
3483 PR target/84988
3484 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3485 (chkp_find_bound_slots_1): Limit number of iterations.
3487 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3489 PR target/84838
3490 * Minor grammar fixes for x86 options.
3492 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3494 PR debug/84875
3495 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3496 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3498 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3500 PR target/83789
3501 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3502 (altivec_lvx_<mode>_1op): Likewise.
3503 (altivec_stvx_<mode>_2op): Likewise.
3504 (altivec_stvx_<mode>_1op): Likewise.
3505 (altivec_lvx_<VM2:mode>): New define_expand.
3506 (altivec_stvx_<VM2:mode>): Likewise.
3507 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3508 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3509 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3510 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3511 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3512 (rs6000_gen_lvx): Likewise.
3513 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3514 (altivec_expand_stv_builtin): Likewise.
3515 (altivec_expand_builtin): Likewise.
3516 * config/rs6000/vector.md: Likewise.
3518 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3520 PR target/82518
3521 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3522 BYTES_BIG_ENDIAN.
3524 2018-03-20 Richard Biener <rguenther@suse.de>
3526 PR target/84986
3527 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3528 sign-conversions as zero, fall back to standard scalar_stmt
3529 cost for the rest.
3531 2018-03-20 Martin Liska <mliska@suse.cz>
3533 PR ipa/84825
3534 * predict.c (rebuild_frequencies): Handle case when we have
3535 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3537 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3539 PR target/84990
3540 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3541 flag_section_anchors.
3542 * varasm.c (use_blocks_for_decl_p): Remove hack for
3543 dw2_force_const_mem.
3545 PR target/84845
3546 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3547 to ...
3548 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3549 be created, use lowpart_subreg of operands[0] rather than operands[0]
3550 itself.
3551 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3552 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3553 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3554 and n constraint instead of aarch64_shift_imm_di and Usd.
3555 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3556 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3558 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3560 PR target/82989
3561 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3562 to favor GPR over NEON registers.
3563 (<shift>di3_neon): Likewise.
3565 2018-03-20 Tom de Vries <tom@codesourcery.com>
3567 PR target/84952
3568 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3569 (nvptx_process_pars): Emit bar.sync asap and alap.
3571 2018-03-20 Tom de Vries <tom@codesourcery.com>
3573 PR target/84954
3574 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3575 seen_label if seen_label is already set.
3577 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3579 PR target/84945
3580 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3581 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3582 Use 1U instead of 1. Formatting fixes.
3584 PR c/84953
3585 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3586 instead of TREE_TYPE (s1) for the return value.
3588 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3590 PR tree-optimization/84946
3591 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3592 bitsize + bitsize in poly_uint64 rather than poly_int64.
3594 PR sanitizer/78651
3595 * dwarf2asm.c: Include fold-const.c.
3596 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3597 of decl rather than decl itself.
3599 PR rtl-optimization/84643
3600 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3602 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3604 PR sanitizer/78651
3605 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3606 calling assemble_variable.
3608 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3610 PR target/81647
3611 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3612 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3614 2018-03-19 Jim Wilson <jimw@sifive.com>
3616 PR bootstrap/84856
3617 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3618 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3619 (riscv_first_stack_step): Likewise.
3620 (riscv_option_override): Use STACK_BOUNDARY instead of
3621 MIN_STACK_BOUNDARY.
3622 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3623 MIN_STACK_BOUNDARY.
3624 (BIGGEST_ALIGNMENT): Set to 128.
3625 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3626 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3627 STACK_BOUNDARY.
3629 2018-03-19 Richard Biener <rguenther@suse.de>
3631 PR tree-optimization/84933
3632 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3633 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3635 2018-03-19 Richard Biener <rguenther@suse.de>
3637 PR tree-optimization/84859
3638 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3639 (cond_if_else_store_replacement): Perform sinking operation on
3640 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3641 Generalize what a BB with a single eligible store is.
3643 2018-03-19 Richard Biener <rguenther@suse.de>
3645 PR tree-optimization/84929
3646 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3647 chrec_is_positive against non-chrec arg.
3649 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3651 PR target/84711
3652 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3654 2018-03-18 Martin Liska <mliska@suse.cz>
3656 PR rtl-optimization/84635
3657 * regrename.c (build_def_use): Use matches_mode only when
3658 matches >= 0.
3660 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3662 PR tree-optimization/84913
3663 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3664 vectorize chains of COND_EXPRs.
3666 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3668 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3670 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3672 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3674 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3676 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3678 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3679 Kito Cheng <kito.cheng@gmail.com>
3681 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3682 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3683 (nds32_adjust_reg_alloc_order): New function.
3684 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3686 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3688 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3689 nds32_print_operand, nds32_print_operand_address): Use
3690 HOST_WIDE_INT_PRINT_DEC instead.
3692 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3694 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3696 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3698 PR target/84902
3699 * config/i386/i386.c (initial_ix86_tune_features,
3700 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3701 unsigned long long.
3702 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3703 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3704 rather than 1u << ix86_tune. Formatting fix.
3705 (ix86_option_override_internal): Change ix86_arch_mask from
3706 unsigned int to unsigned HOST_WIDE_INT, initialize to
3707 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3708 (ix86_function_specific_restore): Likewise.
3710 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3712 PR target/84899
3713 * postreload.c (reload_combine_recognize_pattern): Perform
3714 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3715 truncate_int_for_mode the result for the destination's mode.
3717 PR c/84909
3718 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3719 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3720 Likewise.
3721 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3722 Likewise.
3724 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3726 PR target/84876
3727 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3728 regno_allocno_class_array and sorted_pseudos.
3729 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3730 insns where regno is used.
3732 2018-03-16 Martin Liska <mliska@suse.cz>
3734 PR ipa/84833
3735 * multiple_target.c (create_dispatcher_calls): Redirect
3736 reference in the symbol table.
3738 2018-03-16 Martin Liska <mliska@suse.cz>
3740 PR ipa/84722
3741 * multiple_target.c (create_dispatcher_calls): Redirect also
3742 an alias.
3744 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3746 PR c++/79937
3747 PR c++/82410
3748 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3749 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3750 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3752 2018-03-16 Julia Koval <julia.koval@intel.com>
3754 * doc/invoke.texi (Skylake Server): Add CLWB.
3755 Cannonlake): Remove CLWB.
3757 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3759 PR tree-optimization/84841
3760 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3761 1 << 3.
3762 (FLOAT_ONE_CONST_TYPE): Define.
3763 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3764 (sort_by_operand_rank): Put entries with higher constant_type last
3765 rather than first to match comments.
3767 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3769 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3770 split predicate.
3772 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3774 PR c++/79085
3775 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3776 check and use address of target always.
3778 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3780 PR target/84574
3781 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3782 (indirect_thunk_bnd_needed): Likewise.
3783 (indirect_thunks_used): Likewise.
3784 (indirect_thunks_bnd_used): Likewise.
3785 (indirect_return_needed): New.
3786 (indirect_return_bnd_needed): Likewise.
3787 (output_indirect_thunk_function): Add a bool argument for
3788 function return.
3789 (output_indirect_thunk_function): Don't generate alias for
3790 function return thunk.
3791 (ix86_code_end): Call output_indirect_thunk_function to generate
3792 function return thunks.
3793 (ix86_output_function_return): Set indirect_return_bnd_needed
3794 and indirect_return_needed instead of indirect_thunk_bnd_needed
3795 and indirect_thunk_needed.
3797 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3799 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3800 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3801 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3803 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3804 Paul Hua <paul.hua.gm@gmail.com>
3806 PR c/84852
3807 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3809 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3811 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3812 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3813 resp. SFmode cases.
3815 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3817 PR target/84711
3818 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3819 instead of GET_MODE_SIZE when comparing Units.
3821 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3823 PR target/68256
3824 * varasm.c (hash_section): Return an unchangeble hash value
3825 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3826 Return !aarch64_can_use_per_function_literal_pools_p ().
3828 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3830 PR target/84860
3831 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3832 rather than address of cmode as last argument to prepare_cmp_insn.
3834 2018-03-15 Julia Koval <julia.koval@intel.com>
3836 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3837 F_AVX512VNNI, F_AVX512BITALG): New.
3839 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3841 PR target/83451
3842 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3843 insn for floating-point loads and stores.
3845 2018-03-14 Carl Love <cel@us.ibm.com>
3847 * config/rs6000/rs6000-c.c: Add macro definitions for
3848 ALTIVEC_BUILTIN_VEC_PERMXOR.
3849 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3850 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3851 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3852 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3853 UNSPEC_VPERMXOR.
3854 * config/doc/extend.texi: Add prototypes for vec_permxor.
3856 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3858 PR c/84852
3859 * diagnostic-show-locus.c (class layout_point): Convert m_line
3860 from int to linenum_type.
3861 (line_span::comparator): Use linenum "compare" function when
3862 comparing line numbers.
3863 (test_line_span): New function.
3864 (layout_range::contains_point): Convert param "row" from int to
3865 linenum_type.
3866 (layout_range::intersects_line_p): Likewise.
3867 (layout::will_show_line_p): Likewise.
3868 (layout::print_source_line): Likewise.
3869 (layout::should_print_annotation_line_p): Likewise.
3870 (layout::print_annotation_line): Likewise.
3871 (layout::print_leading_fixits): Likewise.
3872 (layout::annotation_line_showed_range_p): Likewise.
3873 (struct line_corrections): Likewise for field m_row.
3874 (line_corrections::line_corrections): Likewise for param "row".
3875 (layout::print_trailing_fixits): Likewise.
3876 (layout::get_state_at_point): Likewise.
3877 (layout::get_x_bound_for_row): Likewise.
3878 (layout::print_line): Likewise.
3879 (diagnostic_show_locus): Likewise for locals "last_line" and
3880 "row".
3881 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3882 * input.c (selftest::test_linenum_comparisons): New function.
3883 (selftest::input_c_tests): Call it.
3884 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3885 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3886 * selftest.h (ASSERT_GT): New macro.
3887 (ASSERT_GT_AT): New macro.
3888 (ASSERT_LT): New macro.
3889 (ASSERT_LT_AT): New macro.
3891 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3893 PR rtl-optimization/84780
3894 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3896 2018-03-14 Martin Liska <mliska@suse.cz>
3898 * tree.c (record_node_allocation_statistics): Use
3899 get_stats_node_kind.
3900 (get_stats_node_kind): New function extracted from
3901 record_node_allocation_statistics.
3902 (free_node): Use get_stats_node_kind.
3904 2018-03-14 Richard Biener <rguenther@suse.de>
3906 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3907 that the value-set of ANTIC_IN doesn't grow.
3909 Revert
3910 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3911 member.
3912 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3913 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3915 2018-03-14 Julia Koval <julia.koval@intel.com>
3917 * config.gcc (icelake-client, icelake-server): New.
3918 (icelake): Remove.
3919 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3920 (initial_ix86_arch_features): Ditto.
3921 (PTA_SKYLAKE): Add SGX.
3922 (PTA_ICELAKE): Remove.
3923 (PTA_ICELAKE_CLIENT): New.
3924 (PTA_ICELAKE_SERVER): New.
3925 (ix86_option_override_internal): Split up icelake on icelake client and
3926 icelake server.
3927 (get_builtin_code_for_version): Ditto.
3928 (fold_builtin_cpu): Ditto.
3929 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3930 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3931 * config/i386/i386.h (processor_type): Ditto.
3932 * doc/invoke.texi: Ditto.
3934 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3936 PR sanitizer/83392
3937 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3938 INTEGER_CST offset, add it together with bitpos / 8 and
3939 sign extend based on POINTER_SIZE.
3941 PR target/84844
3942 Revert
3943 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3945 PR target/78090
3946 * config/i386/constraints.md (Yc): New register constraint.
3947 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3948 Use Yc constraint for alternative 2 of operand 0. Remove
3949 preferred_for_speed attribute.
3951 2018-03-14 Richard Biener <rguenther@suse.de>
3953 PR tree-optimization/84830
3954 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3955 with the old one to avoid oscillations.
3957 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3959 PR target/83712
3960 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3961 pseudos.
3962 (assign_by_spills): Return a flag of reload assignment failure.
3963 Do not process the reload assignment failures. Do not spill other
3964 reload pseudos if they has the same reg class. Update n if
3965 necessary.
3966 (lra_assign): Add a return arg. Set up from the result of
3967 assign_by_spills call.
3968 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3969 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3970 usage_insns if it is not NULL.
3971 (spill_hard_reg_in_range): New function.
3972 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3973 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3974 function prototypes.
3975 (lra_assign): Change prototype.
3976 * lra.c (lra): Add code to deal with fails by splitting hard reg
3977 live ranges.
3979 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3981 * config/riscv/riscv.opt (mrelax): New option.
3982 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3983 "norelax" when riscv_mrelax is disabled.
3984 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3986 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3988 PR target/84743
3989 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3990 reassociation for int modes.
3992 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3994 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3995 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3996 for big-endian.
3997 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3998 * config/aarch64/aarch64-sve.md
3999 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4000 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4001 (*extend<mode><Vwide>2): Rename to...
4002 (aarch64_sve_extend<mode><Vwide>2): ...this.
4003 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4004 renaming the old pattern to...
4005 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4006 unsigned packs.
4007 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4008 define_expand, renaming the old pattern to...
4009 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4010 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4011 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4012 account when deciding which SVE instruction the optab should use.
4013 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4015 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4017 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4018 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4019 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4020 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4021 (tlsdesc_small_<mode>): Turn a define_expand and use
4022 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4023 (tlsdesc_small_advsimd_<mode>): ...this.
4024 (tlsdesc_small_sve_<mode>): New pattern.
4026 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4028 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4029 (UNSPEC_UMUL_HIGHPART): New constants.
4030 (MUL_HIGHPART): New int iteraor.
4031 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4032 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4033 define_expand.
4034 (*<su>mul<mode>3_highpart): New define_insn.
4036 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4038 PR lto/84805
4039 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4040 incomplete types.
4042 2018-03-13 Martin Liska <mliska@suse.cz>
4044 PR ipa/84658.
4045 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4046 vector.
4047 (sem_item_optimizer::~sem_item_optimizer): Release it.
4048 (sem_item_optimizer::merge_classes): Register variable aliases.
4049 (sem_item_optimizer::fixup_pt_set): New function.
4050 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4051 * ipa-icf.h: Declare new variables and functions.
4053 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4055 PR middle-end/84834
4056 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4057 integer_pow2p@2 and test integer_pow2p in condition.
4058 (A < 0 ? C : 0): Similarly for @1.
4060 PR middle-end/84831
4061 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4062 characters starting at p contain '\0' character, don't look beyond
4063 that.
4065 PR target/84827
4066 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4067 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4069 PR target/84828
4070 * reg-stack.c (change_stack): Change update_end var from int to
4071 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4072 also call set_block_for_insn on the newly added insns and rescan.
4074 PR target/84786
4075 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4076 on the last operand.
4078 PR c++/84704
4079 * tree.c (stabilize_reference_1): Return save_expr (e) for
4080 STATEMENT_LIST even if it doesn't have side-effects.
4082 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4084 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4086 2018-03-12 Renlin Li <renlin.li@arm.com>
4088 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4089 aarch64_output_scalar_simd_mov_immediate.
4091 2018-03-12 Martin Sebor <msebor@redhat.com>
4093 PR tree-optimization/83456
4094 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4095 for perfectly overlapping calls to memcpy.
4096 (gimple_fold_builtin_memory_chk): Same.
4097 (gimple_fold_builtin_strcpy): Handle no-warning.
4098 (gimple_fold_builtin_stxcpy_chk): Same.
4099 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4101 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4103 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4104 parameter. Use it for SFmode.
4105 (rs6000_function_arg_advance_1): Adjust.
4106 (rs6000_function_arg): Adjust.
4107 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4109 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4111 PR rtl-optimization/84169
4112 PR rtl-optimization/84780
4113 * combine.c (can_combine_p): Check for a 2-insn combination whether
4114 the destination register is used between the two insns, too.
4116 2018-03-12 Richard Biener <rguenther@suse.de>
4118 PR tree-optimization/84803
4119 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4120 for refs DR analysis didn't process.
4122 2018-03-12 Richard Biener <rguenther@suse.de>
4124 PR tree-optimization/84777
4125 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4126 force-vectorize loops ignore whether we are optimizing for size.
4128 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4130 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4131 (TARGET_MD_ASM_ADJUST): Define.
4133 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4134 Kito Cheng <kito.cheng@gmail.com>
4135 Chung-Ju Wu <jasonwucj@gmail.com>
4137 * config/nds32/nds32.c (nds32_compute_stack_frame,
4138 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4139 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4140 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4141 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4142 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4143 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4144 * config/nds32/nds32.md (prologue, epilogue): Use macro
4145 NDS32_V3PUSH_AVAILABLE_P to do checking.
4147 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4149 PR debug/58150
4150 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4151 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4152 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4153 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4154 addition of most attributes on !orig_type_die or the attribute not
4155 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4157 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4158 Chung-Ju Wu <jasonwucj@gmail.com>
4160 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4161 __NDS32_VH__ macro.
4162 * config/nds32/nds32.opt (mvh): New option.
4164 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4165 Chung-Ju Wu <jasonwucj@gmail.com>
4167 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4168 function.
4169 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4170 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4171 definition.
4173 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4174 Chung-Ju Wu <jasonwucj@gmail.com>
4176 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4177 function.
4178 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4179 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4181 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4182 Kito Cheng <kito.cheng@gmail.com>
4183 Chung-Ju Wu <jasonwucj@gmail.com>
4185 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4186 UNSPEC_FFMISM and UNSPEC_FLMISM.
4187 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4188 for ffb, ffmism and flmism.
4189 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4190 (unspec_ffmism): Ditto.
4191 (unspec_flmism): Ditto.
4192 (nds32_expand_builtin_impl): Check if string extension is available.
4193 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4194 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4196 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4198 Reverting patch:
4199 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4201 PR target/83712
4202 * lra-assigns.c (assign_by_spills): Return a flag of reload
4203 assignment failure. Do not process the reload assignment
4204 failures. Do not spill other reload pseudos if they has the same
4205 reg class.
4206 (lra_assign): Add a return arg. Set up from the result of
4207 assign_by_spills call.
4208 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4209 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4210 usage_insns if it is not NULL.
4211 (spill_hard_reg_in_range): New function.
4212 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4213 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4214 function prototypes.
4215 (lra_assign): Change prototype.
4216 * lra.c (lra): Add code to deal with fails by splitting hard reg
4217 live ranges.
4219 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4221 PR target/84807
4222 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4224 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4226 PR debug/84620
4227 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4228 (dw_val_node): Add val_symbolic_view.
4229 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4230 (symview_upper_bound): New.
4231 (new_line_info_table): Initialize symviews_since_reset.
4232 (dwarf2out_source_line): Count symviews_since_reset and set
4233 symview_upper_bound.
4234 (dw_val_equal_p): Handle symview.
4235 (add_AT_symview): New.
4236 (print_dw_val): Handle symview.
4237 (attr_checksum, attr_checksum_ordered): Likewise.
4238 (same_dw_val_p, size_of_die): Likewise.
4239 (value_format, output_die): Likewise.
4240 (add_high_low_attributes): Use add_AT_symview for entry_view.
4241 (dwarf2out_finish): Reset symview_upper_bound, clear
4242 zero_view_p.
4244 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4246 PR target/83969
4247 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4248 Add strict argument and use it.
4249 (rs6000_split_multireg_move): Update for new strict argument.
4250 (mem_operand_gpr): Disallow all non-offsettable addresses.
4251 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4253 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4255 PR target/84772
4256 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4257 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4258 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4260 PR c++/84767
4261 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4262 decl, use remap_type if we want to use the type.
4264 2018-03-09 Martin Sebor <msebor@redhat.com>
4266 PR tree-optimization/84526
4267 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4268 Remove dead code.
4269 (builtin_access::generic_overlap): Be prepared to handle non-array
4270 base objects.
4272 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4274 PR rtl-optimization/84682
4275 * lra-constraints.c (process_address_1): Check is_address flag
4276 for address constraints.
4277 (process_alt_operands): Likewise.
4278 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4279 preprocess_constraints.
4280 * recog.h (preprocess_constraints): Add oploc parameter.
4281 Adjust callers.
4282 * recog.c (preprocess_constraints): Test address_operand for
4283 CT_ADDRESS constraints.
4285 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4287 PR target/83712
4288 * lra-assigns.c (assign_by_spills): Return a flag of reload
4289 assignment failure. Do not process the reload assignment
4290 failures. Do not spill other reload pseudos if they has the same
4291 reg class.
4292 (lra_assign): Add a return arg. Set up from the result of
4293 assign_by_spills call.
4294 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4295 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4296 usage_insns if it is not NULL.
4297 (spill_hard_reg_in_range): New function.
4298 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4299 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4300 function prototypes.
4301 (lra_assign): Change prototype.
4302 * lra.c (lra): Add code to deal with fails by splitting hard reg
4303 live ranges.
4305 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4307 PR target/83193
4308 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4309 Accept complain bool parameter. Only emit errors if it is true.
4310 (arm_parse_cpu_option_name): Likewise.
4311 (arm_target_thumb_only): Adjust callers of the above.
4312 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4313 prototype to take a default true bool parameter.
4314 (arm_parse_arch_option_name): Likewise.
4316 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4317 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4319 PR jit/64089
4320 PR jit/84288
4321 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4322 * configure: Regenerate.
4323 * configure.ac ("linker --version-script option"): New.
4324 ("linker soname option"): New.
4326 2018-03-09 Richard Biener <rguenther@suse.de>
4328 PR tree-optimization/84775
4329 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4330 immediate uses of predicate stmts and mark them modified.
4332 Revert
4333 PR tree-optimization/84178
4334 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4335 to caller.
4336 (version_loop_for_if_conversion): Delay update_ssa call.
4337 (tree_if_conversion): Delay update_ssa until after predicate
4338 insertion.
4340 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4342 PR target/84763
4343 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4344 when the function accesses prior frames.
4346 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4348 PR debug/84456
4349 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4350 gen_llsym, otherwise call maybe_gen_llsym.
4352 PR inline-asm/84742
4353 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4354 has ',' character inside of it.
4356 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4358 PR target/84748
4359 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4360 as clobbering CC_REGNUM.
4362 2018-03-08 Richard Biener <rguenther@suse.de>
4364 PR middle-end/84552
4365 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4366 (follow_copies_to_constant): Do not follow SSA names registered
4367 for update.
4369 2018-03-08 Richard Biener <rguenther@suse.de>
4371 PR tree-optimization/84178
4372 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4373 to caller.
4374 (version_loop_for_if_conversion): Delay update_ssa call.
4375 (tree_if_conversion): Delay update_ssa until after predicate
4376 insertion.
4378 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4380 PR tree-optimization/84178
4381 * tree-if-conv.c (release_bb_predicate): Remove the
4382 the assertion that the stmts have NULL use_ops.
4383 Discard the statements, asserting that they haven't
4384 yet been added to a BB.
4386 2018-03-08 Richard Biener <rguenther@suse.de>
4388 PR tree-optimization/84746
4389 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4390 (phi_translate): Pass in destination ANTIC_OUT set.
4391 (phi_translate_1): Likewise. For a simplified result lookup
4392 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4393 (phi_translate_set): Adjust.
4394 (do_pre_regular_insertion): Likewise.
4395 (do_pre_partial_partial_insertion): Likewise.
4397 2018-03-08 Martin Liska <mliska@suse.cz>
4399 PR gcov-profile/84735
4400 * doc/gcov.texi: Document usage of profile files.
4401 * gcov-io.h: Document changes in the format.
4403 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4405 PR debug/84404
4406 PR debug/84408
4407 * dwarf2out.c (struct dw_line_info_table): Update comments for
4408 view == -1.
4409 (FORCE_RESET_NEXT_VIEW): New.
4410 (FORCE_RESETTING_VIEW_P): New.
4411 (RESETTING_VIEW_P): Check for -1 too.
4412 (ZERO_VIEW_P): Likewise.
4413 (new_line_info_table): Force-reset next view.
4414 (dwarf2out_begin_function): Likewise.
4415 (dwarf2out_source_line): Simplify zero_view_p initialization.
4416 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4417 view directly. Omit view when omitting .loc at line 0.
4419 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4421 PR tree-optimization/84740
4422 * tree-switch-conversion.c (process_switch): Call build_constructors
4423 only if info.phi_count is non-zero.
4425 PR tree-optimization/84739
4426 * tree-tailcall.c (find_tail_calls): Check call arguments against
4427 DECL_ARGUMENTS (current_function_decl) rather than
4428 DECL_ARGUMENTS (func) when checking for tail recursion.
4430 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4432 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4433 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4434 Volker Reichelt's entry and add entries for people that perform
4435 GCC fuzzy testing and report numerous bugs.
4437 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4439 PR target/82411
4440 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4441 readonly data in sdata, if that is disabled.
4442 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4443 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4444 -mreadonly-in-sdata option.
4446 2018-03-07 Martin Sebor <msebor@redhat.com>
4448 PR tree-optimization/84468
4449 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4450 basic block when looking for nul assignment.
4452 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4454 PR target/84277
4455 * except.h (output_function_exception_table): Adjust prototype.
4456 * except.c (output_function_exception_table): Remove FNNAME parameter
4457 and add SECTION parameter. Ouput one part of the table at a time.
4458 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4459 the first part of the exception table and emit unwind directives.
4460 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4461 (i386_pe_seh_cold_init): Likewise.
4462 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4463 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4464 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4465 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4466 * config/i386/winnt.c: Include except.h.
4467 (struct seh_frame_state): Add reg_offset, after_prologue and
4468 in_cold_section fields.
4469 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4470 (i386_pe_seh_cold_init): New function.
4471 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4472 to seh->in_cold_section.
4473 (seh_emit_push): Record the offset of the push.
4474 (seh_emit_save): Record the offet of the save.
4475 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4476 Test seh->after_prologue to disregard the epilogue.
4477 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4478 (i386_pe_end_cold_function): New function.
4480 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4482 PR fortran/84565
4483 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4484 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4486 PR c++/84704
4487 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4488 on tmp_var.
4489 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4490 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4492 PR middle-end/84723
4493 * multiple_target.c: Include tree-inline.h and intl.h.
4494 (expand_target_clones): Diagnose and fail if node->definition and
4495 !tree_versionable_function_p (node->decl).
4497 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4499 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4500 sprint_ul.
4501 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4502 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4503 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4505 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4507 PR target/84710
4508 * combine.c (try_combine): Use reg_or_subregno instead of handling
4509 just paradoxical SUBREGs and REGs.
4511 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4513 * config/arc/arc.c (arc_finalize_pic): Remove function.
4514 (arc_must_save_register): We use single base PIC register, remove
4515 checks to save/restore the PIC register.
4516 (arc_expand_prologue): Likewise.
4517 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4518 Remove.
4519 (arc_verify_short): Likewise.
4520 (arc_attr_type): Likewise.
4521 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4522 (walk_stores): Likewise.
4523 (arc_address_cost): Make it static.
4524 (arc_verify_short): Likewise.
4525 (branch_dest): Likewise.
4526 (arc_attr_type): Likewise.
4527 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4528 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4529 (arc_final_prescan_insn): Remove inserting the nops due to
4530 hardware hazards. It is done in reorg step.
4531 (insn_length_variant_t): Remove.
4532 (insn_length_parameters_t): Likewise.
4533 (arc_insn_length_parameters): Likewise.
4534 (arc_get_insn_variants): Likewise.
4535 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4537 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4539 PR inline-asm/84683
4540 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4541 assertion failure.
4543 PR tree-optimization/84687
4544 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4545 on new_node->decl.
4546 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4548 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4550 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4551 Rename to ppc_speculation_barrier.
4552 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4553 __builtin_ppc_speculation_barrier.
4555 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4557 PR target/84700
4558 * combine.c (combine_simplify_rtx): Don't try to simplify if
4559 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4560 are equal to x.
4562 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4564 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4565 to 32 bytes when compiling for POWER9.
4567 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4569 PR target/84564
4570 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4571 regparm >= 3 with no arg reg available also for calls with
4572 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4574 PR target/84524
4575 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4576 orig,vex.
4577 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4579 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4581 PR target/84264
4582 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4584 2018-03-05 Richard Biener <rguenther@suse.de>
4586 PR tree-optimization/84486
4587 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4588 When inserting a __builtin_assume_aligned call set the LHS
4589 SSA name alignment info accordingly.
4591 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4593 PR tree-optimization/84114
4594 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4595 Avoid reassociation of FLOAT_MODE addition.
4597 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4599 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4600 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4601 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4602 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4603 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4604 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4605 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4606 and -mwbnoinvd.
4607 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4608 __builtin_ia32_wbinvd): New builtins.
4609 (SPECIAL_ARGS2): New.
4610 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4611 (SPECIAL_ARGS2): New.
4612 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4613 (ix86_valid_target_attribute_inner_p): Ditto.
4614 (ix86_init_mmx_sse_builtins): Add special_args2.
4615 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4616 TARGET_WBNOINVD_P): New.
4617 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4618 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4619 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4620 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4621 * config/i386/pconfigintrin.h: New file.
4622 * config/i386/wbnoinvdintrin.h: Ditto.
4623 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4624 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4626 2018-03-05 Richard Biener <rguenther@suse.de>
4628 PR tree-optimization/84670
4629 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4630 member.
4631 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4632 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4633 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4634 doesn't grow if all successors (recursively) were visited at least
4635 once.
4637 2018-03-05 Richard Biener <rguenther@suse.de>
4639 PR tree-optimization/84650
4640 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4641 if executed in the loop pipeline.
4643 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4645 * doc/configfiles.texi (Configuration Files): Move info about
4646 conditionalizing $target-protos.h to...
4647 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4648 differs from $target-protos.h.
4650 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4651 Chung-Ju Wu <jasonwucj@gmail.com>
4653 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4654 * config/nds32/nds32-multiple.md (setmemsi): Define.
4655 * config/nds32/nds32-memory-manipulation.c
4656 (nds32_gen_dup_4_byte_to_word_value): New.
4657 (emit_setmem_word_loop): New.
4658 (emit_setmem_byte_loop): New.
4659 (nds32_expand_setmem_loop): New.
4660 (nds32_expand_setmem_loop_v3m): New.
4661 (nds32_expand_setmem_unroll): New.
4662 (nds32_expand_setmem): New.
4664 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4665 Chung-Ju Wu <jasonwucj@gmail.com>
4667 * config/nds32/nds32-memory-manipulation.c
4668 (nds32_emit_load_store): New.
4669 (nds32_emit_post_inc_load_store): New.
4670 (nds32_emit_mem_move): New.
4671 (nds32_emit_mem_move_block): New.
4672 (nds32_expand_movmemsi_loop_unknown_size): New.
4673 (nds32_expand_movmemsi_loop_known_size): New.
4674 (nds32_expand_movmemsi_loop): New.
4675 (nds32_expand_movmemsi_unroll): New.
4676 (nds32_expand_movmemqi): Rename ...
4677 (nds32_expand_movmemsi): ... to this.
4678 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4679 (movmemsi): ... to this.
4680 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4681 (nds32_expand_movmemsi): ... to this.
4683 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4684 Monk Chiang <sh.chiang04@gmail.com>
4685 Chung-Ju Wu <jasonwucj@gmail.com>
4687 * config/nds32/nds32-protos.h
4688 (nds32_expand_load_multiple): New arguments.
4689 (nds32_expand_store_multiple): Ditto.
4690 (nds32_valid_multiple_load_store): Rename ...
4691 (nds32_valid_multiple_load_store_p): ... to this.
4692 * config/nds32/nds32-memory-manipulation.c
4693 (nds32_expand_load_multiple): Refine implementation.
4694 (nds32_expand_store_multiple): Ditto.
4695 * config/nds32/nds32-multiple.md
4696 (load_multiple): Update nds32_expand_load_multiple interface.
4697 (store_multiple): Update nds32_expand_store_multiple interface.
4698 * config/nds32/nds32-predicates.c
4699 (nds32_valid_multiple_load_store): Rename ...
4700 (nds32_valid_multiple_load_store_p): ... to this and refine
4701 implementation.
4702 * config/nds32/predicates.md
4703 (nds32_load_multiple_and_update_address_operation): New predicate.
4704 (nds32_store_multiple_and_update_address_operation): New predicate.
4706 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4707 Chung-Ju Wu <jasonwucj@gmail.com>
4709 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4710 (combo): New attribute.
4711 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4713 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4715 * config/nds32/nds32.opt: Change -mcmodel= default value.
4717 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4718 Monk Chiang <sh.chiang04@gmail.com>
4719 Chung-Ju Wu <jasonwucj@gmail.com>
4721 * config/nds32/constants.md (unspec_element): New enum.
4722 * config/nds32/constraints.md (Umw): New constraint.
4723 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4724 * config/nds32/nds32-intrinsic.md: Likewise.
4725 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4726 (nds32_valid_smw_lwm_base_p): New.
4727 (nds32_output_smw_single_word): New.
4728 (nds32_output_lmw_single_word): New.
4729 (nds32_expand_unaligned_load): New.
4730 (nds32_expand_unaligned_store): New.
4731 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4732 (nds32_output_smw_single_word): Declare.
4733 (nds32_output_lmw_single_word): Declare.
4734 (nds32_expand_unaligned_load): Declare.
4735 (nds32_expand_unaligned_store): Declare.
4736 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4737 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4738 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4739 NDS32_BUILTIN_UASTORE_DW.
4740 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4741 predicate.
4743 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4744 Kito Cheng <kito.cheng@gmail.com>
4745 Chung-Ju Wu <jasonwucj@gmail.com>
4747 * config/nds32/nds32-intrinsic.c
4748 (nds32_expand_builtin_null_ftype_reg): Delete.
4749 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4750 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4751 (nds32_read_argument): New.
4752 (nds32_legitimize_target): Ditto.
4753 (nds32_legitimize_argument): Ditto.
4754 (nds32_check_constant_argument): Ditto.
4755 (nds32_expand_unop_builtin): Ditto.
4756 (nds32_expand_unopimm_builtin): Ditto.
4757 (nds32_expand_binop_builtin): Ditto.
4758 (nds32_builtin_decl_impl): Ditto.
4759 (builtin_description): Ditto.
4760 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4761 (nds32_init_builtins_impl): Ditto.
4762 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4763 (nds32_builtin_decl): New.
4764 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4765 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4767 2018-03-02 Jeff Law <law@redhat.com>
4769 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4770 (redundant_insn, fill_simple_delay_slots): Likewise.
4771 (fill_slots_from_thread): Likewise.
4772 * resource.c (mark_referenced_resources): Likewise.
4773 (mark_set_resources, find_dead_or_set_registers): Likewise.
4775 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4777 * substring-locations.h (format_warning_va): Formatting fix for
4778 ATTRIBUTE_GCC_DIAG.
4779 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4780 argument.
4781 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4782 * substring-locations.c: Include intl.h.
4783 (format_warning_va): Turned into small wrapper around
4784 format_warning_n_va, renamed to ...
4785 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4786 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4787 use ngettext.
4788 (format_warning_at_substring_n): New function.
4789 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4790 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4791 format_warning_at_substring with just a shorter name instead of
4792 const function pointer.
4793 (fmtwarn_n): New function.
4794 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4795 appropriate, get rid of all the fmtstr temporaries, move conditionals
4796 with G_() wrapped string literals directly into fmtwarn arguments,
4797 cast dir.len to (int), formatting fixes.
4799 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4801 * doc/invoke.texi: Remove "Cilk Plus" references.
4803 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4804 Richard Biener <rguenther@suse.de>
4806 PR ipa/84628
4807 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4808 for error or warning attributes if CALL_FROM_THUNK_P is set.
4809 Formatting fixes.
4811 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4813 PR target/56540
4814 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4815 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4817 PR target/56540
4818 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4819 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4821 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4822 instead of -1U in last predictors element's probability member.
4824 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4826 PR ipa/83983
4827 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4828 arguments if they are comparable.
4830 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4832 PR tree-optimization/84634
4833 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4834 masks and masked_loop_p with a single loop_masks, making sure it's
4835 null for bb vectorization.
4837 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4839 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4840 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4841 loop->force_vectorize to check whether there is no alias.
4843 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4845 PR target/84614
4846 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4847 prototypes.
4848 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4849 comments.
4850 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4851 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4852 instead of a loop around prev_real_insn.
4853 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4854 prev_real_insn.
4856 PR inline-asm/84625
4857 * config/i386/i386.c (ix86_print_operand): Use conditional
4858 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4859 zero vector.
4861 2018-03-02 Richard Biener <rguenther@suse.de>
4863 PR tree-optimization/84427
4864 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4865 (bitmap_set_subtract_values): Rewrite to handle multiple
4866 exprs per value.
4867 (clean): Likewise.
4868 (prune_clobbered_mems): Likewise.
4869 (phi_translate): Take edge instead of pred/phiblock.
4870 (phi_translate_1): Likewise.
4871 (phi_translate_set): Likewise. Insert all translated
4872 exprs for a value into the set, keeping possibly multiple
4873 expressions per value.
4874 (compute_antic_aux): Adjust for phi_translate changes.
4875 When intersecting union the expressions and prune those
4876 not in the final value set, keeping possibly multiple
4877 expressions per value. Do not use value-insertion
4878 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4879 all expressions. Add verification that the value-sets
4880 only shrink during iteration.
4881 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4882 (do_pre_regular_insertion): Likewise.
4883 (do_pre_partial_partial_insertion): Likewise.
4885 2018-03-02 Richard Biener <rguenther@suse.de>
4887 PR target/82005
4888 * config/darwin.c (saved_debug_info_level): New static global.
4889 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4890 (darwin_asm_lto_end): Restore debug info generation settings.
4892 2018-03-01 Martin Liska <mliska@suse.cz>
4894 PR sanitizer/82484
4895 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4896 volatile arguments.
4898 2018-03-01 Richard Biener <rguenther@suse.de>
4900 PR debug/84645
4901 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4902 type annotation with LTO when debug was disabled at compile-time.
4904 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4906 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4907 XINT with INTVAL.
4908 (mips_final_postscan_insn): Likewise.
4910 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4912 PR rtl-optimization/84528
4913 * alias.c (init_alias_target): Add commentary.
4914 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4915 a unique base value if the frame pointer is not eliminated
4916 to the stack pointer.
4918 2018-03-01 Tom de Vries <tom@codesourcery.com>
4920 PR rtl-optimization/83327
4921 * lra-int.h (hard_regs_spilled_into): Declare.
4922 * lra.c (hard_regs_spilled_into): Define.
4923 (init_reg_info): Init hard_regs_spilled_into.
4924 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4925 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4926 (process_bb_lives): Handle hard_regs_spilled_into.
4927 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4928 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4930 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4932 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4933 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4934 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4935 * config/rs6000/aix72.h: New file.
4937 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4939 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4940 instead of warning_at with conditional singular and plural messages
4941 where possible.
4943 PR target/52991
4944 * stor-layout.c (update_alignment_for_field): For
4945 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4946 && !DECL_PACKED (field), do the alignment update, just use
4947 only desired_align instead of MAX (type_align, desired_align)
4948 as the alignment.
4949 (place_field): Don't do known_align < desired_align handling
4950 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4951 is non-NULL, instead do it after rli->prev_field handling and
4952 only if not within a bitfield word. For DECL_PACKED (field)
4953 use type_align of BITS_PER_UNIT.
4955 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4957 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4958 superfluous parentheses and trailing spaces.
4960 2018-02-28 Richard Biener <rguenther@suse.de>
4962 PR tree-optimization/84584
4963 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4964 SCoPs with fake exit edge.
4966 2018-02-28 Martin Liska <mliska@suse.cz>
4968 PR testsuite/84597
4969 * timevar.c (timer::print): Fix format to properly print 100%
4970 values.
4972 2018-02-28 Richard Biener <rguenther@suse.de>
4974 PR middle-end/84607
4975 * genmatch.c (capture_info::walk_match): Do not mark
4976 captured expressions without operands as expr_p given
4977 they act more like predicates and should be subject to
4978 "lost tail" side-effect preserving.
4980 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4982 PR rtl-optimization/81611
4983 * auto-inc-dec.c (attempt_change): Move dead note from
4984 mem_insn if it's the next use of regno
4985 (find_address): Take address use of reg holding
4986 non-incremented value. Add parm to limit search to the named
4987 reg only.
4988 (merge_in_block): Attempt to use a mem insn that is the next
4989 use of the original regno.
4991 2018-02-27 Martin Sebor <msebor@redhat.com>
4993 PR c++/83871
4994 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4995 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4997 2018-02-27 Martin Sebor <msebor@redhat.com>
4999 PR translation/84207
5000 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5001 n argument to unsigned HOST_WIDE_INT.
5002 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5003 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5004 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5005 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5007 2018-02-27 Richard Biener <rguenther@suse.de>
5009 PR tree-optimization/84512
5010 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5011 Do not use the estimate returned from record_stmt_cost for
5012 the scalar iteration cost but sum properly using add_stmt_cost.
5014 2018-02-27 Richard Biener <rguenther@suse.de>
5016 PR tree-optimization/84466
5017 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5018 Adjust last change to less strictly validate use operands.
5020 2018-02-27 Martin Liska <mliska@suse.cz>
5022 PR gcov-profile/84548
5023 * gcov.c (process_file): Allow partial overlap and consider it
5024 also as group functions.
5025 (output_lines): Properly calculate range of lines for a group.
5027 2018-02-27 Martin Liska <mliska@suse.cz>
5029 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5030 'ggc' suffixes. Change first column width.
5031 (timer::print): Fix formatting of the column.
5033 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5035 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5036 preserve inline entry blocks for the sake of debug inline
5037 entry point markers alone.
5038 (remove_unused_locals): Suggest in comments a better place to
5039 force the preservation of inline entry blocks that are
5040 otherwise unused, but do not preserve them.
5042 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5044 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5046 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5048 PR target/84039
5049 * config/i386/constraints.md (Bs): Replace
5050 ix86_indirect_branch_register with
5051 TARGET_INDIRECT_BRANCH_REGISTER.
5052 (Bw): Likewise.
5053 * config/i386/i386.md (indirect_jump): Likewise.
5054 (tablejump): Likewise.
5055 (*sibcall_memory): Likewise.
5056 (*sibcall_value_memory): Likewise.
5057 Peepholes of indirect call and jump via memory: Likewise.
5058 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5059 (*sibcall_value_GOT_32): Likewise.
5060 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5061 (GOT_memory_operand): Likewise.
5062 (call_insn_operand): Likewise.
5063 (sibcall_insn_operand): Likewise.
5064 (GOT32_symbol_operand): Likewise.
5065 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5067 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5069 PR rtl-optimization/83496
5070 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5071 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5072 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5073 redundant insn, if any.
5074 (relax_delay_slots): Likewise.
5075 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5077 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5079 PR tree-optimization/83965
5080 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5081 that grouped statements are part of a reduction chain. Return
5082 true if the statement is not marked as a reduction itself but
5083 is part of a group.
5084 (vect_recog_dot_prod_pattern): Don't check whether the statement
5085 is part of a group here.
5086 (vect_recog_sad_pattern): Likewise.
5087 (vect_recog_widen_sum_pattern): Likewise.
5089 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5091 PR debug/84545
5092 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5094 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5096 PR target/84530
5097 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5098 the bool argument.
5099 (ix86_output_indirect_function_return): New prototype.
5100 (ix86_split_simple_return_pop_internal): Likewise.
5101 * config/i386/i386.c (indirect_return_via_cx): New.
5102 (indirect_return_via_cx_bnd): Likewise.
5103 (indirect_thunk_name): Handle return va CX_REG.
5104 (output_indirect_thunk_function): Create alias for
5105 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5106 (ix86_output_indirect_jmp): Remove the bool argument.
5107 (ix86_output_indirect_function_return): New function.
5108 (ix86_split_simple_return_pop_internal): Likewise.
5109 * config/i386/i386.md (*indirect_jump): Don't pass false
5110 to ix86_output_indirect_jmp.
5111 (*tablejump_1): Likewise.
5112 (simple_return_pop_internal): Change it to define_insn_and_split.
5113 Call ix86_split_simple_return_pop_internal to split it for
5114 -mfunction-return=.
5115 (simple_return_indirect_internal): Call
5116 ix86_output_indirect_function_return instead of
5117 ix86_output_indirect_jmp.
5119 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5121 PR bootstrap/84405
5122 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5123 memset and value initialization afterwards.
5125 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5127 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5129 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5131 PR target/84521
5132 * common/config/aarch64/aarch64-common.c
5133 (aarch_option_optimization_table[]): Switch
5134 off fomit-frame-pointer
5136 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5137 Chung-Ju Wu <jasonwucj@gmail.com>
5139 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5140 volatile memory.
5141 (store_multiple): Ditto.
5143 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5145 * config.gcc: Add --with-cpu support for nds32 target.
5146 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5147 * config/nds32/nds32.opt: Add -mcpu= option.
5149 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5151 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5152 isel=yes): Warn for these deprecated options.
5154 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5156 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5157 ISA_2_5_MASKS_EMBEDDED.
5159 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5161 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5162 p->max as pointers rather than using iterative_hash_expr.
5164 2018-02-23 Carl Love <cel@us.ibm.com>
5166 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5167 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5168 BU_P8V_OVERLOAD_2.
5169 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5170 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5171 P8V_BUILTIN_VEC_VUNSIGNED2.
5173 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5175 PR target/81572
5176 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5177 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5178 LRA_UNKNOWN_ALT.
5179 * lra-constraints.c (curr_insn_transform): Set up
5180 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5181 LRA_UNKNOWN_ALT.
5182 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5183 * lra-eliminations.c (spill_pseudos): Ditto.
5184 (process_insn_for_elimination): Ditto.
5185 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5186 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5187 LRA_NON_CLOBBERED_ALT.
5189 2018-02-22 Martin Sebor <msebor@redhat.com>
5191 PR tree-optimization/84480
5192 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5193 to maybe_diag_stxncpy_trunc. Call it.
5194 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5195 from gimple_fold_builtin_strcpy. Print inlining stack.
5196 (handle_builtin_stxncpy): Print inlining stack.
5197 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5199 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5201 PR target/84176
5202 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5203 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5204 and -fcheck-pointer-bounds are used together.
5205 (indirect_thunk_prefix): New enum.
5206 (indirect_thunk_need_prefix): New function.
5207 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5208 "_nt" instead of "_bnd" for NOTRACK prefix.
5209 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5210 (output_indirect_thunk_function): Likewise.
5211 (): Likewise.
5212 (ix86_code_end): Update output_indirect_thunk_function calls.
5213 (ix86_output_indirect_branch_via_reg): Replace
5214 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5215 (ix86_output_indirect_branch_via_push): Likewise.
5216 (ix86_output_function_return): Likewise.
5217 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5218 incompatible with -fcf-protection=branch and
5219 -fcheck-pointer-bounds.
5221 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5223 PR target/83335
5224 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5225 Change gcc_assert call to output_operand_lossage.
5227 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5229 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5231 2018-02-22 DJ Delorie <dj@redhat.com>
5232 Sebastian Perta <sebastian.perta@renesas.com>
5233 Oleg Endo <olegendo@gcc.gnu.org>
5235 * config/rx/rx.c (rx_rtx_costs): New function.
5236 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5238 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5240 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5242 2018-02-22 Martin Liska <mliska@suse.cz>
5244 PR driver/83193
5245 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5246 Add "native" as a possible value.
5248 2018-02-22 Martin Liska <mliska@suse.cz>
5250 PR driver/83193
5251 * config/i386/i386.c (ix86_option_override_internal):
5252 Add "native" as a possible value for -march and -mtune.
5254 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5256 PR target/84502
5257 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5258 to all type variants.
5260 PR tree-optimization/84503
5261 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5262 width as info->bitpos + info->bitsize - start.
5263 (merged_store_group::merge_overlapping): Simplify width computation.
5264 (check_no_overlap): New function.
5265 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5266 start + width and last_order of the group, fail if check_no_overlap
5267 fails.
5268 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5269 to group if check_no_overlap fails.
5271 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5273 * config/rs6000/altivec.md: Delete contraint arguments to
5274 define_expand, define_split, and define_peephole2, and in
5275 define_insn_and_split if always unused.
5276 * config/rs6000/darwin.md: Ditto.
5277 * config/rs6000/dfp.md: Ditto.
5278 * config/rs6000/rs6000.md: Ditto.
5279 * config/rs6000/sync.md: Ditto.
5280 * config/rs6000/vector.md: Ditto.
5281 * config/rs6000/vsx.md: Ditto.
5283 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5285 * config/rs6000/altivec.md: Write output control strings as braced
5286 blocks instead of double-quoted strings.
5287 * config/rs6000/darwin.md: Ditto.
5288 * config/rs6000/rs6000.md: Ditto.
5289 * config/rs6000/vector.md: Ditto.
5290 * config/rs6000/vsx.md: Ditto.
5292 2018-02-21 Jason Merrill <jason@redhat.com>
5294 PR c++/84314 - ICE with templates and fastcall attribute.
5295 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5297 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5299 * ipa-cp.c (determine_versionability): Fix comment typos.
5301 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5303 PR c/84229
5304 * ipa-cp.c (determine_versionability): Do not version functions caling
5305 va_arg_pack.
5307 2018-02-21 Martin Liska <mliska@suse.cz>
5309 PR driver/83193
5310 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5311 Add "native" as a possible value.
5312 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5313 the macro when native cpu detection is available.
5315 2018-02-21 Martin Liska <mliska@suse.cz>
5317 PR driver/83193
5318 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5319 Add "native" as a possible value.
5320 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5321 when native cpu detection is available.
5323 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5324 Martin Sebor <msebor@redhat.com>
5326 PR tree-optimization/84478
5327 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5328 false.
5329 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5330 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5331 support which is conservatively correct, for 2 only stay conservative
5332 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5333 argument to the 2 argument get_range_strlen, adjust 6 arg
5334 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5335 false.
5336 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5337 (gimple_fold_builtin_strlen): Pass true as last argument to
5338 get_range_strlen.
5340 2018-02-20 Martin Sebor <msebor@redhat.com>
5342 PR middle-end/84095
5343 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5344 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5345 (builtin_memref::builtin_memref): Factor out parts into
5346 set_base_and_offset and call it.
5348 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5350 PR middle-end/84406
5351 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5352 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5353 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5354 search at the associated MODE_INT.
5356 2018-02-20 Jeff Law <law@redhat.com>
5358 PR middle-end/82123
5359 PR tree-optimization/81592
5360 PR middle-end/79257
5361 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5362 for range data rather than using global data.
5363 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5364 range data rather than using global data.
5365 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5366 pass it to children as needed.
5367 (struct directive::fmtresult): Similarly.
5368 (struct directive::set_width): Similarly.
5369 (struct directive::set_precision): Similarly.
5370 (format_integer, format_directive, parse_directive): Similarly.
5371 (format_none): Accept unnamed vr_values parameter.
5372 (format_percent, format_floating, format_character): Similarly.
5373 (format_string, format_plain): Similarly.
5374 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5375 the EVRP range analyzer for range data rather than using global data.
5376 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5377 gimple-ssa-evrp-analyze.h
5378 (class sprintf_dom_walker): Add after_dom_children member function.
5379 Add evrp_range_analyzer member.
5380 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5381 range analyzer as needed.
5382 (sprintf_dom_walker::after_dom_children): New member function.
5383 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5384 if not optimizing.
5385 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5386 (evrp_range_analyzer::pop_to_marker): Likewise.
5388 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5390 PR tree-optimization/84419
5391 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5392 with the required type if its current type is compatible but
5393 different.
5395 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5397 PR middle-end/82004
5398 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5399 after vectorization.
5401 2018-02-20 Martin Liska <mliska@suse.cz>
5403 PR driver/83193
5404 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5405 possible values if we don't have a hint.
5407 2018-02-20 Martin Liska <mliska@suse.cz>
5409 PR c/84310
5410 PR target/79747
5411 * final.c (shorten_branches): Build align_tab array with one
5412 more element.
5413 * opts.c (finish_options): Add alignment option limit check.
5414 (MAX_CODE_ALIGN): Likewise.
5415 (MAX_CODE_ALIGN_VALUE): Likewise.
5416 * doc/invoke.texi: Document maximum allowed option value for
5417 all -falign-* options.
5419 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5421 PR target/84146
5422 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5423 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5424 * var-tracking.c (emit_note_insn_var_location): Remove all references
5425 to NOTE_INSN_CALL_ARG_LOCATION.
5426 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5427 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5428 Use copy_rtx_if_shared.
5429 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5430 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5431 (dwarf2out_var_location): Remove handling of
5432 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5433 on call_insn.
5434 * final.c (final_scan_insn): Remove all references to
5435 NOTE_INSN_CALL_ARG_LOCATION.
5436 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5437 before dumping final insns.
5438 * except.c (emit_note_eh_region_end): Remove all references to
5439 NOTE_INSN_CALL_ARG_LOCATION.
5440 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5441 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5442 * config/arc/arc.c (hwloop_optimize): Likewise.
5443 * config/arm/arm.c (create_fix_barrier): Likewise.
5444 * config/s390/s390.c (s390_chunkify_start): Likewise.
5445 * config/sh/sh.c (find_barrier): Likewise.
5446 * config/i386/i386.c (rest_of_insert_endbranch,
5447 ix86_seh_fixup_eh_fallthru): Likewise.
5448 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5449 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5450 * config/frv/frv.c (frv_function_prologue): Likewise.
5451 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5452 reg note.
5453 (note_outside_basic_block_p): Remove all references to
5454 NOTE_INSN_CALL_ARG_LOCATION.
5455 * gengtype.c (adjust_field_rtx_def): Likewise.
5456 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5457 Likewise.
5458 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5459 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5461 PR c++/84444
5462 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5463 is ADDR_EXPR.
5465 PR tree-optimization/84452
5466 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5467 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5468 is NULL.
5470 2018-02-19 Martin Liska <mliska@suse.cz>
5472 PR sanitizer/82183
5473 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5475 2018-02-19 Martin Liska <mliska@suse.cz>
5476 Richard Sandiford <richard.sandiford@linaro.org>
5478 PR tree-optimization/82491
5479 * gimple-fold.c (get_base_constructor): Make earlier bail out
5480 to prevent ubsan.
5482 2018-02-19 Carl Love <cel@us.ibm.com>
5484 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5485 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5486 BU_P8V_OVERLOAD_1.
5487 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5488 P8V_BUILTIN_VEC_NEG.
5490 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5492 * config/rl78/rl78.md (movdf): New define expand.
5494 2018-02-19 Martin Liska <mliska@suse.cz>
5496 PR other/80589
5497 * doc/invoke.texi: Fix typo.
5498 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5500 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5502 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5503 handle rs6000_single_float and rs6000_double_float specially for
5504 e500 family CPUs.
5506 2018-02-16 Jeff Law <law@redhat.com>
5508 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5509 (pop_regs): Use it.
5511 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5513 PR ipa/84425
5514 * ipa-inline.c (inline_small_functions): Fix a typo.
5516 2018-02-16 Nathan Sidwell <nathan@acm.org>
5518 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5520 2018-02-16 Carl Love <cel@us.ibm.com>
5522 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5523 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5524 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5525 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5526 expansion to P8V_BUILTIN_VEC_FLOAT2.
5528 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5530 PR rtl-optimization/70023
5531 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5532 src_regno into account.
5534 2018-02-16 Carl Love <cel@us.ibm.com>
5536 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5537 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5538 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5539 * config/rs6000/rs6000.c: Remove case statements for
5540 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5541 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5542 and P9V_BUILTIN_VEC_VINSERT4B.
5543 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5544 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5545 * config/rs6000/vsx.md:
5546 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5547 vec_insert4b.
5549 2018-02-16 Carl Love <cel@us.ibm.com>
5551 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5552 vec_insert4b.
5553 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5554 definitions.
5555 * config/rs6000/rs6000-c.c: Add the definitions for
5556 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5557 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5558 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5559 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5560 definition for insert4b and define insn *insert3b_internal.
5561 * doc/extend.texi: Add documentation for vec_extract4b.
5563 2018-02-16 Nathan Sidwell <nathan@acm.org>
5565 * doc/extend.texi (Backwards Compatibility): Mention friend
5566 injection. Note for-scope is deprecated.
5567 * doc/invoke.texi (-ffriend-injection): Deprecate.
5569 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5571 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5572 that moved to I2, also allow destinations that are a paradoxical
5573 subreg (instead of a normal reg).
5575 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5577 PR target/83831
5578 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5579 to QImode.
5581 2018-02-16 Richard Biener <rguenther@suse.de>
5583 PR tree-optimization/84037
5584 PR tree-optimization/84016
5585 PR target/82862
5586 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5587 Adjust vec_construct for the fact we need additional higher latency
5588 128bit inserts for AVX256 and AVX512 vector builds.
5589 (ix86_add_stmt_cost): Scale vector construction cost for
5590 elementwise loads.
5592 2018-02-16 Richard Biener <rguenther@suse.de>
5594 PR tree-optimization/84417
5595 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5596 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5597 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5599 2018-02-16 Martin Liska <mliska@suse.cz>
5601 PR sanitizer/84307
5602 * internal-fn.def (ASAN_CHECK): Set proper flags.
5603 (ASAN_MARK): Likewise.
5605 2018-02-16 Julia Koval <julia.koval@intel.com>
5607 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5608 from PTA_CANNONLAKE.
5610 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5612 PR target/84272
5613 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5614 Use ++iter rather than iter++ for std::list iterators.
5615 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5616 defer deleting them until all nodes in the forest are processed. Do
5617 free even leaf nodes. Change to_process into auto_vec.
5619 PR bootstrap/84405
5620 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5621 * vec.h (vec_default_construct): Use memset instead of placement new
5622 if BROKEN_VALUE_INITIALIZATION is defined.
5623 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5624 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5625 is defined.
5627 PR rtl-optimization/83723
5628 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5629 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5630 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5631 recursive calls.
5632 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5633 callers.
5634 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5636 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5638 PR rtl-optimization/81443
5639 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5640 from inner REGs to paradoxical SUBREGs.
5642 2018-02-16 Richard Biener <rguenther@suse.de>
5644 PR tree-optimization/84399
5645 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5646 For operands we can analyze at their definition make sure we can
5647 analyze them at each use as well.
5649 2018-02-16 Richard Biener <rguenther@suse.de>
5651 PR tree-optimization/84190
5652 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5653 volatile accesses if the decl isn't volatile.
5655 2018-02-15 Jason Merrill <jason@redhat.com>
5657 PR c++/84314 - ICE with templates and fastcall attribute.
5658 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5659 TYPE_CANONICAL on an existing type.
5661 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5663 PR tree-optimization/84383
5664 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5665 dstoff nor call operand_equal_p if dstbase is NULL.
5667 PR tree-optimization/84334
5668 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5669 also a CONSTANT_CLASS_P, punt.
5671 2018-02-14 Jim Wilson <jimw@sifive.com>
5673 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5674 first SMALL_OPERAND check. New local min_second_step. Move assert
5675 to where locals are set. Add TARGET_RVC support.
5676 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5678 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5680 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5682 2018-02-14 Martin Sebor <msebor@redhat.com>
5684 PR tree-optimization/83698
5685 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5686 arrays constrain the offset range to their bounds.
5687 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5688 (builtin_access::overlap): Avoid setting the size of overlap if it's
5689 already been set.
5690 (maybe_diag_overlap): Also consider arrays when deciding what values
5691 of offsets to include in diagnostics.
5693 2018-02-14 Martin Sebor <msebor@redhat.com>
5695 PR c/84108
5696 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5697 that correspond to the kind of a declaration.
5699 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5701 PR target/83984
5702 * config/pa/pa.md: Load address of PIC label using the linkage table
5703 if the label is nonlocal.
5705 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5707 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5708 warning message if user requests -maltivec=be.
5709 * doc/invoke.texi: Document deprecation of -maltivec=be.
5711 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5713 PR target/84220
5714 * config/rs6000/rs6000-c.c: Update definitions for
5715 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5716 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5718 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5720 PR target/84239
5721 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5722 add _get_ssp intrinsics. Remove argument from
5723 __builtin_ia32_rdssp[d|q].
5724 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5725 * config/i386/i386-builtin.def: Remove argument from
5726 __builtin_ia32_rdssp[d|q].
5727 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5728 ix86_expand_special_args_builtin for _rdssp[d|q].
5729 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5730 Clear register before usage.
5731 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5732 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5734 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5736 PR tree-optimization/84357
5737 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5738 operand 1 of an ARRAY_REF too.
5740 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5742 PR target/83831
5743 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5744 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5745 declarations.
5746 (set_of_reg): New struct.
5747 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5748 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5749 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5750 functions.
5751 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5752 Split into bitclr, bitset, bitinvert patterns if appropriate.
5753 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5754 use rx_fuse_in_memory_bitop.
5755 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5756 to named insn, correct maximum insn length.
5758 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5760 PR target/79242
5761 * machmode.def: Define a complex mode for PARTIAL_INT.
5762 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5763 MODE_PARTIAL_INT.
5764 * doc/rtl.texi: Document CSPImode.
5765 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5766 handling.
5767 (msp430_hard_regno_nregs_with_padding): Likewise.
5769 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5771 PR target/84279
5772 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5774 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5776 PR rtl-optimization/84169
5777 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5778 we generated a parallel as new i3 and we split that to new i2 and i3
5779 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5780 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5781 those to i2, not i1. Partially rewrite this scan code.
5783 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5785 PR c/82210
5786 * stor-layout.c (place_field): For variable length fields, adjust
5787 offset_align afterwards not just based on the field's alignment,
5788 but also on the size.
5790 PR middle-end/84309
5791 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5792 of exps and logs in the use_exp2 case.
5794 2018-02-13 Jeff Law <law@redhat.com>
5796 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5797 entry for "vector".
5799 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5800 ARGS as unused.
5802 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5804 PR debug/84342
5805 PR debug/84319
5806 * common.opt (gas-loc-support, gas-locview-support): New.
5807 (ginline-points, ginternal-reset-location-views): New.
5808 * doc/invoke.texi: Document them. Use @itemx where intended.
5809 (gvariable-location-views): Adjust.
5810 * target.def (reset_location_view): New.
5811 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5812 (TARGET_RESET_LOCATION_VIEW): New.
5813 * doc/tm.texi: Rebuilt.
5814 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5815 (dwarf2out_default_as_locview_support): New.
5816 (output_asm_line_debug_info): Use option variables.
5817 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5818 (output_loc_list): Likewise.
5819 (add_high_low_attributes): Check option variables.
5820 Don't output entry view attribute in strict mode.
5821 (gen_inlined_subroutine_die): Check option variables.
5822 (dwarf2out_inline_entry): Likewise.
5823 (init_sections_and_labels): Likewise.
5824 (dwarf2out_early_finish): Likewise.
5825 (maybe_reset_location_view): New, from...
5826 (dwarf2out_var_location): ... here. Call it.
5827 * debug.h (dwarf2out_default_as_loc_support): Declare.
5828 (dwarf2out_default_as_locview_support): Declare.
5829 * hooks.c (hook_int_rtx_insn_0): New.
5830 * hooks.h (hook_int_rtx_insn_0): Declare.
5831 * toplev.c (process_options): Take -gas-loc-support and
5832 -gas-locview-support from dwarf2out. Enable
5833 -gvariable-location-views by default only with locview
5834 assembler support. Enable -ginternal-reset-location-views by
5835 default only if the target defines the corresponding hook.
5836 Enable -ginline-points by default if location views are
5837 enabled; force it disabled if statement frontiers are
5838 disabled.
5839 * tree-inline.c (expand_call_inline): Check option variables.
5840 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5842 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5844 PR tree-optimization/84321
5845 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5846 handling. Also check whether the anti-range contains any values
5847 that satisfy the mask; switch to a VR_RANGE if not.
5849 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5851 PR sanitizer/84340
5852 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5854 2018-02-13 Martin Jambor <mjambor@suse.cz>
5856 PR c++/83990
5857 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5858 of call statements, also set location of a load to a temporary.
5860 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5862 * config/rl78/rl78.c (add_vector_labels): New function.
5863 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5864 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5865 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5866 which checks that no arguments are passed.
5867 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5868 * doc/extend.texi: Documentation for the new attribute.
5870 2018-02-13 Andreas Schwab <schwab@suse.de>
5872 * config/riscv/linux.h (CPP_SPEC): Define.
5874 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5876 PR target/84335
5877 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5878 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5879 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5880 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5881 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5882 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5883 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5884 temporarily for AES and PCLMUL builtins.
5886 PR tree-optimization/84339
5887 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5888 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5889 Formatting fixes.
5891 PR middle-end/84309
5892 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5893 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5894 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5895 inline function.
5896 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5897 inline function.
5898 * omp-simd-clone.h: New file.
5899 * omp-simd-clone.c: Include omp-simd-clone.h.
5900 (expand_simd_clones): No longer static.
5901 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5902 cgraph.h and omp-simd-clone.h.
5903 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5904 (vect_recog_widen_shift_pattern): Formatting fix.
5905 (vect_pattern_recog_1): Don't check optab for calls.
5907 PR target/84336
5908 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5909 operands[2] into a REG before using gen_lowpart on it.
5911 2018-02-12 Jeff Law <law@redhat.com>
5913 PR target/83760
5914 * config/sh/sh.c (find_barrier): Consider a sibling call
5915 a barrier as well.
5917 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5918 successfully back substituting a reg.
5920 2018-02-12 Richard Biener <rguenther@suse.de>
5922 PR tree-optimization/84037
5923 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5924 parameter, move visited init to caller.
5925 (vect_slp_analyze_operations): Separate cost from validity
5926 check, initialize visited once for all instances.
5927 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5928 for all instances.
5929 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5930 out an assert.
5931 (vect_model_promotion_demotion_cost): Likewise.
5932 (vectorizable_bswap): Guard cost modeling with !slp_node
5933 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5934 SLP stmts.
5935 (vectorizable_call): Likewise.
5936 (vectorizable_conversion): Likewise.
5937 (vectorizable_assignment): Likewise.
5938 (vectorizable_shift): Likewise.
5939 (vectorizable_operation): Likewise.
5940 (vectorizable_store): Likewise.
5941 (vectorizable_load): Likewise.
5942 (vectorizable_condition): Likewise.
5943 (vectorizable_comparison): Likewise.
5945 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5947 PR sanitizer/84307
5948 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5949 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5950 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5952 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5954 PR middle-end/83665
5955 * params.def (inline-min-speedup): Increase from 8 to 15.
5956 (max-inline-insns-auto): Decrease from 40 to 30.
5957 * ipa-split.c (consider_split): Add some buffer for function to
5958 be considered inlining candidate.
5959 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5960 default values.
5962 2018-02-12 Richard Biener <rguenther@suse.de>
5964 PR tree-optimization/84037
5965 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5966 matched stmts if we cannot swap the non-matched ones.
5968 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5970 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5971 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5972 _mm_maskz_scalef_round_ss): New intrinsics.
5973 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5974 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5975 __builtin_ia32_scalefss_round): Remove.
5976 (__builtin_ia32_scalefsd_mask_round,
5977 __builtin_ia32_scalefss_mask_round): New intrinsics.
5978 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5979 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5980 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5981 "<round_constraint>")): Changed to ...
5982 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5983 "<round_scalar_constraint>")): ... this.
5984 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5985 %0, %1, %2<round_op3>}"): Changed to ...
5986 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5987 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5988 %2<round_scalar_mask_op3>}"): ... this.
5989 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5991 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5993 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5994 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5995 (_mm_maskz_sqrt_round_ss): New intrinsics.
5996 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5997 (__builtin_ia32_sqrtsd_mask_round)
5998 (__builtin_ia32_sqrtss_mask_round): New builtins.
5999 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6000 (__builtin_ia32_sqrtss_round): Remove.
6001 (__builtin_ia32_sqrtsd_mask_round)
6002 (__builtin_ia32_sqrtss_mask_round): New builtins.
6003 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6004 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6005 ((match_operand:VF_128 1 "vector_operand"
6006 "xBm,<round_constraint>")): Changed to ...
6007 ((match_operand:VF_128 1 "vector_operand"
6008 "xBm,<round_scalar_constraint>")): ... this.
6009 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6010 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6011 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6012 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6013 %<iptr>1<round_scalar_mask_op3>}): ... this.
6014 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6015 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6017 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6019 PR target/84266
6020 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6021 Cast vec_cmpeq result to correct type.
6022 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6023 Cast vec_cmpgt result to correct type.
6025 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6027 * final.c (final_scan_insn_1): Renamed from...
6028 (final_scan_insn): ... this. New wrapper, to recover
6029 seen from the outermost call in recursive ones.
6030 * config/sparc/sparc.c (output_return): Drop seen from call.
6031 (output_sibcall): Likewise.
6032 * config/visium/visium.c (output_branch): Likewise.
6034 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6036 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6037 function label.
6039 2018-02-10 Alan Modra <amodra@gmail.com>
6041 PR target/84300
6042 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6043 Specify LR as an input.
6045 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6047 PR sanitizer/83987
6048 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6049 remove_member_access_dummy_vars): New functions.
6050 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6051 lower_omp_1, execute_lower_omp): Use them.
6053 PR rtl-optimization/84308
6054 * shrink-wrap.c (spread_components): Release todo vector.
6056 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6058 PR rtl-optimization/57193
6059 * ira-color.c (struct allocno_color_data): Add member
6060 conflict_allocno_hard_prefs.
6061 (update_conflict_allocno_hard_prefs): New.
6062 (bucket_allocno_compare_func): Add a preference based on
6063 conflict_allocno_hard_prefs.
6064 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6065 (color_allocnos): Remove a dead code. Initiate
6066 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6068 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6070 PR target/84226
6071 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6072 constraint from =wa to wa. Avoid a subreg on the output operand,
6073 instead use a pseudo and subreg it in a move.
6074 (p9_xxbrd_<mode>): Changed to ...
6075 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6076 (p9_xxbrd_v2df): New expander.
6077 (p9_xxbrw_<mode>): Changed to ...
6078 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6079 (p9_xxbrw_v4sf): New expander.
6081 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6083 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6085 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6087 PR target/83926
6088 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6089 multiply in 32-bit mode.
6090 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6091 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6092 mode.
6094 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6096 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6097 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6098 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6099 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6101 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6103 PR lto/84213
6104 * dwarf2out.c (is_trivial_indirect_ref): New function.
6105 (dwarf2out_late_global_decl): Do not generate a location
6106 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6107 and that are not defined in the current unit.
6109 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6111 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6112 instead of a libcall for UNORDERED.
6114 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6116 PR target/82641
6117 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6118 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6120 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6122 PR target/PR84295
6123 * config/s390/s390.c (s390_set_current_function): Invoke
6124 s390_indirect_branch_settings also if fndecl didn't change.
6126 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6128 * config/rs6000/rs6000.md (blockage): Set length to zero.
6130 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6132 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6134 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6136 PR sanitizer/84285
6137 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6138 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6139 -static-lib*san.
6141 PR debug/84252
6142 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6143 PARALLEL incoming that failed vt_get_decl_and_offset check.
6145 PR middle-end/84237
6146 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6147 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6148 TREE_READONLY bit.
6149 (get_variable_section): For decls in named .bss* sections pass true as
6150 second argument to bss_initializer_p.
6152 2018-02-09 Marek Polacek <polacek@redhat.com>
6153 Jakub Jelinek <jakub@redhat.com>
6155 PR c++/83659
6156 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6157 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6158 Sync some changes from cxx_fold_indirect_ref.
6160 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6162 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6163 markers.
6164 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6165 (BLOCK_INLINE_ENTRY_LABEL): New.
6166 (dwarf2out_var_location): Disregard inline entry markers.
6167 (inline_entry_data): New struct.
6168 (inline_entry_data_hasher): New hashtable type.
6169 (inline_entry_data_hasher::hash): New.
6170 (inline_entry_data_hasher::equal): New.
6171 (inline_entry_data_table): New variable.
6172 (add_high_low_attributes): Add DW_AT_entry_pc and
6173 DW_AT_GNU_entry_view attributes if a pending entry is found
6174 in inline_entry_data_table. Add old entry_pc attribute only
6175 if debug nonbinding markers are disabled.
6176 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6177 markers are enabled.
6178 (block_within_block_p, dwarf2out_inline_entry): New.
6179 (dwarf2out_finish): Check that no entries remained in
6180 inline_entry_data_table.
6181 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6182 (final_scan_insn, notice_source_line): Likewise.
6183 (rest_of_clean_state): Skip inline entry markers.
6184 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6185 markers.
6186 * gimple.c (gimple_build_debug_inline_entry): New.
6187 * gimple.h (enum gimple_debug_subcode): Add
6188 GIMPLE_DEBUG_INLINE_ENTRY.
6189 (gimple_build_debug_inline_entry): Declare.
6190 (gimple_debug_inline_entry_p): New.
6191 (gimple_debug_nonbind_marker_p): Adjust.
6192 * insn-notes.def (INLINE_ENTRY): New.
6193 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6194 inline entry marker notes.
6195 (print_insn): Likewise.
6196 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6197 (INSN_DEBUG_MARKER_KIND): Likewise.
6198 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6199 * tree-inline.c (expand_call_inline): Build and insert
6200 debug_inline_entry stmt.
6201 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6202 inline entry blocks early, if nonbind markers are enabled.
6203 (dump_scope_block): Dump fragment info.
6204 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6205 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6206 (gimple_build_debug_inline_entry): New.
6207 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6208 Enable/disable inline entry points too.
6209 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6210 (DEBUG_INSN): Describe inline entry markers.
6212 * common.opt (gvariable-location-views): New.
6213 (gvariable-location-views=incompat5): New.
6214 * config.in: Rebuilt.
6215 * configure: Rebuilt.
6216 * configure.ac: Test assembler for view support.
6217 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6218 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6219 * dwarf2out.c (var_loc_view): New typedef.
6220 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6221 (dwarf2out_locviews_in_attribute): New.
6222 (dwarf2out_locviews_in_loclist): New.
6223 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6224 (enum dw_line_info_opcode): Add LI_adv_address.
6225 (struct dw_line_info_table): Add view.
6226 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6227 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6228 (zero_view_p): New variable.
6229 (ZERO_VIEW_P): New macro.
6230 (output_asm_line_debug_info): New.
6231 (struct var_loc_node): Add view.
6232 (add_AT_view_list, AT_loc_list): New.
6233 (add_var_loc_to_decl): Add view param. Test it against last.
6234 (new_loc_list): Add view params. Record them.
6235 (AT_loc_list_ptr): Handle loc and view lists.
6236 (view_list_to_loc_list_val_node): New.
6237 (print_dw_val): Handle dw_val_class_view_list.
6238 (size_of_die): Likewise.
6239 (value_format): Likewise.
6240 (loc_list_has_views): New.
6241 (gen_llsym): Set vl_symbol too.
6242 (maybe_gen_llsym, skip_loc_list_entry): New.
6243 (dwarf2out_maybe_output_loclist_view_pair): New.
6244 (output_loc_list): Output view list or entries too.
6245 (output_view_list_offset): New.
6246 (output_die): Handle dw_val_class_view_list.
6247 (output_dwarf_version): New.
6248 (output_compilation_unit_header): Use it.
6249 (output_skeleton_debug_sections): Likewise.
6250 (output_rnglists, output_line_info): Likewise.
6251 (output_pubnames, output_aranges): Update version comments.
6252 (output_one_line_info_table): Output view numbers in asm comments.
6253 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6254 Call maybe_gen_llsym.
6255 (loc_list_from_tree_1): Adjust.
6256 (add_AT_location_description): Create view list attribute if
6257 needed, check it's absent otherwise.
6258 (convert_cfa_to_fb_loc_list): Adjust.
6259 (maybe_emit_file): Call output_asm_line_debug_info for test.
6260 (dwarf2out_var_location): Reset views as needed. Precompute
6261 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6262 attribute. Set view.
6263 (new_line_info_table): Reset next view.
6264 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6265 (dwarf2out_source_line): Likewise. Output view resets and labels to
6266 the assembler, or select appropriate line info opcodes.
6267 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6268 (optimize_string_length): Catch it. Adjust.
6269 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6270 dw_val_class_view_list, and remove it if no longer needed.
6271 (hash_loc_list): Hash view numbers.
6272 (loc_list_hasher::equal): Compare them.
6273 (optimize_location_lists): Check whether a view list symbol is
6274 needed, and whether the locview attribute is present, and
6275 whether they match. Remove the locview attribute if no longer
6276 needed.
6277 (index_location_lists): Call skip_loc_list_entry for test.
6278 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6279 Use output_dwarf_version.
6280 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6281 (struct dw_val_node): Add val_view_list.
6282 * final.c (SEEN_NEXT_VIEW): New.
6283 (set_next_view_needed): New.
6284 (clear_next_view_needed): New.
6285 (maybe_output_next_view): New.
6286 (final_start_function): Rename to...
6287 (final_start_function_1): ... this. Take pointer to FIRST,
6288 add SEEN parameter. Emit param bindings in the initial view.
6289 (final_start_function): Reintroduce SEEN-less interface.
6290 (final): Rename to...
6291 (final_1): ... this. Take SEEN parameter. Output final pending
6292 next view at the end.
6293 (final): Reintroduce seen-less interface.
6294 (final_scan_insn): Output pending next view before switching
6295 sections or ending a block. Mark the next view as needed when
6296 outputting variable locations. Notify debug backend of section
6297 changes, and of location view changes.
6298 (rest_of_handle_final): Adjust.
6299 * toplev.c (process_options): Autodetect value for debug variable
6300 location views option. Warn on incompat5 without -gdwarf-5.
6301 * doc/invoke.texi (gvariable-location-views): New.
6302 (gvariable-location-views=incompat5): New.
6303 (gno-variable-location-views): New.
6305 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6307 PR tree-optimization/84136
6308 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6309 that the result of find_edge is non-NULL.
6311 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6313 PR target/83008
6314 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6315 storing integer register in SImode. Fix cost of 256 and 512
6316 byte aligned SSE register store.
6318 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6320 * config/i386/i386.c (ix86_multiplication_cost): Fix
6321 multiplication cost for TARGET_AVX512DQ.
6323 2018-02-08 Marek Polacek <polacek@redhat.com>
6325 PR tree-optimization/84238
6326 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6327 get_range_strlen.
6329 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6331 PR tree-optimization/84265
6332 * tree-vect-stmts.c (vectorizable_store): Don't treat
6333 VMAT_CONTIGUOUS accesses as grouped.
6334 (vectorizable_load): Likewise.
6336 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6338 PR tree-optimization/81635
6339 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6340 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6341 (test_round_for_mask): New functions.
6342 (wide_int_cc_tests): Call test_round_for_mask.
6343 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6344 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6345 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6346 range returned by get_range_info.
6348 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6350 PR ipa/81360
6351 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6352 * symtab.c: Include builtins.h
6353 (symtab_node::output_to_lto_symbol_table_p): Move here
6354 from lto-streamer-out.c:output_symbol_p.
6355 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6356 (output_symbol_p): Move all logic to symtab.c
6357 (produce_symtab): Update.
6359 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6361 * config/s390/s390-opts.h (enum indirect_branch): Define.
6362 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6363 (s390_indirect_branch_via_thunk)
6364 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6365 (enum s390_indirect_branch_type): Define.
6366 * config/s390/s390.c (struct s390_frame_layout, struct
6367 machine_function): Remove.
6368 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6369 (indirect_branch_table_label_no, indirect_branch_table_name):
6370 Define variables.
6371 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6372 (enum s390_indirect_branch_option): Define.
6373 (s390_return_addr_from_memory): New function.
6374 (s390_handle_string_attribute): New function.
6375 (s390_attribute_table): Add new attribute handler.
6376 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6377 (s390_indirect_branch_via_thunk): New function.
6378 (s390_indirect_branch_via_inline_thunk): New function.
6379 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6380 sibling call optimization for non z10 compiles.
6381 (s390_emit_call): Force indirect branch target to be a single
6382 register. Add r1 clobber for non-z10 compiles.
6383 (s390_emit_epilogue): Emit return jump via return_use expander.
6384 (s390_reorg): Handle JUMP_INSNs as execute targets.
6385 (s390_option_override_internal): Perform validity checks for the
6386 new command line options.
6387 (s390_indirect_branch_attrvalue): New function.
6388 (s390_indirect_branch_settings): New function.
6389 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6390 (s390_output_indirect_thunk_function): New function.
6391 (s390_code_end): Implement target hook.
6392 (s390_case_values_threshold): Implement target hook.
6393 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6394 macros.
6395 * config/s390/s390.h (struct s390_frame_layout)
6396 (struct machine_function): Move here from s390.c.
6397 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6398 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6399 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6400 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6401 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6402 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6403 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6404 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6405 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6406 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6407 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6408 (mnemonic attribute): Add values which aren't recognized
6409 automatically.
6410 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6411 pattern for branch conversion. Fix mnemonic attribute.
6412 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6413 indirect branch via thunk if requested.
6414 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6415 ("*indirect_jump"): Disable for branch conversion using out of
6416 line thunks.
6417 ("indirect_jump_via_thunk<mode>_z10")
6418 ("indirect_jump_via_thunk<mode>")
6419 ("indirect_jump_via_inlinethunk<mode>_z10")
6420 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6421 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6422 ("casesi_jump_via_inlinethunk<mode>_z10")
6423 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6424 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6425 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6426 ("*indirect2_jump"): Disable for branch conversion.
6427 ("casesi_jump"): Turn into expander and expand patterns for branch
6428 conversion.
6429 ("return_use"): New expander.
6430 ("*return"): Emit return via thunk and rename it to ...
6431 ("*return<mode>"): ... this one.
6432 * config/s390/s390.opt: Add new options and and enum for the
6433 option values.
6435 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6437 * lra-constraints.c (match_reload): Unconditionally use
6438 gen_lowpart_SUBREG, rather than selecting between that
6439 and equivalent gen_rtx_SUBREG code.
6441 2018-02-08 Richard Biener <rguenther@suse.de>
6443 PR tree-optimization/84233
6444 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6445 changed flag instead of boguously re-using phi_inserted.
6447 2018-02-08 Martin Jambor <mjambor@suse.cz>
6449 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6450 static local variables.
6452 2018-02-08 Richard Biener <rguenther@suse.de>
6454 PR tree-optimization/84278
6455 * tree-vect-stmts.c (vectorizable_store): When looking for
6456 smaller vector types to perform grouped strided loads/stores
6457 make sure the mode is supported by the target.
6458 (vectorizable_load): Likewise.
6460 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6462 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6463 Increase LDP/STP opportunities by adding adjacent callee-saves.
6465 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6467 PR rtl-optimization/84068
6468 PR rtl-optimization/83459
6469 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6471 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6473 PR tree-optimization/84224
6474 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6475 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6476 non-zero arguments.
6478 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6480 PR target/84113
6481 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6482 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6483 count, remove one USE.
6485 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6487 * doc/install.texi (Configuration): Document the
6488 --with-long-double-format={ibm,ieee} PowerPC configuration
6489 options.
6491 PR target/84154
6492 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6493 Convert from define_expand to be define_insn_and_split. Rework
6494 float/double/_Float128 conversions to QI/HI/SImode to work with
6495 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6496 conversions to QI/HImode types did a store and then a load to
6497 truncate the value. For conversions to VSX registers, don't split
6498 the insn, instead emit the code directly. Use the code iterator
6499 any_fix to combine signed and unsigned conversions.
6500 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6501 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6502 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6503 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6504 (fix_<mode>di2_hw): Likewise.
6505 (fixuns_<mode>di2_hw): Likewise.
6506 (fix_<mode>si2_hw): Likewise.
6507 (fixuns_<mode>si2_hw): Likewise.
6508 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6509 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6510 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6511 fix<uns>_trunc<SFDF:mode>si2_p8.
6512 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6513 used.
6514 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6515 (fix<uns>_<mode>_mem): Likewise.
6516 (fctiw<u>z_<mode>_mem): Likewise.
6517 (fix<uns>_<mode>_mem): Likewise.
6518 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6519 the register allocator from doing a direct move to the GPRs to do
6520 a store, and instead use the ISA 3.0 store byte/half-word from
6521 vector register instruction. For IEEE 128-bit floating point,
6522 also optimize stores of 32-bit ints.
6523 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6525 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6527 * genextract.c (push_pathstr_operand): New function to support
6528 [a-zA-Z].
6529 (walk_rtx): Call push_pathstr_operand.
6530 (print_path): Support [a-zA-Z].
6532 2018-02-07 Richard Biener <rguenther@suse.de>
6534 PR tree-optimization/84037
6535 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6536 (cse_and_gimplify_to_preheader): Declare.
6537 (vect_get_place_in_interleaving_chain): Likewise.
6538 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6539 ivexpr_map.
6540 (_loop_vec_info::~_loop_vec_info): Delete it.
6541 (cse_and_gimplify_to_preheader): New function.
6542 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6543 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6544 (vectorizable_load): Likewise. For grouped stores always base
6545 the IV on the first element.
6546 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6547 condition before gimplifying.
6549 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6551 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6552 *DIV_EXPR and *MOD_EXPR.
6554 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6556 PR target/84248
6557 * config/i386/i386.c (ix86_option_override_internal): Mask out
6558 the CF_SET bit when checking -fcf-protection.
6560 2018-02-07 Tom de Vries <tom@codesourcery.com>
6562 PR libgomp/84217
6563 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6564 enough.
6566 2018-02-07 Richard Biener <rguenther@suse.de>
6568 PR tree-optimization/84204
6569 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6570 this place.
6572 PR tree-optimization/84205
6573 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6574 special-case isl_ast_op_zdiv_r.
6576 PR tree-optimization/84223
6577 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6578 Only add conditions from within the region.
6579 (gather_bbs::after_dom_children): Adjust.
6581 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6583 PR target/84209
6584 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6585 * config/avr/avr.md: Only post-reload split REG-REG moves if
6586 either register is GENERAL_REG_P.
6588 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6590 PR tree-optimization/84235
6591 * tree-ssa-scopedtables.c
6592 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6593 if the subtraction is performed in floating point type where NaNs are
6594 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6595 build 1. Formatting fix.
6597 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6599 PR target/84146
6600 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6601 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6602 and skip it regardless of bb boundaries. Use CALL_P macro,
6603 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6604 unnecessarily, formatting fix.
6606 2018-02-06 Michael Collison <michael.collison@arm.com>
6608 * config/arm/thumb2.md:
6609 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6610 (*thumb_mov_notscc): Ditto.
6612 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6614 PR target/84154
6615 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6616 unsigned_fix, not "s".
6618 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6620 * configure.ac (gcc_fn_eh_frame_ro): New function.
6621 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6622 correct .eh_frame permissions.
6623 * configure: Regenerate.
6625 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6627 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6628 irrelevant options.
6630 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6632 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6633 Display warning message for -mno-speculate-indirect-jumps.
6635 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6637 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6638 Undocumented.
6639 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6641 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6643 PR tree-optimization/84225
6644 * tree-eh.c (find_trapping_overflow): Only call
6645 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6647 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6649 PR target/84145
6650 * config/i386/i386.c: Reimplement the check of possible options
6651 -mibt/-mshstk conbination. Change error messages.
6652 * doc/invoke.texi: Fix a typo: remove extra '='.
6654 2018-02-06 Marek Polacek <polacek@redhat.com>
6656 PR tree-optimization/84228
6657 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6659 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6661 PR target/82641
6662 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6663 emitted arch directives.
6664 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6665 __ARM_FEATURE_COPROC before changing architectures.
6667 2018-02-06 Richard Biener <rguenther@suse.de>
6669 * config/i386/i386.c (print_reg): Fix typo.
6670 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6672 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6674 * configure: Regenerate.
6676 2018-02-05 Martin Sebor <msebor@redhat.com>
6678 PR tree-optimization/83369
6679 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6680 inlining context.
6682 2018-02-05 Martin Liska <mliska@suse.cz>
6684 * doc/invoke.texi: Cherry-pick upstream r323995.
6686 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6688 * ira.c (ira_init_register_move_cost): Adjust comment.
6690 2018-02-05 Martin Liska <mliska@suse.cz>
6692 PR gcov-profile/84137
6693 * doc/gcov.texi: Fix typo in documentation.
6695 2018-02-05 Martin Liska <mliska@suse.cz>
6697 PR gcov-profile/83879
6698 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6699 using dlopen functionality.
6701 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6703 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6704 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6705 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6706 _mm_maskz_range_round_ss): New intrinsics.
6707 (__builtin_ia32_rangesd128_round)
6708 (__builtin_ia32_rangess128_round): Remove.
6709 (__builtin_ia32_rangesd128_mask_round,
6710 __builtin_ia32_rangess128_mask_round): New builtins.
6711 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6712 __builtin_ia32_rangess128_round): Remove.
6713 (__builtin_ia32_rangesd128_mask_round,
6714 __builtin_ia32_rangess128_mask_round): New builtins.
6715 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6716 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6717 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6718 "<round_saeonly_constraint>")): Changed to ...
6719 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6720 "<round_saeonly_scalar_constraint>")): ... this.
6721 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6722 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6723 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6724 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6725 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6727 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6729 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6730 options.
6731 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6732 Remove all values except native, 8540 and 8548.
6734 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6736 * config/i386/i386.c (ix86_output_function_return): Pass
6737 INVALID_REGNUM, instead of -1, as invalid register number to
6738 indirect_thunk_name and output_indirect_thunk.
6740 2018-02-02 Julia Koval <julia.koval@intel.com>
6742 * config.gcc: Add -march=icelake.
6743 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6744 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6745 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6746 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6747 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6748 (processor_target_table): Add icelake.
6749 (ix86_option_override_internal): Handle new PTAs.
6750 (get_builtin_code_for_version): Handle icelake.
6751 (M_INTEL_COREI7_ICELAKE): New.
6752 (fold_builtin_cpu): Handle icelake.
6753 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6754 * doc/invoke.texi: Add -march=icelake.
6756 2018-02-02 Julia Koval <julia.koval@intel.com>
6758 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6759 to wide_int_bitmask.
6760 * wide-int-bitmask.h: New.
6762 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6764 PR target/84066
6765 * config/i386/i386.md: Replace Pmode with word_mode in
6766 builtin_setjmp_setup and builtin_longjmp to support x32.
6768 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6770 PR target/56010
6771 PR target/83743
6772 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6773 #include "opts.h".
6774 (rs6000_supported_cpu_names): New static variable.
6775 (linux_cpu_translation_table): Likewise.
6776 (elf_platform) <cpu>: Define new static variable and use it.
6777 Translate kernel AT_PLATFORM name to canonical name if needed.
6778 Error if platform name is unknown.
6780 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6782 PR target/84089
6783 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6785 2018-02-01 Jeff Law <law@redhat.com>
6787 PR target/84128
6788 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6789 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6790 the scratch if RELEASE_VIA_POP is false.
6791 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6792 If we have to save a temporary register, decrement SIZE appropriately.
6793 Pass new arguments to release_scratch_register_on_entry.
6794 (ix86_adjust_stack_and_probe): Likewise.
6795 (ix86_emit_probe_stack_range): Pass new arguments to
6796 release_scratch_register_on_entry.
6798 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6800 PR rtl-optimization/84157
6801 * combine.c (change_zero_ext): Use REG_P predicate in
6802 front of HARD_REGISTER_P predicate.
6804 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6806 * config/avr/avr.c (avr_option_override): Move disabling of
6807 -fdelete-null-pointer-checks to...
6808 * common/config/avr/avr-common.c (avr_option_optimization_table):
6809 ...here.
6811 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6813 PR tree-optimization/81635
6814 * tree-data-ref.c (split_constant_offset_1): For types that
6815 wrap on overflow, try to use range info to prove that wrapping
6816 cannot occur.
6818 2018-02-01 Renlin Li <renlin.li@arm.com>
6820 PR target/83370
6821 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6822 TAILCALL_ADDR_REGS.
6823 (aarch64_register_move_cost): Likewise.
6824 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6825 TAILCALL_ADDR_REGS.
6826 (REG_CLASS_NAMES): Likewise.
6827 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6828 TAILCALL_ADDR_REGS. Remove IP registers.
6829 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6831 2018-02-01 Richard Biener <rguenther@suse.de>
6833 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6834 for specifying RPO order and allow NULL for that.
6835 * domwalk.c (dom_walker::dom_walker): Likewise.
6836 (dom_walker::walk): Handle NULL RPO order.
6837 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6838 in RPO order.
6839 (rewrite_update_dom_walker): Likewise.
6840 (mark_def_dom_walker): Likewise.
6842 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6844 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6845 (aarch64_maybe_expand_sve_subreg_move): Declare.
6846 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6847 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6848 predicate.
6849 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6850 that are semantically a reverse operation.
6851 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6852 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6853 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6854 functions.
6855 (aarch64_can_change_mode_class): For big-endian, forbid changes
6856 between two SVE modes if they have different element sizes.
6858 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6860 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6861 the TImode handling for big-endian targets.
6863 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6865 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6866 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6867 not just bytes.
6868 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6869 Remove BSWAP handing for big-endian targets and use the form of
6870 LD1RQ appropariate for the mode.
6872 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6874 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6875 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6876 duplicated element.
6878 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6880 PR tearget/83845
6881 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6882 check for operands that need to go through aarch64_sve_reload_be.
6884 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6886 PR tree-optimization/81661
6887 PR tree-optimization/84117
6888 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6889 * tree-eh.c: Include gimplify.h.
6890 (find_trapping_overflow, replace_trapping_overflow,
6891 rewrite_to_non_trapping_overflow): New functions.
6892 * tree-vect-loop.c: Include tree-eh.h.
6893 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6894 * tree-data-ref.c: Include tree-eh.h.
6895 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6897 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6899 PR rtl-optimization/84123
6900 * combine.c (change_zero_ext): Check if hard register satisfies
6901 can_change_dest_mode before calling gen_lowpart_SUBREG.
6903 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6905 PR target/82444
6906 * ira.c (ira_init_register_move_cost): Remove assert.
6908 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6910 PR rtl-optimization/84071
6911 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6912 * doc/tm.texi: Regenerate.
6914 2018-01-31 Richard Biener <rguenther@suse.de>
6916 PR tree-optimization/84132
6917 * tree-data-ref.c (analyze_miv_subscript): Properly
6918 check whether evolution_function_is_affine_multivariate_p
6919 before calling gcd_of_steps_may_divide_p.
6921 2018-01-31 Julia Koval <julia.koval@intel.com>
6923 PR target/83618
6924 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6925 * config/i386/i386.md (rdpid_rex64) New.
6926 (rdpid): Make 32bit only.
6928 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6930 PR lto/84105
6931 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6932 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6934 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6936 Revert
6937 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6939 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6941 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6943 PR rtl-optimization/84071
6944 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6945 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6947 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6949 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6950 (arc_attribute_table): Add 'aux' attribute.
6951 (arc_in_small_data_p): Consider aux like variables.
6952 (arc_is_aux_reg_p): New function.
6953 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6954 (arc_get_aux_arg): New function.
6955 (prepare_move_operands): Handle aux-register access.
6956 (arc_handle_aux_attribute): New function.
6957 * doc/extend.texi (ARC Variable attributes): Add subsection.
6959 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6961 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6962 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6963 (arc_attribute_table): Add 'uncached' attribute.
6964 (arc_print_operand): Print '.di' flag for uncached memory
6965 accesses.
6966 (arc_in_small_data_p): Do not consider for small data the uncached
6967 types.
6968 (arc_is_uncached_mem_p): New function.
6969 * config/arc/predicates.md (compact_store_memory_operand): Check
6970 for uncached memory accesses.
6971 (nonvol_nonimm_operand): Likewise.
6972 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6974 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6976 PR c/84100
6977 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6978 falign-loops=): Add Optimization flag.
6980 2018-01-30 Jeff Law <law@redhat.com>
6982 PR target/84064
6983 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6984 INT_REGISTERS_SAVED. Check it prior to calling
6985 get_scratch_register_on_entry.
6986 (ix86_adjust_stack_and_probe): Similarly.
6987 (ix86_emit_probe_stack_range): Similarly.
6988 (ix86_expand_prologue): Corresponding changes.
6990 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6992 PR target/40411
6993 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6994 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6996 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6998 PR target/84112
6999 * lra-constraints.c (curr_insn_transform): Process AND in the
7000 address.
7002 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7004 PR rtl-optimization/83986
7005 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7006 dependence against last_pending_memory_flush in addition to
7007 pending_jump_insns.
7009 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7011 PR tree-optimization/81611
7012 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7013 copies.
7015 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7017 PR target/83758
7018 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7019 a reg rtx.
7021 2018-01-30 Richard Biener <rguenther@suse.de>
7022 Jakub Jelinek <jakub@redhat.com>
7024 PR tree-optimization/84111
7025 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7026 inner loops added during recursion, as they don't have up-to-date
7027 SSA form.
7029 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7031 PR ipa/81360
7032 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7033 (can_inline_edge_by_limits_p): ... here.
7034 (can_early_inline_edge_p, check_callers,
7035 update_caller_keys, update_callee_keys, recursive_inlining,
7036 add_new_edges_to_heap, speculation_useful_p,
7037 inline_small_functions,
7038 inline_small_functions, flatten_function,
7039 inline_to_all_callers_1): Update.
7041 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7043 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7044 zeros correctly.
7046 2018-01-30 Richard Biener <rguenther@suse.de>
7048 PR tree-optimization/83008
7049 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7050 invariant and constant vector uses in stmts when they need
7051 more than one stmt.
7053 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7055 PR bootstrap/84017
7056 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7057 * configure: Regenerate.
7059 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7061 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7062 pattern.
7063 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7064 Use gen_rtx_REG rather than gen_lowpart.
7066 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7068 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7069 rather than 0 when creating partial subregs.
7071 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7073 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7074 of usage.
7076 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7078 PR target/81550
7079 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7080 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7081 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7082 flags. This restores the settings used before the 2017-07-24.
7083 Turning off pre increment/decrement/modify allows IVOPTS to
7084 optimize DF/SF loops where the index is an int.
7086 2018-01-29 Richard Biener <rguenther@suse.de>
7087 Kelvin Nilsen <kelvin@gcc.gnu.org>
7089 PR bootstrap/80867
7090 * tree-vect-stmts.c (vectorizable_call): Don't call
7091 targetm.vectorize_builtin_md_vectorized_function if callee is
7092 NULL.
7094 2018-01-22 Carl Love <cel@us.ibm.com>
7096 * doc/extend.tex: Fix typo in second arg in
7097 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7099 2018-01-29 Richard Biener <rguenther@suse.de>
7101 PR tree-optimization/84086
7102 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7103 (flush_ssaname_freelist): When SSA names were released reset
7104 the SCEV hash table.
7106 2018-01-29 Richard Biener <rguenther@suse.de>
7108 PR tree-optimization/84057
7109 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7110 removed paths when removing edges.
7112 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7114 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7115 -mfunction-return=@var{choice}.
7117 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7119 PR diagnostic/84034
7120 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7121 Handle CR like TAB.
7122 (layout::print_source_line): Likewise.
7123 (test_get_line_width_without_trailing_whitespace): Add test cases.
7125 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7127 PR middle-end/84040
7128 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7129 debug insns.
7131 2018-01-26 Jim Wilson <jimw@sifive.com>
7133 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7135 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7136 specified.
7138 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7140 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7141 and CMP + SUB-immediate -> SUBS.
7143 2018-01-26 Martin Sebor <msebor@redhat.com>
7145 PR tree-optimization/83896
7146 * tree-ssa-strlen.c (get_string_len): Rename...
7147 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7148 Avoid assuming length is constant.
7149 (handle_char_store): Use HOST_WIDE_INT for string length.
7151 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7153 PR target/81763
7154 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7155 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7157 2018-01-26 Richard Biener <rguenther@suse.de>
7159 PR rtl-optimization/84003
7160 * dse.c (record_store): Only record redundant stores when
7161 the earlier store aliases at least all accesses the later one does.
7163 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7165 PR rtl-optimization/83985
7166 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7167 REG_CFA_RESTORE insns.
7168 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7169 REG_CFA_RESTORE insns here.
7171 PR c/83989
7172 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7173 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7175 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7177 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7178 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7179 (arc_init): Likewise.
7180 (arc_override_options): Likewise.
7181 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7182 value.
7183 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7184 support.
7185 * config/arc/arc.h (TARGET_DBNZ): Define.
7186 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7187 properly set the tune attribute.
7188 (dbnz): Use TARGET_DBNZ guard.
7189 * config/arc/arc.opt (mtune): Add core3 option.
7191 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7193 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7194 recognize new pic like addresses.
7195 (arc_delegitimize_address): Clean up.
7197 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7199 * config/arc/arc-arches.def: Option mrf16 valid for all
7200 architectures.
7201 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7202 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7203 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7204 * config/arc/arc-tables.opt: Regenerate.
7205 * config/arc/arc.c (arc_conditional_register_usage): Handle
7206 reduced register file case.
7207 (arc_file_start): Set must have build attributes.
7208 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7209 mrf16 option value.
7210 * config/arc/arc.opt (mrf16): Add new option.
7211 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7212 * config/arc/genmultilib.awk: Handle new mrf16 option.
7213 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7214 * config/arc/t-multilib: Regenerate.
7215 * doc/invoke.texi (ARC Options): Document mrf16 option.
7217 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7219 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7220 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7221 (arc_attribute_table): Add 'secure_call' attribute.
7222 (arc_print_operand): Print secure call operand.
7223 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7224 secure.
7225 (arc_is_secure_call_p): New function. * config/arc/arc.md
7226 (call_i): Add support for sjli instruction.
7227 (call_value_i): Likewise.
7228 * config/arc/constraints.md (Csc): New constraint.
7230 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7231 John Eric Martin <John.Martin@emmicro-us.com>
7233 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7234 * config/arc/arc.c (_arc_jli_section): New struct.
7235 (arc_jli_section): New type.
7236 (rc_jli_sections): New static variable.
7237 (arc_handle_jli_attribute): New function.
7238 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7239 (arc_file_end): New function.
7240 (TARGET_ASM_FILE_END): Define.
7241 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7242 (arc_add_jli_section): New function.
7243 (jli_call_scan): Likewise.
7244 (arc_reorg): Call jli_call_scan.
7245 (arc_output_addsi): Remove 'S' from printing asm operand.
7246 (arc_is_jli_call_p): New function.
7247 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7248 operand.
7249 (movhi_insn): Likewise.
7250 (movsi_insn): Likewise.
7251 (movsi_set_cc_insn): Likewise.
7252 (loadqi_update): Likewise.
7253 (load_zeroextendqisi_update): Likewise.
7254 (load_signextendqisi_update): Likewise.
7255 (loadhi_update): Likewise.
7256 (load_zeroextendhisi_update): Likewise.
7257 (load_signextendhisi_update): Likewise.
7258 (loadsi_update): Likewise.
7259 (loadsf_update): Likewise.
7260 (movsicc_insn): Likewise.
7261 (bset_insn): Likewise.
7262 (bxor_insn): Likewise.
7263 (bclr_insn): Likewise.
7264 (bmsk_insn): Likewise.
7265 (bicsi3_insn): Likewise.
7266 (cmpsi_cc_c_insn): Likewise.
7267 (movsi_ne): Likewise.
7268 (movsi_cond_exec): Likewise.
7269 (clrsbsi2): Likewise.
7270 (norm_f): Likewise.
7271 (normw): Likewise.
7272 (swap): Likewise.
7273 (divaw): Likewise.
7274 (flag): Likewise.
7275 (sr): Likewise.
7276 (kflag): Likewise.
7277 (ffs): Likewise.
7278 (ffs_f): Likewise.
7279 (fls): Likewise.
7280 (call_i): Remove 'S' asm letter, add jli instruction.
7281 (call_value_i): Likewise.
7282 * config/arc/arc.op (mjli-always): New option.
7283 * config/arc/constraints.md (Cji): New constraint.
7284 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7285 operand.
7286 (subsf3_fpx): Likewise.
7287 (mulsf3_fpx): Likewise.
7288 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7289 asm operand.
7290 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7291 function attrbutes.
7292 * doc/invoke.texi (ARC): Document mjli-always option.
7294 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7296 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7297 avoid addition with 0 and use incw and decw where possible.
7299 2018-01-26 Richard Biener <rguenther@suse.de>
7301 PR tree-optimization/81082
7302 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7303 association if it requires casting to unsigned.
7304 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7305 from fold_plusminus_mult_expr to catch important cases late when
7306 range info is available.
7308 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7310 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7311 * configure.ac (hidden_linkonce): New test.
7312 * configure: Regenerate.
7313 * config.in: Regenerate.
7315 2018-01-26 Julia Koval <julia.koval@intel.com>
7317 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7318 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7319 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7320 _mm_mask_bitshuffle_epi64_mask): Fix type.
7321 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7322 USI_FTYPE_V4DI_V4DI_USI): Remove.
7323 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7324 __builtin_ia32_vpshufbitqmb256_mask,
7325 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7326 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7327 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7329 2018-01-26 Alan Modra <amodra@gmail.com>
7331 PR target/84033
7332 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7333 UNSPEC_VBPERMQ. Sort other unspecs.
7335 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7337 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7339 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7341 PR middle-end/83055
7342 * predict.c (drop_profile): Do not push/pop cfun; update also
7343 node->count.
7344 (handle_missing_profiles): Fix logic looking for zero profiles.
7346 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7348 PR middle-end/83977
7349 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7350 on functions with #pragma omp declare simd or functions with simd
7351 attribute.
7352 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7353 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7354 Remove trailing \n from warning_at calls.
7356 2018-01-25 Tom de Vries <tom@codesourcery.com>
7358 PR target/84028
7359 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7360 for neutered workers.
7362 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7364 PR target/68467
7365 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7366 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7368 2018-01-24 Jeff Law <law@redhat.com>
7370 PR target/83994
7371 * i386.c (get_probe_interval): Move to earlier point.
7372 (ix86_compute_frame_layout): If -fstack-clash-protection and
7373 the frame is larger than the probe interval, then use pushes
7374 to save registers rather than reg->mem moves.
7375 (ix86_expand_prologue): Remove conditional for int_registers_saved
7376 assertion.
7378 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7380 PR target/84014
7381 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7382 min/max for never referenced object.
7384 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7386 PR middle-end/83977
7387 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7388 here.
7389 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7390 attributes from DECL_ATTRIBUTES (decl) without affecting
7391 DECL_ATTRIBUTES (current_function_decl).
7392 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7393 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7395 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7397 PR tree-optimization/83979
7398 * fold-const.c (fold_comparison): Use constant_boolean_node
7399 instead of boolean_{true,false}_node.
7401 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7403 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7404 with zero counts.
7406 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7408 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7409 Simplify the clause that sets the length attribute.
7410 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7411 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7412 clause that sets the length attribute.
7413 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7415 2018-01-24 Tom de Vries <tom@codesourcery.com>
7417 PR target/83589
7418 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7419 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7420 Add strict parameter.
7421 (prevent_branch_around_nothing): Insert dummy insn between branch to
7422 label and label with no ptx insn inbetween.
7423 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7425 2018-01-24 Tom de Vries <tom@codesourcery.com>
7427 PR target/81352
7428 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7429 for neutered threads in warp.
7430 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7432 2018-01-24 Richard Biener <rguenther@suse.de>
7434 PR tree-optimization/83176
7435 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7436 operands.
7438 2018-01-24 Richard Biener <rguenther@suse.de>
7440 PR tree-optimization/82819
7441 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7442 code generating pluses that are no-ops in the target precision.
7444 2018-01-24 Richard Biener <rguenther@suse.de>
7446 PR middle-end/84000
7447 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7449 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7451 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7452 to merge probabilities.
7453 * predict.c (probably_never_executed): Also mark as cold functions
7454 with global 0 profile and guessed local profile.
7455 * profile-count.c (profile_probability::combine_with_count): New
7456 member function.
7457 * profile-count.h (profile_probability::operator*,
7458 profile_probability::operator*=, profile_probability::operator/,
7459 profile_probability::operator/=): Reduce precision to adjusted
7460 and set value to guessed on contradictory divisions.
7461 (profile_probability::combine_with_freq): Remove.
7462 (profile_probability::combine_wiht_count): Declare.
7463 (profile_count::force_nonzero):: Set to adjusted.
7464 (profile_count::probability_in):: Set quality to adjusted.
7465 * tree-ssa-tail-merge.c (replace_block_by): Use
7466 combine_with_count.
7468 2018-01-23 Andrew Waterman <andrew@sifive.com>
7469 Jim Wilson <jimw@sifive.com>
7471 * config/riscv/riscv.c (riscv_stack_boundary): New.
7472 (riscv_option_override): Set riscv_stack_boundary. Handle
7473 riscv_preferred_stack_boundary_arg.
7474 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7475 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7476 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7477 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7478 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7479 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7481 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7483 PR target/83905
7484 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7485 of struct ix86_frame.
7486 (ix86_expand_epilogue): Likewise. Add a local variable for
7487 the reg_save_offset field in struct ix86_frame.
7489 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7491 PR tree-optimization/82604
7492 * tree-loop-distribution.c (enum partition_kind): New enum item
7493 PKIND_PARTIAL_MEMSET.
7494 (partition_builtin_p): Support above new enum item.
7495 (generate_code_for_partition): Ditto.
7496 (compute_access_range): Differentiate cases that equality can be
7497 proven at all loops, the innermost loops or no loops.
7498 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7499 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7500 (finalize_partitions, distribute_loop): Don't fuse partition of
7501 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7502 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7503 parloop is enabled.
7505 2018-01-23 Martin Liska <mliska@suse.cz>
7507 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7508 order to ignore the predictor.
7509 (PRED_POLYMORPHIC_CALL): Likewise.
7510 (PRED_RECURSIVE_CALL): Likewise.
7512 2018-01-23 Martin Liska <mliska@suse.cz>
7514 * tree-profile.c (tree_profiling): Print function header to
7515 aware reader which function we are working on.
7516 * value-prof.c (gimple_find_values_to_profile): Do not print
7517 not interesting value histograms.
7519 2018-01-23 Martin Liska <mliska@suse.cz>
7521 * profile-count.h (enum profile_quality): Add
7522 profile_uninitialized as the first value. Do not number values
7523 as they are zero based.
7524 (profile_count::verify): Update sanity check.
7525 (profile_probability::verify): Likewise.
7527 2018-01-23 Nathan Sidwell <nathan@acm.org>
7529 * doc/invoke.texi (ffor-scope): Deprecate.
7531 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7533 PR tree-optimization/83510
7534 * domwalk.c (set_all_edges_as_executable): New function.
7535 (dom_walker::dom_walker): Convert bool param
7536 "skip_unreachable_blocks" to enum reachability. Move setup of
7537 edge flags to set_all_edges_as_executable and only do it when
7538 reachability is REACHABLE_BLOCKS.
7539 * domwalk.h (enum dom_walker::reachability): New enum.
7540 (dom_walker::dom_walker): Convert bool param
7541 "skip_unreachable_blocks" to enum reachability.
7542 (set_all_edges_as_executable): New decl.
7543 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7544 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7545 "reachability".
7546 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7547 but converting true to REACHABLE_BLOCKS.
7548 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7549 * tree-vrp.c
7550 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7551 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7552 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7553 REACHABLE_BLOCKS.
7554 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7555 if check_all_array_refs will be called.
7557 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7559 * tree.c (selftest::test_location_wrappers): Add more test
7560 coverage.
7562 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7564 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7565 (selftest::test_bit_in_range): Likewise.
7567 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7569 PR testsuite/83888
7570 * doc/sourcebuild.texi (vect_float): Say that the selector
7571 only describes the situation when -funsafe-math-optimizations is on.
7572 (vect_float_strict): Document.
7574 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7576 PR tree-optimization/83965
7577 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7578 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7579 instead of checking only for a reduction.
7580 (vect_recog_widen_sum_pattern): Likewise.
7582 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7584 * predict.c (probably_never_executed): Only use precise profile info.
7585 (compute_function_frequency): Skip after inlining hack since we now
7586 have quality checking.
7588 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7590 * profile-count.h (profile_probability::very_unlikely,
7591 profile_probability::unlikely, profile_probability::even): Set
7592 precision to guessed.
7594 2018-01-23 Richard Biener <rguenther@suse.de>
7596 PR tree-optimization/83963
7597 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7598 Properly terminate dominator walk when crossing the exit edge not
7599 when visiting its source block.
7601 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7603 PR c++/83918
7604 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7605 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7607 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7609 PR tree-optimization/83957
7610 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7611 semicolon after for body surrounded by braces.
7613 PR tree-optimization/83081
7614 * profile-count.h (profile_probability::split): New method.
7615 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7616 Use profile_probability::split.
7617 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7618 when splitting a single conditional jump into 2.
7620 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7622 PR tree-optimization/69452
7623 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7624 decl.
7626 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7628 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7629 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7630 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7632 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7634 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7635 * config/rl78/rl78.md (movdi): New define_expand.
7636 * config/rl78/rl78.c (rl78_split_movdi): New function.
7638 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7640 PR target/83862
7641 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7642 no longer used.
7643 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7644 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7645 128-bit to produce an UNSPEC move to get the double word with the
7646 signbit and then a shift directly to do signbit.
7647 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7648 implementation with a new version that just does either a direct
7649 move or a regular move. Move memory interface to separate insns.
7650 Move insns so they are next to the expander.
7651 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7652 with signbit move. Split big and little endian case.
7653 (signbit<mode>2_dm_mem_le): Likewise.
7654 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7655 (signbit<mode>2_dm2): Likewise.
7657 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7659 * config/rl78/rl78.md (anddi3): New define_expand.
7661 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7663 * config/rl78/rl78.md (umindi3): New define_expand.
7665 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7667 * config/rl78/rl78.md (smindi3): New define_expand.
7669 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7671 * config/rl78/rl78.md (smaxdi3): New define_expand.
7673 2018-01-22 Carl Love <cel@us.ibm.com>
7675 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7676 LVX_V1TI): Add macro expansion.
7677 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7678 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7679 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7680 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7681 Change check to determine if the instruction is a byte reversing
7682 entry. Fix typo in comment.
7683 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7684 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7685 Add def_builtin calls for new builtins.
7686 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7687 Add define_insn expansion.
7689 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7691 * config/rl78/rl78.md (umaxdi3): New define_expand.
7693 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7695 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7696 for non-QImode registers.
7698 2018-01-22 Richard Biener <rguenther@suse.de>
7700 PR tree-optimization/83963
7701 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7702 including the loop exit block.
7703 (scop_detection::merge_sese): Likewise.
7704 (scop_detection::add_scop): Do it here instead.
7706 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7708 * doc/sourcebuild.texi (arm_softfloat): Document.
7710 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7712 PR gcc/77734
7713 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7714 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7715 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7717 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7718 David Edelsohn <dje.gcc@gmail.com>
7720 PR target/83946
7721 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7722 Change "crset eq" to "crset 2".
7723 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7724 (*call_indirect_aix<mode>_nospec): Likewise.
7725 (*call_value_indirect_aix<mode>_nospec): Likewise.
7726 (*call_indirect_elfv2<mode>_nospec): Likewise.
7727 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7728 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7729 change assembly output from . to $.
7730 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7731 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7732 (*tablejump<mode>_internal1_nospec): Likewise.
7734 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7736 PR target/80870
7737 * config/sh/sh_optimize_sett_clrt.cc:
7738 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7740 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7742 PR tree-optimization/83940
7743 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7744 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7745 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7746 use it to pass back the definition type.
7747 (vect_check_store_rhs): Likewise rhs_dt_out.
7748 (vect_build_gather_load_calls): Add a mask_dt argument and use
7749 it instead of a call to vect_is_simple_use.
7750 (vectorizable_store): Update calls to vect_check_load_store_mask
7751 and vect_check_store_rhs. Use the dt returned by the latter instead
7752 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7753 instead of calls to vect_is_simple_use. Pass the scalar rather
7754 than the vector operand to vect_is_simple_use when handling
7755 second and subsequent copies of an rhs value.
7756 (vectorizable_load): Update calls to vect_check_load_store_mask
7757 and vect_build_gather_load_calls. Use the cached mask_dt and
7758 gs_info.offset_dt instead of calls to vect_is_simple_use.
7760 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7762 PR middle-end/83945
7763 * tree-emutls.c: Include gimplify.h.
7764 (lower_emutls_2): New function.
7765 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7766 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7767 it before further processing.
7769 PR target/83930
7770 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7771 UINTVAL (trueop1) instead of INTVAL (op1).
7773 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7775 PR debug/81570
7776 PR debug/83728
7777 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7778 INCOMING_FRAME_SP_OFFSET if not defined.
7779 (scan_trace): Add ENTRY argument. If true and
7780 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7781 emit a note to adjust the CFA offset.
7782 (create_cfi_notes): Adjust scan_trace callers.
7783 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7784 INCOMING_FRAME_SP_OFFSET in the CIE.
7785 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7786 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7787 Likewise.
7788 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7789 * doc/tm.texi: Regenerated.
7791 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7793 PR rtl-optimization/83147
7794 * lra-constraints.c (remove_inheritance_pseudos): Use
7795 lra_substitute_pseudo_within_insn.
7797 2018-01-19 Tom de Vries <tom@codesourcery.com>
7798 Cesar Philippidis <cesar@codesourcery.com>
7800 PR target/83920
7801 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7803 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7805 PR target/83790
7806 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7807 spaces for function labels.
7809 2018-01-19 Martin Liska <mliska@suse.cz>
7811 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7812 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7813 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7814 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7815 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7816 (PRED_CONST_RETURN): Change from 69 to 65.
7817 (PRED_NULL_RETURN): Change from 91 to 71.
7818 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7819 (PRED_LOOP_GUARD): Change from 66 to 73.
7821 2018-01-19 Martin Liska <mliska@suse.cz>
7823 * predict.c (predict_insn_def): Add new assert.
7824 (struct branch_predictor): Change type to signed integer.
7825 (test_prediction_value_range): Amend test to cover
7826 PROB_UNINITIALIZED.
7827 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7828 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7829 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7830 (PRED_LOOP_IV_COMPARE): Likewise.
7831 * predict.h (PROB_UNINITIALIZED): Define new constant.
7833 2018-01-19 Martin Liska <mliska@suse.cz>
7835 * predict.c (dump_prediction): Add new format for
7836 analyze_brprob.py script which is enabled with -details
7837 suboption.
7838 * profile-count.h (precise_p): New function.
7840 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7842 PR tree-optimization/83922
7843 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7844 there are no statements that need masking.
7845 (vect_active_double_reduction_p): New function.
7846 (vect_analyze_loop_operations): Use it when handling phis that
7847 are not in the loop header.
7849 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7851 PR tree-optimization/83914
7852 * tree-vect-loop.c (vectorizable_induction): Don't convert
7853 init_expr or apply the peeling adjustment for inductions
7854 that are nested within the vectorized loop.
7856 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7858 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7859 instead of NEG.
7861 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7863 PR sanitizer/81715
7864 PR testsuite/83882
7865 * function.h (gimplify_parameters): Add gimple_seq * argument.
7866 * function.c: Include gimple.h and options.h.
7867 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7868 for the added local temporaries if needed.
7869 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7870 if there are any parameter cleanups, wrap whole body into a
7871 try/finally with the cleanups.
7873 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7875 PR target/82964
7876 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7877 Use GET_MODE_CLASS for scalar floating point.
7879 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7881 PR ipa/82256
7882 patch by PaX Team
7883 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7884 Fix call of call_cgraph_insertion_hooks.
7886 2018-01-18 Martin Sebor <msebor@redhat.com>
7888 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7890 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7892 PR ipa/83619
7893 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7894 frequencies.
7896 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7898 PR other/70268
7899 * common.opt: (-ffile-prefix-map): New option.
7900 * opts.c (common_handle_option): Defer it.
7901 * opts-global.c (handle_common_deferred_options): Handle it.
7902 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7903 * file-prefix-map.h: New file.
7904 (remap_debug_filename, add_debug_prefix_map): ...here.
7905 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7906 * final.c (debug_prefix_map, add_debug_prefix_map
7907 remap_debug_filename): Move to...
7908 * file-prefix-map.c: New file.
7909 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7910 generalize, get rid of alloca(), use strrchr() instead of strchr().
7911 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7912 Implement in terms of add_prefix_map().
7913 (remap_macro_filename, remap_debug_filename): Implement in term of
7914 remap_filename().
7915 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7916 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7917 * dbxout.c: Include file-prefix-map.h.
7918 * varasm.c: Likewise.
7919 * vmsdbgout.c: Likewise.
7920 * xcoffout.c: Likewise.
7921 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7922 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7923 * doc/invoke.texi (-ffile-prefix-map): Document.
7924 (-fdebug-prefix-map): Update description.
7926 2018-01-18 Martin Liska <mliska@suse.cz>
7928 * config/i386/i386.c (indirect_thunk_name): Document that also
7929 lfence is emitted.
7930 (output_indirect_thunk): Document why both instructions
7931 (pause and lfence) are generated.
7933 2018-01-18 Richard Biener <rguenther@suse.de>
7935 PR tree-optimization/83887
7936 * graphite-scop-detection.c
7937 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7938 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7939 (scop_detection::merge_sese): Re-implement with a flood-fill
7940 algorithm that properly finds a SESE region if it exists.
7942 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7944 PR c/61240
7945 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7946 pointer_diff optimizations use view_convert instead of convert.
7948 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7950 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7951 Generate different code for -mno-speculate-indirect-jumps.
7952 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7953 (*call_indirect_aix<mode>): Disable for
7954 -mno-speculate-indirect-jumps.
7955 (*call_indirect_aix<mode>_nospec): New define_insn.
7956 (*call_value_indirect_aix<mode>): Disable for
7957 -mno-speculate-indirect-jumps.
7958 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7959 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7960 -mno-speculate-indirect-jumps.
7961 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7963 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7965 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7966 long double type, set the flags for noting the default long double
7967 type, even if we don't pass or return a long double type.
7969 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7971 PR ipa/83051
7972 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7973 failure.
7975 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7977 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7978 support for merge[hl].
7979 (fold_mergehl_helper): New helper function.
7980 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7981 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7982 (altivec_vmrglw_direct): Add xxmrglw insn.
7984 2018-01-17 Andrew Waterman <andrew@sifive.com>
7986 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7987 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7989 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7991 PR lto/83121
7992 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7993 call the lto_location_cache before reading the
7994 DECL_SOURCE_LOCATION of the types.
7996 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7997 Richard Sandiford <richard.sandiford@linaro.org>
7999 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8000 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8001 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8002 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8003 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8004 Add declaration.
8005 * config/aarch64/constraints.md (aarch64_movti_operand):
8006 Limit immediates.
8007 * config/aarch64/predicates.md (Uti): Add new constraint.
8009 2018-01-17 Carl Love <cel@us.ibm.com>
8011 * config/rs6000/vsx.md (define_expand xl_len_r,
8012 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8013 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8014 lxvll.
8015 (define_expand, define_insn): Move the shift left from the
8016 define_insn to the define_expand for lxvl and stxvl instructions.
8017 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8018 and XL_LEN_R definitions to PURE.
8020 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8022 * config/i386/i386.c (indirect_thunk_name): Declare regno
8023 as unsigned int. Compare regno with INVALID_REGNUM.
8024 (output_indirect_thunk): Ditto.
8025 (output_indirect_thunk_function): Ditto.
8026 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8027 in the call to output_indirect_thunk_function.
8029 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8031 PR middle-end/83884
8032 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8033 rather than the size of inner_type to determine the stack slot size
8034 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8036 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8038 PR target/83546
8039 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8040 to PTA_SILVERMONT.
8042 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8044 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8045 endian Linux systems to optionally enable multilibs for selecting
8046 the long double type if the user configured an explicit type.
8047 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8048 have no long double multilibs if not defined.
8049 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8050 warn if the user used -mabi={ieee,ibm}longdouble and we built
8051 multilibs for long double.
8052 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8053 appropriate multilib option.
8054 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8055 multilib options.
8056 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8057 for building long double multilibs.
8058 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8060 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8062 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8063 copies.
8065 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8066 64 bits.
8067 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8068 128 bits.
8070 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8071 variables.
8073 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8074 return value.
8076 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8078 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8079 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8081 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8083 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8084 different rtl trees depending on TARGET_64BIT.
8085 (rs6000_gen_lvx): Likewise.
8087 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8089 * config/visium/visium.md (nop): Tweak comment.
8090 (hazard_nop): Likewise.
8092 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8094 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8095 -mspeculate-indirect-jumps.
8096 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8097 for -mno-speculate-indirect-jumps.
8098 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8099 (*call_value_indirect_elfv2<mode>): Disable for
8100 -mno-speculate-indirect-jumps.
8101 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8102 (indirect_jump): Emit different RTL for
8103 -mno-speculate-indirect-jumps.
8104 (*indirect_jump<mode>): Disable for
8105 -mno-speculate-indirect-jumps.
8106 (*indirect_jump<mode>_nospec): New define_insn.
8107 (tablejump): Emit different RTL for
8108 -mno-speculate-indirect-jumps.
8109 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8110 (tablejumpsi_nospec): New define_expand.
8111 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8112 (tablejumpdi_nospec): New define_expand.
8113 (*tablejump<mode>_internal1): Disable for
8114 -mno-speculate-indirect-jumps.
8115 (*tablejump<mode>_internal1_nospec): New define_insn.
8116 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8117 option.
8119 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8121 * caller-save.c (insert_save): Drop unnecessary parameter. All
8122 callers updated.
8124 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8125 Richard Biener <rguenth@suse.de>
8127 PR libgomp/83590
8128 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8129 return early, inline manually is_gimple_sizepos. Make sure if we
8130 call gimplify_expr we don't end up with a gimple constant.
8131 * tree.c (variably_modified_type_p): Don't return true for
8132 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8133 * gimplify.h (is_gimple_sizepos): Remove.
8135 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8137 PR tree-optimization/83857
8138 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8139 vectorizable_live_operation for pure SLP statements.
8140 (vectorizable_live_operation): Handle PHIs.
8142 2018-01-16 Richard Biener <rguenther@suse.de>
8144 PR tree-optimization/83867
8145 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8146 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8148 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8150 PR c/83844
8151 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8152 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8153 If off is not INTEGER_CST, issue a may not be aligned warning
8154 rather than isn't aligned. Use isn%'t rather than isn't.
8155 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8156 into MULT_EXPR.
8157 <case MULT_EXPR>: Improve the case when bottom and one of the
8158 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8159 operand, in that case check if the other operand is multiple of
8160 bottom divided by the INTEGER_CST operand.
8162 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8164 PR target/83858
8165 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8166 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8167 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8168 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8169 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8170 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8171 (pa_function_arg_size): New function.
8173 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8175 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8176 in a separate statement.
8178 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8180 PR tree-optimization/83847
8181 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8182 group gathers and scatters.
8184 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8186 PR rtl-optimization/86620
8187 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8189 PR rtl-optimization/83213
8190 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8191 to last if both are JUMP_INSNs.
8193 PR tree-optimization/83843
8194 * gimple-ssa-store-merging.c
8195 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8196 store_immediate_info for bswap/nop orig_stores.
8198 2018-01-15 Andrew Waterman <andrew@sifive.com>
8200 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8201 !TARGET_MUL.
8202 <UDIV>: Increase cost if !TARGET_DIV.
8204 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8206 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8207 (define_attr "cr_logical_3op"): New.
8208 (cceq_ior_compare): Adjust.
8209 (cceq_ior_compare_complement): Adjust.
8210 (*cceq_rev_compare): Adjust.
8211 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8212 (is_cracked_insn): Adjust.
8213 (insn_must_be_first_in_group): Adjust.
8214 * config/rs6000/40x.md: Adjust.
8215 * config/rs6000/440.md: Adjust.
8216 * config/rs6000/476.md: Adjust.
8217 * config/rs6000/601.md: Adjust.
8218 * config/rs6000/603.md: Adjust.
8219 * config/rs6000/6xx.md: Adjust.
8220 * config/rs6000/7450.md: Adjust.
8221 * config/rs6000/7xx.md: Adjust.
8222 * config/rs6000/8540.md: Adjust.
8223 * config/rs6000/cell.md: Adjust.
8224 * config/rs6000/e300c2c3.md: Adjust.
8225 * config/rs6000/e500mc.md: Adjust.
8226 * config/rs6000/e500mc64.md: Adjust.
8227 * config/rs6000/e5500.md: Adjust.
8228 * config/rs6000/e6500.md: Adjust.
8229 * config/rs6000/mpc.md: Adjust.
8230 * config/rs6000/power4.md: Adjust.
8231 * config/rs6000/power5.md: Adjust.
8232 * config/rs6000/power6.md: Adjust.
8233 * config/rs6000/power7.md: Adjust.
8234 * config/rs6000/power8.md: Adjust.
8235 * config/rs6000/power9.md: Adjust.
8236 * config/rs6000/rs64.md: Adjust.
8237 * config/rs6000/titan.md: Adjust.
8239 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8241 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8242 ix86_indirect_branch_register logic.
8244 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8246 * config/i386/constraints.md (Bs): Update
8247 ix86_indirect_branch_register check. Don't check
8248 ix86_indirect_branch_register with GOT_memory_operand.
8249 (Bw): Likewise.
8250 * config/i386/predicates.md (GOT_memory_operand): Don't check
8251 ix86_indirect_branch_register here.
8252 (GOT32_symbol_operand): Likewise.
8254 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8256 * config/i386/predicates.md (constant_call_address_operand):
8257 Rewrite ix86_indirect_branch_register logic.
8258 (sibcall_insn_operand): Likewise.
8260 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8262 * config/i386/constraints.md (Bs): Replace
8263 ix86_indirect_branch_thunk_register with
8264 ix86_indirect_branch_register.
8265 (Bw): Likewise.
8266 * config/i386/i386.md (indirect_jump): Likewise.
8267 (tablejump): Likewise.
8268 (*sibcall_memory): Likewise.
8269 (*sibcall_value_memory): Likewise.
8270 Peepholes of indirect call and jump via memory: Likewise.
8271 * config/i386/i386.opt: Likewise.
8272 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8273 (GOT_memory_operand): Likewise.
8274 (call_insn_operand): Likewise.
8275 (sibcall_insn_operand): Likewise.
8276 (GOT32_symbol_operand): Likewise.
8278 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8280 PR middle-end/83837
8281 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8282 type rather than type addr's type points to.
8283 (expand_omp_atomic_mutex): Likewise.
8284 (expand_omp_atomic): Likewise.
8286 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8288 PR target/83839
8289 * config/i386/i386.c (output_indirect_thunk_function): Use
8290 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8291 for __x86_return_thunk.
8293 2018-01-15 Richard Biener <rguenther@suse.de>
8295 PR middle-end/83850
8296 * expmed.c (extract_bit_field_1): Fix typo.
8298 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8300 PR target/83687
8301 * config/arm/iterators.md (VF): New mode iterator.
8302 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8303 Remove integer-related logic from pattern.
8304 (neon_vabd<mode>_3): Likewise.
8306 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8308 PR middle-end/82694
8309 * common.opt (fstrict-overflow): No longer an alias.
8310 (fwrapv-pointer): New option.
8311 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8312 also for pointer types based on flag_wrapv_pointer.
8313 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8314 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8315 opts->x_flag_wrapv got set.
8316 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8317 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8318 POINTER_TYPE_OVERFLOW_UNDEFINED.
8319 * match.pd: Likewise in address comparison pattern.
8320 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8322 2018-01-15 Richard Biener <rguenther@suse.de>
8324 PR lto/83804
8325 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8326 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8327 Reset type names to their identifier if their TYPE_DECL doesn't
8328 have linkage (and thus is used for ODR and devirt).
8329 (save_debug_info_for_decl): Remove.
8330 (save_debug_info_for_type): Likewise.
8331 (add_tree_to_fld_list): Adjust.
8332 * tree-pretty-print.c (dump_generic_node): Make dumping of
8333 type names more robust.
8335 2018-01-15 Richard Biener <rguenther@suse.de>
8337 * BASE-VER: Bump to 8.0.1.
8339 2018-01-14 Martin Sebor <msebor@redhat.com>
8341 PR other/83508
8342 * builtins.c (check_access): Avoid warning when the no-warning bit
8343 is set.
8345 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8347 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8348 * ira-color (allocno_hard_regs_compare): Likewise.
8350 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8352 PR target/83013
8353 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8354 Use .pushsection/.popsection.
8356 2018-01-14 Martin Sebor <msebor@redhat.com>
8358 PR c++/81327
8359 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8361 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8363 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8364 entry from extra_headers.
8365 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8366 extra_headers, make the list bitwise identical to the i?86-*-* one.
8368 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8370 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8371 -mcmodel=large with -mindirect-branch=thunk,
8372 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8373 -mfunction-return=thunk-extern.
8374 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8375 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8376 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8378 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8380 * config/i386/i386.c (print_reg): Print the name of the full
8381 integer register without '%'.
8382 (ix86_print_operand): Handle 'V'.
8383 * doc/extend.texi: Document 'V' modifier.
8385 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8387 * config/i386/constraints.md (Bs): Disallow memory operand for
8388 -mindirect-branch-register.
8389 (Bw): Likewise.
8390 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8391 (GOT_memory_operand): Likewise.
8392 (call_insn_operand): Likewise.
8393 (sibcall_insn_operand): Likewise.
8394 (GOT32_symbol_operand): Likewise.
8395 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8396 for -mindirect-branch-register.
8397 (tablejump): Likewise.
8398 (*sibcall_memory): Likewise.
8399 (*sibcall_value_memory): Likewise.
8400 Disallow peepholes of indirect call and jump via memory for
8401 -mindirect-branch-register.
8402 (*call_pop): Replace m with Bw.
8403 (*call_value_pop): Likewise.
8404 (*sibcall_pop_memory): Replace m with Bs.
8405 * config/i386/i386.opt (mindirect-branch-register): New option.
8406 * doc/invoke.texi: Document -mindirect-branch-register option.
8408 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8410 * config/i386/i386-protos.h (ix86_output_function_return): New.
8411 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8412 set function_return_type.
8413 (indirect_thunk_name): Add ret_p to indicate thunk for function
8414 return.
8415 (output_indirect_thunk_function): Pass false to
8416 indirect_thunk_name.
8417 (ix86_output_indirect_branch_via_reg): Likewise.
8418 (ix86_output_indirect_branch_via_push): Likewise.
8419 (output_indirect_thunk_function): Create alias for function
8420 return thunk if regno < 0.
8421 (ix86_output_function_return): New function.
8422 (ix86_handle_fndecl_attribute): Handle function_return.
8423 (ix86_attribute_table): Add function_return.
8424 * config/i386/i386.h (machine_function): Add
8425 function_return_type.
8426 * config/i386/i386.md (simple_return_internal): Use
8427 ix86_output_function_return.
8428 (simple_return_internal_long): Likewise.
8429 * config/i386/i386.opt (mfunction-return=): New option.
8430 (indirect_branch): Mention -mfunction-return=.
8431 * doc/extend.texi: Document function_return function attribute.
8432 * doc/invoke.texi: Document -mfunction-return= option.
8434 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8436 * config/i386/i386-opts.h (indirect_branch): New.
8437 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8438 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8439 with local indirect jump when converting indirect call and jump.
8440 (ix86_set_indirect_branch_type): New.
8441 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8442 (indirectlabelno): New.
8443 (indirect_thunk_needed): Likewise.
8444 (indirect_thunk_bnd_needed): Likewise.
8445 (indirect_thunks_used): Likewise.
8446 (indirect_thunks_bnd_used): Likewise.
8447 (INDIRECT_LABEL): Likewise.
8448 (indirect_thunk_name): Likewise.
8449 (output_indirect_thunk): Likewise.
8450 (output_indirect_thunk_function): Likewise.
8451 (ix86_output_indirect_branch_via_reg): Likewise.
8452 (ix86_output_indirect_branch_via_push): Likewise.
8453 (ix86_output_indirect_branch): Likewise.
8454 (ix86_output_indirect_jmp): Likewise.
8455 (ix86_code_end): Call output_indirect_thunk_function if needed.
8456 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8457 needed.
8458 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8459 (ix86_attribute_table): Add indirect_branch.
8460 * config/i386/i386.h (machine_function): Add indirect_branch_type
8461 and has_local_indirect_jump.
8462 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8463 to true.
8464 (tablejump): Likewise.
8465 (*indirect_jump): Use ix86_output_indirect_jmp.
8466 (*tablejump_1): Likewise.
8467 (simple_return_indirect_internal): Likewise.
8468 * config/i386/i386.opt (mindirect-branch=): New option.
8469 (indirect_branch): New.
8470 (keep): Likewise.
8471 (thunk): Likewise.
8472 (thunk-inline): Likewise.
8473 (thunk-extern): Likewise.
8474 * doc/extend.texi: Document indirect_branch function attribute.
8475 * doc/invoke.texi: Document -mindirect-branch= option.
8477 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8479 PR ipa/83051
8480 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8482 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8484 * ipa-inline.c (want_inline_small_function_p): Return false if
8485 inlining has already failed with CIF_FINAL_ERROR.
8486 (update_caller_keys): Call want_inline_small_function_p before
8487 can_inline_edge_p.
8488 (update_callee_keys): Likewise.
8490 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8492 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8493 New function.
8494 (rs6000_quadword_masked_address_p): Likewise.
8495 (quad_aligned_load_p): Likewise.
8496 (quad_aligned_store_p): Likewise.
8497 (const_load_sequence_p): Add comment to describe the outer-most loop.
8498 (mimic_memory_attributes_and_flags): New function.
8499 (rs6000_gen_stvx): Likewise.
8500 (replace_swapped_aligned_store): Likewise.
8501 (rs6000_gen_lvx): Likewise.
8502 (replace_swapped_aligned_load): Likewise.
8503 (replace_swapped_load_constant): Capitalize argument name in
8504 comment describing this function.
8505 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8506 and stores that access quad-word aligned addresses and replace
8507 with stvx or lvx instructions when appropriate.
8508 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8509 New function prototype.
8510 (rs6000_quadword_masked_address_p): Likewise.
8511 (rs6000_gen_lvx): Likewise.
8512 (rs6000_gen_stvx): Likewise.
8513 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8514 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8515 when memory address is aligned.
8516 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8517 this split to select lvx instruction when memory address is aligned.
8518 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8519 instruction when memory address is aligned.
8520 (*vsx_le_perm_load_v16qi): Likewise.
8521 (four unnamed splitters): Modify to select the stvx instruction
8522 when memory is aligned.
8524 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8526 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8527 which appears in the queue multiple times.
8529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8530 Alan Hayward <alan.hayward@arm.com>
8531 David Sherwood <david.sherwood@arm.com>
8533 * tree-vectorizer.h (vec_lower_bound): New structure.
8534 (_loop_vec_info): Add check_nonzero and lower_bounds.
8535 (LOOP_VINFO_CHECK_NONZERO): New macro.
8536 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8537 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8538 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8539 fields. Make seg_len the distance travelled, not including the
8540 access size.
8541 (dr_direction_indicator): Declare.
8542 (dr_zero_step_indicator): Likewise.
8543 (dr_known_forward_stride_p): Likewise.
8544 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8545 tree-ssanames.h.
8546 (runtime_alias_check_p): Allow runtime alias checks with
8547 variable strides.
8548 (operator ==): Compare access_size and align.
8549 (prune_runtime_alias_test_list): Rework for new distinction between
8550 the access_size and seg_len.
8551 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8552 segment lengths.
8553 (get_segment_min_max): New function.
8554 (create_intersect_range_checks): Use it.
8555 (dr_step_indicator): New function.
8556 (dr_direction_indicator): Likewise.
8557 (dr_zero_step_indicator): Likewise.
8558 (dr_known_forward_stride_p): Likewise.
8559 * tree-loop-distribution.c (data_ref_segment_size): Return
8560 DR_STEP * (niters - 1).
8561 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8562 constructor.
8563 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8564 (vect_preserves_scalar_order_p): New function, split out from...
8565 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8566 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8567 (vect_vfa_access_size): New function.
8568 (vect_vfa_align): Likewise.
8569 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8570 (dump_lower_bound): New function.
8571 (vect_check_lower_bound): Likewise.
8572 (vect_small_gap_p): Likewise.
8573 (vectorizable_with_step_bound_p): Likewise.
8574 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8575 depencies if the vectorization factor is 1. Convert the checks
8576 for nonzero steps into checks on the bounds of DR_STEP. Try using
8577 a bunds check for variable steps if the minimum required step is
8578 relatively small. Update calls to the dr_with_seg_len
8579 constructor and to vect_compile_time_alias.
8580 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8581 function.
8582 (vect_loop_versioning): Call it.
8583 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8584 when retrying.
8585 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8588 Alan Hayward <alan.hayward@arm.com>
8589 David Sherwood <david.sherwood@arm.com>
8591 * doc/sourcebuild.texi (vect_scatter_store): Document.
8592 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8593 optabs.
8594 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8595 Document.
8596 * genopinit.c (main): Add supports_vec_scatter_store and
8597 supports_vec_scatter_store_cached to target_optabs.
8598 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8599 IFN_MASK_SCATTER_STORE.
8600 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8601 functions.
8602 * internal-fn.h (internal_store_fn_p): Declare.
8603 (internal_fn_stored_value_index): Likewise.
8604 * internal-fn.c (scatter_store_direct): New macro.
8605 (expand_scatter_store_optab_fn): New function.
8606 (direct_scatter_store_optab_supported_p): New macro.
8607 (internal_store_fn_p): New function.
8608 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8609 IFN_MASK_SCATTER_STORE.
8610 (internal_fn_mask_index): Likewise.
8611 (internal_fn_stored_value_index): New function.
8612 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8613 for scatter stores.
8614 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8615 * optabs-query.c (supports_vec_scatter_store_p): New function.
8616 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8617 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8618 true for scatter stores.
8619 (vect_gather_scatter_fn_p): Handle scatter stores too.
8620 (vect_check_gather_scatter): Consider using scatter stores if
8621 supports_vec_scatter_store_p.
8622 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8623 scatter stores too.
8624 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8625 internal_fn_stored_value_index.
8626 (check_load_store_masking): Handle scatter stores too.
8627 (vect_get_store_rhs): Make public.
8628 (vectorizable_call): Use internal_store_fn_p.
8629 (vectorizable_store): Handle scatter store internal functions.
8630 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8631 when deciding whether the end of the group has been reached.
8632 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8633 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8634 (mask_scatter_store<mode>): New insns.
8636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8637 Alan Hayward <alan.hayward@arm.com>
8638 David Sherwood <david.sherwood@arm.com>
8640 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8641 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8642 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8643 function.
8644 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8645 Use vect_truncate_gather_scatter_offset if we can't treat the
8646 operation as a normal gather load or scatter store.
8647 (get_group_load_store_type): Take the gather_scatter_info
8648 as argument. Try using a gather load or scatter store for
8649 single-element groups.
8650 (get_load_store_type): Update calls to get_group_load_store_type
8651 and vect_use_strided_gather_scatters_p.
8653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8654 Alan Hayward <alan.hayward@arm.com>
8655 David Sherwood <david.sherwood@arm.com>
8657 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8658 optional tree argument.
8659 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8660 null target hooks.
8661 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8662 but continue to use the current value as a fallback.
8663 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8664 to compare the updates.
8665 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8666 (get_load_store_type): Use it when handling a strided access.
8667 (vect_get_strided_load_store_ops): New function.
8668 (vect_get_data_ptr_increment): Likewise.
8669 (vectorizable_load): Handle strided gather loads. Always pass
8670 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8672 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8673 Alan Hayward <alan.hayward@arm.com>
8674 David Sherwood <david.sherwood@arm.com>
8676 * doc/md.texi (gather_load@var{m}): Document.
8677 (mask_gather_load@var{m}): Likewise.
8678 * genopinit.c (main): Add supports_vec_gather_load and
8679 supports_vec_gather_load_cached to target_optabs.
8680 * optabs-tree.c (init_tree_optimization_optabs): Use
8681 ggc_cleared_alloc to allocate target_optabs.
8682 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8683 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8684 functions.
8685 * internal-fn.h (internal_load_fn_p): Declare.
8686 (internal_gather_scatter_fn_p): Likewise.
8687 (internal_fn_mask_index): Likewise.
8688 (internal_gather_scatter_fn_supported_p): Likewise.
8689 * internal-fn.c (gather_load_direct): New macro.
8690 (expand_gather_load_optab_fn): New function.
8691 (direct_gather_load_optab_supported_p): New macro.
8692 (direct_internal_fn_optab): New function.
8693 (internal_load_fn_p): Likewise.
8694 (internal_gather_scatter_fn_p): Likewise.
8695 (internal_fn_mask_index): Likewise.
8696 (internal_gather_scatter_fn_supported_p): Likewise.
8697 * optabs-query.c (supports_at_least_one_mode_p): New function.
8698 (supports_vec_gather_load_p): Likewise.
8699 * optabs-query.h (supports_vec_gather_load_p): Declare.
8700 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8701 and memory_type field.
8702 (NUM_PATTERNS): Bump to 15.
8703 * tree-vect-data-refs.c: Include internal-fn.h.
8704 (vect_gather_scatter_fn_p): New function.
8705 (vect_describe_gather_scatter_call): Likewise.
8706 (vect_check_gather_scatter): Try using internal functions for
8707 gather loads. Recognize existing calls to a gather load function.
8708 (vect_analyze_data_refs): Consider using gather loads if
8709 supports_vec_gather_load_p.
8710 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8711 (vect_get_gather_scatter_offset_type): Likewise.
8712 (vect_convert_mask_for_vectype): Likewise.
8713 (vect_add_conversion_to_patterm): Likewise.
8714 (vect_try_gather_scatter_pattern): Likewise.
8715 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8716 (vect_vect_recog_func_ptrs): Add it.
8717 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8718 internal_fn_mask_index and internal_gather_scatter_fn_p.
8719 (check_load_store_masking): Take the gather_scatter_info as an
8720 argument and handle gather loads.
8721 (vect_get_gather_scatter_ops): New function.
8722 (vectorizable_call): Check internal_load_fn_p.
8723 (vectorizable_load): Likewise. Handle gather load internal
8724 functions.
8725 (vectorizable_store): Update call to check_load_store_masking.
8726 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8727 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8728 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8729 (aarch64_gather_scale_operand_d): New predicates.
8730 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8731 (mask_gather_load<mode>): New insns.
8733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8734 Alan Hayward <alan.hayward@arm.com>
8735 David Sherwood <david.sherwood@arm.com>
8737 * optabs.def (fold_left_plus_optab): New optab.
8738 * doc/md.texi (fold_left_plus_@var{m}): Document.
8739 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8740 * internal-fn.c (fold_left_direct): Define.
8741 (expand_fold_left_optab_fn): Likewise.
8742 (direct_fold_left_optab_supported_p): Likewise.
8743 * fold-const-call.c (fold_const_fold_left): New function.
8744 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8745 * tree-parloops.c (valid_reduction_p): New function.
8746 (gather_scalar_reductions): Use it.
8747 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8748 (vect_finish_replace_stmt): Declare.
8749 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8750 (needs_fold_left_reduction_p): New function, split out from...
8751 (vect_is_simple_reduction): ...here. Accept reductions that
8752 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8753 (vect_force_simple_reduction): Also store the reduction type in
8754 the assignment's STMT_VINFO_REDUC_TYPE.
8755 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8756 (merge_with_identity): New function.
8757 (vect_expand_fold_left): Likewise.
8758 (vectorize_fold_left_reduction): Likewise.
8759 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8760 scalar phi in place for it. Check for target support and reject
8761 cases that would reassociate the operation. Defer the transform
8762 phase to vectorize_fold_left_reduction.
8763 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8764 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8765 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8769 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8770 call to ifc_temp_var.
8772 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8773 Alan Hayward <alan.hayward@arm.com>
8774 David Sherwood <david.sherwood@arm.com>
8776 * target.def (legitimize_address_displacement): Take the original
8777 offset as a poly_int.
8778 * targhooks.h (default_legitimize_address_displacement): Update
8779 accordingly.
8780 * targhooks.c (default_legitimize_address_displacement): Likewise.
8781 * doc/tm.texi: Regenerate.
8782 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8783 as an argument, moving assert of ad->disp == ad->disp_term to...
8784 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8785 Try calling targetm.legitimize_address_displacement before expanding
8786 the address rather than afterwards, and adjust for the new interface.
8787 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8788 Match the new hook interface. Handle SVE addresses.
8789 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8790 new hook interface.
8792 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8794 * Makefile.in (OBJS): Add early-remat.o.
8795 * target.def (select_early_remat_modes): New hook.
8796 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8797 * doc/tm.texi: Regenerate.
8798 * targhooks.h (default_select_early_remat_modes): Declare.
8799 * targhooks.c (default_select_early_remat_modes): New function.
8800 * timevar.def (TV_EARLY_REMAT): New timevar.
8801 * passes.def (pass_early_remat): New pass.
8802 * tree-pass.h (make_pass_early_remat): Declare.
8803 * early-remat.c: New file.
8804 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8805 function.
8806 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8809 Alan Hayward <alan.hayward@arm.com>
8810 David Sherwood <david.sherwood@arm.com>
8812 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8813 vfm1 with a bound_epilog parameter.
8814 (vect_do_peeling): Update calls accordingly, and move the prologue
8815 call earlier in the function. Treat the base bound_epilog as 0 for
8816 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8817 this base when peeling for gaps.
8818 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8819 with fully-masked loops.
8820 (vect_estimate_min_profitable_iters): Handle the single peeled
8821 iteration in that case.
8823 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8824 Alan Hayward <alan.hayward@arm.com>
8825 David Sherwood <david.sherwood@arm.com>
8827 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8828 single-element interleaving even if the size is not a power of 2.
8829 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8830 accesses for single-element interleaving if the group size is
8831 not a power of 2.
8833 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8834 Alan Hayward <alan.hayward@arm.com>
8835 David Sherwood <david.sherwood@arm.com>
8837 * doc/md.texi (fold_extract_last_@var{m}): Document.
8838 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8839 * optabs.def (fold_extract_last_optab): New optab.
8840 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8841 * internal-fn.c (fold_extract_direct): New macro.
8842 (expand_fold_extract_optab_fn): Likewise.
8843 (direct_fold_extract_optab_supported_p): Likewise.
8844 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8845 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8846 EXTRACT_LAST_REDUCTION.
8847 (get_initial_def_for_reduction): Do not create an initial vector
8848 for EXTRACT_LAST_REDUCTION reductions.
8849 (vectorizable_reduction): Leave the scalar phi in place for
8850 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8851 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8852 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8853 transform phase to vectorizable_condition.
8854 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8855 split out from...
8856 (vect_finish_stmt_generation): ...here.
8857 (vect_finish_replace_stmt): New function.
8858 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8859 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8860 pattern.
8861 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8863 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8864 Alan Hayward <alan.hayward@arm.com>
8865 David Sherwood <david.sherwood@arm.com>
8867 * doc/md.texi (extract_last_@var{m}): Document.
8868 * optabs.def (extract_last_optab): New optab.
8869 * internal-fn.def (EXTRACT_LAST): New internal function.
8870 * internal-fn.c (cond_unary_direct): New macro.
8871 (expand_cond_unary_optab_fn): Likewise.
8872 (direct_cond_unary_optab_supported_p): Likewise.
8873 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8874 loops using EXTRACT_LAST.
8875 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8876 (extract_last_<mode>): ...this optab.
8877 (vec_extract<mode><Vel>): Update accordingly.
8879 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8880 Alan Hayward <alan.hayward@arm.com>
8881 David Sherwood <david.sherwood@arm.com>
8883 * target.def (empty_mask_is_expensive): New hook.
8884 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8885 * doc/tm.texi: Regenerate.
8886 * targhooks.h (default_empty_mask_is_expensive): Declare.
8887 * targhooks.c (default_empty_mask_is_expensive): New function.
8888 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8889 if the target says that empty masks are expensive.
8890 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8891 New function.
8892 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8894 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8895 Alan Hayward <alan.hayward@arm.com>
8896 David Sherwood <david.sherwood@arm.com>
8898 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8899 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8900 (vect_use_loop_mask_for_alignment_p): New function.
8901 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8902 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8903 niters_skip argument. Make sure that the first niters_skip elements
8904 of the first iteration are inactive.
8905 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8906 Update call to vect_set_loop_masks_directly.
8907 (get_misalign_in_elems): New function, split out from...
8908 (vect_gen_prolog_loop_niters): ...here.
8909 (vect_update_init_of_dr): Take a code argument that specifies whether
8910 the adjustment should be added or subtracted.
8911 (vect_update_init_of_drs): Likewise.
8912 (vect_prepare_for_masked_peels): New function.
8913 (vect_do_peeling): Skip prologue peeling if we're using a mask
8914 instead. Update call to vect_update_inits_of_drs.
8915 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8916 mask_skip_niters.
8917 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8918 alignment. Do not include the number of peeled iterations in
8919 the minimum threshold in that case.
8920 (vectorizable_induction): Adjust the start value down by
8921 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8922 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8923 Take the number of skipped iterations into account when calculating
8924 the loop bounds.
8925 * tree-vect-stmts.c (vect_gen_while_not): New function.
8927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8928 Alan Hayward <alan.hayward@arm.com>
8929 David Sherwood <david.sherwood@arm.com>
8931 * doc/sourcebuild.texi (vect_fully_masked): Document.
8932 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8933 default value to 0.
8934 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8935 split out from...
8936 (vect_analyze_loop_2): ...here. Don't check the vectorization
8937 factor against the number of loop iterations if the loop is
8938 fully-masked.
8940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8941 Alan Hayward <alan.hayward@arm.com>
8942 David Sherwood <david.sherwood@arm.com>
8944 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8945 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8946 (dump_groups): Update accordingly.
8947 (iv_use::mem_type): New member variable.
8948 (address_p): New function.
8949 (record_use): Add a mem_type argument and initialize the new
8950 mem_type field.
8951 (record_group_use): Add a mem_type argument. Use address_p.
8952 Remove obsolete null checks of base_object. Update call to record_use.
8953 (find_interesting_uses_op): Update call to record_group_use.
8954 (find_interesting_uses_cond): Likewise.
8955 (find_interesting_uses_address): Likewise.
8956 (get_mem_type_for_internal_fn): New function.
8957 (find_address_like_use): Likewise.
8958 (find_interesting_uses_stmt): Try find_address_like_use before
8959 calling find_interesting_uses_op.
8960 (addr_offset_valid_p): Use the iv mem_type field as the type
8961 of the addressed memory.
8962 (add_autoinc_candidates): Likewise.
8963 (get_address_cost): Likewise.
8964 (split_small_address_groups_p): Use address_p.
8965 (split_address_groups): Likewise.
8966 (add_iv_candidate_for_use): Likewise.
8967 (autoinc_possible_for_pair): Likewise.
8968 (rewrite_groups): Likewise.
8969 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8970 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8971 (get_alias_ptr_type_for_ptr_address): New function.
8972 (rewrite_use_address): Rewrite address uses in calls that were
8973 identified by find_address_like_use.
8975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8976 Alan Hayward <alan.hayward@arm.com>
8977 David Sherwood <david.sherwood@arm.com>
8979 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8980 TARGET_MEM_REFs.
8981 * gimple-expr.h (is_gimple_addressable: Likewise.
8982 * gimple-expr.c (is_gimple_address): Likewise.
8983 * internal-fn.c (expand_call_mem_ref): New function.
8984 (expand_mask_load_optab_fn): Use it.
8985 (expand_mask_store_optab_fn): Likewise.
8987 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8988 Alan Hayward <alan.hayward@arm.com>
8989 David Sherwood <david.sherwood@arm.com>
8991 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8992 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8993 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8994 (cond_umax@var{mode}): Document.
8995 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8996 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8997 (cond_umin_optab, cond_umax_optab): New optabs.
8998 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8999 (COND_IOR, COND_XOR): New internal functions.
9000 * internal-fn.h (get_conditional_internal_fn): Declare.
9001 * internal-fn.c (cond_binary_direct): New macro.
9002 (expand_cond_binary_optab_fn): Likewise.
9003 (direct_cond_binary_optab_supported_p): Likewise.
9004 (get_conditional_internal_fn): New function.
9005 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9006 Cope with reduction statements that are vectorized as calls rather
9007 than assignments.
9008 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9009 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9010 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9011 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9012 (UNSPEC_COND_EOR): New unspecs.
9013 (optab): Add mappings for them.
9014 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9015 (sve_int_op, sve_fp_op): New int attributes.
9017 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9018 Alan Hayward <alan.hayward@arm.com>
9019 David Sherwood <david.sherwood@arm.com>
9021 * optabs.def (while_ult_optab): New optab.
9022 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9023 * internal-fn.def (WHILE_ULT): New internal function.
9024 * internal-fn.h (direct_internal_fn_supported_p): New override
9025 that takes two types as argument.
9026 * internal-fn.c (while_direct): New macro.
9027 (expand_while_optab_fn): New function.
9028 (convert_optab_supported_p): Likewise.
9029 (direct_while_optab_supported_p): New macro.
9030 * wide-int.h (wi::udiv_ceil): New function.
9031 * tree-vectorizer.h (rgroup_masks): New structure.
9032 (vec_loop_masks): New typedef.
9033 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9034 and fully_masked_p.
9035 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9036 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9037 (vect_max_vf): New function.
9038 (slpeel_make_loop_iterate_ntimes): Delete.
9039 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9040 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9041 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9042 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9043 internal-fn.h, stor-layout.h and optabs-query.h.
9044 (vect_set_loop_mask): New function.
9045 (add_preheader_seq): Likewise.
9046 (add_header_seq): Likewise.
9047 (interleave_supported_p): Likewise.
9048 (vect_maybe_permute_loop_masks): Likewise.
9049 (vect_set_loop_masks_directly): Likewise.
9050 (vect_set_loop_condition_masked): Likewise.
9051 (vect_set_loop_condition_unmasked): New function, split out from
9052 slpeel_make_loop_iterate_ntimes.
9053 (slpeel_make_loop_iterate_ntimes): Rename to..
9054 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9055 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9056 (vect_do_peeling): Update call accordingly.
9057 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9058 loops.
9059 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9060 mask_compare_type, can_fully_mask_p and fully_masked_p.
9061 (release_vec_loop_masks): New function.
9062 (_loop_vec_info): Use it to free the loop masks.
9063 (can_produce_all_loop_masks_p): New function.
9064 (vect_get_max_nscalars_per_iter): Likewise.
9065 (vect_verify_full_masking): Likewise.
9066 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9067 retries, and free the mask rgroups before retrying. Check loop-wide
9068 reasons for disallowing fully-masked loops. Make the final decision
9069 about whether use a fully-masked loop or not.
9070 (vect_estimate_min_profitable_iters): Do not assume that peeling
9071 for the number of iterations will be needed for fully-masked loops.
9072 (vectorizable_reduction): Disable fully-masked loops.
9073 (vectorizable_live_operation): Likewise.
9074 (vect_halve_mask_nunits): New function.
9075 (vect_double_mask_nunits): Likewise.
9076 (vect_record_loop_mask): Likewise.
9077 (vect_get_loop_mask): Likewise.
9078 (vect_transform_loop): Handle the case in which the final loop
9079 iteration might handle a partial vector. Call vect_set_loop_condition
9080 instead of slpeel_make_loop_iterate_ntimes.
9081 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9082 (check_load_store_masking): New function.
9083 (prepare_load_store_mask): Likewise.
9084 (vectorizable_store): Handle fully-masked loops.
9085 (vectorizable_load): Likewise.
9086 (supportable_widening_operation): Use vect_halve_mask_nunits for
9087 booleans.
9088 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9089 (vect_gen_while): New function.
9090 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9091 (aarch64_uqdec<mode>): New insn.
9093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9094 Alan Hayward <alan.hayward@arm.com>
9095 David Sherwood <david.sherwood@arm.com>
9097 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9098 (reduc_xor_scal_optab): New optabs.
9099 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9100 (reduc_xor_scal_@var{m}): Document.
9101 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9102 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9103 internal functions.
9104 * fold-const-call.c (fold_const_call): Handle them.
9105 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9106 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9107 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9108 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9109 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9110 (UNSPEC_XORV): New unspecs.
9111 (optab): Add entries for them.
9112 (BITWISEV): New int iterator.
9113 (bit_reduc_op): New int attributes.
9115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9116 Alan Hayward <alan.hayward@arm.com>
9117 David Sherwood <david.sherwood@arm.com>
9119 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9120 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9121 * optabs.def (vec_shl_insert_optab): New optab.
9122 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9123 (duplicate_and_interleave): Likewise.
9124 * tree-vect-loop.c: Include internal-fn.h.
9125 (neutral_op_for_slp_reduction): New function, split out from
9126 get_initial_defs_for_reduction.
9127 (get_initial_def_for_reduction): Handle option 2 for variable-length
9128 vectors by loading the neutral value into a vector and then shifting
9129 the initial value into element 0.
9130 (get_initial_defs_for_reduction): Replace the code argument with
9131 the neutral value calculated by neutral_op_for_slp_reduction.
9132 Use gimple_build_vector for constant-length vectors.
9133 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9134 but the first group_size elements have a neutral value.
9135 Use duplicate_and_interleave otherwise.
9136 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9137 Update call to get_initial_defs_for_reduction. Handle SLP
9138 reductions for variable-length vectors by creating one vector
9139 result for each scalar result, with the elements associated
9140 with other scalar results stubbed out with the neutral value.
9141 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9142 Require IFN_VEC_SHL_INSERT for double reductions on
9143 variable-length vectors, or SLP reductions that have
9144 a neutral value. Require can_duplicate_and_interleave_p
9145 support for variable-length unchained SLP reductions if there
9146 is no neutral value, such as for MIN/MAX reductions. Also require
9147 the number of vector elements to be a multiple of the number of
9148 SLP statements when doing variable-length unchained SLP reductions.
9149 Update call to vect_create_epilog_for_reduction.
9150 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9151 and remove initial values.
9152 (duplicate_and_interleave): Make public.
9153 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9154 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9156 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9157 Alan Hayward <alan.hayward@arm.com>
9158 David Sherwood <david.sherwood@arm.com>
9160 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9161 (can_duplicate_and_interleave_p): New function.
9162 (vect_get_and_check_slp_defs): Take the vector of statements
9163 rather than just the current one. Remove excess parentheses.
9164 Restriction rejectinon of vect_constant_def and vect_external_def
9165 for variable-length vectors to boolean types, or types for which
9166 can_duplicate_and_interleave_p is false.
9167 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9168 (duplicate_and_interleave): New function.
9169 (vect_get_constant_vectors): Use gimple_build_vector for
9170 constant-length vectors and suitable variable-length constant
9171 vectors. Use duplicate_and_interleave for other variable-length
9172 vectors. Don't defer the update when inserting new statements.
9174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9175 Alan Hayward <alan.hayward@arm.com>
9176 David Sherwood <david.sherwood@arm.com>
9178 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9179 min_profitable_iters doesn't go negative.
9181 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9182 Alan Hayward <alan.hayward@arm.com>
9183 David Sherwood <david.sherwood@arm.com>
9185 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9186 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9187 * optabs.def (vec_mask_load_lanes_optab): New optab.
9188 (vec_mask_store_lanes_optab): Likewise.
9189 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9190 (MASK_STORE_LANES): Likewise.
9191 * internal-fn.c (mask_load_lanes_direct): New macro.
9192 (mask_store_lanes_direct): Likewise.
9193 (expand_mask_load_optab_fn): Handle masked operations.
9194 (expand_mask_load_lanes_optab_fn): New macro.
9195 (expand_mask_store_optab_fn): Handle masked operations.
9196 (expand_mask_store_lanes_optab_fn): New macro.
9197 (direct_mask_load_lanes_optab_supported_p): Likewise.
9198 (direct_mask_store_lanes_optab_supported_p): Likewise.
9199 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9200 parameter.
9201 (vect_load_lanes_supported): Likewise.
9202 * tree-vect-data-refs.c (strip_conversion): New function.
9203 (can_group_stmts_p): Likewise.
9204 (vect_analyze_data_ref_accesses): Use it instead of checking
9205 for a pair of assignments.
9206 (vect_store_lanes_supported): Take a masked_p parameter.
9207 (vect_load_lanes_supported): Likewise.
9208 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9209 vect_store_lanes_supported and vect_load_lanes_supported.
9210 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9211 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9212 parameter. Don't allow gaps for masked accesses.
9213 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9214 and vect_load_lanes_supported.
9215 (get_load_store_type): Take a masked_p parameter and update
9216 call to get_group_load_store_type.
9217 (vectorizable_store): Update call to get_load_store_type.
9218 Handle IFN_MASK_STORE_LANES.
9219 (vectorizable_load): Update call to get_load_store_type.
9220 Handle IFN_MASK_LOAD_LANES.
9222 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9223 Alan Hayward <alan.hayward@arm.com>
9224 David Sherwood <david.sherwood@arm.com>
9226 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9227 modes for SVE.
9228 * config/aarch64/aarch64-protos.h
9229 (aarch64_sve_struct_memory_operand_p): Declare.
9230 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9231 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9232 (VPRED, vpred): Handle SVE structure modes.
9233 * config/aarch64/constraints.md (Utx): New constraint.
9234 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9235 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9236 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9237 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9238 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9239 structure modes. Split into pieces after RA.
9240 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9241 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9242 New patterns.
9243 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9244 SVE structure modes.
9245 (aarch64_classify_address): Likewise.
9246 (sizetochar): Move earlier in file.
9247 (aarch64_print_operand): Handle SVE register lists.
9248 (aarch64_array_mode): New function.
9249 (aarch64_sve_struct_memory_operand_p): Likewise.
9250 (TARGET_ARRAY_MODE): Redefine.
9252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9253 Alan Hayward <alan.hayward@arm.com>
9254 David Sherwood <david.sherwood@arm.com>
9256 * target.def (array_mode): New target hook.
9257 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9258 * doc/tm.texi: Regenerate.
9259 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9260 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9261 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9262 targetm.array_mode.
9263 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9264 type sizes.
9266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9267 Alan Hayward <alan.hayward@arm.com>
9268 David Sherwood <david.sherwood@arm.com>
9270 * fold-const.c (fold_binary_loc): Check the argument types
9271 rather than the result type when testing for a vector operation.
9273 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9275 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9276 * doc/tm.texi: Regenerate.
9278 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9279 Alan Hayward <alan.hayward@arm.com>
9280 David Sherwood <david.sherwood@arm.com>
9282 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9283 (sve): Document new AArch64 extension.
9284 * doc/md.texi (w): Extend the description of the AArch64
9285 constraint to include SVE vectors.
9286 (Upl, Upa): Document new AArch64 predicate constraints.
9287 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9288 enum.
9289 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9290 (msve-vector-bits=): New option.
9291 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9292 SVE when these are disabled.
9293 (sve): New extension.
9294 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9295 modes. Adjust their number of units based on aarch64_sve_vg.
9296 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9297 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9298 aarch64_addr_query_type.
9299 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9300 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9301 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9302 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9303 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9304 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9305 (aarch64_simd_imm_zero_p): Delete.
9306 (aarch64_check_zero_based_sve_index_immediate): Declare.
9307 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9308 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9309 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9310 (aarch64_sve_float_mul_immediate_p): Likewise.
9311 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9312 rather than an rtx.
9313 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9314 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9315 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9316 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9317 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9318 (aarch64_regmode_natural_size): Likewise.
9319 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9320 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9321 left one place.
9322 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9323 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9324 for VG and the SVE predicate registers.
9325 (V_ALIASES): Add a "z"-prefixed alias.
9326 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9327 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9328 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9329 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9330 (REG_CLASS_NAMES): Add entries for them.
9331 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9332 and the predicate registers.
9333 (aarch64_sve_vg): Declare.
9334 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9335 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9336 (REGMODE_NATURAL_SIZE): Define.
9337 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9338 SVE macros.
9339 * config/aarch64/aarch64.c: Include cfgrtl.h.
9340 (simd_immediate_info): Add a constructor for series vectors,
9341 and an associated step field.
9342 (aarch64_sve_vg): New variable.
9343 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9344 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9345 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9346 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9347 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9348 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9349 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9350 (aarch64_get_mask_mode): New functions.
9351 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9352 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9353 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9354 predicate modes and predicate registers. Explicitly restrict
9355 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9356 to store a vector mode if it is recognized by
9357 aarch64_classify_vector_mode.
9358 (aarch64_regmode_natural_size): New function.
9359 (aarch64_hard_regno_caller_save_mode): Return the original mode
9360 for predicates.
9361 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9362 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9363 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9364 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9365 functions.
9366 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9367 does not overlap dest if the function is frame-related. Handle
9368 SVE constants.
9369 (aarch64_split_add_offset): New function.
9370 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9371 them aarch64_add_offset.
9372 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9373 and update call to aarch64_sub_sp.
9374 (aarch64_add_cfa_expression): New function.
9375 (aarch64_expand_prologue): Pass extra temporary registers to the
9376 functions above. Handle the case in which we need to emit new
9377 DW_CFA_expressions for registers that were originally saved
9378 relative to the stack pointer, but now have to be expressed
9379 relative to the frame pointer.
9380 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9381 functions above.
9382 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9383 IP0 and IP1 values for SVE frames.
9384 (aarch64_expand_vec_series): New function.
9385 (aarch64_expand_sve_widened_duplicate): Likewise.
9386 (aarch64_expand_sve_const_vector): Likewise.
9387 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9388 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9389 into the register, rather than emitting a SET directly.
9390 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9391 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9392 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9393 (offset_9bit_signed_scaled_p): New functions.
9394 (aarch64_replicate_bitmask_imm): New function.
9395 (aarch64_bitmask_imm): Use it.
9396 (aarch64_cannot_force_const_mem): Reject expressions involving
9397 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9398 (aarch64_classify_index): Handle SVE indices, by requiring
9399 a plain register index with a scale that matches the element size.
9400 (aarch64_classify_address): Handle SVE addresses. Assert that
9401 the mode of the address is VOIDmode or an integer mode.
9402 Update call to aarch64_classify_symbol.
9403 (aarch64_classify_symbolic_expression): Update call to
9404 aarch64_classify_symbol.
9405 (aarch64_const_vec_all_in_range_p): New function.
9406 (aarch64_print_vector_float_operand): Likewise.
9407 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9408 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9409 and the FP immediates 1.0 and 0.5.
9410 (aarch64_print_address_internal): Handle SVE addresses.
9411 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9412 (aarch64_regno_regclass): Handle predicate registers.
9413 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9414 data modes.
9415 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9416 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9417 (aarch64_convert_sve_vector_bits): New function.
9418 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9419 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9420 rather than an rtx.
9421 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9422 Handle SVE vector and predicate modes. Accept VL-based constants
9423 that need only one temporary register, and VL offsets that require
9424 no temporary registers.
9425 (aarch64_conditional_register_usage): Mark the predicate registers
9426 as fixed if SVE isn't available.
9427 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9428 Return true for SVE vector and predicate modes.
9429 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9430 rather than an unsigned int. Handle SVE modes.
9431 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9432 SVE modes.
9433 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9434 if SVE is enabled.
9435 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9436 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9437 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9438 (aarch64_sve_float_mul_immediate_p): New functions.
9439 (aarch64_sve_valid_immediate): New function.
9440 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9441 Explicitly reject structure modes. Check for INDEX constants.
9442 Handle PTRUE and PFALSE constants.
9443 (aarch64_check_zero_based_sve_index_immediate): New function.
9444 (aarch64_simd_imm_zero_p): Delete.
9445 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9446 vector modes. Accept constants in the range of CNT[BHWD].
9447 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9448 ask for an Advanced SIMD mode.
9449 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9450 (aarch64_simd_vector_alignment): Handle SVE predicates.
9451 (aarch64_vectorize_preferred_vector_alignment): New function.
9452 (aarch64_simd_vector_alignment_reachable): Use it instead of
9453 the vector size.
9454 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9455 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9456 functions.
9457 (MAX_VECT_LEN): Delete.
9458 (expand_vec_perm_d): Add a vec_flags field.
9459 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9460 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9461 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9462 for SVE modes.
9463 (aarch64_evpc_rev): Rename to...
9464 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9465 (aarch64_evpc_rev_global): New function.
9466 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9467 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9468 MAX_VECT_LEN.
9469 (aarch64_evpc_sve_tbl): New function.
9470 (aarch64_expand_vec_perm_const_1): Update after rename of
9471 aarch64_evpc_rev. Handle SVE permutes too, trying
9472 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9473 than aarch64_evpc_tbl.
9474 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9475 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9476 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9477 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9478 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9479 (aarch64_expand_sve_vcond): New functions.
9480 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9481 of aarch64_vector_mode_p.
9482 (aarch64_dwarf_poly_indeterminate_value): New function.
9483 (aarch64_compute_pressure_classes): Likewise.
9484 (aarch64_can_change_mode_class): Likewise.
9485 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9486 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9487 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9488 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9489 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9490 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9491 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9492 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9493 constraints.
9494 (Dn, Dl, Dr): Accept const as well as const_vector.
9495 (Dz): Likewise. Compare against CONST0_RTX.
9496 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9497 of "vector" where appropriate.
9498 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9499 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9500 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9501 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9502 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9503 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9504 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9505 (v_int_equiv): Extend to SVE modes.
9506 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9507 mode attributes.
9508 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9509 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9510 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9511 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9512 (SVE_COND_FP_CMP): New int iterators.
9513 (perm_hilo): Handle the new unpack unspecs.
9514 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9515 attributes.
9516 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9517 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9518 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9519 (aarch64_equality_operator, aarch64_constant_vector_operand)
9520 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9521 (aarch64_sve_nonimmediate_operand): Likewise.
9522 (aarch64_sve_general_operand): Likewise.
9523 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9524 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9525 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9526 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9527 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9528 (aarch64_sve_float_arith_immediate): Likewise.
9529 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9530 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9531 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9532 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9533 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9534 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9535 (aarch64_sve_float_arith_operand): Likewise.
9536 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9537 (aarch64_sve_float_mul_operand): Likewise.
9538 (aarch64_sve_vec_perm_operand): Likewise.
9539 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9540 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9541 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9542 as well as const_vector.
9543 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9544 in file. Use CONST0_RTX and CONSTM1_RTX.
9545 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9546 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9547 Use aarch64_simd_imm_zero.
9548 * config/aarch64/aarch64-sve.md: New file.
9549 * config/aarch64/aarch64.md: Include it.
9550 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9551 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9552 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9553 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9554 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9555 (sve): New attribute.
9556 (enabled): Disable instructions with the sve attribute unless
9557 TARGET_SVE.
9558 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9559 aarch64_expand_mov_immediate.
9560 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9561 CNT[BHSD] immediates.
9562 (movti): Split CONST_POLY_INT moves into two halves.
9563 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9564 Split additions that need a temporary here if the destination
9565 is the stack pointer.
9566 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9567 (*add<mode>3_poly_1): New instruction.
9568 (set_clobber_cc): New expander.
9570 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9572 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9573 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9574 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9575 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9576 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9577 Change innermode from fixed_mode_size to machine_mode.
9578 (simplify_subreg): Update call accordingly. Handle a constant-sized
9579 subreg of a variable-length CONST_VECTOR.
9581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9582 Alan Hayward <alan.hayward@arm.com>
9583 David Sherwood <david.sherwood@arm.com>
9585 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9586 (add_offset_to_base): New function, split out from...
9587 (create_mem_ref): ...here. When handling a scale other than 1,
9588 check first whether the address is valid without the offset.
9589 Add it into the base if so, leaving the index and scale as-is.
9591 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9593 PR c++/83778
9594 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9595 fold_for_warn before checking if arg2 is INTEGER_CST.
9597 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9599 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9600 (store_multiple_operation): Delete.
9601 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9602 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9603 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9604 guarded by TARGET_STRING.
9605 (rs6000_output_load_multiple): Delete.
9606 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9607 OPTION_MASK_STRING / TARGET_STRING handling.
9608 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9609 (const rs6000_opt_masks) <"string">: Change mask to 0.
9610 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9611 (MASK_STRING): Delete.
9612 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9613 parts. Simplify.
9614 (load_multiple): Delete.
9615 (*ldmsi8): Delete.
9616 (*ldmsi7): Delete.
9617 (*ldmsi6): Delete.
9618 (*ldmsi5): Delete.
9619 (*ldmsi4): Delete.
9620 (*ldmsi3): Delete.
9621 (store_multiple): Delete.
9622 (*stmsi8): Delete.
9623 (*stmsi7): Delete.
9624 (*stmsi6): Delete.
9625 (*stmsi5): Delete.
9626 (*stmsi4): Delete.
9627 (*stmsi3): Delete.
9628 (movmemsi_8reg): Delete.
9629 (corresponding unnamed define_insn): Delete.
9630 (movmemsi_6reg): Delete.
9631 (corresponding unnamed define_insn): Delete.
9632 (movmemsi_4reg): Delete.
9633 (corresponding unnamed define_insn): Delete.
9634 (movmemsi_2reg): Delete.
9635 (corresponding unnamed define_insn): Delete.
9636 (movmemsi_1reg): Delete.
9637 (corresponding unnamed define_insn): Delete.
9638 * config/rs6000/rs6000.opt (mno-string): New.
9639 (mstring): Replace by deprecation warning stub.
9640 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9642 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9644 * regrename.c (regrename_do_replace): If replacing the same
9645 reg multiple times, try to reuse last created gen_raw_REG.
9647 PR debug/81155
9648 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9649 main to workaround a bug in GDB.
9651 2018-01-12 Tom de Vries <tom@codesourcery.com>
9653 PR target/83737
9654 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9656 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9658 PR rtl-optimization/80481
9659 * ira-color.c (get_cap_member): New function.
9660 (allocnos_conflict_by_live_ranges_p): Use it.
9661 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9662 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9664 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9666 PR target/83628
9667 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9668 (*saddl_se_1): Ditto.
9669 (*ssubsi_1): Ditto.
9670 (*ssubl_se_1): Ditto.
9672 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9674 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9675 rather than wi::to_widest for DR_INITs.
9676 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9677 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9678 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9679 INTEGER_CSTs.
9680 (vect_analyze_group_access_1): Note that here.
9682 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9684 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9685 polynomial type sizes.
9687 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9689 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9690 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9691 (gimple_add_tmp_var): Likewise.
9693 2018-01-12 Martin Liska <mliska@suse.cz>
9695 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9696 (gimple_alloc_sizes): Likewise.
9697 (dump_gimple_statistics): Use PRIu64 in printf format.
9698 * gimple.h: Change uint64_t to int.
9700 2018-01-12 Martin Liska <mliska@suse.cz>
9702 * tree-core.h: Use uint64_t instead of int.
9703 * tree.c (tree_node_counts): Likewise.
9704 (tree_node_sizes): Likewise.
9705 (dump_tree_statistics): Use PRIu64 in printf format.
9707 2018-01-12 Martin Liska <mliska@suse.cz>
9709 * Makefile.in: As qsort_chk is implemented in vec.c, add
9710 vec.o to linkage of gencfn-macros.
9711 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9712 passing the info to record_node_allocation_statistics.
9713 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9714 and pass the info.
9715 * ggc-common.c (struct ggc_usage): Add operator== and use
9716 it in operator< and compare function.
9717 * mem-stats.h (struct mem_usage): Likewise.
9718 * vec.c (struct vec_usage): Remove operator< and compare
9719 function. Can be simply inherited.
9721 2018-01-12 Martin Jambor <mjambor@suse.cz>
9723 PR target/81616
9724 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9725 * tree-ssa-math-opts.c: Include domwalk.h.
9726 (convert_mult_to_fma_1): New function.
9727 (fma_transformation_info): New type.
9728 (fma_deferring_state): Likewise.
9729 (cancel_fma_deferring): New function.
9730 (result_of_phi): Likewise.
9731 (last_fma_candidate_feeds_initial_phi): Likewise.
9732 (convert_mult_to_fma): Added deferring logic, split actual
9733 transformation to convert_mult_to_fma_1.
9734 (math_opts_dom_walker): New type.
9735 (math_opts_dom_walker::after_dom_children): New method, body moved
9736 here from pass_optimize_widening_mul::execute, added deferring logic
9737 bits.
9738 (pass_optimize_widening_mul::execute): Moved most of code to
9739 math_opts_dom_walker::after_dom_children.
9740 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9741 * config/i386/i386.c (ix86_option_override_internal): Added
9742 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9744 2018-01-12 Richard Biener <rguenther@suse.de>
9746 PR debug/83157
9747 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9748 inline instance vars.
9750 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9752 PR target/81819
9753 * config/rx/rx.c (rx_is_restricted_memory_address):
9754 Handle SUBREG case.
9756 2018-01-12 Richard Biener <rguenther@suse.de>
9758 PR tree-optimization/80846
9759 * target.def (split_reduction): New target hook.
9760 * targhooks.c (default_split_reduction): New function.
9761 * targhooks.h (default_split_reduction): Declare.
9762 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9763 target requests first reduce vectors by combining low and high
9764 parts.
9765 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9766 (get_vectype_for_scalar_type_and_size): Export.
9767 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9768 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9769 * doc/tm.texi: Regenerate.
9770 * config/i386/i386.c (ix86_split_reduction): Implement
9771 TARGET_VECTORIZE_SPLIT_REDUCTION.
9773 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9775 PR target/83368
9776 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9777 in PIC mode except for TARGET_VXWORKS_RTP.
9778 * config/sparc/sparc.c: Include cfgrtl.h.
9779 (TARGET_INIT_PIC_REG): Define.
9780 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9781 (sparc_pic_register_p): New predicate.
9782 (sparc_legitimate_address_p): Use it.
9783 (sparc_legitimize_pic_address): Likewise.
9784 (sparc_delegitimize_address): Likewise.
9785 (sparc_mode_dependent_address_p): Likewise.
9786 (gen_load_pcrel_sym): Remove 4th parameter.
9787 (load_got_register): Adjust call to above. Remove obsolete stuff.
9788 (sparc_expand_prologue): Do not call load_got_register here.
9789 (sparc_flat_expand_prologue): Likewise.
9790 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9791 (sparc_use_pseudo_pic_reg): New function.
9792 (sparc_init_pic_reg): Likewise.
9793 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9794 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9796 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9798 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9799 Add item for branch_cost.
9801 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9803 PR rtl-optimization/83565
9804 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9805 not extend the result to a larger mode for rotate operations.
9806 (num_sign_bit_copies1): Likewise.
9808 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9810 PR target/40411
9811 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9812 -symbolic.
9813 Use values-Xc.o for -pedantic.
9814 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9816 2018-01-12 Martin Liska <mliska@suse.cz>
9818 PR ipa/83054
9819 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9820 New function.
9821 (possible_polymorphic_call_targets): Use it.
9822 (ipa_devirt): Likewise.
9824 2018-01-12 Martin Liska <mliska@suse.cz>
9826 * profile-count.h (enum profile_quality): Use 0 as invalid
9827 enum value of profile_quality.
9829 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9831 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9832 -mext-string options.
9834 2018-01-12 Richard Biener <rguenther@suse.de>
9836 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9837 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9838 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9839 Likewise.
9840 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9842 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9844 * configure.ac (--with-long-double-format): Add support for the
9845 configuration option to change the default long double format on
9846 PowerPC systems.
9847 * config.gcc (powerpc*-linux*-*): Likewise.
9848 * configure: Regenerate.
9849 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9850 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9851 used without modification.
9853 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9855 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9856 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9857 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9858 MISC_BUILTIN_SPEC_BARRIER.
9859 (rs6000_init_builtins): Likewise.
9860 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9861 enum value.
9862 (speculation_barrier): New define_insn.
9863 * doc/extend.texi: Document __builtin_speculation_barrier.
9865 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9867 PR target/83203
9868 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9869 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9870 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9871 iterators.
9872 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9873 integral modes instead of "ss" and "sd".
9874 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9875 vectors with 32-bit and 64-bit elements.
9876 (vecdupssescalarmodesuffix): New mode attribute.
9877 (vec_dup<mode>): Use it.
9879 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9881 PR target/83330
9882 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9883 frame if argument is passed on stack.
9885 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9887 PR target/82682
9888 * ree.c (combine_reaching_defs): Optimize also
9889 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9890 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9892 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9894 PR middle-end/83189
9895 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9897 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9899 PR middle-end/83718
9900 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9901 after they are computed.
9903 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9905 PR tree-optimization/83695
9906 * gimple-loop-linterchange.cc
9907 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9908 reset cached scev information after interchange.
9909 (pass_linterchange::execute): Remove call to scev_reset_htab.
9911 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9913 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9914 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9915 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9916 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9917 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9918 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9919 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9920 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9921 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9922 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9923 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9924 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9925 (V_lane_reg): Likewise.
9926 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9927 New define_expand.
9928 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9929 (vfmal_lane_low<mode>_intrinsic,
9930 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9931 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9932 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9933 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9934 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9935 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9937 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9939 * config/arm/arm-cpus.in (fp16fml): New feature.
9940 (ALL_SIMD): Add fp16fml.
9941 (armv8.2-a): Add fp16fml as an option.
9942 (armv8.3-a): Likewise.
9943 (armv8.4-a): Add fp16fml as part of fp16.
9944 * config/arm/arm.h (TARGET_FP16FML): Define.
9945 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9946 when appropriate.
9947 * config/arm/arm-modes.def (V2HF): Define.
9948 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9949 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9950 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9951 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9952 vfmsl_low, vfmsl_high): New set of builtins.
9953 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9954 (vfml_op): New code attribute.
9955 (VFMLHALVES): New int iterator.
9956 (VFML, VFMLSEL): New mode attributes.
9957 (V_reg): Define mapping for V2HF.
9958 (V_hi, V_lo): New mode attributes.
9959 (VF_constraint): Likewise.
9960 (vfml_half, vfml_half_selector): New int attributes.
9961 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9962 define_expand.
9963 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9964 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9965 New define_insn.
9966 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9967 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9968 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9969 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9970 documentation.
9971 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9972 Document new effective target and option set.
9974 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9976 * config/arm/arm-cpus.in (armv8_4): New feature.
9977 (ARMv8_4a): New fgroup.
9978 (armv8.4-a): New arch.
9979 * config/arm/arm-tables.opt: Regenerate.
9980 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9981 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9982 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9983 Add matching rules for -march=armv8.4-a and extensions.
9984 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9986 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9988 PR target/81821
9989 * config/rx/rx.md (BW): New mode attribute.
9990 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9992 2018-01-11 Richard Biener <rguenther@suse.de>
9994 PR tree-optimization/83435
9995 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9996 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9997 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9999 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10000 Alan Hayward <alan.hayward@arm.com>
10001 David Sherwood <david.sherwood@arm.com>
10003 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10004 field.
10005 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10006 (aarch64_print_address_internal): Use it to check for a zero offset.
10008 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10009 Alan Hayward <alan.hayward@arm.com>
10010 David Sherwood <david.sherwood@arm.com>
10012 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10013 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10014 Return a poly_int64 rather than a HOST_WIDE_INT.
10015 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10016 rather than a HOST_WIDE_INT.
10017 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10018 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10019 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10020 final_offset from HOST_WIDE_INT to poly_int64.
10021 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10022 to_constant when getting the number of units in an Advanced SIMD
10023 mode.
10024 (aarch64_builtin_vectorized_function): Check for a constant number
10025 of units.
10026 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10027 GET_MODE_SIZE.
10028 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10029 attribute instead of GET_MODE_NUNITS.
10030 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10031 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10032 GET_MODE_SIZE for fixed-size registers.
10033 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10034 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10035 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10036 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10037 (aarch64_print_operand, aarch64_print_address_internal)
10038 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10039 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10040 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10041 Handle polynomial GET_MODE_SIZE.
10042 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10043 wider than SImode without modification.
10044 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10045 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10046 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10047 passing and returning SVE modes.
10048 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10049 rather than GEN_INT.
10050 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10051 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10052 (aarch64_allocate_and_probe_stack_space): Likewise.
10053 (aarch64_layout_frame): Cope with polynomial offsets.
10054 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10055 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10056 polynomial offsets.
10057 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10058 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10059 poly_int64 rather than a HOST_WIDE_INT.
10060 (aarch64_get_separate_components, aarch64_process_components)
10061 (aarch64_expand_prologue, aarch64_expand_epilogue)
10062 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10063 (aarch64_anchor_offset): New function, split out from...
10064 (aarch64_legitimize_address): ...here.
10065 (aarch64_builtin_vectorization_cost): Handle polynomial
10066 TYPE_VECTOR_SUBPARTS.
10067 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10068 GET_MODE_NUNITS.
10069 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10070 number of elements from the PARALLEL rather than the mode.
10071 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10072 rather than GET_MODE_BITSIZE.
10073 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10074 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10075 (aarch64_expand_vec_perm_const_1): Handle polynomial
10076 d->perm.length () and d->perm elements.
10077 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10078 Apply to_constant to d->perm elements.
10079 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10080 polynomial CONST_VECTOR_NUNITS.
10081 (aarch64_move_pointer): Take amount as a poly_int64 rather
10082 than an int.
10083 (aarch64_progress_pointer): Avoid temporary variable.
10084 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10085 the mode attribute instead of GET_MODE.
10087 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10088 Alan Hayward <alan.hayward@arm.com>
10089 David Sherwood <david.sherwood@arm.com>
10091 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10092 x exists before using it.
10093 (aarch64_add_constant_internal): Rename to...
10094 (aarch64_add_offset_1): ...this. Replace regnum with separate
10095 src and dest rtxes. Handle the case in which they're different,
10096 including when the offset is zero. Replace scratchreg with an rtx.
10097 Use 2 additions if there is no spare register into which we can
10098 move a 16-bit constant.
10099 (aarch64_add_constant): Delete.
10100 (aarch64_add_offset): Replace reg with separate src and dest
10101 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10102 Use aarch64_add_offset_1.
10103 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10104 an rtx rather than an int. Take the delta as a poly_int64
10105 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10106 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10107 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10108 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10109 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10110 and aarch64_add_sp.
10111 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10112 aarch64_add_constant.
10114 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10116 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10117 Use scalar_float_mode.
10119 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10121 * config/aarch64/aarch64-simd.md
10122 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10123 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10124 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10125 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10126 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10127 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10128 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10129 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10130 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10131 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10133 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10135 PR target/83514
10136 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10137 targ_options->x_arm_arch_string is non NULL.
10139 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10141 * config/aarch64/aarch64.h
10142 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10144 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10146 PR target/82096
10147 * expmed.c (emit_store_flag_force): Swap if const op0
10148 and change VOIDmode to mode of op0.
10150 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10152 PR rtl-optimization/83761
10153 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10154 than bytes to mode_for_size.
10156 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10158 PR middle-end/83189
10159 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10160 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10161 profile.
10163 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10165 PR middle-end/83575
10166 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10167 when in layout mode.
10168 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10169 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10170 partition fixup.
10172 2018-01-10 Michael Collison <michael.collison@arm.com>
10174 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10175 * config/aarch64/aarch64-option-extension.def: Add
10176 AARCH64_OPT_EXTENSION of 'fp16fml'.
10177 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10178 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10179 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10180 * config/aarch64/constraints.md (Ui7): New constraint.
10181 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10182 (VFMLA_SEL_W): Ditto.
10183 (f16quad): Ditto.
10184 (f16mac1): Ditto.
10185 (VFMLA16_LOW): New int iterator.
10186 (VFMLA16_HIGH): Ditto.
10187 (UNSPEC_FMLAL): New unspec.
10188 (UNSPEC_FMLSL): Ditto.
10189 (UNSPEC_FMLAL2): Ditto.
10190 (UNSPEC_FMLSL2): Ditto.
10191 (f16mac): New code attribute.
10192 * config/aarch64/aarch64-simd-builtins.def
10193 (aarch64_fmlal_lowv2sf): Ditto.
10194 (aarch64_fmlsl_lowv2sf): Ditto.
10195 (aarch64_fmlalq_lowv4sf): Ditto.
10196 (aarch64_fmlslq_lowv4sf): Ditto.
10197 (aarch64_fmlal_highv2sf): Ditto.
10198 (aarch64_fmlsl_highv2sf): Ditto.
10199 (aarch64_fmlalq_highv4sf): Ditto.
10200 (aarch64_fmlslq_highv4sf): Ditto.
10201 (aarch64_fmlal_lane_lowv2sf): Ditto.
10202 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10203 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10204 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10205 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10206 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10207 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10208 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10209 (aarch64_fmlal_lane_highv2sf): Ditto.
10210 (aarch64_fmlsl_lane_highv2sf): Ditto.
10211 (aarch64_fmlal_laneq_highv2sf): Ditto.
10212 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10213 (aarch64_fmlalq_lane_highv4sf): Ditto.
10214 (aarch64_fmlsl_lane_highv4sf): Ditto.
10215 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10216 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10217 * config/aarch64/aarch64-simd.md:
10218 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10219 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10220 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10221 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10222 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10223 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10224 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10225 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10226 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10227 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10228 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10229 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10230 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10231 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10232 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10233 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10234 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10235 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10236 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10237 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10238 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10239 (vfmlsl_low_u32): Ditto.
10240 (vfmlalq_low_u32): Ditto.
10241 (vfmlslq_low_u32): Ditto.
10242 (vfmlal_high_u32): Ditto.
10243 (vfmlsl_high_u32): Ditto.
10244 (vfmlalq_high_u32): Ditto.
10245 (vfmlslq_high_u32): Ditto.
10246 (vfmlal_lane_low_u32): Ditto.
10247 (vfmlsl_lane_low_u32): Ditto.
10248 (vfmlal_laneq_low_u32): Ditto.
10249 (vfmlsl_laneq_low_u32): Ditto.
10250 (vfmlalq_lane_low_u32): Ditto.
10251 (vfmlslq_lane_low_u32): Ditto.
10252 (vfmlalq_laneq_low_u32): Ditto.
10253 (vfmlslq_laneq_low_u32): Ditto.
10254 (vfmlal_lane_high_u32): Ditto.
10255 (vfmlsl_lane_high_u32): Ditto.
10256 (vfmlal_laneq_high_u32): Ditto.
10257 (vfmlsl_laneq_high_u32): Ditto.
10258 (vfmlalq_lane_high_u32): Ditto.
10259 (vfmlslq_lane_high_u32): Ditto.
10260 (vfmlalq_laneq_high_u32): Ditto.
10261 (vfmlslq_laneq_high_u32): Ditto.
10262 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10263 (AARCH64_FL_FOR_ARCH8_4): New.
10264 (AARCH64_ISA_F16FML): New ISA flag.
10265 (TARGET_F16FML): New feature flag for fp16fml.
10266 (doc/invoke.texi): Document new fp16fml option.
10268 2018-01-10 Michael Collison <michael.collison@arm.com>
10270 * config/aarch64/aarch64-builtins.c:
10271 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10272 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10273 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10274 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10275 (AARCH64_ISA_SHA3): New ISA flag.
10276 (TARGET_SHA3): New feature flag for sha3.
10277 * config/aarch64/iterators.md (sha512_op): New int attribute.
10278 (CRYPTO_SHA512): New int iterator.
10279 (UNSPEC_SHA512H): New unspec.
10280 (UNSPEC_SHA512H2): Ditto.
10281 (UNSPEC_SHA512SU0): Ditto.
10282 (UNSPEC_SHA512SU1): Ditto.
10283 * config/aarch64/aarch64-simd-builtins.def
10284 (aarch64_crypto_sha512hqv2di): New builtin.
10285 (aarch64_crypto_sha512h2qv2di): Ditto.
10286 (aarch64_crypto_sha512su0qv2di): Ditto.
10287 (aarch64_crypto_sha512su1qv2di): Ditto.
10288 (aarch64_eor3qv8hi): Ditto.
10289 (aarch64_rax1qv2di): Ditto.
10290 (aarch64_xarqv2di): Ditto.
10291 (aarch64_bcaxqv8hi): Ditto.
10292 * config/aarch64/aarch64-simd.md:
10293 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10294 (aarch64_crypto_sha512su0qv2di): Ditto.
10295 (aarch64_crypto_sha512su1qv2di): Ditto.
10296 (aarch64_eor3qv8hi): Ditto.
10297 (aarch64_rax1qv2di): Ditto.
10298 (aarch64_xarqv2di): Ditto.
10299 (aarch64_bcaxqv8hi): Ditto.
10300 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10301 (vsha512h2q_u64): Ditto.
10302 (vsha512su0q_u64): Ditto.
10303 (vsha512su1q_u64): Ditto.
10304 (veor3q_u16): Ditto.
10305 (vrax1q_u64): Ditto.
10306 (vxarq_u64): Ditto.
10307 (vbcaxq_u16): Ditto.
10308 * config/arm/types.md (crypto_sha512): New type attribute.
10309 (crypto_sha3): Ditto.
10310 (doc/invoke.texi): Document new sha3 option.
10312 2018-01-10 Michael Collison <michael.collison@arm.com>
10314 * config/aarch64/aarch64-builtins.c:
10315 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10316 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10317 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10318 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10319 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10320 (AARCH64_ISA_SM4): New ISA flag.
10321 (TARGET_SM4): New feature flag for sm4.
10322 * config/aarch64/aarch64-simd-builtins.def
10323 (aarch64_sm3ss1qv4si): Ditto.
10324 (aarch64_sm3tt1aq4si): Ditto.
10325 (aarch64_sm3tt1bq4si): Ditto.
10326 (aarch64_sm3tt2aq4si): Ditto.
10327 (aarch64_sm3tt2bq4si): Ditto.
10328 (aarch64_sm3partw1qv4si): Ditto.
10329 (aarch64_sm3partw2qv4si): Ditto.
10330 (aarch64_sm4eqv4si): Ditto.
10331 (aarch64_sm4ekeyqv4si): Ditto.
10332 * config/aarch64/aarch64-simd.md:
10333 (aarch64_sm3ss1qv4si): Ditto.
10334 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10335 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10336 (aarch64_sm4eqv4si): Ditto.
10337 (aarch64_sm4ekeyqv4si): Ditto.
10338 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10339 (sm3part_op): Ditto.
10340 (CRYPTO_SM3TT): Ditto.
10341 (CRYPTO_SM3PART): Ditto.
10342 (UNSPEC_SM3SS1): New unspec.
10343 (UNSPEC_SM3TT1A): Ditto.
10344 (UNSPEC_SM3TT1B): Ditto.
10345 (UNSPEC_SM3TT2A): Ditto.
10346 (UNSPEC_SM3TT2B): Ditto.
10347 (UNSPEC_SM3PARTW1): Ditto.
10348 (UNSPEC_SM3PARTW2): Ditto.
10349 (UNSPEC_SM4E): Ditto.
10350 (UNSPEC_SM4EKEY): Ditto.
10351 * config/aarch64/constraints.md (Ui2): New constraint.
10352 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10353 * config/arm/types.md (crypto_sm3): New type attribute.
10354 (crypto_sm4): Ditto.
10355 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10356 (vsm3tt1aq_u32): Ditto.
10357 (vsm3tt1bq_u32): Ditto.
10358 (vsm3tt2aq_u32): Ditto.
10359 (vsm3tt2bq_u32): Ditto.
10360 (vsm3partw1q_u32): Ditto.
10361 (vsm3partw2q_u32): Ditto.
10362 (vsm4eq_u32): Ditto.
10363 (vsm4ekeyq_u32): Ditto.
10364 (doc/invoke.texi): Document new sm4 option.
10366 2018-01-10 Michael Collison <michael.collison@arm.com>
10368 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10369 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10370 (AARCH64_FL_FOR_ARCH8_4): New.
10371 (AARCH64_FL_V8_4): New flag.
10372 (doc/invoke.texi): Document new armv8.4-a option.
10374 2018-01-10 Michael Collison <michael.collison@arm.com>
10376 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10377 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10378 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10379 * config/aarch64/aarch64-option-extension.def: Add
10380 AARCH64_OPT_EXTENSION of 'sha2'.
10381 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10382 (crypto): Disable sha2 and aes if crypto disabled.
10383 (crypto): Enable aes and sha2 if enabled.
10384 (simd): Disable sha2 and aes if simd disabled.
10385 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10386 New flags.
10387 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10388 (TARGET_SHA2): New feature flag for sha2.
10389 (TARGET_AES): New feature flag for aes.
10390 * config/aarch64/aarch64-simd.md:
10391 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10392 conditional on TARGET_AES.
10393 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10394 (aarch64_crypto_sha1hsi): Make pattern conditional
10395 on TARGET_SHA2.
10396 (aarch64_crypto_sha1hv4si): Ditto.
10397 (aarch64_be_crypto_sha1hv4si): Ditto.
10398 (aarch64_crypto_sha1su1v4si): Ditto.
10399 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10400 (aarch64_crypto_sha1su0v4si): Ditto.
10401 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10402 (aarch64_crypto_sha256su0v4si): Ditto.
10403 (aarch64_crypto_sha256su1v4si): Ditto.
10404 (doc/invoke.texi): Document new aes and sha2 options.
10406 2018-01-10 Martin Sebor <msebor@redhat.com>
10408 PR tree-optimization/83781
10409 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10410 as string arrays.
10412 2018-01-11 Martin Sebor <msebor@gmail.com>
10413 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10415 PR tree-optimization/83501
10416 PR tree-optimization/81703
10418 * tree-ssa-strlen.c (get_string_cst): Rename...
10419 (get_string_len): ...to this. Handle global constants.
10420 (handle_char_store): Adjust.
10422 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10423 Jim Wilson <jimw@sifive.com>
10425 * config/riscv/riscv-protos.h (riscv_output_return): New.
10426 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10427 (riscv_attribute_table, riscv_output_return),
10428 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10429 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10430 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10431 function.
10432 (riscv_expand_prologue): Add early return for naked function.
10433 (riscv_expand_epilogue): Likewise.
10434 (riscv_function_ok_for_sibcall): Return false for naked function.
10435 (riscv_set_current_function): New.
10436 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10437 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10438 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10439 * doc/extend.texi (RISC-V Function Attributes): New.
10441 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10443 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10444 check for 128-bit long double before checking TCmode.
10445 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10446 128-bit long doubles before checking TFmode or TCmode.
10447 (FLOAT128_IBM_P): Likewise.
10449 2018-01-10 Martin Sebor <msebor@redhat.com>
10451 PR tree-optimization/83671
10452 * builtins.c (c_strlen): Unconditionally return zero for the empty
10453 string.
10454 Use -Warray-bounds for warnings.
10455 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10456 for non-constant array indices with COMPONENT_REF, arrays of
10457 arrays, and pointers to arrays.
10458 (gimple_fold_builtin_strlen): Determine and set length range for
10459 non-constant character arrays.
10461 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10463 PR middle-end/81897
10464 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10465 empty blocks.
10467 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10469 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10471 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10473 PR target/83399
10474 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10475 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10476 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10477 indexed_or_indirect_operand predicate.
10478 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10479 (*vsx_le_perm_load_v8hi): Likewise.
10480 (*vsx_le_perm_load_v16qi): Likewise.
10481 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10482 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10483 (*vsx_le_perm_store_v8hi): Likewise.
10484 (*vsx_le_perm_store_v16qi): Likewise.
10485 (eight unnamed splitters): Likewise.
10487 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10489 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10490 * config/rs6000/emmintrin.h: Likewise.
10491 * config/rs6000/mmintrin.h: Likewise.
10492 * config/rs6000/xmmintrin.h: Likewise.
10494 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10496 PR c++/43486
10497 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10498 "public_flag".
10499 * tree.c (tree_nop_conversion): Return true for location wrapper
10500 nodes.
10501 (maybe_wrap_with_location): New function.
10502 (selftest::check_strip_nops): New function.
10503 (selftest::test_location_wrappers): New function.
10504 (selftest::tree_c_tests): Call it.
10505 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10506 (maybe_wrap_with_location): New decl.
10507 (EXPR_LOCATION_WRAPPER_P): New macro.
10508 (location_wrapper_p): New inline function.
10509 (tree_strip_any_location_wrapper): New inline function.
10511 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10513 PR target/83735
10514 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10515 stack_realign_offset for the largest alignment of stack slot
10516 actually used.
10517 (ix86_find_max_used_stack_alignment): New function.
10518 (ix86_finalize_stack_frame_flags): Use it. Set
10519 max_used_stack_alignment if we don't realign stack.
10520 * config/i386/i386.h (machine_function): Add
10521 max_used_stack_alignment.
10523 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10525 * config/arm/arm.opt (-mbranch-cost): New option.
10526 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10527 account.
10529 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10531 PR target/83629
10532 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10533 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10535 2018-01-10 Richard Biener <rguenther@suse.de>
10537 PR debug/83765
10538 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10539 early out so it also covers the case where we have a non-NULL
10540 origin.
10542 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10544 PR tree-optimization/83753
10545 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10546 for non-strided grouped accesses if the number of elements is 1.
10548 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10550 PR target/81616
10551 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10552 * i386.h (TARGET_USE_GATHER): Define.
10553 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10555 2018-01-10 Martin Liska <mliska@suse.cz>
10557 PR bootstrap/82831
10558 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10559 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10560 partitioning.
10561 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10562 CLEANUP_NO_PARTITIONING is not set.
10564 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10566 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10567 for vectors, as a partial revert of r254296.
10568 * rtl.h (const_vec_p): Delete.
10569 (const_vec_duplicate_p): Don't test for vector CONSTs.
10570 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10571 * expmed.c (make_tree): Likewise.
10573 Revert:
10574 * common.md (E, F): Use CONSTANT_P instead of checking for
10575 CONST_VECTOR.
10576 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10577 checking for CONST_VECTOR.
10579 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10581 PR middle-end/83575
10582 * predict.c (force_edge_cold): Handle in more sane way edges
10583 with no prediction.
10585 2018-01-09 Carl Love <cel@us.ibm.com>
10587 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10588 V4SI, V4SF types.
10589 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10590 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10591 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10592 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10593 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10594 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10595 * config/rs6000/rs6000-protos.h: Add extern defition for
10596 rs6000_generate_float2_double_code.
10597 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10598 function.
10599 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10600 (float2_v2df): Add define_expand.
10602 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10604 PR target/83628
10605 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10606 op_mode in the force_to_mode call.
10608 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10610 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10611 instead of checking each element individually.
10612 (aarch64_evpc_uzp): Likewise.
10613 (aarch64_evpc_zip): Likewise.
10614 (aarch64_evpc_ext): Likewise.
10615 (aarch64_evpc_rev): Likewise.
10616 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10617 instead of checking each element individually. Return true without
10618 generating rtl if
10619 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10620 whether all selected elements come from the same input, instead of
10621 checking each element individually. Remove calls to gen_rtx_REG,
10622 start_sequence and end_sequence and instead assert that no rtl is
10623 generated.
10625 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10627 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10628 order of HIGH and CONST checks.
10630 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10632 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10633 if the destination isn't an SSA_NAME.
10635 2018-01-09 Richard Biener <rguenther@suse.de>
10637 PR tree-optimization/83668
10638 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10639 move prologue...
10640 (canonicalize_loop_form): ... here, renamed from ...
10641 (canonicalize_loop_closed_ssa_form): ... this and amended to
10642 swap successor edges for loop exit blocks to make us use
10643 the RPO order we need for initial schedule generation.
10645 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10647 PR tree-optimization/64811
10648 * match.pd: When optimizing comparisons with Inf, avoid
10649 introducing or losing exceptions from comparisons with NaN.
10651 2018-01-09 Martin Liska <mliska@suse.cz>
10653 PR sanitizer/82517
10654 * asan.c (shadow_mem_size): Add gcc_assert.
10656 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10658 Don't save registers in main().
10660 PR target/83738
10661 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10662 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10663 * config/avr/avr.c (avr_set_current_function): Don't error if
10664 naked, OS_task or OS_main are specified at the same time.
10665 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10666 OS_main.
10667 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10668 attribute.
10669 * common/config/avr/avr-common.c (avr_option_optimization_table):
10670 Switch on -mmain-is-OS_task for optimizing compilations.
10672 2018-01-09 Richard Biener <rguenther@suse.de>
10674 PR tree-optimization/83572
10675 * graphite.c: Include cfganal.h.
10676 (graphite_transform_loops): Connect infinite loops to exit
10677 and remove fake edges at the end.
10679 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10681 * ipa-inline.c (edge_badness): Revert accidental checkin.
10683 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10685 PR ipa/80763
10686 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10687 symbols; not inline clones.
10689 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10691 PR target/83507
10692 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10693 hard registers. Formatting fixes.
10695 PR preprocessor/83722
10696 * gcc.c (try_generate_repro): Pass
10697 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10698 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10699 do_report_bug.
10701 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10702 Kito Cheng <kito.cheng@gmail.com>
10704 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10705 (riscv_leaf_function_p): Delete.
10706 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10708 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10710 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10711 function.
10712 (do_ifelse): New function.
10713 (do_isel): New function.
10714 (do_sub3): New function.
10715 (do_add3): New function.
10716 (do_load_mask_compare): New function.
10717 (do_overlap_load_compare): New function.
10718 (expand_compare_loop): New function.
10719 (expand_block_compare): Call expand_compare_loop() when appropriate.
10720 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10721 option description.
10722 (-mblock-compare-inline-loop-limit): New option.
10724 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10726 PR target/83677
10727 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10728 Reverse order of second and third operands in first alternative.
10729 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10730 of first and second elements in UNSPEC_VPERMR vector.
10731 (altivec_expand_vec_perm_le): Likewise.
10733 2018-01-08 Jeff Law <law@redhat.com>
10735 PR rtl-optimizatin/81308
10736 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10737 (process_switch): If group_case_labels makes a change, then set
10738 cfg_altered.
10739 (pass_convert_switch::execute): If a switch is converted, then
10740 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10742 PR rtl-optimization/81308
10743 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10744 splitting insns.
10746 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10748 PR target/83663 - Revert r255946
10749 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10750 generation for cases where splatting a value is not useful.
10751 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10752 across a vec_duplicate and a paradoxical subreg forming a vector
10753 mode to a vec_concat.
10755 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10757 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10758 -march=armv8.3-a variants.
10759 * config/arm/t-multilib: Likewise.
10760 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10762 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10764 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10765 to generate rtl.
10766 (cceq_ior_compare_complement): Give it a name so I can use it, and
10767 change boolean_or_operator predicate to boolean_operator so it can
10768 be used to generate a crand.
10769 (eqne): New code iterator.
10770 (bd/bd_neg): New code_attrs.
10771 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10772 a single define_insn.
10773 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10774 decrement (bdnzt/bdnzf/bdzt/bdzf).
10775 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10776 with the new names of the branch decrement patterns, and added the
10777 names of the branch decrement conditional patterns.
10779 2018-01-08 Richard Biener <rguenther@suse.de>
10781 PR tree-optimization/83563
10782 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10783 cache.
10785 2018-01-08 Richard Biener <rguenther@suse.de>
10787 PR middle-end/83713
10788 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10790 2018-01-08 Richard Biener <rguenther@suse.de>
10792 PR tree-optimization/83685
10793 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10794 references to abnormals.
10796 2018-01-08 Richard Biener <rguenther@suse.de>
10798 PR lto/83719
10799 * dwarf2out.c (output_indirect_strings): Handle empty
10800 skeleton_debug_str_hash.
10801 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10803 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10805 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10806 (emit_store_direct): Likewise.
10807 (arc_trampoline_adjust_address): Likewise.
10808 (arc_asm_trampoline_template): New function.
10809 (arc_initialize_trampoline): Use asm_trampoline_template.
10810 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10811 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10812 * config/arc/arc.md (flush_icache): Delete pattern.
10814 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10816 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10817 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10818 munaligned-access.
10820 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10822 PR target/83681
10823 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10824 by not USED_FOR_TARGET.
10825 (make_pass_resolve_sw_modes): Likewise.
10827 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10829 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10830 USED_FOR_TARGET.
10832 2018-01-08 Richard Biener <rguenther@suse.de>
10834 PR middle-end/83580
10835 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10837 2018-01-08 Richard Biener <rguenther@suse.de>
10839 PR middle-end/83517
10840 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10842 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10844 PR middle-end/81897
10845 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10846 basic blocks with a small number of successors.
10847 (convert_control_dep_chain_into_preds): Improve handling of
10848 forwarder blocks.
10849 (dump_predicates): Split apart into...
10850 (dump_pred_chain): ...here...
10851 (dump_pred_info): ...and here.
10852 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10853 (can_chain_union_be_invalidated_p): Improve check for invalidation
10854 of paths.
10855 (uninit_uses_cannot_happen): Avoid unnecessary if
10856 convert_control_dep_chain_into_preds yielded nothing.
10858 2018-01-06 Martin Sebor <msebor@redhat.com>
10860 PR tree-optimization/83640
10861 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10862 subtracting negative offset from size.
10863 (builtin_access::overlap): Adjust offset bounds of the access to fall
10864 within the size of the object if possible.
10866 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10868 PR rtl-optimization/83699
10869 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10870 extract_bit_field_as_subreg to cases in which the extracted
10871 value is also a vector.
10873 * lra-constraints.c (process_alt_operands): Test for the equivalence
10874 substitutions when detecting a possible reload cycle.
10876 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10878 PR debug/83480
10879 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10880 by default if flag_selective_schedling{,2}. Formatting fixes.
10882 PR rtl-optimization/83682
10883 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10884 if it has non-VECTOR_MODE element mode.
10885 (vec_duplicate_p): Likewise.
10887 PR middle-end/83694
10888 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10889 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10891 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10893 PR target/83604
10894 * config/i386/i386-builtin.def
10895 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10896 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10897 Require also OPTION_MASK_ISA_AVX512F in addition to
10898 OPTION_MASK_ISA_GFNI.
10899 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10900 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10901 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10902 to OPTION_MASK_ISA_GFNI.
10903 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10904 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10905 OPTION_MASK_ISA_AVX512BW.
10906 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10907 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10908 addition to OPTION_MASK_ISA_GFNI.
10909 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10910 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10911 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10912 to OPTION_MASK_ISA_GFNI.
10913 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10914 a requirement for all ISAs rather than any of them with a few
10915 exceptions.
10916 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10917 processing.
10918 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10919 bitmasks to be enabled with 3 exceptions, instead of requiring any
10920 enabled ISA with lots of exceptions.
10921 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10922 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10923 Change avx512bw in isa attribute to avx512f.
10924 * config/i386/sgxintrin.h: Add license boilerplate.
10925 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10926 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10927 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10928 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10929 defined.
10930 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10931 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10932 temporarily sse2 rather than sse if not enabled already.
10934 PR target/83604
10935 * config/i386/sse.md (VI248_VLBW): Rename to ...
10936 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10937 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10938 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10939 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10940 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10941 mode iterator instead of VI248_VLBW.
10943 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10945 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10946 (record_modified): Skip clobbers; add debug output.
10947 (param_change_prob): Use sreal frequencies.
10949 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10951 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10952 punt for user-aligned variables.
10954 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10956 * tree-chrec.c (chrec_contains_symbols): Return true for
10957 POLY_INT_CST.
10959 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10961 PR target/82439
10962 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10963 of (x|y) == x for BICS pattern.
10965 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10967 PR tree-optimization/83605
10968 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10969 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10970 can throw.
10972 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10974 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10975 * config/epiphany/rtems.h: New file.
10977 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10978 Uros Bizjak <ubizjak@gmail.com>
10980 PR target/83554
10981 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10982 QIreg_operand instead of register_operand predicate.
10983 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10984 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10985 comments instead of -fmitigate[-_]rop.
10987 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10989 PR bootstrap/81926
10990 * cgraphunit.c (symbol_table::compile): Switch to text_section
10991 before calling assembly_start debug hook.
10992 * run-rtl-passes.c (run_rtl_passes): Likewise.
10993 Include output.h.
10995 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10997 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10998 range_int_cst_p rather than !symbolic_range_p before calling
10999 extract_range_from_multiplicative_op_1.
11001 2018-01-04 Jeff Law <law@redhat.com>
11003 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11004 redundant test in assertion.
11006 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11008 * doc/rtl.texi: Document machine_mode wrapper classes.
11010 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11012 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11013 using tree_to_uhwi.
11015 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11017 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11018 the VEC_PERM_EXPR fold to fail.
11020 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11022 PR debug/83585
11023 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11024 to switched_sections.
11026 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11028 PR target/83680
11029 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11030 test for d.testing.
11032 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11034 PR target/83387
11035 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11036 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11038 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11040 PR debug/83666
11041 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11042 is BLKmode and bitpos not zero or mode change is needed.
11044 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11046 PR target/83675
11047 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11048 TARGET_VIS2.
11050 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11052 PR target/83628
11053 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11054 instead of MULT rtx. Update all corresponding splitters.
11055 (*saddl_se): Ditto.
11056 (*ssub<modesuffix>): Ditto.
11057 (*ssubl_se): Ditto.
11058 (*cmp_sadd_di): Update split patterns.
11059 (*cmp_sadd_si): Ditto.
11060 (*cmp_sadd_sidi): Ditto.
11061 (*cmp_ssub_di): Ditto.
11062 (*cmp_ssub_si): Ditto.
11063 (*cmp_ssub_sidi): Ditto.
11064 * config/alpha/predicates.md (const23_operand): New predicate.
11065 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11066 Look for ASHIFT, not MULT inner operand.
11067 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11069 2018-01-04 Martin Liska <mliska@suse.cz>
11071 PR gcov-profile/83669
11072 * gcov.c (output_intermediate_file): Add version to intermediate
11073 gcov file.
11074 * doc/gcov.texi: Document new field 'version' in intermediate
11075 file format. Fix location of '-k' option of gcov command.
11077 2018-01-04 Martin Liska <mliska@suse.cz>
11079 PR ipa/82352
11080 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11082 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11084 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11086 2018-01-03 Martin Sebor <msebor@redhat.com>
11088 PR tree-optimization/83655
11089 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11090 checking calls with invalid arguments.
11092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11094 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11095 (vectorizable_mask_load_store): Delete.
11096 (vectorizable_call): Return false for masked loads and stores.
11097 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11098 instead of gimple_assign_rhs1.
11099 (vectorizable_load): Handle IFN_MASK_LOAD.
11100 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11104 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11105 split out from..,
11106 (vectorizable_mask_load_store): ...here.
11107 (vectorizable_load): ...and here.
11109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11111 * tree-vect-stmts.c (vect_build_all_ones_mask)
11112 (vect_build_zero_merge_argument): New functions, split out from...
11113 (vectorizable_load): ...here.
11115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11117 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11118 split out from...
11119 (vectorizable_mask_load_store): ...here.
11120 (vectorizable_store): ...and here.
11122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11124 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11125 split out from...
11126 (vectorizable_mask_load_store): ...here.
11128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11130 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11131 (vect_model_store_cost): Take a vec_load_store_type instead of a
11132 vect_def_type.
11133 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11134 (vect_model_store_cost): Take a vec_load_store_type instead of a
11135 vect_def_type.
11136 (vectorizable_mask_load_store): Update accordingly.
11137 (vectorizable_store): Likewise.
11138 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11142 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11143 IFN_MASK_LOAD calls here rather than...
11144 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11147 Alan Hayward <alan.hayward@arm.com>
11148 David Sherwood <david.sherwood@arm.com>
11150 * expmed.c (extract_bit_field_1): For vector extracts,
11151 fall back to extract_bit_field_as_subreg if vec_extract
11152 isn't available.
11154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11155 Alan Hayward <alan.hayward@arm.com>
11156 David Sherwood <david.sherwood@arm.com>
11158 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11159 they are variable or constant sized.
11160 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11161 slots for constant-sized data.
11163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11164 Alan Hayward <alan.hayward@arm.com>
11165 David Sherwood <david.sherwood@arm.com>
11167 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11168 handling COND_EXPRs with boolean comparisons, try to find a better
11169 basis for the mask type than the boolean itself.
11171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11173 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11174 is calculated and how it can be overridden.
11175 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11176 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11177 if defined.
11178 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11179 if nonzero.
11181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11182 Alan Hayward <alan.hayward@arm.com>
11183 David Sherwood <david.sherwood@arm.com>
11185 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11186 Remove the mode argument.
11187 (aarch64_simd_valid_immediate): Remove the mode and inverse
11188 arguments.
11189 * config/aarch64/iterators.md (bitsize): New iterator.
11190 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11191 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11192 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11193 aarch64_simd_valid_immediate.
11194 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11195 (aarch64_reg_or_bic_imm): Likewise.
11196 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11197 with an insn_type enum and msl with a modifier_type enum.
11198 Replace element_width with a scalar_mode. Change the shift
11199 to unsigned int. Add constructors for scalar_float_mode and
11200 scalar_int_mode elements.
11201 (aarch64_vect_float_const_representable_p): Delete.
11202 (aarch64_can_const_movi_rtx_p)
11203 (aarch64_simd_scalar_immediate_valid_for_move)
11204 (aarch64_simd_make_constant): Update call to
11205 aarch64_simd_valid_immediate.
11206 (aarch64_advsimd_valid_immediate_hs): New function.
11207 (aarch64_advsimd_valid_immediate): Likewise.
11208 (aarch64_simd_valid_immediate): Remove mode and inverse
11209 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11210 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11211 and aarch64_float_const_representable_p on the result.
11212 (aarch64_output_simd_mov_immediate): Remove mode argument.
11213 Update call to aarch64_simd_valid_immediate and use of
11214 simd_immediate_info.
11215 (aarch64_output_scalar_simd_mov_immediate): Update call
11216 accordingly.
11218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11219 Alan Hayward <alan.hayward@arm.com>
11220 David Sherwood <david.sherwood@arm.com>
11222 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11223 (mode_nunits): Likewise CONST_MODE_NUNITS.
11224 * machmode.def (ADJUST_NUNITS): Document.
11225 * genmodes.c (mode_data::need_nunits_adj): New field.
11226 (blank_mode): Update accordingly.
11227 (adj_nunits): New variable.
11228 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11229 parameter.
11230 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11231 listed in adj_nunits.
11232 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11233 listed in adj_nunits. Don't emit case statements for such modes.
11234 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11235 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11236 nothing if adj_nunits is nonnull.
11237 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11238 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11239 (emit_mode_fbit): Update use of print_maybe_const_decl.
11240 (emit_move_size): Likewise. Treat the array as non-const
11241 if adj_nunits.
11242 (emit_mode_adjustments): Handle adj_nunits.
11244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11246 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11247 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11248 (VECTOR_MODES): Use it.
11249 (make_vector_modes): Take the prefix as an argument.
11251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11252 Alan Hayward <alan.hayward@arm.com>
11253 David Sherwood <david.sherwood@arm.com>
11255 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11256 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11257 for MODE_VECTOR_BOOL.
11258 * machmode.def (VECTOR_BOOL_MODE): Document.
11259 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11260 (make_vector_bool_mode): New function.
11261 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11262 MODE_VECTOR_BOOL.
11263 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11264 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11265 Likewise.
11266 * stor-layout.c (int_mode_for_mode): Likewise.
11267 * tree.c (build_vector_type_for_mode): Likewise.
11268 * varasm.c (output_constant_pool_2): Likewise.
11269 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11270 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11271 for MODE_VECTOR_BOOL.
11272 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11273 of mode class checks.
11274 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11275 instead of a list of mode class checks.
11276 (expand_vector_scalar_condition): Likewise.
11277 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11280 Alan Hayward <alan.hayward@arm.com>
11281 David Sherwood <david.sherwood@arm.com>
11283 * machmode.h (mode_size): Change from unsigned short to
11284 poly_uint16_pod.
11285 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11286 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11287 or if measurement_type is not polynomial.
11288 (fixed_size_mode::includes_p): Check for constant-sized modes.
11289 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11290 return a poly_uint16 rather than an unsigned short.
11291 (emit_mode_size): Change the type of mode_size from unsigned short
11292 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11293 (emit_mode_adjustments): Cope with polynomial vector sizes.
11294 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11295 for GET_MODE_SIZE.
11296 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11297 for GET_MODE_SIZE.
11298 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11299 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11300 * caller-save.c (setup_save_areas): Likewise.
11301 (replace_reg_with_saved_mem): Likewise.
11302 * calls.c (emit_library_call_value_1): Likewise.
11303 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11304 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11305 (gen_lowpart_for_combine): Likewise.
11306 * convert.c (convert_to_integer_1): Likewise.
11307 * cse.c (equiv_constant, cse_insn): Likewise.
11308 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11309 (cselib_subst_to_values): Likewise.
11310 * dce.c (word_dce_process_block): Likewise.
11311 * df-problems.c (df_word_lr_mark_ref): Likewise.
11312 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11313 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11314 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11315 (rtl_for_decl_location): Likewise.
11316 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11317 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11318 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11319 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11320 (expand_expr_real_1): Likewise.
11321 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11322 (pad_below): Likewise.
11323 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11324 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11325 * ira.c (get_subreg_tracking_sizes): Likewise.
11326 * ira-build.c (ira_create_allocno_objects): Likewise.
11327 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11328 (ira_sort_regnos_for_alter_reg): Likewise.
11329 * ira-costs.c (record_operand_costs): Likewise.
11330 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11331 (resolve_simple_move): Likewise.
11332 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11333 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11334 (lra_constraints): Likewise.
11335 (CONST_POOL_OK_P): Reject variable-sized modes.
11336 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11337 (add_pseudo_to_slot, lra_spill): Likewise.
11338 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11339 * optabs-query.c (get_best_extraction_insn): Likewise.
11340 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11341 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11342 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11343 * recog.c (offsettable_address_addr_space_p): Likewise.
11344 * regcprop.c (maybe_mode_change): Likewise.
11345 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11346 * regrename.c (build_def_use): Likewise.
11347 * regstat.c (dump_reg_info): Likewise.
11348 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11349 (find_reloads, find_reloads_subreg_address): Likewise.
11350 * reload1.c (eliminate_regs_1): Likewise.
11351 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11352 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11353 (simplify_binary_operation_1, simplify_subreg): Likewise.
11354 * targhooks.c (default_function_arg_padding): Likewise.
11355 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11356 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11357 (verify_gimple_assign_ternary): Likewise.
11358 * tree-inline.c (estimate_move_cost): Likewise.
11359 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11360 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11361 (get_address_cost_ainc): Likewise.
11362 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11363 (vect_supportable_dr_alignment): Likewise.
11364 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11365 (vectorizable_reduction): Likewise.
11366 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11367 (vectorizable_operation, vectorizable_load): Likewise.
11368 * tree.c (build_same_sized_truth_vector_type): Likewise.
11369 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11370 * var-tracking.c (emit_note_insn_var_location): Likewise.
11371 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11372 (ADDR_VEC_ALIGN): Likewise.
11374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11375 Alan Hayward <alan.hayward@arm.com>
11376 David Sherwood <david.sherwood@arm.com>
11378 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11379 unsigned short.
11380 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11381 or if measurement_type is polynomial.
11382 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11383 * combine.c (make_extraction): Likewise.
11384 * dse.c (find_shift_sequence): Likewise.
11385 * dwarf2out.c (mem_loc_descriptor): Likewise.
11386 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11387 (extract_bit_field, extract_low_bits): Likewise.
11388 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11389 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11390 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11391 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11392 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11393 * reload.c (find_reloads): Likewise.
11394 * reload1.c (alter_reg): Likewise.
11395 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11396 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11397 * tree-if-conv.c (predicate_mem_writes): Likewise.
11398 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11399 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11400 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11401 * valtrack.c (dead_debug_insert_temp): Likewise.
11402 * varasm.c (mergeable_constant_section): Likewise.
11403 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11406 Alan Hayward <alan.hayward@arm.com>
11407 David Sherwood <david.sherwood@arm.com>
11409 * expr.c (expand_assignment): Cope with polynomial mode sizes
11410 when assigning to a CONCAT.
11412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11413 Alan Hayward <alan.hayward@arm.com>
11414 David Sherwood <david.sherwood@arm.com>
11416 * machmode.h (mode_precision): Change from unsigned short to
11417 poly_uint16_pod.
11418 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11419 short.
11420 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11421 or if measurement_type is not polynomial.
11422 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11423 in which the mode is already known to be a scalar_int_mode.
11424 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11425 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11426 initializer.
11427 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11428 for GET_MODE_PRECISION.
11429 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11430 for GET_MODE_PRECISION.
11431 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11432 as polynomial.
11433 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11434 (expand_field_assignment, make_extraction): Likewise.
11435 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11436 (get_last_value): Likewise.
11437 * convert.c (convert_to_integer_1): Likewise.
11438 * cse.c (cse_insn): Likewise.
11439 * expr.c (expand_expr_real_1): Likewise.
11440 * lra-constraints.c (simplify_operand_subreg): Likewise.
11441 * optabs-query.c (can_atomic_load_p): Likewise.
11442 * optabs.c (expand_atomic_load): Likewise.
11443 (expand_atomic_store): Likewise.
11444 * ree.c (combine_reaching_defs): Likewise.
11445 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11446 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11447 * tree.h (type_has_mode_precision_p): Likewise.
11448 * ubsan.c (instrument_si_overflow): Likewise.
11450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11451 Alan Hayward <alan.hayward@arm.com>
11452 David Sherwood <david.sherwood@arm.com>
11454 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11455 polynomial numbers of units.
11456 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11457 (valid_vector_subparts_p): New function.
11458 (build_vector_type): Remove temporary shim and take the number
11459 of units as a poly_uint64 rather than an int.
11460 (build_opaque_vector_type): Take the number of units as a
11461 poly_uint64 rather than an int.
11462 * tree.c (build_vector_from_ctor): Handle polynomial
11463 TYPE_VECTOR_SUBPARTS.
11464 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11465 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11466 (build_vector_from_val): If the number of units is variable,
11467 use build_vec_duplicate_cst for constant operands and
11468 VEC_DUPLICATE_EXPR otherwise.
11469 (make_vector_type): Remove temporary is_constant ().
11470 (build_vector_type, build_opaque_vector_type): Take the number of
11471 units as a poly_uint64 rather than an int.
11472 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11473 VECTOR_CST_NELTS.
11474 * cfgexpand.c (expand_debug_expr): Likewise.
11475 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11476 (store_constructor, expand_expr_real_1): Likewise.
11477 (const_scalar_mask_from_tree): Likewise.
11478 * fold-const-call.c (fold_const_reduction): Likewise.
11479 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11480 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11481 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11482 (fold_relational_const): Likewise.
11483 (native_interpret_vector): Likewise. Change the size from an
11484 int to an unsigned int.
11485 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11486 TYPE_VECTOR_SUBPARTS.
11487 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11488 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11489 duplicating a non-constant operand into a variable-length vector.
11490 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11491 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11492 * ipa-icf.c (sem_variable::equals): Likewise.
11493 * match.pd: Likewise.
11494 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11495 * print-tree.c (print_node): Likewise.
11496 * stor-layout.c (layout_type): Likewise.
11497 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11498 * tree-cfg.c (verify_gimple_comparison): Likewise.
11499 (verify_gimple_assign_binary): Likewise.
11500 (verify_gimple_assign_ternary): Likewise.
11501 (verify_gimple_assign_single): Likewise.
11502 * tree-pretty-print.c (dump_generic_node): Likewise.
11503 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11504 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11505 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11506 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11507 (vect_shift_permute_load_chain): Likewise.
11508 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11509 (expand_vector_condition, optimize_vector_constructor): Likewise.
11510 (lower_vec_perm, get_compute_type): Likewise.
11511 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11512 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11513 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11514 (vect_recog_mask_conversion_pattern): Likewise.
11515 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11516 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11517 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11518 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11519 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11520 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11521 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11522 (supportable_widening_operation): Likewise.
11523 (supportable_narrowing_operation): Likewise.
11524 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11525 Likewise.
11526 * varasm.c (output_constant): Likewise.
11528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11529 Alan Hayward <alan.hayward@arm.com>
11530 David Sherwood <david.sherwood@arm.com>
11532 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11533 so that both the length == 3 and length != 3 cases set up their
11534 own permute vectors. Add comments explaining why we know the
11535 number of elements is constant.
11536 (vect_permute_load_chain): Likewise.
11538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11539 Alan Hayward <alan.hayward@arm.com>
11540 David Sherwood <david.sherwood@arm.com>
11542 * machmode.h (mode_nunits): Change from unsigned char to
11543 poly_uint16_pod.
11544 (ONLY_FIXED_SIZE_MODES): New macro.
11545 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11546 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11547 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11548 New typedefs.
11549 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11550 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11551 or if measurement_type is not polynomial.
11552 * genmodes.c (ZERO_COEFFS): New macro.
11553 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11554 poly_uint16.
11555 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11556 Use ZERO_COEFFS when emitting initializers.
11557 * data-streamer.h (bp_pack_poly_value): New function.
11558 (bp_unpack_poly_value): Likewise.
11559 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11560 for GET_MODE_NUNITS.
11561 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11562 for GET_MODE_NUNITS.
11563 * tree.c (make_vector_type): Remove temporary shim and make
11564 the real function take the number of units as a poly_uint64
11565 rather than an int.
11566 (build_vector_type_for_mode): Handle polynomial nunits.
11567 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11568 * emit-rtl.c (const_vec_series_p_1): Likewise.
11569 (gen_rtx_CONST_VECTOR): Likewise.
11570 * fold-const.c (test_vec_duplicate_folding): Likewise.
11571 * genrecog.c (validate_pattern): Likewise.
11572 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11573 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11574 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11575 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11576 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11577 * rtlanal.c (subreg_get_info): Likewise.
11578 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11579 (vect_grouped_load_supported): Likewise.
11580 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11581 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11582 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11583 (simplify_const_unary_operation, simplify_binary_operation_1)
11584 (simplify_const_binary_operation, simplify_ternary_operation)
11585 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11586 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11587 instead of CONST_VECTOR_NUNITS.
11588 * varasm.c (output_constant_pool_2): Likewise.
11589 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11590 explicit-encoded elements in the XVEC for variable-length vectors.
11592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11594 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11597 Alan Hayward <alan.hayward@arm.com>
11598 David Sherwood <david.sherwood@arm.com>
11600 * coretypes.h (fixed_size_mode): Declare.
11601 (fixed_size_mode_pod): New typedef.
11602 * builtins.h (target_builtins::x_apply_args_mode)
11603 (target_builtins::x_apply_result_mode): Change type to
11604 fixed_size_mode_pod.
11605 * builtins.c (apply_args_size, apply_result_size, result_vector)
11606 (expand_builtin_apply_args_1, expand_builtin_apply)
11607 (expand_builtin_return): Update accordingly.
11609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11611 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11612 * cselib.c (cselib_hash_rtx): Likewise.
11613 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11614 CONST_VECTOR encoding.
11616 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11617 Jeff Law <law@redhat.com>
11619 PR target/83641
11620 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11621 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11622 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11623 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11625 PR target/83641
11626 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11627 explicitly probe *sp in a noreturn function if there were any callee
11628 register saves or frame pointer is needed.
11630 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11632 PR debug/83621
11633 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11634 BLKmode for ternary, binary or unary expressions.
11636 PR debug/83645
11637 * var-tracking.c (delete_vta_debug_insn): New inline function.
11638 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11639 insns from get_insns () to NULL instead of each bb separately.
11640 Use delete_vta_debug_insn. No longer static.
11641 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11642 delete_vta_debug_insns callers.
11643 * rtl.h (delete_vta_debug_insns): Declare.
11644 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11645 instead of variable_tracking_main.
11647 2018-01-03 Martin Sebor <msebor@redhat.com>
11649 PR tree-optimization/83603
11650 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11651 arguments past the endof the argument list in functions declared
11652 without a prototype.
11653 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11654 Avoid checking when arguments are null.
11656 2018-01-03 Martin Sebor <msebor@redhat.com>
11658 PR c/83559
11659 * doc/extend.texi (attribute const): Fix a typo.
11660 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11661 issuing -Wsuggest-attribute for void functions.
11663 2018-01-03 Martin Sebor <msebor@redhat.com>
11665 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11666 offset_int::from instead of wide_int::to_shwi.
11667 (maybe_diag_overlap): Remove assertion.
11668 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11669 * gimple-ssa-sprintf.c (format_directive): Same.
11670 (parse_directive): Same.
11671 (sprintf_dom_walker::compute_format_length): Same.
11672 (try_substitute_return_value): Same.
11674 2018-01-03 Jeff Law <law@redhat.com>
11676 PR middle-end/83654
11677 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11678 non-constant residual for zero at runtime and avoid probing in
11679 that case. Reorganize code for trailing problem to mirror handling
11680 of the residual.
11682 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11684 PR tree-optimization/83501
11685 * tree-ssa-strlen.c (get_string_cst): New.
11686 (handle_char_store): Call get_string_cst.
11688 2018-01-03 Martin Liska <mliska@suse.cz>
11690 PR tree-optimization/83593
11691 * tree-ssa-strlen.c: Include tree-cfg.h.
11692 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11693 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11694 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11695 to false.
11696 (strlen_dom_walker::before_dom_children): Call
11697 gimple_purge_dead_eh_edges. Dump tranformation with details
11698 dump flags.
11699 (strlen_dom_walker::before_dom_children): Update call by adding
11700 new argument cleanup_eh.
11701 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11703 2018-01-03 Martin Liska <mliska@suse.cz>
11705 PR ipa/83549
11706 * cif-code.def (VARIADIC_THUNK): New enum value.
11707 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11708 thunks.
11710 2018-01-03 Jan Beulich <jbeulich@suse.com>
11712 * sse.md (mov<mode>_internal): Tighten condition for when to use
11713 vmovdqu<ssescalarsize> for TI and OI modes.
11715 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11717 Update copyright years.
11719 2018-01-03 Martin Liska <mliska@suse.cz>
11721 PR ipa/83594
11722 * ipa-visibility.c (function_and_variable_visibility): Skip
11723 functions with noipa attribure.
11725 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11727 * gcc.c (process_command): Update copyright notice dates.
11728 * gcov-dump.c (print_version): Ditto.
11729 * gcov.c (print_version): Ditto.
11730 * gcov-tool.c (print_version): Ditto.
11731 * gengtype.c (create_file): Ditto.
11732 * doc/cpp.texi: Bump @copying's copyright year.
11733 * doc/cppinternals.texi: Ditto.
11734 * doc/gcc.texi: Ditto.
11735 * doc/gccint.texi: Ditto.
11736 * doc/gcov.texi: Ditto.
11737 * doc/install.texi: Ditto.
11738 * doc/invoke.texi: Ditto.
11740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11742 * vector-builder.h (vector_builder::m_full_nelts): Change from
11743 unsigned int to poly_uint64.
11744 (vector_builder::full_nelts): Update prototype accordingly.
11745 (vector_builder::new_vector): Likewise.
11746 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11747 (vector_builder::operator ==): Likewise.
11748 (vector_builder::finalize): Likewise.
11749 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11750 Take the number of elements as a poly_uint64 rather than an
11751 unsigned int.
11752 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11753 from unsigned int to poly_uint64.
11754 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11755 (vec_perm_indices::new_vector): Likewise.
11756 (vec_perm_indices::length): Likewise.
11757 (vec_perm_indices::nelts_per_input): Likewise.
11758 (vec_perm_indices::input_nelts): Likewise.
11759 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11760 number of elements per input as a poly_uint64 rather than an
11761 unsigned int. Use the original encoding for variable-length
11762 vectors, rather than clamping each individual element.
11763 For the second and subsequent elements in each pattern,
11764 clamp the step and base before clamping their sum.
11765 (vec_perm_indices::series_p): Handle polynomial element counts.
11766 (vec_perm_indices::all_in_range_p): Likewise.
11767 (vec_perm_indices_to_tree): Likewise.
11768 (vec_perm_indices_to_rtx): Likewise.
11769 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11770 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11771 (tree_vector_builder::new_binary_operation): Handle polynomial
11772 element counts. Return false if we need to know the number
11773 of elements at compile time.
11774 * fold-const.c (fold_vec_perm): Punt if the number of elements
11775 isn't known at compile time.
11777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11779 * vec-perm-indices.h (vec_perm_builder): Change element type
11780 from HOST_WIDE_INT to poly_int64.
11781 (vec_perm_indices::element_type): Update accordingly.
11782 (vec_perm_indices::clamp): Handle polynomial element_types.
11783 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11784 (vec_perm_indices::all_in_range_p): Likewise.
11785 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11786 than shwi trees.
11787 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11788 polynomial vec_perm_indices element types.
11789 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11790 * fold-const.c (fold_vec_perm): Likewise.
11791 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11792 * tree-vect-generic.c (lower_vec_perm): Likewise.
11793 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11794 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11795 element type to HOST_WIDE_INT.
11797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11798 Alan Hayward <alan.hayward@arm.com>
11799 David Sherwood <david.sherwood@arm.com>
11801 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11802 rather than an int. Use plus_constant.
11803 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11804 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11807 Alan Hayward <alan.hayward@arm.com>
11808 David Sherwood <david.sherwood@arm.com>
11810 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11811 a HOST_WIDE_INT to a poly_int64.
11813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11814 Alan Hayward <alan.hayward@arm.com>
11815 David Sherwood <david.sherwood@arm.com>
11817 * calls.c (load_register_parameters): Cope with polynomial
11818 mode sizes. Require a constant size for BLKmode parameters
11819 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11820 forces a parameter to be padded at the lsb end in order to
11821 fill a complete number of words, require the parameter size
11822 to be ordered wrt UNITS_PER_WORD.
11824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11825 Alan Hayward <alan.hayward@arm.com>
11826 David Sherwood <david.sherwood@arm.com>
11828 * reload1.c (spill_stack_slot_width): Change element type
11829 from unsigned int to poly_uint64_pod.
11830 (alter_reg): Treat mode sizes as polynomial.
11832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11833 Alan Hayward <alan.hayward@arm.com>
11834 David Sherwood <david.sherwood@arm.com>
11836 * reload.c (complex_word_subreg_p): New function.
11837 (reload_inner_reg_of_subreg, push_reload): Use it.
11839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11840 Alan Hayward <alan.hayward@arm.com>
11841 David Sherwood <david.sherwood@arm.com>
11843 * lra-constraints.c (process_alt_operands): Reject matched
11844 operands whose sizes aren't ordered.
11845 (match_reload): Refer to this check here.
11847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11848 Alan Hayward <alan.hayward@arm.com>
11849 David Sherwood <david.sherwood@arm.com>
11851 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11852 that the mode size is in the set {1, 2, 4, 8, 16}.
11854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11855 Alan Hayward <alan.hayward@arm.com>
11856 David Sherwood <david.sherwood@arm.com>
11858 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11859 Use plus_constant instead of gen_rtx_PLUS.
11861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11862 Alan Hayward <alan.hayward@arm.com>
11863 David Sherwood <david.sherwood@arm.com>
11865 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11866 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11867 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11868 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11869 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11870 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11871 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11872 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11873 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11874 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11875 a poly_int64.
11876 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11877 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11878 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11879 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11880 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11881 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11882 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11883 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11884 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11885 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11886 function.
11887 * expr.c (emit_move_resolve_push): Treat the input and result
11888 of PUSH_ROUNDING as a poly_int64.
11889 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11890 (emit_push_insn): Likewise.
11891 * lra-eliminations.c (mark_not_eliminable): Likewise.
11892 * recog.c (push_operand): Likewise.
11893 * reload1.c (elimination_effects): Likewise.
11894 * rtlanal.c (nonzero_bits1): Likewise.
11895 * calls.c (store_one_arg): Likewise. Require the padding to be
11896 known at compile time.
11898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11899 Alan Hayward <alan.hayward@arm.com>
11900 David Sherwood <david.sherwood@arm.com>
11902 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11903 Use plus_constant instead of gen_rtx_PLUS.
11905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11906 Alan Hayward <alan.hayward@arm.com>
11907 David Sherwood <david.sherwood@arm.com>
11909 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11910 rather than an int.
11912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11913 Alan Hayward <alan.hayward@arm.com>
11914 David Sherwood <david.sherwood@arm.com>
11916 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11917 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11918 via stack temporaries. Treat the mode size as polynomial too.
11920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11921 Alan Hayward <alan.hayward@arm.com>
11922 David Sherwood <david.sherwood@arm.com>
11924 * expr.c (expand_expr_real_2): When handling conversions involving
11925 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11926 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11927 as a poly_uint64 too.
11929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11930 Alan Hayward <alan.hayward@arm.com>
11931 David Sherwood <david.sherwood@arm.com>
11933 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11936 Alan Hayward <alan.hayward@arm.com>
11937 David Sherwood <david.sherwood@arm.com>
11939 * combine.c (can_change_dest_mode): Handle polynomial
11940 REGMODE_NATURAL_SIZE.
11941 * expmed.c (store_bit_field_1): Likewise.
11942 * expr.c (store_constructor): Likewise.
11943 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11944 and polynomial REGMODE_NATURAL_SIZE.
11945 (gen_lowpart_common): Likewise.
11946 * reginfo.c (record_subregs_of_mode): Likewise.
11947 * rtlanal.c (read_modify_subreg_p): Likewise.
11949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11950 Alan Hayward <alan.hayward@arm.com>
11951 David Sherwood <david.sherwood@arm.com>
11953 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11954 numbers of elements.
11956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11957 Alan Hayward <alan.hayward@arm.com>
11958 David Sherwood <david.sherwood@arm.com>
11960 * match.pd: Cope with polynomial numbers of vector elements.
11962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11963 Alan Hayward <alan.hayward@arm.com>
11964 David Sherwood <david.sherwood@arm.com>
11966 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11967 in a POINTER_PLUS_EXPR.
11969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11970 Alan Hayward <alan.hayward@arm.com>
11971 David Sherwood <david.sherwood@arm.com>
11973 * omp-simd-clone.c (simd_clone_subparts): New function.
11974 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11975 (ipa_simd_modify_function_body): Likewise.
11977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11978 Alan Hayward <alan.hayward@arm.com>
11979 David Sherwood <david.sherwood@arm.com>
11981 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11982 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11983 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11984 (expand_vector_condition, vector_element): Likewise.
11985 (subparts_gt): New function.
11986 (get_compute_type): Use subparts_gt.
11987 (count_type_subparts): Delete.
11988 (expand_vector_operations_1): Use subparts_gt instead of
11989 count_type_subparts.
11991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11992 Alan Hayward <alan.hayward@arm.com>
11993 David Sherwood <david.sherwood@arm.com>
11995 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11996 (vect_compile_time_alias): ...this new function. Do the calculation
11997 on poly_ints rather than trees.
11998 (vect_prune_runtime_alias_test_list): Update call accordingly.
12000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12001 Alan Hayward <alan.hayward@arm.com>
12002 David Sherwood <david.sherwood@arm.com>
12004 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12005 numbers of units.
12006 (vect_schedule_slp_instance): Likewise.
12008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12009 Alan Hayward <alan.hayward@arm.com>
12010 David Sherwood <david.sherwood@arm.com>
12012 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12013 constant and extern definitions for variable-length vectors.
12014 (vect_get_constant_vectors): Note that the number of units
12015 is known to be constant.
12017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12018 Alan Hayward <alan.hayward@arm.com>
12019 David Sherwood <david.sherwood@arm.com>
12021 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12022 of units as polynomial. Choose between WIDE and NARROW based
12023 on multiple_p.
12025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12026 Alan Hayward <alan.hayward@arm.com>
12027 David Sherwood <david.sherwood@arm.com>
12029 * tree-vect-stmts.c (simd_clone_subparts): New function.
12030 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12033 Alan Hayward <alan.hayward@arm.com>
12034 David Sherwood <david.sherwood@arm.com>
12036 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12037 vectors as polynomial. Use build_index_vector for
12038 IFN_GOMP_SIMD_LANE.
12040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12041 Alan Hayward <alan.hayward@arm.com>
12042 David Sherwood <david.sherwood@arm.com>
12044 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12045 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12046 for variable-length vectors.
12047 (vectorizable_mask_load_store): Treat the number of units as
12048 polynomial, asserting that it is constant if the condition has
12049 already been enforced.
12050 (vectorizable_store, vectorizable_load): Likewise.
12052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12053 Alan Hayward <alan.hayward@arm.com>
12054 David Sherwood <david.sherwood@arm.com>
12056 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12057 of units as polynomial. Punt if we can't tell at compile time
12058 which vector contains the final result.
12060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12061 Alan Hayward <alan.hayward@arm.com>
12062 David Sherwood <david.sherwood@arm.com>
12064 * tree-vect-loop.c (vectorizable_induction): Treat the number
12065 of units as polynomial. Punt on SLP inductions. Use an integer
12066 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12067 cast of such a series for variable-length floating-point
12068 reductions.
12070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12071 Alan Hayward <alan.hayward@arm.com>
12072 David Sherwood <david.sherwood@arm.com>
12074 * tree.h (build_index_vector): Declare.
12075 * tree.c (build_index_vector): New function.
12076 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12077 of units as polynomial, forcibly converting it to a constant if
12078 vectorizable_reduction has already enforced the condition.
12079 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12080 to create a {1,2,3,...} vector.
12081 (vectorizable_reduction): Treat the number of units as polynomial.
12082 Choose vectype_in based on the largest scalar element size rather
12083 than the smallest number of units. Enforce the restrictions
12084 relied on above.
12086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12087 Alan Hayward <alan.hayward@arm.com>
12088 David Sherwood <david.sherwood@arm.com>
12090 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12091 number of units as polynomial.
12093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12094 Alan Hayward <alan.hayward@arm.com>
12095 David Sherwood <david.sherwood@arm.com>
12097 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12098 * target.def (autovectorize_vector_sizes): Return the vector sizes
12099 by pointer, using vector_sizes rather than a bitmask.
12100 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12101 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12102 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12103 Likewise.
12104 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12105 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12106 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12107 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12108 * omp-general.c (omp_max_vf): Likewise.
12109 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12110 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12111 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12112 * tree-vect-slp.c (vect_slp_bb): Likewise.
12113 * doc/tm.texi: Regenerate.
12114 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12115 to a poly_uint64.
12116 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12117 the vector size as a poly_uint64 rather than an unsigned int.
12118 (current_vector_size): Change from an unsigned int to a poly_uint64.
12119 (get_vectype_for_scalar_type): Update accordingly.
12120 * tree.h (build_truth_vector_type): Take the size and number of
12121 units as a poly_uint64 rather than an unsigned int.
12122 (build_vector_type): Add a temporary overload that takes
12123 the number of units as a poly_uint64 rather than an unsigned int.
12124 * tree.c (make_vector_type): Likewise.
12125 (build_truth_vector_type): Take the number of units as a poly_uint64
12126 rather than an unsigned int.
12128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12129 Alan Hayward <alan.hayward@arm.com>
12130 David Sherwood <david.sherwood@arm.com>
12132 * target.def (get_mask_mode): Take the number of units and length
12133 as poly_uint64s rather than unsigned ints.
12134 * targhooks.h (default_get_mask_mode): Update accordingly.
12135 * targhooks.c (default_get_mask_mode): Likewise.
12136 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12137 * doc/tm.texi: Regenerate.
12139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12140 Alan Hayward <alan.hayward@arm.com>
12141 David Sherwood <david.sherwood@arm.com>
12143 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12144 * omp-general.c (omp_max_vf): Likewise.
12145 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12146 (expand_omp_simd): Handle polynomial safelen.
12147 * omp-low.c (omplow_simd_context): Add a default constructor.
12148 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12149 (lower_rec_simd_input_clauses): Update accordingly.
12150 (lower_rec_input_clauses): Likewise.
12152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12153 Alan Hayward <alan.hayward@arm.com>
12154 David Sherwood <david.sherwood@arm.com>
12156 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12157 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12158 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12159 (vect_analyze_slp_cost): Likewise.
12160 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12161 (vect_model_load_cost): Likewise.
12163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12164 Alan Hayward <alan.hayward@arm.com>
12165 David Sherwood <david.sherwood@arm.com>
12167 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12168 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12169 from an unsigned int * to a poly_uint64_pod *.
12170 (calculate_unrolling_factor): New function.
12171 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12174 Alan Hayward <alan.hayward@arm.com>
12175 David Sherwood <david.sherwood@arm.com>
12177 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12178 from an unsigned int to a poly_uint64.
12179 (_loop_vec_info::slp_unrolling_factor): Likewise.
12180 (_loop_vec_info::vectorization_factor): Change from an int
12181 to a poly_uint64.
12182 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12183 (vect_get_num_vectors): New function.
12184 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12185 (vect_get_num_copies): Use vect_get_num_vectors.
12186 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12187 to an unsigned int *.
12188 (vect_analyze_data_refs): Change min_vf from an int * to a
12189 poly_uint64 *.
12190 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12191 than an unsigned HOST_WIDE_INT.
12192 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12193 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12194 to an unsigned int *.
12195 (vect_analyze_data_ref_dependences): Likewise.
12196 (vect_compute_data_ref_alignment): Handle polynomial vf.
12197 (vect_enhance_data_refs_alignment): Likewise.
12198 (vect_prune_runtime_alias_test_list): Likewise.
12199 (vect_shift_permute_load_chain): Likewise.
12200 (vect_supportable_dr_alignment): Likewise.
12201 (dependence_distance_ge_vf): Take the vectorization factor as a
12202 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12203 (vect_analyze_data_refs): Change min_vf from an int * to a
12204 poly_uint64 *.
12205 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12206 vfm1 as a poly_uint64 rather than an int. Make the same change
12207 for the returned bound_scalar.
12208 (vect_gen_vector_loop_niters): Handle polynomial vf.
12209 (vect_do_peeling): Likewise. Update call to
12210 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12211 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12212 be constant.
12213 * tree-vect-loop.c (vect_determine_vectorization_factor)
12214 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12215 (vect_get_known_peeling_cost): Likewise.
12216 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12217 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12218 (vect_transform_loop): Likewise. Use the lowest possible VF when
12219 updating the upper bounds of the loop.
12220 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12221 rather than an int.
12222 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12223 polynomial unroll factors.
12224 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12225 (vect_make_slp_decision): Likewise.
12226 (vect_supported_load_permutation_p): Likewise, and polynomial
12227 vf too.
12228 (vect_analyze_slp_cost): Handle polynomial vf.
12229 (vect_slp_analyze_node_operations): Likewise.
12230 (vect_slp_analyze_bb_1): Likewise.
12231 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12232 than an unsigned HOST_WIDE_INT.
12233 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12234 (vectorizable_load): Handle polynomial vf.
12235 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12236 a poly_uint64.
12237 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12240 Alan Hayward <alan.hayward@arm.com>
12241 David Sherwood <david.sherwood@arm.com>
12243 * match.pd: Handle bit operations involving three constants
12244 and try to fold one pair.
12246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12248 * tree-vect-loop-manip.c: Include gimple-fold.h.
12249 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12250 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12251 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12252 Add a path that uses a step of VF instead of 1, but disable it
12253 for now.
12254 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12255 and niters_no_overflow parameters. Update calls to
12256 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12257 Create a new SSA name if the latter choses to use a ste other
12258 than zero, and return it via niters_vector_mult_vf_var.
12259 * tree-vect-loop.c (vect_transform_loop): Update calls to
12260 vect_do_peeling, vect_gen_vector_loop_niters and
12261 slpeel_make_loop_iterate_ntimes.
12262 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12263 (vect_gen_vector_loop_niters): Update declarations after above changes.
12265 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12267 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12268 128-bit round to integer instructions.
12269 (ceil<mode>2): Likewise.
12270 (btrunc<mode>2): Likewise.
12271 (round<mode>2): Likewise.
12273 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12275 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12276 unaligned VSX load/store on P8/P9.
12277 (expand_block_clear): Allow the use of unaligned VSX
12278 load/store on P8/P9.
12280 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12282 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12283 New function.
12284 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12285 swap associated with both a load and a store.
12287 2018-01-02 Andrew Waterman <andrew@sifive.com>
12289 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12290 * config/riscv/riscv.md (clear_cache): Use it.
12292 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12294 * web.c: Remove out-of-date comment.
12296 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12298 * expr.c (fixup_args_size_notes): Check that any existing
12299 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12300 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12301 (emit_single_push_insn): ...here.
12303 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12305 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12306 (const_vector_encoded_nelts): New function.
12307 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12308 (const_vector_int_elt, const_vector_elt): Declare.
12309 * emit-rtl.c (const_vector_int_elt_1): New function.
12310 (const_vector_elt): Likewise.
12311 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12312 of CONST_VECTOR_ELT.
12314 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12316 * expr.c: Include rtx-vector-builder.h.
12317 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12318 directly on the tree encoding.
12319 (const_vector_from_tree): Likewise.
12320 * optabs.c: Include rtx-vector-builder.h.
12321 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12322 sequence of "u" values.
12323 * vec-perm-indices.c: Include rtx-vector-builder.h.
12324 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12325 directly on the vec_perm_indices encoding.
12327 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12329 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12330 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12331 * rtx-vector-builder.h: New file.
12332 * rtx-vector-builder.c: Likewise.
12333 * rtl.h (rtx_def::u2): Add a const_vector field.
12334 (CONST_VECTOR_NPATTERNS): New macro.
12335 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12336 (CONST_VECTOR_DUPLICATE_P): Likewise.
12337 (CONST_VECTOR_STEPPED_P): Likewise.
12338 (CONST_VECTOR_ENCODED_ELT): Likewise.
12339 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12340 (unwrap_const_vec_duplicate): Likewise.
12341 (const_vec_series_p): Check for a non-duplicated vector encoding.
12342 Say that the function only returns true for integer vectors.
12343 * emit-rtl.c: Include rtx-vector-builder.h.
12344 (gen_const_vec_duplicate_1): Delete.
12345 (gen_const_vector): Call gen_const_vec_duplicate instead of
12346 gen_const_vec_duplicate_1.
12347 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12348 (gen_const_vec_duplicate): Use rtx_vector_builder.
12349 (gen_const_vec_series): Likewise.
12350 (gen_rtx_CONST_VECTOR): Likewise.
12351 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12352 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12353 Build a new vector rather than modifying a CONST_VECTOR in-place.
12354 (handle_special_swappables): Update call accordingly.
12355 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12356 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12357 Build a new vector rather than modifying a CONST_VECTOR in-place.
12358 (handle_special_swappables): Update call accordingly.
12360 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12362 * simplify-rtx.c (simplify_const_binary_operation): Use
12363 CONST_VECTOR_ELT instead of XVECEXP.
12365 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12367 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12368 the selector elements to be different from the data elements
12369 if the selector is a VECTOR_CST.
12370 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12371 ssizetype for the selector.
12373 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12375 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12376 before testing each element individually.
12377 * tree-vect-generic.c (lower_vec_perm): Likewise.
12379 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12381 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12382 * selftest-run-tests.c (selftest::run_tests): Call it.
12383 * vector-builder.h (vector_builder::operator ==): New function.
12384 (vector_builder::operator !=): Likewise.
12385 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12386 (vec_perm_indices::all_from_input_p): New function.
12387 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12388 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12389 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12390 instead of reading the VECTOR_CST directly. Detect whether both
12391 vector inputs are the same before constructing the vec_perm_indices,
12392 and update the number of inputs argument accordingly. Use the
12393 utility functions added above. Only construct sel2 if we need to.
12395 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12397 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12398 the broadcast of the low byte.
12399 (expand_mult_highpart): Use an explicit encoding for the permutes.
12400 * optabs-query.c (can_mult_highpart_p): Likewise.
12401 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12402 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12403 (vectorizable_bswap): Likewise.
12404 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12405 explicit encoding for the power-of-2 permutes.
12406 (vect_permute_store_chain): Likewise.
12407 (vect_grouped_load_supported): Likewise.
12408 (vect_permute_load_chain): Likewise.
12410 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12412 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12413 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12414 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12415 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12416 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12417 (vect_gen_perm_mask_any): Likewise.
12419 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12421 * int-vector-builder.h: New file.
12422 * vec-perm-indices.h: Include int-vector-builder.h.
12423 (vec_perm_indices): Redefine as an int_vector_builder.
12424 (auto_vec_perm_indices): Delete.
12425 (vec_perm_builder): Redefine as a stand-alone class.
12426 (vec_perm_indices::vec_perm_indices): New function.
12427 (vec_perm_indices::clamp): Likewise.
12428 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12429 (vec_perm_indices::new_vector): New function.
12430 (vec_perm_indices::new_expanded_vector): Update for new
12431 vec_perm_indices class.
12432 (vec_perm_indices::rotate_inputs): New function.
12433 (vec_perm_indices::all_in_range_p): Operate directly on the
12434 encoded form, without computing elided elements.
12435 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12436 encoding. Update for new vec_perm_indices class.
12437 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12438 the given vec_perm_builder.
12439 (expand_vec_perm_var): Update vec_perm_builder constructor.
12440 (expand_mult_highpart): Use vec_perm_builder instead of
12441 auto_vec_perm_indices.
12442 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12443 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12444 or double series encoding as appropriate.
12445 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12446 vec_perm_indices instead of auto_vec_perm_indices.
12447 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12448 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12449 (vect_permute_store_chain): Likewise.
12450 (vect_grouped_load_supported): Likewise.
12451 (vect_permute_load_chain): Likewise.
12452 (vect_shift_permute_load_chain): Likewise.
12453 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12454 (vect_transform_slp_perm_load): Likewise.
12455 (vect_schedule_slp_instance): Likewise.
12456 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12457 (vectorizable_mask_load_store): Likewise.
12458 (vectorizable_bswap): Likewise.
12459 (vectorizable_store): Likewise.
12460 (vectorizable_load): Likewise.
12461 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12462 vec_perm_indices instead of auto_vec_perm_indices. Use
12463 tree_to_vec_perm_builder to read the vector from a tree.
12464 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12465 vec_perm_builder instead of a vec_perm_indices.
12466 (have_whole_vector_shift): Use vec_perm_builder and
12467 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12468 truncation to calc_vec_perm_mask_for_shift.
12469 (vect_create_epilog_for_reduction): Likewise.
12470 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12471 from auto_vec_perm_indices to vec_perm_indices.
12472 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12473 instead of changing individual elements.
12474 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12475 the vector in d.perm.
12476 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12477 from auto_vec_perm_indices to vec_perm_indices.
12478 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12479 instead of changing individual elements.
12480 (arm_vectorize_vec_perm_const): Use new_vector to install
12481 the vector in d.perm.
12482 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12483 Update vec_perm_builder constructor.
12484 (rs6000_expand_interleave): Likewise.
12485 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12486 (rs6000_expand_interleave): Likewise.
12488 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12490 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12491 to qimode could truncate the indices.
12492 * optabs.c (expand_vec_perm_var): Likewise.
12494 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12496 * Makefile.in (OBJS): Add vec-perm-indices.o.
12497 * vec-perm-indices.h: New file.
12498 * vec-perm-indices.c: Likewise.
12499 * target.h (vec_perm_indices): Replace with a forward class
12500 declaration.
12501 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12502 * optabs.h: Include vec-perm-indices.h.
12503 (expand_vec_perm): Delete.
12504 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12505 (expand_vec_perm_const): Declare.
12506 * target.def (vec_perm_const_ok): Replace with...
12507 (vec_perm_const): ...this new hook.
12508 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12509 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12510 * doc/tm.texi: Regenerate.
12511 * optabs.def (vec_perm_const): Delete.
12512 * doc/md.texi (vec_perm_const): Likewise.
12513 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12514 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12515 expand_vec_perm for constant permutation vectors. Assert that
12516 the mode of variable permutation vectors is the integer equivalent
12517 of the mode that is being permuted.
12518 * optabs-query.h (selector_fits_mode_p): Declare.
12519 * optabs-query.c: Include vec-perm-indices.h.
12520 (selector_fits_mode_p): New function.
12521 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12522 is defined, instead of checking whether the vec_perm_const_optab
12523 exists. Use targetm.vectorize.vec_perm_const instead of
12524 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12525 fit in the vector mode before using a variable permute.
12526 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12527 vec_perm_indices instead of an rtx.
12528 (expand_vec_perm): Replace with...
12529 (expand_vec_perm_const): ...this new function. Take the selector
12530 as a vec_perm_indices rather than an rtx. Also take the mode of
12531 the selector. Update call to shift_amt_for_vec_perm_mask.
12532 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12533 Use vec_perm_indices::new_expanded_vector to expand the original
12534 selector into bytes. Check whether the indices fit in the vector
12535 mode before using a variable permute.
12536 (expand_vec_perm_var): Make global.
12537 (expand_mult_highpart): Use expand_vec_perm_const.
12538 * fold-const.c: Includes vec-perm-indices.h.
12539 * tree-ssa-forwprop.c: Likewise.
12540 * tree-vect-data-refs.c: Likewise.
12541 * tree-vect-generic.c: Likewise.
12542 * tree-vect-loop.c: Likewise.
12543 * tree-vect-slp.c: Likewise.
12544 * tree-vect-stmts.c: Likewise.
12545 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12546 Delete.
12547 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12548 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12549 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12550 (aarch64_vectorize_vec_perm_const): ...this new function.
12551 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12552 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12553 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12554 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12555 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12556 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12557 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12558 into...
12559 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12560 check for NEON modes.
12561 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12562 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12563 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12564 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12565 into...
12566 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12567 the old VEC_PERM_CONST conditions.
12568 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12569 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12570 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12571 (ia64_vectorize_vec_perm_const_ok): Merge into...
12572 (ia64_vectorize_vec_perm_const): ...this new function.
12573 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12574 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12575 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12576 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12577 * config/mips/mips.c (mips_expand_vec_perm_const)
12578 (mips_vectorize_vec_perm_const_ok): Merge into...
12579 (mips_vectorize_vec_perm_const): ...this new function.
12580 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12581 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12582 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12583 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12584 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12585 (rs6000_expand_vec_perm_const): Delete.
12586 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12587 Delete.
12588 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12589 (altivec_expand_vec_perm_const_le): Take each operand individually.
12590 Operate on constant selectors rather than rtxes.
12591 (altivec_expand_vec_perm_const): Likewise. Update call to
12592 altivec_expand_vec_perm_const_le.
12593 (rs6000_expand_vec_perm_const): Delete.
12594 (rs6000_vectorize_vec_perm_const_ok): Delete.
12595 (rs6000_vectorize_vec_perm_const): New function.
12596 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12597 an element count and rtx array.
12598 (rs6000_expand_extract_even): Update call accordingly.
12599 (rs6000_expand_interleave): Likewise.
12600 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12601 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12602 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12603 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12604 (rs6000_expand_vec_perm_const): Delete.
12605 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12606 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12607 (altivec_expand_vec_perm_const_le): Take each operand individually.
12608 Operate on constant selectors rather than rtxes.
12609 (altivec_expand_vec_perm_const): Likewise. Update call to
12610 altivec_expand_vec_perm_const_le.
12611 (rs6000_expand_vec_perm_const): Delete.
12612 (rs6000_vectorize_vec_perm_const_ok): Delete.
12613 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12614 reference to the SPE evmerge intructions.
12615 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12616 an element count and rtx array.
12617 (rs6000_expand_extract_even): Update call accordingly.
12618 (rs6000_expand_interleave): Likewise.
12619 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12620 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12621 new function.
12622 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12624 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12626 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12627 vector mode and that that mode matches the mode of the data
12628 being permuted.
12629 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12630 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12631 directly using expand_vec_perm_1 when forcing selectors into
12632 registers.
12633 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12635 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12637 * optabs-query.h (can_vec_perm_p): Delete.
12638 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12639 * optabs-query.c (can_vec_perm_p): Split into...
12640 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12641 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12642 particular selector is valid.
12643 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12644 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12645 (vect_grouped_load_supported): Likewise.
12646 (vect_shift_permute_load_chain): Likewise.
12647 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12648 (vect_transform_slp_perm_load): Likewise.
12649 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12650 (vectorizable_bswap): Likewise.
12651 (vect_gen_perm_mask_checked): Likewise.
12652 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12653 implementations of variable permutation vectors into account
12654 when deciding which selector to use.
12655 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12656 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12657 with a false third argument.
12658 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12659 to test whether the constant selector is valid and can_vec_perm_var_p
12660 to test whether a variable selector is valid.
12662 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12664 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12665 * optabs-query.c (can_vec_perm_p): Likewise.
12666 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12667 instead of vec_perm_indices.
12668 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12669 (vect_gen_perm_mask_checked): Likewise,
12670 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12671 (vect_gen_perm_mask_checked): Likewise,
12673 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12675 * optabs-query.h (qimode_for_vec_perm): Declare.
12676 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12677 (qimode_for_vec_perm): ...this new function.
12678 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12680 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12682 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12683 does not have a conditional at the top.
12685 2018-01-02 Richard Biener <rguenther@suse.de>
12687 * ipa-inline.c (big_speedup_p): Fix expression.
12689 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12691 PR target/81616
12692 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12693 for generic 4->6.
12695 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12697 PR target/81616
12698 Generic tuning.
12699 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12700 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12701 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12702 cond_taken_branch_cost 3->4.
12704 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12706 PR tree-optimization/83581
12707 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12708 TODO_cleanup_cfg if any changes have been made.
12710 PR middle-end/83608
12711 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12712 convert_modes if target mode has the right side, but different mode
12713 class.
12715 PR middle-end/83609
12716 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12717 last argument when extracting from CONCAT. If either from_real or
12718 from_imag is NULL, use expansion through memory. If result is not
12719 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12720 the parts directly to inner mode, if even that fails, use expansion
12721 through memory.
12723 PR middle-end/83623
12724 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12725 check for bswap in mode rather than HImode and use that in expand_unop
12726 too.
12727 \f
12728 Copyright (C) 2018 Free Software Foundation, Inc.
12730 Copying and distribution of this file, with or without modification,
12731 are permitted in any medium without royalty provided the copyright
12732 notice and this notice are preserved.