1 2018-05-16 Richard Biener <rguenther@suse.de>
3 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4 verify the [VEC_]COND_EXPR embedded comparison.
6 2018-05-15 Martin Sebor <msebor@redhat.com>
8 PR tree-optimization/85753
9 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
10 RECORD_TYPE in addition to ARRAY_TYPE.
12 2018-05-15 Martin Sebor <msebor@redhat.com>
15 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
17 2018-05-15 Richard Biener <rguenther@suse.de>
19 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
20 add by_clobber_p one. Change algorithm to collect all defs
21 representing uses we need to walk and try reducing them to
22 a single one before failing.
23 (dse_dom_walker::dse_optimize_stmt): Adjust.
25 2018-05-13 Mark Wielaard <mark@klomp.org>
27 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
28 (size_of_loc_descr): Likewise.
29 (output_loc_operands): Likewise.
30 (output_loc_operands_raw): Likewise.
31 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
32 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
33 (hash_loc_operands): Likewise.
34 (compare_loc_operands): Likewise.
36 2018-05-14 Mark Wielaard <mark@klomp.org>
38 * dwarf2out.c (count_index_addrs): New function.
39 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
41 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
43 PR tree-optimization/83648
44 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
45 return value as malloc candidate.
47 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
50 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
51 as true in call to suggest_attribute.
53 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
55 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
58 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
60 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
62 (aarch64_crypto_aesd_fused): Likewise.
64 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
66 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
67 (movsi_aarch64): Likewise.
68 (load_pairsi): Likewise.
69 (load_pairdi): Likewise.
70 (store_pairsi): Likewise.
71 (store_pairdi): Likewise.
72 (load_pairsf): Likewise.
73 (load_pairdf): Likewise.
74 (store_pairsf): Likewise.
75 (store_pairdf): Likewise.
76 (zero_extend): Likewise.
77 (trunc): Swap alternatives.
78 (fcvt_target): Add '?' to prefer w over r.
80 2018-05-14 Jakub Jelinek <jakub@redhat.com>
83 * config/i386/i386.md: Disallow non-commutative arithmetics in
84 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
85 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
86 in the peephole2 before it.
88 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
90 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
91 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
92 (ix86_handle_option): Handle -mcldemote.
93 * config.gcc: New header.
94 * config/i386/cldemoteintrin.h: New file.
95 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
96 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
98 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
99 OPTION_MASK_ISA_CLDEMOTE.
100 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
101 (ix86_valid_target_attribute_inner_p): Ditto.
102 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
103 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
104 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
105 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
106 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
108 * config/i386/i386.opt: Add -mcldemote.
109 * config/i386/x86intrin.h: New header.
110 * doc/invoke.texi: Add -mcldemote.
112 2018-05-14 Richard Biener <rguenther@suse.de>
114 * doc/match-and-simplify.texi: Adjust :s documentation.
116 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
118 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
119 intended memcpy size.
120 (REORDER_45): Likewise.
122 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
125 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
126 * vec.c (qsort_chk): Use gcc_qsort.
127 * Makefile.in (OBJS-libcommon): Add sort.o.
128 (build/sort.o): New target. Use it...
129 (BUILD_RTL): ... here, and...
130 (build/gencfn-macros): ... here, and...
131 (build/genmatch): ... here.
133 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
134 Chung-Ju Wu <jasonwucj@gmail.com>
136 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
137 * config/nds32/nds32-graywolf.md: New file.
138 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
139 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
141 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
142 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
143 * config/nds32/nds32.md (pipeline_model): Add graywolf.
144 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
145 * config/nds32/pipelines.md: Include n15 settings.
147 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
148 Chung-Ju Wu <jasonwucj@gmail.com>
150 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
151 * config/nds32/nds32-n13.md: New file.
152 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
153 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
155 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
156 * config/nds32/nds32.md (pipeline_model): Add n13.
157 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
158 * config/nds32/pipelines.md: Include n13 settings.
160 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
161 Chung-Ju Wu <jasonwucj@gmail.com>
163 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
164 * config/nds32/nds32-n10.md: New file.
165 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
166 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
168 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
169 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
170 * config/nds32/nds32.md (pipeline_model): Add n10.
171 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
172 * config/nds32/pipelines.md: Include n10 settings.
174 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
175 Kito Cheng <kito.cheng@gmail.com>
176 Chung-Ju Wu <jasonwucj@gmail.com>
178 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
179 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
180 Add enum values for DSP extension instructions.
181 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
183 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
184 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
186 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
187 * config/nds32/nds32-dspext.md: New file for DSP implementation.
188 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
189 * config/nds32/nds32-intrinsic.md: Likewise.
190 * config/nds32/nds32_intrinsic.h: Likewise.
191 * config/nds32/nds32-md-auxiliary.c: Likewise.
192 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
193 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
194 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
195 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
196 * config/nds32/nds32-protos.h: New declarations for DSP extension.
197 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
198 TYPE_DMAC in switch statement.
199 * config/nds32/nds32.c: New checking and implementation for DSP
200 extension instructions.
201 * config/nds32/nds32.h: Likewise.
202 * config/nds32/nds32.md: Likewise.
203 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
204 * config/nds32/predicates.md: Implement new predicates for DSP
207 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
209 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
210 Reformat alternatives and attributes so it is easier to identify
211 which constraints/attributes go with which instruction.
212 (mov<mode>_hardfloat32, FMOVE64): Likewise.
213 (mov<mode>_softfloat32, FMOVE64): Likewise.
214 (mov<mode>_hardfloat64, FMOVE64): Likewise.
215 (mov<mode>_softfloat64, FMOVE64): Likewise.
217 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
219 * doc/extend.texi (PowerPC Built-in Functions): Rename this
221 (Basic PowerPC Built-in Functions): The new name of the
222 subsection previously known as "PowerPC Built-in Functions".
223 (Basic PowerPC Built-in Functions Available on all Configurations):
225 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
226 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
227 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
228 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
230 2018-05-11 Martin Jambor <mjambor@suse.cz>
233 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
236 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
239 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
241 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
243 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
244 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
245 (ix86_handle_option): Handle -mwaitpkg.
246 * config.gcc: New header.
247 * config/i386/cpuid.h (bit_WAITPKG): New bit.
248 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
249 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
251 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
252 OPTION_MASK_ISA_WAITPKG.
253 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
254 (ix86_option_override_internal): Add PTA_WAITPKG.
255 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
256 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
258 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
259 __builtin_ia32_umwait and __builtin_ia32_tpause.
260 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
261 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
262 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
263 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
264 UNSPECV_TPAUSE): New.
265 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
266 * config/i386/i386.opt: Add -mwaitpkg.
267 * config/i386/waitpkgintrin.h: New file.
268 * config/i386/x86intrin.h: New header.
269 * doc/invoke.texi: Add -mwaitpkg.
271 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
274 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
276 (cortex-m0): Use armv6s-m isa.
277 (cortex-m0plus): Likewise.
278 (cortex-m1): Likewise.
279 (cortex-m0.small-multiply): Likewise.
280 (cortex-m0plus.small-multiply): Likewise.
281 (cortex-m1.small-multiply): Likewise.
283 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
284 Jakub Jelinek <jakub@redhat.com>
286 PR tree-optimization/85692
287 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
288 source permute as well.
290 2018-05-11 Martin Liska <mliska@suse.cz>
293 * doc/extend.texi: Document LLVM style format for no_sanitize
296 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
298 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
299 mode_supports_vsx_dform_quad to mode_supports_dq_form.
300 (mode_supports_vsx_dform_quad): Likewise.
301 (mode_supports_vmx_dform): Move these functions to be next to the
302 other mode_supports functions.
303 (mode_supports_dq_form): Likewise.
304 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
305 mode_supports_dq_form.
306 (reg_offset_addressing_ok_p): Likewise.
307 (offsettable_ok_by_alignment): Likewise.
308 (rs6000_legitimate_offset_address_p): Likewise.
309 (legitimate_lo_sum_address_p): Likewise.
310 (rs6000_legitimize_address): Likewise.
311 (rs6000_legitimize_reload_address): Likewise.
312 (rs6000_secondary_reload_inner): Likewise.
313 (rs6000_preferred_reload_class): Likewise.
314 (rs6000_output_move_128bit): Likewise.
316 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
318 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
319 Generate SImode target register for null target.
320 <case IX86_BUILTIN_XGETBV>: Ditto.
321 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
322 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
324 2018-05-10 Carl Love <cel@us.ibm.com>
326 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
327 dcbtt and dcbtstt if operands[2] is 0.
329 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
332 * config/i386/sse.md (usadv64qi): New expander.
334 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
336 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
337 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
338 -maltivec=be support.
339 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
340 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
341 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
342 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
343 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
344 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
345 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
346 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
347 altivec_vsumsws): Adjust.
348 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
349 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
350 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
352 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
353 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
354 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
355 (altivec_lve<VI_char>x): Delete expand.
356 (*altivec_lve<VI_char>x_internal): Rename to...
357 (altivec_lve<VI_char>x): ... this.
358 (altivec_lvxl_<mode>): Delete expand.
359 (*altivec_lvxl_<mode>_internal): Rename to ...
360 (altivec_lvxl_<mode>): ... this.
361 (altivec_stvxl_<mode>): Delete expand.
362 (*altivec_stvxl_<mode>_internal): Rename to ...
363 (altivec_stvxl_<mode>): ... this.
364 (altivec_stve<VI_char>x): Delete expand.
365 (*altivec_stve<VI_char>x_internal): Rename to ...
366 (altivec_stve<VI_char>x): ... this.
367 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
368 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
369 reduc_plus_scal_<mode>): Adjust.
370 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
372 (rs6000_cpu_cpp_builtins): Adjust.
373 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
374 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
375 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
376 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
377 -maltivec=be support.
378 (rs6000_split_vec_extract_var): Adjust.
379 (rs6000_split_v4si_init): Adjust.
380 (swap_selector_for_mode): Delete.
381 (altivec_expand_lvx_be, altivec_expand_stvx_be,
382 altivec_expand_stvex_be): Delete.
383 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
384 -maltivec=be support.
385 (rs6000_gimple_fold_builtin): Ditto.
386 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
388 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
389 (TARGET_DIRECT_MOVE_64BIT): Adjust.
390 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
391 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
392 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
393 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
394 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
395 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
396 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
397 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
398 anonymous split): Adjust.
399 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
400 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
402 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
404 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
405 when --with-gxx-include-dir is also specified.
406 * configure: Regenerate.
408 2018-05-09 Jim Wilson <jimw@sifive.com>
411 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
412 * config/riscv/t-withmultilib: New.
413 * config/riscv/withmultilib.h: New.
414 * doc/install.texi: Document RISC-V --with-multilib-list support.
416 2018-05-09 Richard Biener <rguenther@suse.de>
418 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
420 (vect_bb_vectorization_profitable_p): Adjust. Compute
421 actual scalar cost using the cost vector and the add_stmt_cost
424 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
426 PR rtl-optimization/85645
427 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
428 in the REG_CFA_REGISTER note for LR, don't leave it empty.
430 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
432 PR rtl-optimization/85645
433 * shrink-wrap.c (spread_components): Return a boolean saying if
434 anything was changed.
435 (try_shrink_wrapping_separate): Iterate spread_components until
436 nothing changes anymore.
438 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
440 PR rtl-optimization/85645
441 * regrename.c (build_def_use): Also kill the chains that include the
442 destination of a REG_CFA_REGISTER note.
444 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
446 PR rtl-optimization/85645
447 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
448 insn that has a REG_CFA_REGISTER note.
450 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
452 * cfgexpand.c (expand_clobber): New function.
453 (expand_gimple_stmt_1): Use it.
454 * tree-vect-stmts.c (vect_clobber_variable): New function,
456 (vectorizable_simd_clone_call): ...here.
457 (vectorizable_store): Emit a clobber either side of an
458 IFN_STORE_LANES sequence.
459 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
461 2018-05-09 Tom de Vries <tom@codesourcery.com>
464 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
465 (define_insn "trap_if_false"): Add exit after trap.
467 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
469 PR rtl-optimization/85638
470 * bb-reorder.c: Include common/common-target.h.
471 (create_forwarder_block): New function extracted from...
472 (fix_up_crossing_landing_pad): ...here. Rename into...
473 (dw2_fix_up_crossing_landing_pad): ...this.
474 (sjlj_fix_up_crossing_landing_pad): New function.
475 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
476 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
477 from both partitions and exit the loop after one iteration.
479 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
482 * doc/extend.texi (PowerPC Built-in Functions): Rename this
484 (Basic PowerPC Built-in Functions): The new name of the
485 subsection previously known as "PowerPC Built-in Functions".
486 (Basic PowerPC Built-in Functions Available on all Configurations):
488 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
490 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
492 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
494 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
497 2018-05-08 Jim Wilson <jimw@sifive.com>
499 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
500 (LD_EMUL_SUFFIX): New.
503 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
505 * doc/extend.texi (PowerPC Built-in Functions): Rename this
507 (Basic PowerPC Built-in Functions): The new name of the
508 subsection previously known as "PowerPC Built-in Functions".
509 (Basic PowerPC Built-in Functions Available on all Configurations):
511 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
513 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
515 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
517 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
520 2018-05-08 Jakub Jelinek <jakub@redhat.com>
523 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
524 after cmpelim optimization.
526 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
528 * config.gcc: Support "goldmont".
529 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
530 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
532 * config/i386/i386.c (m_GOLDMONT): Define.
533 (processor_target_table): Add "goldmont".
534 (PTA_GOLDMONT): Define.
535 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
536 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
537 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
538 (fold_builtin_cpu): Add "goldmont".
539 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
540 (ix86_option_override_internal): Add "goldmont".
541 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
542 (processor_type): Add PROCESSOR_GOLDMONT.
543 * config/i386/i386.md: Add CPU "glm".
544 * config/i386/glm.md: New file.
545 * config/i386/x86-tune.def: Add m_GOLDMONT.
546 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
548 2018-05-08 Jakub Jelinek <jakub@redhat.com>
551 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
553 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
554 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
555 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
558 * config/i386/i386.c (ix86_fold_builtin): Handle
559 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
562 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
563 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
565 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
568 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
569 (check_arch): Likewise.
570 (check_fpu): Return the result rather than printing it.
571 (end arch): Fix operator precedence.
573 (END): Print the result from check_fpu.
575 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
576 Alan Hayward <alan.hayward@arm.com>
577 David Sherwood <david.sherwood@arm.com>
579 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
580 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
581 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
582 (*fcmuo<mode>_and): New patterns.
584 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
586 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
587 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
588 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
589 (cmp_op, sve_imm_con): New code attributes.
590 (SVE_COND_INT_CMP, imm_con): Delete.
591 (cmp_op): Remove above unspecs from int attribute.
592 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
594 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
595 comparison-specific unspecs.
596 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
597 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
598 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
599 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
600 (*vec_fcm<cmp_op><mode>): Rename to...
601 (*fcm<cmp_op><mode>): ...this and adjust likewise.
602 (*vec_fcmuo<mode>): Rename to...
603 (*fcmuo<mode>): ...this and adjust likewise.
604 (*pred_fcm<cmp_op><mode>): New pattern.
605 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
606 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
608 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
610 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
611 (aarch64_emit_sve_predicated_cond): New function.
612 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
613 (aarch64_emit_unspec_cond_or): Replace with...
614 (aarch64_emit_sve_or_conds): ...this new function. Use
615 aarch64_emit_sve_ptrue_op for the individual comparisons and
616 aarch64_emit_binop to OR them together.
617 (aarch64_emit_inverted_unspec_cond): Replace with...
618 (aarch64_emit_sve_inverted_cond): ...this new function. Use
619 aarch64_emit_sve_ptrue_op for the comparison and
620 aarch64_emit_unop to invert the result.
621 (aarch64_expand_sve_vec_cmp_float): Update after the above
622 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
624 2018-05-07 Nathan Sidwell <nathan@acm.org>
626 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
627 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
628 (Backwards Compatibility): Likewise.
630 2018-05-07 Luis Machado <luis.machado@linaro.org>
634 2018-05-07 Luis Machado <luis.machado@linaro.org>
636 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
637 <prefetch_dynamic_strides>: New const bool field.
638 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
639 prefetch_dynamic_strides.
640 (exynosm1_prefetch_tune): Likewise.
641 (thunderxt88_prefetch_tune): Likewise.
642 (thunderx_prefetch_tune): Likewise.
643 (thunderx2t99_prefetch_tune): Likewise.
644 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
645 (aarch64_override_options_internal): Update to set
646 PARAM_PREFETCH_DYNAMIC_STRIDES.
647 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
648 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
649 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
650 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
651 prefetch-dynamic-strides setting.
653 2018-05-07 Luis Machado <luis.machado@linaro.org>
655 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
656 <minimum_stride>: New const int field.
657 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
658 minimum_stride field.
659 (exynosm1_prefetch_tune): Likewise.
660 (thunderxt88_prefetch_tune): Likewise.
661 (thunderx_prefetch_tune): Likewise.
662 (thunderx2t99_prefetch_tune): Likewise.
663 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
664 (aarch64_override_options_internal): Update to set
665 PARAM_PREFETCH_MINIMUM_STRIDE.
666 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
667 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
668 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
669 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
670 stride is constant and is below the minimum stride threshold.
672 2018-05-07 Luis Machado <luis.machado@linaro.org>
674 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
677 2018-05-07 Luis Machado <luis.machado@linaro.org>
679 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
680 <prefetch_dynamic_strides>: New const bool field.
681 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
682 prefetch_dynamic_strides.
683 (exynosm1_prefetch_tune): Likewise.
684 (thunderxt88_prefetch_tune): Likewise.
685 (thunderx_prefetch_tune): Likewise.
686 (thunderx2t99_prefetch_tune): Likewise.
687 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
688 (aarch64_override_options_internal): Update to set
689 PARAM_PREFETCH_DYNAMIC_STRIDES.
690 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
691 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
692 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
693 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
694 prefetch-dynamic-strides setting.
696 2018-05-07 Luis Machado <luis.machado@linaro.org>
698 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
699 <minimum_stride>: New const int field.
700 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
701 minimum_stride field.
702 (exynosm1_prefetch_tune): Likewise.
703 (thunderxt88_prefetch_tune): Likewise.
704 (thunderx_prefetch_tune): Likewise.
705 (thunderx2t99_prefetch_tune): Likewise.
706 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
707 (aarch64_override_options_internal): Update to set
708 PARAM_PREFETCH_MINIMUM_STRIDE.
709 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
710 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
711 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
712 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
713 stride is constant and is below the minimum stride threshold.
715 2018-05-06 Jakub Jelinek <jakub@redhat.com>
718 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
719 the type is addressable. Don't force op into register if it has
722 2018-05-05 Roland McGrath <mcgrathr@google.com>
725 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
726 any section for which we don't know a specific type it should have,
727 regardless of name. Previously this was done only for the exact
728 names ".init_array", ".fini_array", and ".preinit_array".
729 (default_elf_asm_named_section): Add comment about
730 relationship with default_section_type_flags and SECTION_NOTYPE.
731 (get_section): Don't consider it a type conflict if one side has
732 SECTION_NOTYPE and the other doesn't, as long as neither has the
733 SECTION_BSS et al used in the default_section_type_flags logic.
735 2018-05-05 Tom de Vries <tom@codesourcery.com>
738 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
739 (workaround_barsyncs): New function.
740 (nvptx_reorg): Use workaround_barsyncs.
741 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
742 (define_expand "nvptx_membar_cta"): New define_expand.
743 (define_insn "*nvptx_membar_cta"): New insn.
745 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
747 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
748 To improve optimization opportunities.
749 * builtin-types.def: The new needed builtin types for the above.
751 2018-05-04 Richard Biener <rguenther@suse.de>
753 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
754 * gimple-ssa-store-merging.c
755 (imm_store_chain_info::output_merged_store): Remove redundant create,
756 release split_store vector contents on failure.
757 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
758 scalar stmt vector on cache hit.
760 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
762 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
764 * config.gcc (powerpc-xilinx-eabi*): Remove.
765 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
767 (fusion_addis_mem_combo_load): Ditto.
768 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
770 (rs6000_cpu_cpp_builtins): Ditto.
771 * config/rs6000/rs6000-linux.c
772 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
773 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
774 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
776 (rs6000_setup_reg_addr_masks): Ditto.
777 (rs6000_init_hard_regno_mode_ok): Ditto.
778 (rs6000_option_override_internal): Ditto.
779 (legitimate_lo_sum_address_p): Ditto.
780 (rs6000_legitimize_address): Ditto.
781 (rs6000_legitimize_reload_address): Ditto.
782 (rs6000_legitimate_address_p): Ditto.
783 (abi_v4_pass_in_fpr): Ditto.
784 (setup_incoming_varargs): Ditto.
785 (rs6000_gimplify_va_arg): Ditto.
786 (rs6000_split_multireg_move): Ditto.
787 (rs6000_savres_strategy): Ditto.
788 (rs6000_emit_prologue_components): Ditto.
789 (rs6000_emit_epilogue_components): Ditto.
790 (rs6000_emit_prologue): Ditto.
791 (rs6000_emit_epilogue): Ditto.
792 (rs6000_elf_file_end): Ditto.
793 (rs6000_function_value): Ditto.
794 (rs6000_libcall_value): Ditto.
795 * config/rs6000/rs6000.h: Ditto.
796 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
797 (TARGET_MINMAX): ... this. New.
798 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
799 * config/rs6000/rs6000.md: Remove Xilinx FP support.
800 (*movsi_internal1_single): Delete.
801 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
802 mfpu=, mxilinx-fpu): Delete.
803 * config/rs6000/singlefp.h: Delete.
804 * config/rs6000/sysv4.h: Remove Xilinx FP support.
805 * config/rs6000/t-rs6000: Ditto.
806 * config/rs6000/t-xilinx: Delete.
807 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
808 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
809 (VStype_simple): Delete.
810 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
811 * config/rs6000/xfpu.h: Delete.
812 * config/rs6000/xfpu.md: Delete.
813 * config/rs6000/xilinx.h: Delete.
814 * config/rs6000/xilinx.opt: Delete.
815 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
816 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
818 2018-05-04 Tom de Vries <tom@codesourcery.com>
821 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
824 2018-05-04 Richard Biener <rguenther@suse.de>
827 * tree-complex.c (update_complex_assignment): We are always in SSA form.
828 (expand_complex_div_wide): Likewise.
829 (expand_complex_operations_1): Likewise.
830 (expand_complex_libcall): Preserve EH info of the original stmt.
831 (tree_lower_complex): Handle removed blocks.
832 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
833 on complex multiplication and division libcall builtins.
835 2018-05-04 Richard Biener <rguenther@suse.de>
838 * fold-const.c (negate_expr_p): Restrict negation of operand
839 zero of a division to when we know that can happen without
841 (fold_negate_expr_1): Likewise.
843 2018-05-04 Jakub Jelinek <jakub@redhat.com>
846 * real.h (real_nextafter): Declare.
847 * real.c (real_nextafter): New function.
848 * fold-const-call.c (fold_const_nextafter): New function.
849 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
851 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
852 even when arg1_mode is different from arg0_mode.
854 2018-05-03 Nathan Sidwell <nathan@acm.org>
856 * doc/extend.texi (Deprecated Features): Remove
858 (Backwards Compatibility): Likewise.
859 * doc/invoke.texi (C++ Language Options): Likewise.
860 (C++ Dialect Options): Likewise.
862 2018-05-03 Jakub Jelinek <jakub@redhat.com>
865 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
866 _mm512_mask_mullox_epi64): New intrinsics.
868 2018-05-03 Tom de Vries <tom@codesourcery.com>
871 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
872 dump files): Add offload-tree.
874 2018-05-03 Richard Biener <rguenther@suse.de>
876 PR tree-optimization/85615
877 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
878 to loops not nested in BBs loop father to avoid creating multi-entry
881 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
883 PR tree-optimization/70291
884 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
885 arguments. Change return type to tree. Emit libcall as a new
886 statement rather than replacing existing one when inplace_p is true.
887 (expand_complex_multiplication_components): New function.
888 (expand_complex_multiplication): Expand floating-point complex
889 multiplication using the above.
890 (expand_complex_division): Rename inner_type parameter to type.
891 Update expand_complex_libcall call-site.
892 (expand_complex_operations_1): Update expand_complex_multiplication
893 and expand_complex_division call-sites.
895 2018-05-02 Jakub Jelinek <jakub@redhat.com>
898 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
899 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
900 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
901 the highest significant bit of the shift count mask is clear. In
902 check whether and[sq]i3 is needed verify that all significant bits
903 of the shift count other than the highest are set.
905 2018-05-02 Tom de Vries <tom@codesourcery.com>
908 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
909 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
910 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
911 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
912 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
913 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
914 __builtin_goacc_parlevel_size.
916 2018-05-02 Richard Biener <rguenther@suse.de>
918 PR tree-optimization/85597
919 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
920 do not use split vect_get_vec_defs call but call vect_get_slp_defs
923 2018-05-02 Tom de Vries <tom@codesourcery.com>
926 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
927 dump files): Add ltrans-tree.
929 2018-05-02 Tom de Vries <tom@codesourcery.com>
932 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
933 dump files): Add wpa-ipa.
935 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
937 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
938 powerpc*-*-linux*paired* target.
939 * config/rs6000/750cl.h: Delete.
940 * config/rs6000/paired.h: Delete.
941 * config/rs6000/paired.md: Delete.
942 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
944 * config/rs6000/rs6000-builtin.def: Remove paired float support.
945 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
946 comment. Remove paired float support.
947 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
948 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
950 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
951 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
953 * config/rs6000/rs6000.c: Remove paired float support.
954 (paired_expand_vector_init, paired_expand_vector_move,
955 paired_emit_vector_compare, paired_emit_vector_cond_expr,
956 (paired_expand_lv_builtin, paired_expand_stv_builtin,
957 paired_expand_builtin, paired_expand_predicate_builtin,
958 paired_init_builtins): Delete.
959 * config/rs6000/rs6000.h: Remove paired float support.
960 * config/rs6000/rs6000.md: Remove paired float support.
961 (move_from_CR_ov_bit): Delete.
962 * config/rs6000/rs6000.opt (mpaired): Delete.
963 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
964 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
966 2018-05-02 Richard Biener <rguenther@suse.de>
969 * gimplify.c (gimplify_save_expr): When in SSA form allow
970 SAVE_EXPRs to compute to SSA vars.
972 2018-05-02 Jakub Jelinek <jakub@redhat.com>
975 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
976 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
977 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
978 clobber operands[2], instead use a new pseudo. Formatting fixes.
980 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
982 PR tree-optimization/85586
983 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
984 exit early for statements in the same group if the accesses are
987 2018-05-02 Tom de Vries <tom@codesourcery.com>
990 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
993 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
995 PR tree-optimization/85143
996 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
998 2018-05-01 Tom de Vries <tom@codesourcery.com>
1001 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1002 not found" error message.
1004 2018-05-01 Tom de Vries <tom@codesourcery.com>
1007 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1008 * vec.c (test_ordered_remove_if): New function.
1009 (vec_c_tests): Call test_ordered_remove_if.
1010 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1011 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1012 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1013 VEC_ORDERED_REMOVE_IF.
1015 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1017 PR tree-optimization/82665
1018 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1019 pointer subtraction where arguments come from a memchr call.
1021 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1023 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1024 --push-state --as-needed and --pop-state instead of --as-needed and
1025 --no-as-needed if ld supports it.
1026 * configure: Regenerated.
1029 * doc/install.texi2html: Replace _002d with - and _002a with * in
1030 generated html files using sed.
1032 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1035 * gcc-rich-location.c (blank_line_before_p): New function.
1036 (use_new_line): New function.
1037 (gcc_rich_location::add_fixit_insert_formatted): New function.
1038 * gcc-rich-location.h
1039 (gcc_rich_location::add_fixit_insert_formatted): New function.
1041 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1043 * selftest.c (assert_streq): Rename "expected" and "actual" to
1044 "val1" and "val2". Extend NULL-handling to cover both inputs
1045 symmetrically, while still requiring both to be non-NULL for a pass.
1046 * selftest.h (assert_streq): Rename "expected" and "actual" to
1048 (ASSERT_EQ): Likewise.
1049 (ASSERT_EQ_AT): Likewise.
1050 (ASSERT_KNOWN_EQ): Likewise.
1051 (ASSERT_KNOWN_EQ_AT): Likewise.
1052 (ASSERT_NE): Likewise.
1053 (ASSERT_MAYBE_NE): Likewise.
1054 (ASSERT_MAYBE_NE_AT): Likewise.
1055 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1056 the assertion to pass.
1057 (ASSERT_STREQ_AT): Likewise.
1059 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1061 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1062 interaction with -pie.
1064 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1066 * selftest.h: Fix alphabetization of per-source-file selftest
1069 2018-04-30 Jason Merrill <jason@redhat.com>
1071 PR c++/61982 - dead stores to destroyed objects.
1072 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1075 2018-04-30 Jason Merrill <jason@redhat.com>
1077 * tree.c (build_clobber): New.
1078 * tree.h: Declare it.
1079 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1081 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1083 * diagnostic-show-locus.c (layout::layout): Update for
1084 location_get_source_line returning a char_span.
1085 (struct char_span): Move to input.h.
1086 (struct correction): Update for fields in char_span becoming
1088 (struct source_line): Update for location_get_source_line
1089 returning a char_span.
1090 (layout::print_line): Likewise.
1091 * edit-context.c (edited_file::print_content): Likewise.
1092 (edited_file::print_diff_hunk): Likewise.
1093 (edited_file::print_run_of_changed_lines): Likewise.
1094 (edited_file::get_num_lines): Likewise.
1095 (edited_line::edited_line): Likewise.
1096 * final.c (asm_show_source): Likewise.
1097 * input.c (location_get_source_line): Convert return type
1098 from const char * to char_span, losing the final "line_len"
1100 (dump_location_info): Update for the above.
1101 (get_substring_ranges_for_loc): Likewise. Use a char_span
1102 when handling the literal within the line.
1103 (test_reading_source_line): Update for location_get_source_line
1104 returning a char_span.
1105 * input.h (class char_span): Move here from
1106 diagnostic-show-locus.c, converting from a struct to a class.
1107 Make data members private.
1108 (char_span::operator bool): New.
1109 (char_span::length): New.
1110 (char_span::get_buffer): New.
1111 (char_span::operator[]): New.
1112 (char_span::subspan): Make const.
1113 (char_span::xstrdup): New.
1114 (location_get_source_line): Convert return type from const char *
1115 to char_span, losing the final "line_size" param.
1117 2018-04-30 Jan Hubicka <jh@suse.cz>
1119 * lto-wrapper.c (ltrans_priorities): New static var.
1120 (cmp_priority): New.
1121 (run_gcc): Read priorities and if doing parallel build order
1122 the Makefile by them.
1124 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1126 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1128 2018-04-30 Richard Biener <rguenther@suse.de>
1130 * tree-cfg.c (verify_address): Remove base argument, add
1131 flag whether to check TREE_ADDRESSABLE and do that.
1132 (verify_expr): Remove.
1133 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1134 (verify_gimple_assign_single): Likewise.
1135 (verify_gimple_switch): Likewise.
1136 (verify_expr_location_1): Dereference tp once. Add (disabled)
1137 piece from verify_expr.
1138 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1140 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1142 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1144 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1146 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1147 (small_data_pattern): Likewise.
1148 (arc_rewrite_small_data): Likewise.
1149 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1150 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1151 (get_symbol_alignment): New function.
1152 (legitimate_small_data_address_p): Likewise.
1153 (legitimate_scaled_address): Update, call
1154 legitimate_small_data_address_p.
1155 (output_sdata): New static variable.
1156 (arc_print_operand): Update how we handle small data operands.
1157 (arc_print_operand_address): Likewise.
1158 (arc_legitimate_address_p): Update, use
1159 legitimate_small_data_address_p.
1160 (arc_rewrite_small_data_p): Remove.
1161 (arc_rewrite_small_data_1): Likewise.
1162 (arc_rewrite_small_data): Likewise.
1163 (small_data_pattern): Likewise.
1164 (compact_sda_memory_operand): Update to use
1165 legitimate_small_data_address_p and get_symbol_alignment.
1166 (prepare_move_operands): Don't rewite sdata pattern.
1167 (prepare_extend_operands): Remove.
1168 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1170 (zero_extendqisi2): Likewise.
1171 (zero_extendhisi2): Likewise.
1172 (extendqihi2): Likewise.
1173 (extendqisi2): Likewise.
1174 (extendhisi2): Likewise.
1178 * config/arc/constraints.md (Usd): Change it to memory constraint.
1180 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1182 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1183 as source of std instructions.
1184 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1185 allow 6-bit constants as source for store instructions.
1186 (movdi_insn): Update instruction pattern to allow 6-bit constants
1187 as source for store instructions.
1189 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1191 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1193 2018-04-30 Nathan Sidwell <nathan@acm.org>
1194 Sandra Loosemore <sandra@codesourcery.com>
1196 * dumpfile.c (dump_open): Allow '-' for stdout.
1197 * doc/invoke.texi (Developer Options): Document dump filename
1198 determination early. Document stdin/stdout selection.
1200 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1202 Microblaze Target: PIC data text relative
1204 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1205 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1207 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1208 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1209 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1210 New addressing mode for data-text relative position indepenedent code.
1211 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1212 'ADDRESS_SYMBOLIC_TXT_REL'.
1213 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1214 (microblaze_legitimate_pic_operand): Exclude function calls from
1215 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1216 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1218 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1219 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1220 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1221 for 'address + offset'.
1222 (microblaze_expand_prologue): Add new function prologue call for
1224 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1225 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1226 table in case of TARGET_PIC_DATA_TEXT_REL.
1227 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1228 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1229 Add new macros 'UNSPEC_TEXT',
1230 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1231 + exclude function calls from 'UNSPEC_PLT' in case of data text
1233 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1234 new target hook for generating address diff vector tables in case of
1236 * doc/tm.texi : Regenerate.
1237 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1238 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1239 of addr diff vector generation.
1240 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1241 target hook definition.
1242 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1243 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1244 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1245 option for data text relative.
1247 2018-04-30 Richard Biener <rguenther@suse.de>
1249 * tree-chrec.h (evolution_function_is_constant_p): Remove
1251 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1253 2018-04-30 Richard Biener <rguenther@suse.de>
1256 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1258 2018-04-30 Richard Biener <rguenther@suse.de>
1260 PR tree-optimization/28364
1261 PR tree-optimization/85275
1262 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1263 copying first exit test.
1265 2018-04-28 Mark Wielaard <mark@klomp.org>
1267 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1269 (dwarf_AT): Handle DW_AT_addr_base.
1270 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1272 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1275 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1276 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1277 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1278 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1280 2018-04-28 Richard Biener <rguenther@suse.de>
1282 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1283 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1284 to reflect use. Only add interesting stmts.
1286 2018-04-27 Martin Jambor <mjambor@suse.cz>
1289 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1290 the jump function allows for passing through aggregate values.
1292 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1294 * input.h (in_system_header_at): Convert from macro to inline
1296 (from_macro_expansion_at): Likewise.
1297 (from_macro_definition_at): Likewise.
1299 2018-04-27 Jeff Law <law@redhat.com>
1301 * config.gcc: Mark tile* targets as deprecated/obsolete.
1303 2018-04-27 Richard Biener <rguenther@suse.de>
1305 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1308 2018-04-27 Richard Biener <rguenther@suse.de>
1310 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1312 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1314 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1315 with Yd constraint. Set "preferred_for_speed" attribute from
1316 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1318 (*movdi_internal): Ditto.
1319 (movti_interunit splitters): Remove
1320 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1321 (movdi_interunit splitters): Ditto.
1322 * config/i386/constraints.md (Ye): Remove.
1323 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1325 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1328 * config/aarch64/constraints.md (Usg): Limit to 31.
1331 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1333 PR tree-optimization/85529
1334 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1335 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1336 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1337 zero extension or masking of the MSB bit.
1338 (optimize_range_tests): Add FIRST_BB argument, pass it through
1339 to optimize_range_tests_var_bound.
1340 (maybe_optimize_range_tests, reassociate_bb): Adjust
1341 optimize_range_tests callers.
1343 2018-04-26 Richard Biener <rguenther@suse.de>
1344 Jakub Jelinek <jakub@redhat.com>
1346 * cgraph.h (symbol_table): Just declare debug method here.
1347 * symtab.c (symbol_table::debug): Define.
1349 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1351 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1353 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1355 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1356 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1357 (*movdi_internal): Substitute Yi and Yj constraint with x
1358 and Ym and Yn constraint with y constraint. Update "isa"
1359 attribute and set "preferred_for_speed" attribute from
1360 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1361 (*movsi_internal): Ditto.
1362 (*movdf_internal): Ditto.
1363 (*movsf_internal): Ditto.
1364 (*zero_extendsidi2): Ditto.
1365 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1366 (sse2_loadld): Ditto.
1367 (*vec_extract<ssevecmodelower>_0): Ditto.
1368 (*vec_extractv4si_0_zext_sse4): Ditto.
1369 (vec_concatv2di): Ditto.
1370 (*vec_dup<mode>): Ditto.
1371 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1372 * config/i386/constraints.md (Yi): Remove.
1377 2018-04-26 Nathan Sidwell <nathan@acm.org>
1379 * dumpfile.c (dump_open): New.
1380 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1381 (dump_finish): Detect stdio/stderr by value not name.
1383 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1385 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1387 2018-04-26 Tom de Vries <tom@codesourcery.com>
1390 * config/nvptx/nvptx.c (verify_neutering_jumps)
1391 (verify_neutering_labels): New function
1392 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1394 2018-04-26 Tom de Vries <tom@codesourcery.com>
1397 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1398 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1401 2018-04-26 Richard Biener <rguenther@suse.de>
1402 Tom de Vries <tom@codesourcery.com>
1405 * lto-streamer-out.c (output_function): Fixup loops if required to match
1406 discovery done in the reader.
1408 2018-04-26 Richard Biener <rguenther@suse.de>
1410 PR tree-optimization/85116
1411 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1412 have a loop exit from the single latch predecessor. Remove
1413 case of header with just condition.
1414 (ch_base::copy_headers): Exclude infinite loops from any
1416 (pass_ch::execute): Record exits.
1418 2018-04-26 Richard Biener <rguenther@suse.de>
1420 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1421 prologue cost vector and pass it to vect_get_load_cost.
1422 (vect_get_peeling_costs_all_drs): Likewise.
1423 (vect_peeling_hash_get_lowest_cost): Likewise.
1424 (vect_enhance_data_refs_alignment): Likewise.
1426 2018-04-26 Richard Biener <rguenther@suse.de>
1429 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1430 checking of integer<->pointer conversions.
1431 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1432 sign-/zero-extending pointer types.
1433 (expand_omp_for_static_chunk): Likewise.
1435 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1436 Jean Lee <xiaoyur347@gmail.com>
1438 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1439 (TARGET_ASAN_SHADOW_OFFSET): Define.
1440 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1441 true for -fsanitize=address.
1443 2018-04-25 Mark Wielaard <mark@klomp.org>
1445 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1448 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1450 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1451 than "alu", remove explicit "memory" and "imm_disp" attributes.
1452 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1455 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1456 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1457 gen_lowpart_no_emit.
1459 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1462 * config/i386/i386.c (ix86_expand_builtin): Change memory
1463 operand to XI, extend p0 to Pmode.
1464 * config/i386/i386.md: Change unspec volatile and operand
1465 1 mode to XI, change operand 0 mode to P.
1467 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1469 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1470 GET_MODE_MASK before any checking.
1471 (nds32_can_use_bset_p): Likewise.
1472 (nds32_can_use_btgl_p): Likewise.
1474 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1476 * config/nds32/nds32-doubleword.md: New define_split pattern for
1477 illegal register number.
1479 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1481 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1483 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1485 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1487 2018-04-25 Richard Biener <rguenther@suse.de>
1489 * lto-streamer.h (LTO_major_version): Bump to 8.
1491 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1493 * BASE-VER: Set to 9.0.0.
1495 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1497 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1498 in __abskf2 and __powikf2.
1500 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1503 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1504 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1505 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1506 Use the above on operand 2. Reindent.
1507 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1509 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1512 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1514 * config/i386/i386.opt (mcet): Removed.
1515 * doc/install.texi: Remove -mcet documentation.
1516 * doc/invoke.texi: Likewise.
1518 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1521 * doc/install.texi: Remove -mcet from bootstrap-cet.
1523 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1526 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1527 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1531 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1532 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1533 containing a CONST_VECTOR.
1535 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1537 * doc/install.texi: Update newlib dependency for nvptx.
1539 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1542 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1543 instead of INTVAL when shifting x left.
1545 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1547 PR tree-optimization/85478
1548 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1549 vect_grouped_store_supported for single element vectors.
1551 2018-04-24 Richard Biener <rguenther@suse.de>
1554 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1555 load cost increase to the case of non-constant step.
1557 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1560 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1561 destination if any_malformed_asm.
1563 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1566 * expr.c (store_field): In the bitfield case, if the value comes from
1567 a function call and is returned in registers by means of a PARALLEL,
1568 do not change the mode of the temporary unless BLKmode and VOIDmode.
1570 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1572 PR rtl-optimization/85423
1573 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1574 dependencies to debug insns when the previous insn is non-debug.
1576 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1578 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1579 enums into a single definition.
1580 (fls): Fix predicates and printing.
1583 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1585 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1586 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1587 and short u6 immediate.
1588 (check_if_valid_sleep_operand): Remove.
1589 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1591 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1593 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1594 flag_always_save_lp condition.
1595 * config/nds32/nds32.opt (malways-save-lp): New option.
1597 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1599 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1600 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1601 * config/nds32/nds32.h
1602 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1603 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1605 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1607 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1608 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1610 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1611 Chung-Ju Wu <jasonwucj@gmail.com>
1613 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1615 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1616 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1618 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1620 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1622 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1624 * config/nds32/nds32-protos.h (nds32_data_alignment,
1625 nds32_local_alignment): Declare.
1626 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1627 nds32_local_alignment): New functions.
1628 (TARGET_CONSTANT_ALIGNMENT): Define.
1629 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1631 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1633 * config/nds32/nds32.c
1634 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1635 (TARGET_MODES_TIEABLE_P): Likewise.
1637 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1639 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1642 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1643 Chung-Ju Wu <jasonwucj@gmail.com>
1645 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1646 for unaligned access.
1647 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1649 * config/nds32/nds32-intrinsic.md: Likewise.
1650 * config/nds32/nds32_intrinsic.h: Likewise.
1651 * config/nds32/nds32.h (nds32_builtins): Likewise.
1652 * config/nds32/nds32.opt (munaligned-access): New option.
1653 * config/nds32/nds32.c (nds32_asm_file_start): Display
1654 flag_unaligned_access status.
1656 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1658 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1659 -mno-relax is present.
1660 * config/riscv/linux.h (LINK_SPEC): Ditto.
1662 2018-04-20 Martin Sebor <msebor@redhat.com>
1665 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1667 (gimple_fold_builtin_stxcpy_chk): Same.
1668 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1670 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1673 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1674 __powikf2 when long double is IEEE 128-bit.
1676 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1678 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1679 step to make sure stack always aligned.
1681 2018-04-20 Carl Love <cel@us.ibm.com>
1684 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1685 size check for arg0.
1687 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1688 Tom de Vries <tom@codesourcery.com>
1691 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1692 Emit insns for calls too.
1693 (nvptx_find_par): Always look for worker-level predecessor insn.
1694 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1696 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1697 (nvptx_process_pars): Propagate frames for calls.
1699 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1702 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1704 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1705 (ix86_handle_option): Don't handle OPT_mibt.
1706 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1708 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1710 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1711 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1712 (ix86_target_macros): Define __CET__ with flag_cf_protection
1713 for -fcf-protection.
1714 * config/i386/i386.c (isa2_opts): Remove -mibt.
1715 * config/i386/i386.h (TARGET_IBT): Removed.
1716 (TARGET_IBT_P): Likewise.
1717 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1718 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1719 * config/i386/i386.opt (mcet): Update help message.
1722 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1723 -mcet as an alias for -mshstk.
1725 2018-04-20 Richard Biener <rguenther@suse.de>
1728 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1729 complexity by forcing a single use of the multiply operand.
1731 2018-04-20 Martin Jambor <mjambor@suse.cz>
1734 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1735 recursion dependency to only apply to non-clones.
1737 2018-04-20 Martin Jambor <mjambor@suse.cz>
1740 * ipa-cp.c (create_specialized_node): Check that clones of
1741 self-recursive edges exist during IPA-CP.
1743 2018-04-19 Toon Moene <toon@moene.org>
1745 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1748 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1750 PR tree-optimization/85467
1751 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1752 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1753 VECTOR_CST element to type.
1755 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1758 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1759 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1760 (builtin_longjmp): Likewise.
1761 (save_stack_nonlocal): New pattern.
1762 (restore_stack_nonlocal): Likewise.
1764 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1767 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1768 Replace ASM_OUTPUT_LABEL with fprintf.
1770 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1773 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1774 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1775 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1776 define __IBT__ and __SHSTK__ for -fcf-protection.
1777 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1779 (ix86_trampoline_init): Likewise.
1780 (x86_output_mi_thunk): Likewise.
1781 (ix86_notrack_prefixed_insn_p): Likewise.
1782 (ix86_option_override_internal): Don't disallow -fcf-protection.
1783 * config/i386/i386.md (rdssp<mode>): Also enable for
1785 (incssp<mode>): Likewise.
1786 (nop_endbr): Likewise.
1787 * config/i386/i386.opt (mcet): Change help message to built-in
1791 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1792 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1793 enable CET built-in functions.
1795 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1797 * common/config/i386/i386-common.c
1798 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1799 OPTION_MASK_ISA_MOVDIRI_UNSET,
1800 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1801 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1802 * config.gcc (movdirintrin.h): New header.
1803 * config/i386/cpuid.h (bit_MOVDIRI,
1804 bit_MOVDIR64B): New bits.
1805 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1807 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1808 (VOID, PVOID, PCVOID)): New function types.
1809 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1810 __builtin_ia32_directstoreu_u64,
1811 __builtin_ia32_movdir64b): New builtins.
1812 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1813 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1815 (ix86_valid_target_attribute_inner_p): Ditto.
1816 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1817 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1818 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1819 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1820 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1821 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1822 (movdiri<mode>, movdir64b_<mode>): New.
1823 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1824 * config/i386/immintrin.h: Include movdirintrin.h.
1825 * config/i386/movdirintrin.h: New file.
1826 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1828 2018-04-19 Richard Biener <rguenther@suse.de>
1831 * cfg.c (clear_bb_flags): When loop state says we have
1832 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1834 2018-04-19 Richard Biener <rguenther@suse.de>
1836 PR tree-optimization/84737
1837 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1838 copying restrict info.
1839 (vect_setup_realignment): Use it.
1840 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1841 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1842 the first DR to all generated stores.
1843 (vectorizable_load): Likewise for loads.
1845 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1847 PR tree-optimization/85446
1848 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1849 the integral and pointer types to have the same precision.
1851 * doc/install.texi: Document --disable-cet being the default and
1854 2018-04-18 Martin Liska <mliska@suse.cz>
1856 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1859 2018-04-18 Martin Liska <mliska@suse.cz>
1862 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1865 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1866 arguments if they are comparable.
1868 2018-04-18 Martin Liska <mliska@suse.cz>
1871 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1874 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1877 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1880 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1881 ENDBR after calling __morestack.
1883 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1886 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1887 by using gcc_base_ver to generate a gcc_driver_version, and use
1888 it when generating GCC_DRIVER_NAME.
1889 * configure: Regenerate.
1891 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1894 * config.gcc: Obsolete powerpc*-*-*spe*.
1896 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1899 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1900 (stabstr_D): Change type of unum from unsigned int to
1901 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1904 2018-04-17 Jim Wilson <jimw@sifive.com>
1907 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1908 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1909 Set arg_pointer_offset after using pretend_args_size.
1911 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1913 PR rtl-optimization/85431
1914 * dse.c (record_store): Ignore zero width stores.
1917 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1918 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1919 __builtin_stack_restore rather than after it.
1920 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1921 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1922 argument instead of virtual_dynamic_stack_rtx.
1924 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1926 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1928 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1929 Add note to error message to explain internal mapping of overloaded
1930 built-in function name to non-overloaded built-in function name.
1931 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1934 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1937 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1938 where the inputs overlap with the output.
1940 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1942 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1943 (=v, v) alternative and explicit "memory" attribute.
1944 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1945 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1947 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1948 "sselog1" type instead of "sselog".
1949 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1950 "sselog". Remove explicit "memory" attribute.
1951 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1952 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1954 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1955 "sselog1" type instead of "sselog". Remove explicit "memory"
1957 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1958 use "sselog1" type instead of "sselog". Remove explicit "memory"
1960 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1961 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1963 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1964 "sselog1" type instead of "sselog". Remove explicit "memory"
1966 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1967 use "sselog1" type instead of "sselog". Remove explicit "memory"
1971 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1974 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1977 2018-04-17 Martin Jambor <mjambor@suse.cz>
1980 * ipa-cp.c (create_specialized_node): Call
1981 expand_all_artificial_thunks if necessary.
1983 2018-04-17 Martin Liska <mliska@suse.cz>
1986 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1987 in message, remote space in between '_G' and '('.
1989 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1992 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1993 avx512f_vmcmp<mode>3<round_saeonly_name>,
1994 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1995 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1996 avx512f_rndscale<mode><round_saeonly_name>,
1997 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1998 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1999 Use %<iptr>2 instead of %2 for -masm=intel.
2000 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2001 avx512f_vcvttss2usi<round_saeonly_name>,
2002 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2004 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2005 avx512f_vcvttsd2usi<round_saeonly_name>,
2006 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2007 Use %q1 instead of %1 for -masm=intel.
2008 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2009 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2010 of %3 for -masm=intel.
2011 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2013 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2015 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2017 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2018 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2020 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2022 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2023 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2024 %g1 and one with %0 and %1.
2025 (avx512er_vmrcp28<mode><round_saeonly_name>,
2026 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2028 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2029 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2030 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2031 of %0 and %{%4%} for -masm=intel.
2032 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2033 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2034 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2035 order of %0 and %{%5%}%{z%} for -masm=intel.
2037 2018-04-17 Jan Hubicka <jh@suse.cz>
2040 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2042 2018-04-17 Martin Liska <mliska@suse.cz>
2045 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2046 for target_clone error message. Make default implementation
2047 clone to be a local declaration.
2048 (separate_attrs): Add new argument and check for an empty
2050 (expand_target_clones): Handle it.
2051 (ipa_target_clone): Make redirection just for target_clones
2054 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2055 Tom de Vries <tom@codesourcery.com>
2058 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2059 tiled basic blocks without omp continue statements.
2061 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2064 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2065 vec_extract expression as having side effects to make sure it gets
2068 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2071 * config/i386/i386.c (get_builtin_code_for_version): Check
2074 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2077 * gcc/config.gcc: Support "skylake".
2078 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2080 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2081 (processor_target_table): Add "skylake".
2082 (ix86_option_override_internal): Add "skylake".
2083 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2084 PROCESSOR_CANNONLAKE.
2085 (get_builtin_code_for_version): Fix priority for
2086 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2087 PROCESSOR_SKYLAKE-AVX512.
2088 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2089 (processor_type): Add PROCESSOR_SKYLAKE.
2091 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2092 Jason Merrill <jason@redhat.com>
2095 * convert.c (convert_to_integer_1): Use direct recursion for
2096 enumeral types and types with a precision less than the number
2097 of bits in their mode.
2099 2018-04-16 Julia Koval <julia.koval@intel.com>
2102 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2103 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2105 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2108 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2109 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2110 and -mno-direct-move.
2112 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2115 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2116 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2117 Ensure negative shifts result in {0}.
2119 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2121 PR rtl-optimization/79916
2122 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2123 regs (if any) to define how to gnerate SD moves when LRA is in
2126 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2128 PR rtl-optimization/85393
2129 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2130 * except.c (expand_dw2_landing_pad_for_region): Make static.
2131 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2132 a label and unconditional jump to old_bb, rather than
2133 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2136 PR rtl-optimization/85376
2137 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2138 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2139 instead of a specific value.
2141 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2142 Bin Cheng <bin.cheng@arm.com>
2144 PR tree-optimization/82965
2145 PR tree-optimization/83991
2146 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2147 by_profile_only parameter.
2148 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2149 information if the loop was predicted to iterate too many times.
2150 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2152 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2155 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2158 2018-04-13 Martin Liska <mliska@suse.cz>
2159 Jakub Jelinek <jakub@redhat.com>
2162 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2163 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2164 * builtins.c (expand_builtin_memory_copy_args): Use
2165 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2166 handle dest_addr == pc_rtx.
2168 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2171 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2172 asked to not generate direct moves.
2173 (fix_trunc<mode>si2_stfiwx): Similar.
2174 (fix_trunc<mode>si2_internal): Similar.
2176 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2179 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2180 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2181 lookup if dest in some wider mode is known to be const0_rtx and
2182 if so, record permanent equivalence for it to be ZERO_EXTEND of
2183 the narrower mode destination.
2185 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2187 * lto-streamer-out.c (output_function): Revert 259346.
2188 * omp-expand.c (expand_oacc_for): Likewise.
2190 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2192 PR rtl-optimization/85354
2193 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2194 * sel-sched.c (sel_global_init): ... here.
2196 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2199 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2200 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2201 mode for PE-COFF targets.
2202 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2203 (i386_pe_asm_lto_end): Likewise.
2204 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2205 (TARGET_ASM_LTO_END): Likewise.
2206 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2207 (i386_pe_asm_lto_start): New function.
2208 (i386_pe_asm_lto_end): Likewise.
2210 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2211 Richard Biener <rguenther@suse.de>
2214 * lto-streamer-out.c (output_function): Fix CFG loop state before
2216 * omp-expand.c (expand_oacc_for): Handle calls to internal
2217 functions like regular functions.
2219 2018-04-12 Richard Biener <rguenther@suse.de>
2222 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2223 for the early LTO debug to properly generate references to it
2224 during DIE emission. Do not re-use that for the skeleton for
2226 (dwarf2out_early_finish): Likewise.
2228 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2231 * config/i386/sse.md
2232 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2233 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2234 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2235 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2236 and output is a reg, avoid creating invalid lowpart subreg, but
2237 instead split into a 512-bit move. Don't split if not AVX512VL,
2238 input is xmm16+ reg and output is a mem.
2239 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2240 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2241 xmm16+ reg and output is a mem.
2243 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2245 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2246 also for flag_dwarf2_cfi_asm.
2248 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2250 PR rtl-optimization/85342
2251 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2252 a bool scalar var inside of the loop instead. Don't try to update
2253 recog_data.operand after failed apply_change_group.
2255 2018-04-12 Tom de Vries <tom@codesourcery.com>
2258 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2259 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2260 array with flexible array member as array without given dimension.
2261 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2262 argument for undefined param to true.
2264 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2267 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2268 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2269 from PowerPC section.
2270 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2271 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2272 help text that is too long.
2273 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2274 help text that is too long.
2275 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2276 help text that is too long.
2278 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2280 * config/alpha/alpha.md (stack_probe_internal): Rename
2281 from "probe_stack". Update all callers.
2283 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2285 PR rtl-optimization/84566
2286 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2287 sched_macro_fuse_insns.
2289 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2292 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2293 (compute_block_dependences): ... from here.
2295 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2297 PR tree-optimization/85331
2298 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2299 from int to HOST_WIDE_INT.
2301 2018-04-11 Martin Jambor <mjambor@suse.cz>
2304 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2305 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2306 not the same as the source val.
2307 (cgraph_edge_brings_value_p): New parameter.
2308 (gather_edges_for_value): Pass destination value to
2309 cgraph_edge_brings_value_p.
2310 (perhaps_add_new_callers): Likewise.
2311 (get_info_about_necessary_edges): Likewise and exclude values brought
2312 only by self-recursive edges.
2313 (create_specialized_node): Redirect only clones of self-calling edges.
2314 (+self_recursive_pass_through_p): New function.
2315 (find_more_scalar_values_for_callers_subset): Use it.
2316 (find_aggregate_values_for_callers_subset): Likewise.
2317 (known_aggs_to_agg_replacement_list): Removed.
2318 (decide_whether_version_node): Re-calculate known constants for all
2319 remaining context clones.
2321 2018-04-11 Richard Biener <rguenther@suse.de>
2324 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2325 from early DWARF output.
2326 (dwarf2out_early_finish): Output line info unconditionally into
2327 early DWARF and add reference to it.
2329 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2332 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2333 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2334 other than V2DFmode using iptr mode attribute.
2335 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2337 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2339 PR rtl-optimization/84659
2340 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2342 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2345 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2347 (output_loc_list): Pass address of a dummy size variable even in the
2348 locview handling loop.
2349 (index_location_lists): Add comment on why skip_loc_list_entry can't
2352 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2355 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2358 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2361 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2362 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2363 and -mstring-compare-inline-limit.
2365 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2368 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2369 for stack clash protection in a register whenever we need it to be in
2372 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2374 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2375 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2377 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2380 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2382 (mlong-double-): Ditto.
2383 * config/rs6000/sysv4.opt (msdata=): Ditto.
2384 (mtls-size=): Ditto.
2386 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2388 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2389 erroneous entries for
2390 "vector int vec_ldl (int, long int *)", and
2391 "vector unsigned int vec_ldl (int, unsigned long int *)".
2392 Add comments and entries for
2393 "vector bool char vec_ldl (int, bool char *)",
2394 "vector bool short vec_ldl (int, bool short *)",
2395 "vector bool int vec_ldl (int, bool int *)",
2396 "vector bool long long vec_ldl (int, bool long long *)",
2397 "vector pixel vec_ldl (int, pixel *)",
2398 "vector long long vec_ldl (int, long long *)",
2399 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2400 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2401 type tree bool_long_long_type_node and correct definition of
2402 bool_V2DI_type_node to make reference to this new type tree.
2403 (rs6000_mangle_type): Replace erroneous reference to
2404 bool_long_type_node with bool_long_long_type_node.
2405 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2406 comments to emphasize sign distinctions for char and int types and
2407 replace RS6000_BTI_bool_long constant with
2408 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2409 use of RS6000_BTI_pixel.
2410 (bool_long_type_node): Remove this macro definition.
2411 (bool_long_long_type_node): New macro definition
2413 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2415 PR rtl-optimization/85300
2416 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2417 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2418 simplify_unary_operation fails.
2420 2018-04-10 Martin Liska <mliska@suse.cz>
2422 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2423 cgraph_edge and ipa_ref.
2425 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2429 * config/i386/sse.md
2430 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2431 computation of the VEC_MERGE selector from mask.
2432 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2433 Fix decoding of the VEC_MERGE selector into mask.
2435 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2437 PR tree-optimization/85286
2438 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2440 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2442 * final.c (final_1): Set insn_last_address as well as
2443 insn_current_address.
2445 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2448 * explow.c (emit_stack_probe): Call validize_mem on memory location
2449 before passing it to gen_probe_stack. Create address operand and
2450 legitimize it for the probe_stack_address case.
2452 2018-04-09 Jan Hubicka <jh@suse.cz>
2455 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2456 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2457 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2459 (free_lang_data): Rebuild type inheritance graph.
2461 2018-04-09 Martin Sebor <msebor@redhat.com>
2463 * invoke.texi (-finline-small-functions): Mention other optimization
2465 (-findirect-inlining, -fpartial-inlining): Same.
2466 (-finline-functions-called-once): Same.
2467 (-freorder-blocks-and-partition): Same.
2469 2018-04-09 Jan Hubicka <jh@suse.cz>
2472 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2473 jumps; choose last target that matches the criteria (i.e.
2474 no partition changes for non-crossing jumps).
2475 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2476 support for redirecting crossing jumps to non-crossing.
2478 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2480 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2481 also for naked functions.
2483 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2485 * config/arc/arc.md (add_shift): New pattern.
2486 (add_shift2): Likewise.
2487 (sub_shift): Likewise.
2488 (sub_shift_cmp0_noout): Likewise.
2489 (compare_si_ashiftsi): Likewise.
2490 (xbfu_cmp0_noout): New combine pattern.
2491 (xbfu_cmp0"): Likewise.
2492 (movsi_set_cc_insn): Place the predicable variant first.
2493 (commutative_binary_cmp0_noout): Remove clobber.
2494 (commutative_binary_cmp0): New pattern.
2495 (noncommutative_binary_cmp0): Likewise.
2496 (noncommutative_binary_cmp0_noout): Likewise.
2497 (noncommutative_binary_comparison_result_used): Removed.
2498 (rsub_cmp0): New pattern.
2499 (rsub_cmp0_noout): Likewise.
2500 (extzvsi): Changed, keep only meaningful variants.
2501 (SQH, SEZ): New iterators.
2502 (SQH_postfix): New mode attribute.
2503 (SEZ_prefix): New code attribute.
2504 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2505 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2506 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2508 (noncommutative_operator): Check the availability of barrel
2511 2018-04-09 Richard Biener <rguenther@suse.de>
2513 PR tree-optimization/85284
2514 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2515 Only use the niter constraining form of simple_iv when the exit
2518 2018-04-09 Tom de Vries <tom@codesourcery.com>
2521 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2522 (define_expand "*memory_barrier"): New define_expand.
2523 (define_insn "memory_barrier"): New insn.
2525 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2527 PR rtl-optimization/80463
2528 PR rtl-optimization/83972
2529 PR rtl-optimization/83480
2531 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2532 correct producer for the insn.
2533 (tidy_control_flow): Fixup seqnos in case of debug insns.
2535 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2537 PR rtl-optimization/83913
2539 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2540 different sched-times when merging exprs.
2542 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2544 PR rtl-optimization/83962
2546 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2547 tidy_fallthru_edge and tidy_control_flow.
2549 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2551 PR rtl-optimization/83530
2553 * sel-sched.c (force_next_insn): New global variable.
2554 (remove_insn_for_debug): When force_next_insn is true, also leave only
2555 next insn in the ready list.
2556 (sel_sched_region): When the region wasn't scheduled, make another pass
2557 over it with force_next_insn set to 1.
2559 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2561 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2563 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2564 for interrupt control.
2565 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2566 functions for interrupt control.
2567 * config/nds32/nds32-intrinsic.md: Likewise.
2568 * config/nds32/nds32_intrinsic.h: Likewise.
2569 * config/nds32/nds32.h (nds32_builtins): Likewise.
2571 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2573 * config/nds32/nds32.c (nds32_init_machine_status,
2574 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2575 strict_aligned_p field.
2576 (nds32_expand_to_rtl_hook): New function.
2577 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2578 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2580 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2581 Chung-Ju Wu <jasonwucj@gmail.com>
2583 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2584 * config/nds32/nds32-n7.md: New file.
2585 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2586 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2588 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2589 * config/nds32/nds32.md (pipeline_model): Add n7.
2590 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2591 * config/nds32/pipelines.md: Include n7 settings.
2593 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2594 Chung-Ju Wu <jasonwucj@gmail.com>
2596 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2597 * config/nds32/nds32-e8.md: New file.
2598 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2599 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2601 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2602 * config/nds32/nds32.md (pipeline_model): Add e8.
2603 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2604 * config/nds32/pipelines.md: Include e8 settings.
2606 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2607 Chung-Ju Wu <jasonwucj@gmail.com>
2609 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2610 * config/nds32/nds32-n8.md: New file.
2611 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2612 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2614 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2615 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2616 * config/nds32/nds32.md (pipeline_model): Add n8.
2617 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2618 * config/nds32/pipelines.md: Include n8 settings.
2620 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2621 Chung-Ju Wu <jasonwucj@gmail.com>
2623 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2624 * config/nds32/nds32-n9-2r1w.md: New file.
2625 * config/nds32/nds32-n9-3r2w.md: New file.
2626 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2627 nds32_register_ports): New or modify for cpu n9.
2628 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2630 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2631 * config/nds32/nds32-utils.c: New file.
2632 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2633 TARGET_MUL_SLOW): Define.
2634 * config/nds32/nds32.md (pipeline_model): New attribute.
2635 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2636 New options that support cpu n9.
2637 * config/nds32/pipelines.md: Include n9 settings.
2638 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2640 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2642 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2643 information if necessary.
2644 (output_cond_branch_compare_zero): Likewise.
2645 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2646 (nds32_target_alignment): Refine for alignment.
2647 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2648 (FUNCTION_BOUNDARY): Modify.
2649 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2651 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2653 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2655 * config/nds32/constants.md (unspec_volatile_element): Add values for
2656 TLB operation and data prefetch.
2657 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2658 functions for TLB operation and data prefetch.
2659 * config/nds32/nds32-intrinsic.md: Likewise.
2660 * config/nds32/nds32_intrinsic.h: Likewise.
2661 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2662 (nds32_print_operand): Likewise.
2663 * config/nds32/nds32.h (nds32_builtins): Likewise.
2665 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2666 Andrew Pinski <pinsika@gcc.gnu.org>
2669 * match.pd: Use constant_boolean_node of correct type instead of
2670 boolean_true_node or boolean_false_node for simplifying
2671 pointer comparisons to zero.
2673 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2675 PR tree-optimization/80021
2676 * tree.c (verify_type_variant): Make error call in verify_variant_match
2677 translatable and remove final full stop.
2679 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2681 * config/nds32/constants.md (unspec_volatile_element): Add
2682 UNSPEC_VOLATILE_EH_RETURN.
2683 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2684 nds32_output_stack_pop): Support dwarf exception handling process.
2685 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2686 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2687 exception handling process.
2688 (nds32_compute_stack_frame): Likewise.
2689 (nds32_return_addr_rtx): Likewise.
2690 (nds32_initial_elimination_offset): Likewise.
2691 (nds32_expand_prologue): Likewise.
2692 (nds32_expand_epilogue): Likewise.
2693 (nds32_dynamic_chain_address): New function.
2694 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2696 (DYNAMIC_CHAIN_ADDRESS): Define.
2697 (EH_RETURN_DATA_REGNO): Define.
2698 (EH_RETURN_STACKADJ_RTX): Define.
2699 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2700 patterns for dwarf exception handling.
2702 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2704 * config/nds32/nds32.h: Clean up obsolete macros.
2706 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2708 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2709 Add enum values for particular instructions.
2710 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2711 particular intrinsic functions.
2712 * config/nds32/nds32-intrinsic.md: Likewise.
2713 * config/nds32/nds32_intrinsic.h: Likewise.
2714 * config/nds32/nds32.h (nds32_builtins): Likewise.
2715 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2716 (btst, ave): New patterns for particular instructions.
2718 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2720 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2721 Add enum values for atomic load/store and memory sync.
2722 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2724 * config/nds32/nds32-intrinsic.md: Likewise.
2725 * config/nds32/nds32_intrinsic.h: Likewise.
2726 * config/nds32/nds32.h (nds32_builtins): Likewise.
2728 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2730 PR tree-optimization/85257
2731 * fold-const.c (native_encode_vector): If not all elts could fit
2732 and off is -1, return 0 rather than offset.
2733 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2734 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2735 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2736 adjust buffer in native_interpret_expr call.
2738 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2740 * config/nds32/constants.md (unspec_volatile_element): Add cache
2741 control enum values.
2742 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2743 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2744 * config/nds32/nds32.c (nds32_cctl_names): New.
2745 (nds32_print_operand): Handle cache control register names.
2746 * config/nds32/nds32.h (nds32_builtins): New enum values.
2747 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2749 * config/nds32/nds32.md (type): Add mmu.
2750 * config/nds32/pipelines.md (simple_insn): Add mmu.
2752 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2754 * config/nds32/nds32.md (type): Remove call.
2755 * config/nds32/pipelines.md (simple_insn): Likewise.
2757 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2759 * config/nds32/constants.md (unspec_volatile_element): Add
2760 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2761 UNSPEC_VOLATILE_FMFCFG.
2762 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2763 description for fmfcfg and fmfcsr.
2764 (bdesc_1arg): Add fmtcsr.
2765 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2766 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2767 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2768 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2769 unspec_fmfcfg): New patterns.
2770 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2771 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2772 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2773 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2774 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2775 __nds32__fmfcfg): Define.
2777 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2779 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2780 intrinsic register names.
2781 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2782 intrinsic register enum values and macros.
2784 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2786 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2787 for load/store addressing form.
2788 (nds32_print_operand_address): Likewise.
2790 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2793 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2794 based on LABEL_REF. Remove useless assertion.
2795 (pic_address_needs_scratch): Fix formatting.
2796 (sparc_legitimize_pic_address): Minor tweaks.
2797 (sparc_delegitimize_address): Adjust assertion accordingly.
2798 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2799 into symbolic_operand.
2800 (movsi_high_pic_label_ref): Likewise.
2801 (movsi_lo_sum_pic_label_ref): Likewise.
2802 (movdi_pic_label_ref): Likewise.
2803 (movdi_high_pic_label_ref): Likewise.
2804 (movdi_lo_sum_pic_label_ref): Likewise.
2806 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2808 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2809 custom LIB_SPEC setup.
2811 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2812 Kito Cheng <kito.cheng@gmail.com>
2814 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2815 * config/riscv/freebsd.h: New.
2817 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2819 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2820 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2823 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2824 Kito Cheng <kito.cheng@gmail.com>
2826 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2827 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2828 * config/nds32/nds32-protos.h (nds32_output_call,
2829 nds32_output_return): Declare.
2830 * config/nds32/nds32.md: Refine all the call and return patterns.
2832 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2835 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2836 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2838 PR rtl-optimization/84872
2839 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2840 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2843 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2845 * expr.c (copy_blkmode_to_reg): Revert 254862.
2846 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2848 2018-04-06 Richard Biener <rguenther@suse.de>
2851 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2852 after seeing a component reference with an adjacent field. Treat
2853 refs to arrays at struct end of external decls similar to
2854 refs to unconstrained commons.
2856 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2859 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2860 look through SAVE_EXPRs with non-side-effects argument. Adjust
2862 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2865 2018-04-06 Richard Biener <rguenther@suse.de>
2868 * alias.c (find_base_term): New wrapper around find_base_term
2869 unwinding CSELIB_VAL_PTR changes.
2870 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2873 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2875 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2877 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2878 constant definitions.
2879 ("nop"): lr 0,0 -> nopr r0
2880 ("nop_lr0", "nop_lr1"): New insn definitions.
2882 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2884 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2885 NDS32_V3PUSH_AVAILABLE_P macro.
2887 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2888 Chung-Ju Wu <jasonwucj@gmail.com>
2890 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2891 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2892 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2893 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2894 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2895 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2896 * config/nds32/constraints.md: New constraints and checking for hard
2897 float configuration.
2898 * config/nds32/iterators.md: New mode iterator and attribute for hard
2899 float configuration.
2900 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2902 * config/nds32/nds32-fpu.md: New file.
2903 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2904 deal with hard float code generation.
2905 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2907 (abi_type, float_reg_number): New enum type.
2908 * config/nds32/nds32-predicates.c: New predicates for hard float.
2909 * config/nds32/nds32-protos.h: Declare functions for hard float.
2910 * config/nds32/nds32.c: Implementation for hard float configuration.
2911 * config/nds32/nds32.h: Definitions for hard float configuration.
2912 * config/nds32/nds32.md: Include hard float machine description and
2913 modify patterns for hard float configuration.
2914 * config/nds32/nds32.opt: New options for hard float configuration.
2915 * config/nds32/predicates.md: New predicates for hard float
2918 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2920 * common/config/nds32/nds32-common.c
2921 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2923 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2926 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2927 CONSTRUCTOR_ELT (ctor, ...)->value.
2929 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2932 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2934 2018-04-05 Tom de Vries <tom@codesourcery.com>
2937 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2940 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2941 Kito Cheng <kito.cheng@gmail.com>
2943 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2944 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2945 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2946 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2948 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2949 Kito Cheng <kito.cheng@gmail.com>
2951 * config/nds32/constraints.md (Ufe): New memory constraint.
2952 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2953 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2954 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2956 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2957 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2959 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2961 * config/nds32/nds32.md: Use optimize_size in the condition for
2962 alu-shift instructions.
2964 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2966 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2968 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2970 * config/nds32/nds32.md (negsi2): Refine pattern.
2972 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2973 Chung-Ju Wu <jasonwucj@gmail.com>
2975 * config/nds32/iterators.md (shift_rotate): New code iterator.
2976 (shift): New code attribute.
2977 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2978 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2979 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2980 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2981 bit-wise operations.
2982 (andsi3, *andsi3): Ditto.
2983 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2984 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2985 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2986 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2987 nds32_ior_operand, nds32_xor_operand): New predicates.
2989 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2991 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2992 (addsi3, subsi3): ... this.
2994 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2996 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2998 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3000 * config/nds32/nds32.md: Adjust indention.
3002 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3004 * config/nds32/nds32.md (feature): New attribute.
3006 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3008 * config/nds32/nds32.md (subtype): New attribute.
3010 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3013 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3014 expansion to perform a bitwise AND of the argument followed by a
3015 boolean negation of the result.
3017 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3019 PR rtl-optimization/84878
3020 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3021 the basic block. Assert the use reference is not artificial and that
3022 it has an associated insn.
3024 2018-04-04 Michael Matz <matz@suse.de>
3026 * builtins.c (compute_objsize): Pass correct operand
3027 to array_at_struct_end_p.
3029 2018-04-04 Richard Biener <rguenther@suse.de>
3032 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3033 from contexts for DINFO_LEVEL_TERSE and below.
3035 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3037 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3038 resiter_operand condition.
3039 * config/nds32/nds32.md (*move<mode>): Ditto.
3041 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3042 Monk Chiang <sh.chiang04@gmail.com>
3044 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3046 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3048 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3050 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3051 Kito Cheng <kito.cheng@gmail.com>
3053 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3054 nds32_cond_code_str, output_cond_branch,
3055 output_cond_branch_compare_zero, nds32_expand_cbranch,
3056 nds32_expand_cstore, nds32_expand_movcc,
3057 nds32_output_cbranchsi4_equality_zero,
3058 nds32_output_cbranchsi4_equality_reg,
3059 nds32_output_cbranchsi4_equality_reg_or_const_int,
3060 nds32_output_cbranchsi4_greater_less_zero: New functions.
3061 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3062 nds32_expand_cstore, nds32_expand_movcc,
3063 nds32_output_cbranchsi4_equality_zero,
3064 nds32_output_cbranchsi4_equality_reg,
3065 nds32_output_cbranchsi4_equality_reg_or_const_int,
3066 nds32_output_cbranchsi4_greater_less_zero): Declare.
3067 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3068 nds32_rimm11s_operand): New predicates.
3069 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3070 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3073 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3075 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3076 * config/nds32/nds32.md: Ditto.
3077 * config/nds32/pipelines.md: Ditto.
3079 2018-04-04 Richard Biener <rguenther@suse.de>
3081 PR tree-optimization/85168
3082 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3083 propagating abnormals.
3085 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3087 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3089 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3090 Kito Cheng <kito.cheng@gmail.com>
3092 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3093 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3094 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3095 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3096 * config/nds32/nds32.md (sibcall_internal): New.
3097 (sibcall_register): Remove.
3098 (sibcall_immediate): Remove.
3099 (sibcall_value_internal): New.
3100 (sibcall_value_register): Remove.
3101 (sibcall_value_immediate): Remove.
3102 * config/nds32/predicates.md (nds32_general_register_operand): New.
3103 (nds32_call_address_operand): New.
3105 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3107 PR rtl-optimization/85167
3108 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3109 bb_defs if *split_p, instead preinitialize it to NULL.
3111 PR tree-optimization/85156
3112 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3113 evaluating the argument multiple times.
3115 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3117 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3119 (_mm_cvtpd_ps): Likewise.
3120 (_mm_cvttpd_epi32): Likewise.
3121 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3122 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3123 vector, pixel, and bool following altivec.h include.
3125 2018-04-03 Martin Sebor <msebor@redhat.com>
3127 * doc/extend.texi (Common Function Attributes): Clarify.
3128 (const attribute): Likewise.
3129 (pure attribute): Likewise.
3131 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3134 * config/i386/i386.c (ix86_expand_vector_set): Use
3135 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3137 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3139 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3140 instructions when changing rounding bits to preserve precision bits
3141 in the x87 control word.
3143 2018-04-03 Martin Liska <mliska@suse.cz>
3145 PR tree-optimization/82491
3146 * rtl.h (strip_offset_and_add): Replace += suboffset with
3147 poly_uint64 () + suboffset.
3149 2018-03-29 Martin Liska <mliska@suse.cz>
3150 Martin Jambor <mjambor@suse.cz>
3153 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3154 param_type is not an integral or pointer type.
3156 2018-04-03 Richard Biener <rguenther@suse.de>
3158 * sese.h (recompute_all_dominators): Remove.
3160 2018-04-02 Martin Sebor <msebor@redhat.com>
3162 * doc/invoke.texi (-Wrestrict): Fix typos.
3164 2018-04-02 Jim Wilson <jimw@sifive.com>
3166 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3167 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3168 (<optab>di3, <optab>si3_extend): Likewise.
3169 (<optab>si3_mask, <optab>si3_mask_1): New.
3170 (<optab>di3_mask, <optab>di3_mask_1): New.
3171 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3172 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3173 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3175 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3177 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3180 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3182 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3183 (nds32_canonicalize_comparison): New function.
3185 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3186 Kito Cheng <kito.cheng@gmail.com>
3187 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3189 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3190 * config/nds32/constants.md (unspec_volatile_element): Add
3191 UNSPEC_VOLATILE_RELAX_GROUP.
3192 * config/nds32/nds32-relax-opt.c: New file.
3193 * config/nds32/nds32-predicates.c
3194 (nds32_symbol_load_store_p): New function.
3195 * config/nds32/nds32-protos.h
3196 (nds32_symbol_load_store_p): Declare function.
3197 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3198 * config/nds32/nds32.c
3199 (nds32_register_pass): New function to register pass.
3200 (nds32_register_passes): New function to register passes.
3201 * config/nds32/nds32.md (relax_group): New pattern.
3202 * config/nds32/nds32.opt (mrelax-hint): New option.
3203 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3205 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3207 * config/nds32/t-nds32: Modify files dependency.
3209 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3211 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3212 (PROFILE_HOOK): Define its implementation.
3214 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3216 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3217 type and 32-bit size.
3219 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3222 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3223 (V_128_256): New mode iterator.
3224 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3225 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3226 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3228 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3229 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3231 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3234 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3235 NaN inputs correctly.
3237 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3240 * config/rs6000/vsx.md (??r): New mode attribute.
3241 (*vsx_mov<mode>_64bit): Use it.
3242 (*vsx_mov<mode>_32bit): Likewise.
3244 2018-03-30 Martin Sebor <msebor@redhat.com>
3246 PR tree-optimization/84818
3247 * builtins.c (check_access): Use warning_n.
3249 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3252 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3254 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3257 2018-03-30 Julia Koval <julia.koval@intel.com>
3260 * x86-tune.def (movx, partial_reg_dependency): Enable for
3263 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3266 * lra-constraints.c (process_alt_operands): Move setting
3267 this_alternative_matches below.
3269 2018-03-29 Martin Liska <mliska@suse.cz>
3272 * doc/invoke.texi: Document how LTO works with debug info.
3273 Describe auto-load support of binutils. Mention 'x86-64'
3274 as valid option value of -march option.
3276 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3278 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3281 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3282 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3283 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3286 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3289 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3290 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3291 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3292 for RS6000_BTM_POWERPC64.
3293 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3294 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3295 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3300 2018-03-28 Carl Love <cel@us.ibm.com>
3303 2017-09-27 Carl Love <cel@us.ibm.com>
3305 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3306 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3307 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3310 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3312 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3313 instead of __vector bool.
3314 (_mm_max_pu8): Likewise.
3315 (_mm_min_pi16): Likewise.
3317 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3320 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3321 (DIVWEUO): Likewise.
3323 (DIVDEUO): Likewise.
3324 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3325 DIVWEUO and DIVDEUO.
3326 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3327 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3328 (div_extend): Likewise.
3329 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3331 (__builtin_divweuo): Likewise.
3332 (__builtin_divdeo): Likewise.
3333 (__builtin_divdeuo): Likewise.
3335 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3338 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3339 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3341 PR tree-optimization/82004
3342 * gimple-match-head.c (optimize_pow_to_exp): New function.
3343 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3344 Don't fold to exp if optimize_pow_to_exp is false.
3346 2018-03-28 Martin Liska <mliska@suse.cz>
3349 * calls.c (initialize_argument_information): Fix trailing space.
3350 * common.opt: Fix typo and provide better explanation for
3351 -fsanitize-coverage option.
3352 * config/i386/i386.opt: Fix typo.
3354 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3355 Martin Liska <mliska@suse.cz>
3358 * gimplify.c (asan_poison_variable): Don't do the check for
3359 gimplify_omp_ctxp here.
3360 (gimplify_decl_expr): Do it here.
3361 (gimplify_target_expr): Likewise.
3363 2018-03-28 Martin Liska <mliska@suse.cz>
3366 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3367 chkp_type_bounds_count if MPX is not enabled.
3369 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3371 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3373 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3376 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3377 function to create the function decl for complex long double
3378 multiply and divide for -mabi=ieeelongdouble.
3379 (init_float128_ieee): Call it.
3381 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3384 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3385 -fcf-protection=branch -mibt.
3386 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3388 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3391 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3393 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3396 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3397 extern array declarations.
3399 2018-03-27 Richard Biener <rguenther@suse.de>
3402 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3403 explicit single_use checks.
3405 2018-03-27 Richard Biener <rguenther@suse.de>
3407 PR tree-optimization/85082
3408 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3411 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3413 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3414 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3415 Turn on fasynchronous-unwind-tables and funwind-tables.
3417 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3420 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3421 (*bmi_blsr_<mode>_ccz): Ditto.
3423 2018-03-26 Tom de Vries <tom@codesourcery.com>
3425 PR tree-optimization/85063
3426 * omp-general.c (offloading_function_p): New function. Factor out
3428 * omp-offload.c (pass_omp_target_link::gate): ... here.
3429 * omp-general.h (offloading_function_p): Declare.
3430 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3431 with attribute omp declare target for offloading functions.
3433 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3435 PR tree-optimization/84005
3436 * tree-data-ref.h (get_base_for_alignment): Declare.
3437 * tree-data-ref.c (get_base_for_alignment_1): New function.
3438 (get_base_for_alignment): Likewise.
3439 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3440 get_base_for_alignment to find a suitable base object, instead
3441 of always using drb->base_address.
3443 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3446 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3447 known size by default.
3449 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3452 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3453 and non BLKmode operands.
3455 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3458 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3459 Clean up attributes.
3461 2018-03-23 Richard Biener <rguenther@suse.de>
3464 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3465 we are going to emit early debug for LTO.
3467 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3470 * function.c (match_asm_constraints_1): Don't optimize if input
3471 doesn't satisfy general_operand predicate for output's mode.
3474 * alias.c (write_dependence_p): Don't require for x_canonicalized
3475 non-VOIDmode if x has VOIDmode.
3478 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3479 just don't try to optimize it rather than assert it never happens.
3481 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3483 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3484 macro expansions for definition of ST_INTERNAL_<mode> and
3485 LD_INTERNAL_<mode> builtins.
3486 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3488 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3490 (altivec_expand_st_builtin): Likewise.
3491 (altivec_expand_builtin): Remove calls to deleted functions.
3492 (rs6000_address_for_altivec): Delete this function.
3493 * config/rs6000/vector.md: Remove expands for
3494 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3496 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3499 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3500 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3501 re-computing once computed.
3502 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3503 (arm_init_machine_status): Initialize
3504 machine->static_chain_stack_bytes.
3506 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3509 * doc/extend.texi: Add four new prototypes for vec_ld.
3510 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3511 definitions for more logical presentation.
3512 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3513 entries for V1TI variants of __builtin_altivec_ld builtin.
3514 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3515 handling of V1TI variant of LVX icode pattern.
3516 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3517 (rs6000_gimple_fold_builtin): Likewise.
3518 (altivec_init_builtins): Add code to define
3519 __builtin_altivec_lvx_v1ti function.
3521 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3524 * function.c (match_asm_constraints_1): Don't do the optimization
3525 if input isn't a REG, SUBREG, MEM or constant.
3527 2018-03-22 Tom de Vries <tom@codesourcery.com>
3529 PR tree-optimization/84956
3530 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3531 bb_has_abnormal_pred.
3533 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3536 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3537 DECL_INITIAL (decl) to decl at the end.
3538 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3541 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3543 * doc/extend.texi (__builtin_tgmath): Document when complex
3544 integer types are treated as _Complex _Float64.
3546 2018-03-21 Tom de Vries <tom@codesourcery.com>
3548 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3550 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3552 PR tree-optimization/84960
3553 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3554 if it is ENTRY block, move them into single succ of ENTRY in that case.
3556 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3558 PR tree-optimization/84811
3559 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3560 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3561 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3562 (known_subrange_p): Update accordingly. Cast each value involved
3563 in the size comparison, rather than casting the result of the
3566 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3568 PR tree-optimization/84982
3569 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3570 by flipping the least significant bit rather than all bits from
3571 bitpos to bitpos + bitsize - 1.
3573 2018-03-21 Nathan Sidwell <nathan@acm.org>
3575 * doc/extend.texi (Deprecated Features): Remove mention of
3576 long-deleted deprecations.
3578 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3581 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3582 * configure: Regenerate.
3584 2018-03-21 Tom de Vries <tom@codesourcery.com>
3586 PR tree-optimization/83126
3587 * tree-parloops.c (num_phis): New function.
3588 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3590 2018-03-21 Nathan Sidwell <nathan@acm.org>
3592 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3593 mention anon-struct/union members and trailing attributes.
3595 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3597 PR tree-optimization/84969
3598 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3599 builtin memset partitions if they set different rhs values.
3601 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3603 PR rtl-optimization/84989
3604 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3605 VEC_DUPLICATE with scalar result mode.
3607 2018-03-21 Martin Liska <mliska@suse.cz>
3610 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3611 not intended return statement.
3613 2018-03-21 Martin Liska <mliska@suse.cz>
3616 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3617 (chkp_find_bound_slots_1): Limit number of iterations.
3619 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3622 * Minor grammar fixes for x86 options.
3624 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3627 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3628 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3630 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3633 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3634 (altivec_lvx_<mode>_1op): Likewise.
3635 (altivec_stvx_<mode>_2op): Likewise.
3636 (altivec_stvx_<mode>_1op): Likewise.
3637 (altivec_lvx_<VM2:mode>): New define_expand.
3638 (altivec_stvx_<VM2:mode>): Likewise.
3639 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3640 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3641 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3642 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3643 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3644 (rs6000_gen_lvx): Likewise.
3645 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3646 (altivec_expand_stv_builtin): Likewise.
3647 (altivec_expand_builtin): Likewise.
3648 * config/rs6000/vector.md: Likewise.
3650 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3653 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3656 2018-03-20 Richard Biener <rguenther@suse.de>
3659 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3660 sign-conversions as zero, fall back to standard scalar_stmt
3663 2018-03-20 Martin Liska <mliska@suse.cz>
3666 * predict.c (rebuild_frequencies): Handle case when we have
3667 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3669 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3672 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3673 flag_section_anchors.
3674 * varasm.c (use_blocks_for_decl_p): Remove hack for
3675 dw2_force_const_mem.
3678 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3680 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3681 be created, use lowpart_subreg of operands[0] rather than operands[0]
3683 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3684 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3685 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3686 and n constraint instead of aarch64_shift_imm_di and Usd.
3687 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3688 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3690 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3693 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3694 to favor GPR over NEON registers.
3695 (<shift>di3_neon): Likewise.
3697 2018-03-20 Tom de Vries <tom@codesourcery.com>
3700 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3701 (nvptx_process_pars): Emit bar.sync asap and alap.
3703 2018-03-20 Tom de Vries <tom@codesourcery.com>
3706 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3707 seen_label if seen_label is already set.
3709 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3712 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3713 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3714 Use 1U instead of 1. Formatting fixes.
3717 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3718 instead of TREE_TYPE (s1) for the return value.
3720 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3722 PR tree-optimization/84946
3723 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3724 bitsize + bitsize in poly_uint64 rather than poly_int64.
3727 * dwarf2asm.c: Include fold-const.c.
3728 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3729 of decl rather than decl itself.
3731 PR rtl-optimization/84643
3732 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3734 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3737 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3738 calling assemble_variable.
3740 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3743 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3744 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3746 2018-03-19 Jim Wilson <jimw@sifive.com>
3749 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3750 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3751 (riscv_first_stack_step): Likewise.
3752 (riscv_option_override): Use STACK_BOUNDARY instead of
3754 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3756 (BIGGEST_ALIGNMENT): Set to 128.
3757 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3758 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3761 2018-03-19 Richard Biener <rguenther@suse.de>
3763 PR tree-optimization/84933
3764 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3765 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3767 2018-03-19 Richard Biener <rguenther@suse.de>
3769 PR tree-optimization/84859
3770 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3771 (cond_if_else_store_replacement): Perform sinking operation on
3772 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3773 Generalize what a BB with a single eligible store is.
3775 2018-03-19 Richard Biener <rguenther@suse.de>
3777 PR tree-optimization/84929
3778 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3779 chrec_is_positive against non-chrec arg.
3781 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3784 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3786 2018-03-18 Martin Liska <mliska@suse.cz>
3788 PR rtl-optimization/84635
3789 * regrename.c (build_def_use): Use matches_mode only when
3792 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3794 PR tree-optimization/84913
3795 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3796 vectorize chains of COND_EXPRs.
3798 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3800 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3802 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3804 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3806 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3808 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3810 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3811 Kito Cheng <kito.cheng@gmail.com>
3813 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3814 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3815 (nds32_adjust_reg_alloc_order): New function.
3816 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3818 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3820 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3821 nds32_print_operand, nds32_print_operand_address): Use
3822 HOST_WIDE_INT_PRINT_DEC instead.
3824 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3826 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3828 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3831 * config/i386/i386.c (initial_ix86_tune_features,
3832 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3834 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3835 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3836 rather than 1u << ix86_tune. Formatting fix.
3837 (ix86_option_override_internal): Change ix86_arch_mask from
3838 unsigned int to unsigned HOST_WIDE_INT, initialize to
3839 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3840 (ix86_function_specific_restore): Likewise.
3842 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3845 * postreload.c (reload_combine_recognize_pattern): Perform
3846 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3847 truncate_int_for_mode the result for the destination's mode.
3850 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3851 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3853 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3856 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3859 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3860 regno_allocno_class_array and sorted_pseudos.
3861 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3862 insns where regno is used.
3864 2018-03-16 Martin Liska <mliska@suse.cz>
3867 * multiple_target.c (create_dispatcher_calls): Redirect
3868 reference in the symbol table.
3870 2018-03-16 Martin Liska <mliska@suse.cz>
3873 * multiple_target.c (create_dispatcher_calls): Redirect also
3876 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3880 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3881 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3882 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3884 2018-03-16 Julia Koval <julia.koval@intel.com>
3886 * doc/invoke.texi (Skylake Server): Add CLWB.
3887 Cannonlake): Remove CLWB.
3889 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3891 PR tree-optimization/84841
3892 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3894 (FLOAT_ONE_CONST_TYPE): Define.
3895 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3896 (sort_by_operand_rank): Put entries with higher constant_type last
3897 rather than first to match comments.
3899 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3901 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3904 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3907 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3908 check and use address of target always.
3910 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3913 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3914 (indirect_thunk_bnd_needed): Likewise.
3915 (indirect_thunks_used): Likewise.
3916 (indirect_thunks_bnd_used): Likewise.
3917 (indirect_return_needed): New.
3918 (indirect_return_bnd_needed): Likewise.
3919 (output_indirect_thunk_function): Add a bool argument for
3921 (output_indirect_thunk_function): Don't generate alias for
3922 function return thunk.
3923 (ix86_code_end): Call output_indirect_thunk_function to generate
3924 function return thunks.
3925 (ix86_output_function_return): Set indirect_return_bnd_needed
3926 and indirect_return_needed instead of indirect_thunk_bnd_needed
3927 and indirect_thunk_needed.
3929 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3931 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3932 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3933 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3935 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3936 Paul Hua <paul.hua.gm@gmail.com>
3939 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3941 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3943 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3944 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3947 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3950 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3951 instead of GET_MODE_SIZE when comparing Units.
3953 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3956 * varasm.c (hash_section): Return an unchangeble hash value
3957 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3958 Return !aarch64_can_use_per_function_literal_pools_p ().
3960 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3963 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3964 rather than address of cmode as last argument to prepare_cmp_insn.
3966 2018-03-15 Julia Koval <julia.koval@intel.com>
3968 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3969 F_AVX512VNNI, F_AVX512BITALG): New.
3971 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3974 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3975 insn for floating-point loads and stores.
3977 2018-03-14 Carl Love <cel@us.ibm.com>
3979 * config/rs6000/rs6000-c.c: Add macro definitions for
3980 ALTIVEC_BUILTIN_VEC_PERMXOR.
3981 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3982 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3983 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3984 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3986 * config/doc/extend.texi: Add prototypes for vec_permxor.
3988 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3991 * diagnostic-show-locus.c (class layout_point): Convert m_line
3992 from int to linenum_type.
3993 (line_span::comparator): Use linenum "compare" function when
3994 comparing line numbers.
3995 (test_line_span): New function.
3996 (layout_range::contains_point): Convert param "row" from int to
3998 (layout_range::intersects_line_p): Likewise.
3999 (layout::will_show_line_p): Likewise.
4000 (layout::print_source_line): Likewise.
4001 (layout::should_print_annotation_line_p): Likewise.
4002 (layout::print_annotation_line): Likewise.
4003 (layout::print_leading_fixits): Likewise.
4004 (layout::annotation_line_showed_range_p): Likewise.
4005 (struct line_corrections): Likewise for field m_row.
4006 (line_corrections::line_corrections): Likewise for param "row".
4007 (layout::print_trailing_fixits): Likewise.
4008 (layout::get_state_at_point): Likewise.
4009 (layout::get_x_bound_for_row): Likewise.
4010 (layout::print_line): Likewise.
4011 (diagnostic_show_locus): Likewise for locals "last_line" and
4013 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4014 * input.c (selftest::test_linenum_comparisons): New function.
4015 (selftest::input_c_tests): Call it.
4016 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4017 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4018 * selftest.h (ASSERT_GT): New macro.
4019 (ASSERT_GT_AT): New macro.
4020 (ASSERT_LT): New macro.
4021 (ASSERT_LT_AT): New macro.
4023 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4025 PR rtl-optimization/84780
4026 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4028 2018-03-14 Martin Liska <mliska@suse.cz>
4030 * tree.c (record_node_allocation_statistics): Use
4031 get_stats_node_kind.
4032 (get_stats_node_kind): New function extracted from
4033 record_node_allocation_statistics.
4034 (free_node): Use get_stats_node_kind.
4036 2018-03-14 Richard Biener <rguenther@suse.de>
4038 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4039 that the value-set of ANTIC_IN doesn't grow.
4042 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4044 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4045 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4047 2018-03-14 Julia Koval <julia.koval@intel.com>
4049 * config.gcc (icelake-client, icelake-server): New.
4051 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4052 (initial_ix86_arch_features): Ditto.
4053 (PTA_SKYLAKE): Add SGX.
4054 (PTA_ICELAKE): Remove.
4055 (PTA_ICELAKE_CLIENT): New.
4056 (PTA_ICELAKE_SERVER): New.
4057 (ix86_option_override_internal): Split up icelake on icelake client and
4059 (get_builtin_code_for_version): Ditto.
4060 (fold_builtin_cpu): Ditto.
4061 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4062 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4063 * config/i386/i386.h (processor_type): Ditto.
4064 * doc/invoke.texi: Ditto.
4066 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4069 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4070 INTEGER_CST offset, add it together with bitpos / 8 and
4071 sign extend based on POINTER_SIZE.
4075 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4078 * config/i386/constraints.md (Yc): New register constraint.
4079 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4080 Use Yc constraint for alternative 2 of operand 0. Remove
4081 preferred_for_speed attribute.
4083 2018-03-14 Richard Biener <rguenther@suse.de>
4085 PR tree-optimization/84830
4086 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4087 with the old one to avoid oscillations.
4089 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4092 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4094 (assign_by_spills): Return a flag of reload assignment failure.
4095 Do not process the reload assignment failures. Do not spill other
4096 reload pseudos if they has the same reg class. Update n if
4098 (lra_assign): Add a return arg. Set up from the result of
4099 assign_by_spills call.
4100 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4101 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4102 usage_insns if it is not NULL.
4103 (spill_hard_reg_in_range): New function.
4104 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4105 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4106 function prototypes.
4107 (lra_assign): Change prototype.
4108 * lra.c (lra): Add code to deal with fails by splitting hard reg
4111 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4113 * config/riscv/riscv.opt (mrelax): New option.
4114 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4115 "norelax" when riscv_mrelax is disabled.
4116 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4118 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4121 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4122 reassociation for int modes.
4124 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4126 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4127 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4129 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4130 * config/aarch64/aarch64-sve.md
4131 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4132 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4133 (*extend<mode><Vwide>2): Rename to...
4134 (aarch64_sve_extend<mode><Vwide>2): ...this.
4135 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4136 renaming the old pattern to...
4137 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4139 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4140 define_expand, renaming the old pattern to...
4141 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4142 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4143 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4144 account when deciding which SVE instruction the optab should use.
4145 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4147 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4149 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4150 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4151 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4152 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4153 (tlsdesc_small_<mode>): Turn a define_expand and use
4154 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4155 (tlsdesc_small_advsimd_<mode>): ...this.
4156 (tlsdesc_small_sve_<mode>): New pattern.
4158 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4160 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4161 (UNSPEC_UMUL_HIGHPART): New constants.
4162 (MUL_HIGHPART): New int iteraor.
4163 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4164 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4166 (*<su>mul<mode>3_highpart): New define_insn.
4168 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4171 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4174 2018-03-13 Martin Liska <mliska@suse.cz>
4177 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4179 (sem_item_optimizer::~sem_item_optimizer): Release it.
4180 (sem_item_optimizer::merge_classes): Register variable aliases.
4181 (sem_item_optimizer::fixup_pt_set): New function.
4182 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4183 * ipa-icf.h: Declare new variables and functions.
4185 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4188 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4189 integer_pow2p@2 and test integer_pow2p in condition.
4190 (A < 0 ? C : 0): Similarly for @1.
4193 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4194 characters starting at p contain '\0' character, don't look beyond
4198 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4199 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4202 * reg-stack.c (change_stack): Change update_end var from int to
4203 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4204 also call set_block_for_insn on the newly added insns and rescan.
4207 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4208 on the last operand.
4211 * tree.c (stabilize_reference_1): Return save_expr (e) for
4212 STATEMENT_LIST even if it doesn't have side-effects.
4214 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4216 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4218 2018-03-12 Renlin Li <renlin.li@arm.com>
4220 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4221 aarch64_output_scalar_simd_mov_immediate.
4223 2018-03-12 Martin Sebor <msebor@redhat.com>
4225 PR tree-optimization/83456
4226 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4227 for perfectly overlapping calls to memcpy.
4228 (gimple_fold_builtin_memory_chk): Same.
4229 (gimple_fold_builtin_strcpy): Handle no-warning.
4230 (gimple_fold_builtin_stxcpy_chk): Same.
4231 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4233 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4235 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4236 parameter. Use it for SFmode.
4237 (rs6000_function_arg_advance_1): Adjust.
4238 (rs6000_function_arg): Adjust.
4239 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4241 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4243 PR rtl-optimization/84169
4244 PR rtl-optimization/84780
4245 * combine.c (can_combine_p): Check for a 2-insn combination whether
4246 the destination register is used between the two insns, too.
4248 2018-03-12 Richard Biener <rguenther@suse.de>
4250 PR tree-optimization/84803
4251 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4252 for refs DR analysis didn't process.
4254 2018-03-12 Richard Biener <rguenther@suse.de>
4256 PR tree-optimization/84777
4257 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4258 force-vectorize loops ignore whether we are optimizing for size.
4260 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4262 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4263 (TARGET_MD_ASM_ADJUST): Define.
4265 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4266 Kito Cheng <kito.cheng@gmail.com>
4267 Chung-Ju Wu <jasonwucj@gmail.com>
4269 * config/nds32/nds32.c (nds32_compute_stack_frame,
4270 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4271 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4272 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4273 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4274 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4275 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4276 * config/nds32/nds32.md (prologue, epilogue): Use macro
4277 NDS32_V3PUSH_AVAILABLE_P to do checking.
4279 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4282 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4283 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4284 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4285 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4286 addition of most attributes on !orig_type_die or the attribute not
4287 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4289 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4290 Chung-Ju Wu <jasonwucj@gmail.com>
4292 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4294 * config/nds32/nds32.opt (mvh): New option.
4296 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4297 Chung-Ju Wu <jasonwucj@gmail.com>
4299 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4301 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4302 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4305 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4306 Chung-Ju Wu <jasonwucj@gmail.com>
4308 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4310 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4311 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4313 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4314 Kito Cheng <kito.cheng@gmail.com>
4315 Chung-Ju Wu <jasonwucj@gmail.com>
4317 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4318 UNSPEC_FFMISM and UNSPEC_FLMISM.
4319 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4320 for ffb, ffmism and flmism.
4321 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4322 (unspec_ffmism): Ditto.
4323 (unspec_flmism): Ditto.
4324 (nds32_expand_builtin_impl): Check if string extension is available.
4325 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4326 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4328 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4331 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4334 * lra-assigns.c (assign_by_spills): Return a flag of reload
4335 assignment failure. Do not process the reload assignment
4336 failures. Do not spill other reload pseudos if they has the same
4338 (lra_assign): Add a return arg. Set up from the result of
4339 assign_by_spills call.
4340 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4341 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4342 usage_insns if it is not NULL.
4343 (spill_hard_reg_in_range): New function.
4344 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4345 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4346 function prototypes.
4347 (lra_assign): Change prototype.
4348 * lra.c (lra): Add code to deal with fails by splitting hard reg
4351 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4354 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4356 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4359 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4360 (dw_val_node): Add val_symbolic_view.
4361 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4362 (symview_upper_bound): New.
4363 (new_line_info_table): Initialize symviews_since_reset.
4364 (dwarf2out_source_line): Count symviews_since_reset and set
4365 symview_upper_bound.
4366 (dw_val_equal_p): Handle symview.
4367 (add_AT_symview): New.
4368 (print_dw_val): Handle symview.
4369 (attr_checksum, attr_checksum_ordered): Likewise.
4370 (same_dw_val_p, size_of_die): Likewise.
4371 (value_format, output_die): Likewise.
4372 (add_high_low_attributes): Use add_AT_symview for entry_view.
4373 (dwarf2out_finish): Reset symview_upper_bound, clear
4376 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4379 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4380 Add strict argument and use it.
4381 (rs6000_split_multireg_move): Update for new strict argument.
4382 (mem_operand_gpr): Disallow all non-offsettable addresses.
4383 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4385 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4388 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4389 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4390 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4393 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4394 decl, use remap_type if we want to use the type.
4396 2018-03-09 Martin Sebor <msebor@redhat.com>
4398 PR tree-optimization/84526
4399 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4401 (builtin_access::generic_overlap): Be prepared to handle non-array
4404 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4406 PR rtl-optimization/84682
4407 * lra-constraints.c (process_address_1): Check is_address flag
4408 for address constraints.
4409 (process_alt_operands): Likewise.
4410 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4411 preprocess_constraints.
4412 * recog.h (preprocess_constraints): Add oploc parameter.
4414 * recog.c (preprocess_constraints): Test address_operand for
4415 CT_ADDRESS constraints.
4417 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4420 * lra-assigns.c (assign_by_spills): Return a flag of reload
4421 assignment failure. Do not process the reload assignment
4422 failures. Do not spill other reload pseudos if they has the same
4424 (lra_assign): Add a return arg. Set up from the result of
4425 assign_by_spills call.
4426 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4427 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4428 usage_insns if it is not NULL.
4429 (spill_hard_reg_in_range): New function.
4430 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4431 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4432 function prototypes.
4433 (lra_assign): Change prototype.
4434 * lra.c (lra): Add code to deal with fails by splitting hard reg
4437 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4440 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4441 Accept complain bool parameter. Only emit errors if it is true.
4442 (arm_parse_cpu_option_name): Likewise.
4443 (arm_target_thumb_only): Adjust callers of the above.
4444 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4445 prototype to take a default true bool parameter.
4446 (arm_parse_arch_option_name): Likewise.
4448 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4449 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4453 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4454 * configure: Regenerate.
4455 * configure.ac ("linker --version-script option"): New.
4456 ("linker soname option"): New.
4458 2018-03-09 Richard Biener <rguenther@suse.de>
4460 PR tree-optimization/84775
4461 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4462 immediate uses of predicate stmts and mark them modified.
4465 PR tree-optimization/84178
4466 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4468 (version_loop_for_if_conversion): Delay update_ssa call.
4469 (tree_if_conversion): Delay update_ssa until after predicate
4472 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4475 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4476 when the function accesses prior frames.
4478 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4481 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4482 gen_llsym, otherwise call maybe_gen_llsym.
4485 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4486 has ',' character inside of it.
4488 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4491 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4492 as clobbering CC_REGNUM.
4494 2018-03-08 Richard Biener <rguenther@suse.de>
4497 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4498 (follow_copies_to_constant): Do not follow SSA names registered
4501 2018-03-08 Richard Biener <rguenther@suse.de>
4503 PR tree-optimization/84178
4504 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4506 (version_loop_for_if_conversion): Delay update_ssa call.
4507 (tree_if_conversion): Delay update_ssa until after predicate
4510 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4512 PR tree-optimization/84178
4513 * tree-if-conv.c (release_bb_predicate): Remove the
4514 the assertion that the stmts have NULL use_ops.
4515 Discard the statements, asserting that they haven't
4516 yet been added to a BB.
4518 2018-03-08 Richard Biener <rguenther@suse.de>
4520 PR tree-optimization/84746
4521 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4522 (phi_translate): Pass in destination ANTIC_OUT set.
4523 (phi_translate_1): Likewise. For a simplified result lookup
4524 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4525 (phi_translate_set): Adjust.
4526 (do_pre_regular_insertion): Likewise.
4527 (do_pre_partial_partial_insertion): Likewise.
4529 2018-03-08 Martin Liska <mliska@suse.cz>
4531 PR gcov-profile/84735
4532 * doc/gcov.texi: Document usage of profile files.
4533 * gcov-io.h: Document changes in the format.
4535 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4539 * dwarf2out.c (struct dw_line_info_table): Update comments for
4541 (FORCE_RESET_NEXT_VIEW): New.
4542 (FORCE_RESETTING_VIEW_P): New.
4543 (RESETTING_VIEW_P): Check for -1 too.
4544 (ZERO_VIEW_P): Likewise.
4545 (new_line_info_table): Force-reset next view.
4546 (dwarf2out_begin_function): Likewise.
4547 (dwarf2out_source_line): Simplify zero_view_p initialization.
4548 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4549 view directly. Omit view when omitting .loc at line 0.
4551 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4553 PR tree-optimization/84740
4554 * tree-switch-conversion.c (process_switch): Call build_constructors
4555 only if info.phi_count is non-zero.
4557 PR tree-optimization/84739
4558 * tree-tailcall.c (find_tail_calls): Check call arguments against
4559 DECL_ARGUMENTS (current_function_decl) rather than
4560 DECL_ARGUMENTS (func) when checking for tail recursion.
4562 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4564 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4565 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4566 Volker Reichelt's entry and add entries for people that perform
4567 GCC fuzzy testing and report numerous bugs.
4569 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4572 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4573 readonly data in sdata, if that is disabled.
4574 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4575 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4576 -mreadonly-in-sdata option.
4578 2018-03-07 Martin Sebor <msebor@redhat.com>
4580 PR tree-optimization/84468
4581 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4582 basic block when looking for nul assignment.
4584 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4587 * except.h (output_function_exception_table): Adjust prototype.
4588 * except.c (output_function_exception_table): Remove FNNAME parameter
4589 and add SECTION parameter. Ouput one part of the table at a time.
4590 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4591 the first part of the exception table and emit unwind directives.
4592 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4593 (i386_pe_seh_cold_init): Likewise.
4594 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4595 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4596 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4597 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4598 * config/i386/winnt.c: Include except.h.
4599 (struct seh_frame_state): Add reg_offset, after_prologue and
4600 in_cold_section fields.
4601 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4602 (i386_pe_seh_cold_init): New function.
4603 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4604 to seh->in_cold_section.
4605 (seh_emit_push): Record the offset of the push.
4606 (seh_emit_save): Record the offet of the save.
4607 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4608 Test seh->after_prologue to disregard the epilogue.
4609 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4610 (i386_pe_end_cold_function): New function.
4612 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4615 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4616 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4619 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4621 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4622 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4625 * multiple_target.c: Include tree-inline.h and intl.h.
4626 (expand_target_clones): Diagnose and fail if node->definition and
4627 !tree_versionable_function_p (node->decl).
4629 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4631 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4633 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4634 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4635 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4637 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4640 * combine.c (try_combine): Use reg_or_subregno instead of handling
4641 just paradoxical SUBREGs and REGs.
4643 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4645 * config/arc/arc.c (arc_finalize_pic): Remove function.
4646 (arc_must_save_register): We use single base PIC register, remove
4647 checks to save/restore the PIC register.
4648 (arc_expand_prologue): Likewise.
4649 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4651 (arc_verify_short): Likewise.
4652 (arc_attr_type): Likewise.
4653 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4654 (walk_stores): Likewise.
4655 (arc_address_cost): Make it static.
4656 (arc_verify_short): Likewise.
4657 (branch_dest): Likewise.
4658 (arc_attr_type): Likewise.
4659 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4660 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4661 (arc_final_prescan_insn): Remove inserting the nops due to
4662 hardware hazards. It is done in reorg step.
4663 (insn_length_variant_t): Remove.
4664 (insn_length_parameters_t): Likewise.
4665 (arc_insn_length_parameters): Likewise.
4666 (arc_get_insn_variants): Likewise.
4667 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4669 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4672 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4675 PR tree-optimization/84687
4676 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4678 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4680 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4682 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4683 Rename to ppc_speculation_barrier.
4684 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4685 __builtin_ppc_speculation_barrier.
4687 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4690 * combine.c (combine_simplify_rtx): Don't try to simplify if
4691 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4694 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4696 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4697 to 32 bytes when compiling for POWER9.
4699 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4702 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4703 regparm >= 3 with no arg reg available also for calls with
4704 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4707 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4709 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4711 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4714 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4716 2018-03-05 Richard Biener <rguenther@suse.de>
4718 PR tree-optimization/84486
4719 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4720 When inserting a __builtin_assume_aligned call set the LHS
4721 SSA name alignment info accordingly.
4723 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4725 PR tree-optimization/84114
4726 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4727 Avoid reassociation of FLOAT_MODE addition.
4729 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4731 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4732 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4733 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4734 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4735 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4736 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4737 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4739 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4740 __builtin_ia32_wbinvd): New builtins.
4741 (SPECIAL_ARGS2): New.
4742 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4743 (SPECIAL_ARGS2): New.
4744 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4745 (ix86_valid_target_attribute_inner_p): Ditto.
4746 (ix86_init_mmx_sse_builtins): Add special_args2.
4747 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4748 TARGET_WBNOINVD_P): New.
4749 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4750 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4751 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4752 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4753 * config/i386/pconfigintrin.h: New file.
4754 * config/i386/wbnoinvdintrin.h: Ditto.
4755 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4756 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4758 2018-03-05 Richard Biener <rguenther@suse.de>
4760 PR tree-optimization/84670
4761 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4763 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4764 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4765 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4766 doesn't grow if all successors (recursively) were visited at least
4769 2018-03-05 Richard Biener <rguenther@suse.de>
4771 PR tree-optimization/84650
4772 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4773 if executed in the loop pipeline.
4775 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4777 * doc/configfiles.texi (Configuration Files): Move info about
4778 conditionalizing $target-protos.h to...
4779 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4780 differs from $target-protos.h.
4782 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4783 Chung-Ju Wu <jasonwucj@gmail.com>
4785 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4786 * config/nds32/nds32-multiple.md (setmemsi): Define.
4787 * config/nds32/nds32-memory-manipulation.c
4788 (nds32_gen_dup_4_byte_to_word_value): New.
4789 (emit_setmem_word_loop): New.
4790 (emit_setmem_byte_loop): New.
4791 (nds32_expand_setmem_loop): New.
4792 (nds32_expand_setmem_loop_v3m): New.
4793 (nds32_expand_setmem_unroll): New.
4794 (nds32_expand_setmem): New.
4796 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4797 Chung-Ju Wu <jasonwucj@gmail.com>
4799 * config/nds32/nds32-memory-manipulation.c
4800 (nds32_emit_load_store): New.
4801 (nds32_emit_post_inc_load_store): New.
4802 (nds32_emit_mem_move): New.
4803 (nds32_emit_mem_move_block): New.
4804 (nds32_expand_movmemsi_loop_unknown_size): New.
4805 (nds32_expand_movmemsi_loop_known_size): New.
4806 (nds32_expand_movmemsi_loop): New.
4807 (nds32_expand_movmemsi_unroll): New.
4808 (nds32_expand_movmemqi): Rename ...
4809 (nds32_expand_movmemsi): ... to this.
4810 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4811 (movmemsi): ... to this.
4812 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4813 (nds32_expand_movmemsi): ... to this.
4815 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4816 Monk Chiang <sh.chiang04@gmail.com>
4817 Chung-Ju Wu <jasonwucj@gmail.com>
4819 * config/nds32/nds32-protos.h
4820 (nds32_expand_load_multiple): New arguments.
4821 (nds32_expand_store_multiple): Ditto.
4822 (nds32_valid_multiple_load_store): Rename ...
4823 (nds32_valid_multiple_load_store_p): ... to this.
4824 * config/nds32/nds32-memory-manipulation.c
4825 (nds32_expand_load_multiple): Refine implementation.
4826 (nds32_expand_store_multiple): Ditto.
4827 * config/nds32/nds32-multiple.md
4828 (load_multiple): Update nds32_expand_load_multiple interface.
4829 (store_multiple): Update nds32_expand_store_multiple interface.
4830 * config/nds32/nds32-predicates.c
4831 (nds32_valid_multiple_load_store): Rename ...
4832 (nds32_valid_multiple_load_store_p): ... to this and refine
4834 * config/nds32/predicates.md
4835 (nds32_load_multiple_and_update_address_operation): New predicate.
4836 (nds32_store_multiple_and_update_address_operation): New predicate.
4838 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4839 Chung-Ju Wu <jasonwucj@gmail.com>
4841 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4842 (combo): New attribute.
4843 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4845 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4847 * config/nds32/nds32.opt: Change -mcmodel= default value.
4849 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4850 Monk Chiang <sh.chiang04@gmail.com>
4851 Chung-Ju Wu <jasonwucj@gmail.com>
4853 * config/nds32/constants.md (unspec_element): New enum.
4854 * config/nds32/constraints.md (Umw): New constraint.
4855 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4856 * config/nds32/nds32-intrinsic.md: Likewise.
4857 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4858 (nds32_valid_smw_lwm_base_p): New.
4859 (nds32_output_smw_single_word): New.
4860 (nds32_output_lmw_single_word): New.
4861 (nds32_expand_unaligned_load): New.
4862 (nds32_expand_unaligned_store): New.
4863 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4864 (nds32_output_smw_single_word): Declare.
4865 (nds32_output_lmw_single_word): Declare.
4866 (nds32_expand_unaligned_load): Declare.
4867 (nds32_expand_unaligned_store): Declare.
4868 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4869 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4870 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4871 NDS32_BUILTIN_UASTORE_DW.
4872 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4875 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4876 Kito Cheng <kito.cheng@gmail.com>
4877 Chung-Ju Wu <jasonwucj@gmail.com>
4879 * config/nds32/nds32-intrinsic.c
4880 (nds32_expand_builtin_null_ftype_reg): Delete.
4881 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4882 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4883 (nds32_read_argument): New.
4884 (nds32_legitimize_target): Ditto.
4885 (nds32_legitimize_argument): Ditto.
4886 (nds32_check_constant_argument): Ditto.
4887 (nds32_expand_unop_builtin): Ditto.
4888 (nds32_expand_unopimm_builtin): Ditto.
4889 (nds32_expand_binop_builtin): Ditto.
4890 (nds32_builtin_decl_impl): Ditto.
4891 (builtin_description): Ditto.
4892 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4893 (nds32_init_builtins_impl): Ditto.
4894 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4895 (nds32_builtin_decl): New.
4896 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4897 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4899 2018-03-02 Jeff Law <law@redhat.com>
4901 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4902 (redundant_insn, fill_simple_delay_slots): Likewise.
4903 (fill_slots_from_thread): Likewise.
4904 * resource.c (mark_referenced_resources): Likewise.
4905 (mark_set_resources, find_dead_or_set_registers): Likewise.
4907 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4909 * substring-locations.h (format_warning_va): Formatting fix for
4911 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4913 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4914 * substring-locations.c: Include intl.h.
4915 (format_warning_va): Turned into small wrapper around
4916 format_warning_n_va, renamed to ...
4917 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4918 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4920 (format_warning_at_substring_n): New function.
4921 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4922 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4923 format_warning_at_substring with just a shorter name instead of
4924 const function pointer.
4925 (fmtwarn_n): New function.
4926 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4927 appropriate, get rid of all the fmtstr temporaries, move conditionals
4928 with G_() wrapped string literals directly into fmtwarn arguments,
4929 cast dir.len to (int), formatting fixes.
4931 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4933 * doc/invoke.texi: Remove "Cilk Plus" references.
4935 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4936 Richard Biener <rguenther@suse.de>
4939 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4940 for error or warning attributes if CALL_FROM_THUNK_P is set.
4943 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4946 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4947 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4950 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4951 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4953 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4954 instead of -1U in last predictors element's probability member.
4956 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4959 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4960 arguments if they are comparable.
4962 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4964 PR tree-optimization/84634
4965 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4966 masks and masked_loop_p with a single loop_masks, making sure it's
4967 null for bb vectorization.
4969 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4971 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4972 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4973 loop->force_vectorize to check whether there is no alias.
4975 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4978 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4980 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4982 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4983 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4984 instead of a loop around prev_real_insn.
4985 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4989 * config/i386/i386.c (ix86_print_operand): Use conditional
4990 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4993 2018-03-02 Richard Biener <rguenther@suse.de>
4995 PR tree-optimization/84427
4996 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4997 (bitmap_set_subtract_values): Rewrite to handle multiple
5000 (prune_clobbered_mems): Likewise.
5001 (phi_translate): Take edge instead of pred/phiblock.
5002 (phi_translate_1): Likewise.
5003 (phi_translate_set): Likewise. Insert all translated
5004 exprs for a value into the set, keeping possibly multiple
5005 expressions per value.
5006 (compute_antic_aux): Adjust for phi_translate changes.
5007 When intersecting union the expressions and prune those
5008 not in the final value set, keeping possibly multiple
5009 expressions per value. Do not use value-insertion
5010 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5011 all expressions. Add verification that the value-sets
5012 only shrink during iteration.
5013 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5014 (do_pre_regular_insertion): Likewise.
5015 (do_pre_partial_partial_insertion): Likewise.
5017 2018-03-02 Richard Biener <rguenther@suse.de>
5020 * config/darwin.c (saved_debug_info_level): New static global.
5021 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5022 (darwin_asm_lto_end): Restore debug info generation settings.
5024 2018-03-01 Martin Liska <mliska@suse.cz>
5027 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5030 2018-03-01 Richard Biener <rguenther@suse.de>
5033 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5034 type annotation with LTO when debug was disabled at compile-time.
5036 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5038 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5040 (mips_final_postscan_insn): Likewise.
5042 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5044 PR rtl-optimization/84528
5045 * alias.c (init_alias_target): Add commentary.
5046 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5047 a unique base value if the frame pointer is not eliminated
5048 to the stack pointer.
5050 2018-03-01 Tom de Vries <tom@codesourcery.com>
5052 PR rtl-optimization/83327
5053 * lra-int.h (hard_regs_spilled_into): Declare.
5054 * lra.c (hard_regs_spilled_into): Define.
5055 (init_reg_info): Init hard_regs_spilled_into.
5056 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5057 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5058 (process_bb_lives): Handle hard_regs_spilled_into.
5059 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5060 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5062 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5064 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5065 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5066 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5067 * config/rs6000/aix72.h: New file.
5069 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5071 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5072 instead of warning_at with conditional singular and plural messages
5076 * stor-layout.c (update_alignment_for_field): For
5077 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5078 && !DECL_PACKED (field), do the alignment update, just use
5079 only desired_align instead of MAX (type_align, desired_align)
5081 (place_field): Don't do known_align < desired_align handling
5082 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5083 is non-NULL, instead do it after rli->prev_field handling and
5084 only if not within a bitfield word. For DECL_PACKED (field)
5085 use type_align of BITS_PER_UNIT.
5087 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5089 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5090 superfluous parentheses and trailing spaces.
5092 2018-02-28 Richard Biener <rguenther@suse.de>
5094 PR tree-optimization/84584
5095 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5096 SCoPs with fake exit edge.
5098 2018-02-28 Martin Liska <mliska@suse.cz>
5101 * timevar.c (timer::print): Fix format to properly print 100%
5104 2018-02-28 Richard Biener <rguenther@suse.de>
5107 * genmatch.c (capture_info::walk_match): Do not mark
5108 captured expressions without operands as expr_p given
5109 they act more like predicates and should be subject to
5110 "lost tail" side-effect preserving.
5112 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5114 PR rtl-optimization/81611
5115 * auto-inc-dec.c (attempt_change): Move dead note from
5116 mem_insn if it's the next use of regno
5117 (find_address): Take address use of reg holding
5118 non-incremented value. Add parm to limit search to the named
5120 (merge_in_block): Attempt to use a mem insn that is the next
5121 use of the original regno.
5123 2018-02-27 Martin Sebor <msebor@redhat.com>
5126 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5127 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5129 2018-02-27 Martin Sebor <msebor@redhat.com>
5131 PR translation/84207
5132 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5133 n argument to unsigned HOST_WIDE_INT.
5134 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5135 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5136 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5137 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5139 2018-02-27 Richard Biener <rguenther@suse.de>
5141 PR tree-optimization/84512
5142 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5143 Do not use the estimate returned from record_stmt_cost for
5144 the scalar iteration cost but sum properly using add_stmt_cost.
5146 2018-02-27 Richard Biener <rguenther@suse.de>
5148 PR tree-optimization/84466
5149 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5150 Adjust last change to less strictly validate use operands.
5152 2018-02-27 Martin Liska <mliska@suse.cz>
5154 PR gcov-profile/84548
5155 * gcov.c (process_file): Allow partial overlap and consider it
5156 also as group functions.
5157 (output_lines): Properly calculate range of lines for a group.
5159 2018-02-27 Martin Liska <mliska@suse.cz>
5161 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5162 'ggc' suffixes. Change first column width.
5163 (timer::print): Fix formatting of the column.
5165 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5167 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5168 preserve inline entry blocks for the sake of debug inline
5169 entry point markers alone.
5170 (remove_unused_locals): Suggest in comments a better place to
5171 force the preservation of inline entry blocks that are
5172 otherwise unused, but do not preserve them.
5174 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5176 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5178 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5181 * config/i386/constraints.md (Bs): Replace
5182 ix86_indirect_branch_register with
5183 TARGET_INDIRECT_BRANCH_REGISTER.
5185 * config/i386/i386.md (indirect_jump): Likewise.
5186 (tablejump): Likewise.
5187 (*sibcall_memory): Likewise.
5188 (*sibcall_value_memory): Likewise.
5189 Peepholes of indirect call and jump via memory: Likewise.
5190 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5191 (*sibcall_value_GOT_32): Likewise.
5192 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5193 (GOT_memory_operand): Likewise.
5194 (call_insn_operand): Likewise.
5195 (sibcall_insn_operand): Likewise.
5196 (GOT32_symbol_operand): Likewise.
5197 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5199 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5201 PR rtl-optimization/83496
5202 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5203 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5204 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5205 redundant insn, if any.
5206 (relax_delay_slots): Likewise.
5207 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5209 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5211 PR tree-optimization/83965
5212 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5213 that grouped statements are part of a reduction chain. Return
5214 true if the statement is not marked as a reduction itself but
5216 (vect_recog_dot_prod_pattern): Don't check whether the statement
5217 is part of a group here.
5218 (vect_recog_sad_pattern): Likewise.
5219 (vect_recog_widen_sum_pattern): Likewise.
5221 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5224 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5226 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5229 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5231 (ix86_output_indirect_function_return): New prototype.
5232 (ix86_split_simple_return_pop_internal): Likewise.
5233 * config/i386/i386.c (indirect_return_via_cx): New.
5234 (indirect_return_via_cx_bnd): Likewise.
5235 (indirect_thunk_name): Handle return va CX_REG.
5236 (output_indirect_thunk_function): Create alias for
5237 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5238 (ix86_output_indirect_jmp): Remove the bool argument.
5239 (ix86_output_indirect_function_return): New function.
5240 (ix86_split_simple_return_pop_internal): Likewise.
5241 * config/i386/i386.md (*indirect_jump): Don't pass false
5242 to ix86_output_indirect_jmp.
5243 (*tablejump_1): Likewise.
5244 (simple_return_pop_internal): Change it to define_insn_and_split.
5245 Call ix86_split_simple_return_pop_internal to split it for
5247 (simple_return_indirect_internal): Call
5248 ix86_output_indirect_function_return instead of
5249 ix86_output_indirect_jmp.
5251 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5254 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5255 memset and value initialization afterwards.
5257 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5259 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5261 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5264 * common/config/aarch64/aarch64-common.c
5265 (aarch_option_optimization_table[]): Switch
5266 off fomit-frame-pointer
5268 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5269 Chung-Ju Wu <jasonwucj@gmail.com>
5271 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5273 (store_multiple): Ditto.
5275 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5277 * config.gcc: Add --with-cpu support for nds32 target.
5278 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5279 * config/nds32/nds32.opt: Add -mcpu= option.
5281 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5283 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5284 isel=yes): Warn for these deprecated options.
5286 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5288 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5289 ISA_2_5_MASKS_EMBEDDED.
5291 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5293 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5294 p->max as pointers rather than using iterative_hash_expr.
5296 2018-02-23 Carl Love <cel@us.ibm.com>
5298 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5299 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5301 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5302 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5303 P8V_BUILTIN_VEC_VUNSIGNED2.
5305 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5308 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5309 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5311 * lra-constraints.c (curr_insn_transform): Set up
5312 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5314 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5315 * lra-eliminations.c (spill_pseudos): Ditto.
5316 (process_insn_for_elimination): Ditto.
5317 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5318 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5319 LRA_NON_CLOBBERED_ALT.
5321 2018-02-22 Martin Sebor <msebor@redhat.com>
5323 PR tree-optimization/84480
5324 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5325 to maybe_diag_stxncpy_trunc. Call it.
5326 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5327 from gimple_fold_builtin_strcpy. Print inlining stack.
5328 (handle_builtin_stxncpy): Print inlining stack.
5329 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5331 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5334 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5335 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5336 and -fcheck-pointer-bounds are used together.
5337 (indirect_thunk_prefix): New enum.
5338 (indirect_thunk_need_prefix): New function.
5339 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5340 "_nt" instead of "_bnd" for NOTRACK prefix.
5341 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5342 (output_indirect_thunk_function): Likewise.
5344 (ix86_code_end): Update output_indirect_thunk_function calls.
5345 (ix86_output_indirect_branch_via_reg): Replace
5346 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5347 (ix86_output_indirect_branch_via_push): Likewise.
5348 (ix86_output_function_return): Likewise.
5349 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5350 incompatible with -fcf-protection=branch and
5351 -fcheck-pointer-bounds.
5353 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5356 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5357 Change gcc_assert call to output_operand_lossage.
5359 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5361 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5363 2018-02-22 DJ Delorie <dj@redhat.com>
5364 Sebastian Perta <sebastian.perta@renesas.com>
5365 Oleg Endo <olegendo@gcc.gnu.org>
5367 * config/rx/rx.c (rx_rtx_costs): New function.
5368 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5370 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5372 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5374 2018-02-22 Martin Liska <mliska@suse.cz>
5377 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5378 Add "native" as a possible value.
5380 2018-02-22 Martin Liska <mliska@suse.cz>
5383 * config/i386/i386.c (ix86_option_override_internal):
5384 Add "native" as a possible value for -march and -mtune.
5386 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5389 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5390 to all type variants.
5392 PR tree-optimization/84503
5393 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5394 width as info->bitpos + info->bitsize - start.
5395 (merged_store_group::merge_overlapping): Simplify width computation.
5396 (check_no_overlap): New function.
5397 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5398 start + width and last_order of the group, fail if check_no_overlap
5400 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5401 to group if check_no_overlap fails.
5403 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5405 * config/rs6000/altivec.md: Delete contraint arguments to
5406 define_expand, define_split, and define_peephole2, and in
5407 define_insn_and_split if always unused.
5408 * config/rs6000/darwin.md: Ditto.
5409 * config/rs6000/dfp.md: Ditto.
5410 * config/rs6000/rs6000.md: Ditto.
5411 * config/rs6000/sync.md: Ditto.
5412 * config/rs6000/vector.md: Ditto.
5413 * config/rs6000/vsx.md: Ditto.
5415 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5417 * config/rs6000/altivec.md: Write output control strings as braced
5418 blocks instead of double-quoted strings.
5419 * config/rs6000/darwin.md: Ditto.
5420 * config/rs6000/rs6000.md: Ditto.
5421 * config/rs6000/vector.md: Ditto.
5422 * config/rs6000/vsx.md: Ditto.
5424 2018-02-21 Jason Merrill <jason@redhat.com>
5426 PR c++/84314 - ICE with templates and fastcall attribute.
5427 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5429 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5431 * ipa-cp.c (determine_versionability): Fix comment typos.
5433 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5436 * ipa-cp.c (determine_versionability): Do not version functions caling
5439 2018-02-21 Martin Liska <mliska@suse.cz>
5442 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5443 Add "native" as a possible value.
5444 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5445 the macro when native cpu detection is available.
5447 2018-02-21 Martin Liska <mliska@suse.cz>
5450 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5451 Add "native" as a possible value.
5452 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5453 when native cpu detection is available.
5455 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5456 Martin Sebor <msebor@redhat.com>
5458 PR tree-optimization/84478
5459 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5461 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5462 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5463 support which is conservatively correct, for 2 only stay conservative
5464 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5465 argument to the 2 argument get_range_strlen, adjust 6 arg
5466 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5468 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5469 (gimple_fold_builtin_strlen): Pass true as last argument to
5472 2018-02-20 Martin Sebor <msebor@redhat.com>
5475 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5476 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5477 (builtin_memref::builtin_memref): Factor out parts into
5478 set_base_and_offset and call it.
5480 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5483 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5484 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5485 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5486 search at the associated MODE_INT.
5488 2018-02-20 Jeff Law <law@redhat.com>
5491 PR tree-optimization/81592
5493 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5494 for range data rather than using global data.
5495 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5496 range data rather than using global data.
5497 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5498 pass it to children as needed.
5499 (struct directive::fmtresult): Similarly.
5500 (struct directive::set_width): Similarly.
5501 (struct directive::set_precision): Similarly.
5502 (format_integer, format_directive, parse_directive): Similarly.
5503 (format_none): Accept unnamed vr_values parameter.
5504 (format_percent, format_floating, format_character): Similarly.
5505 (format_string, format_plain): Similarly.
5506 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5507 the EVRP range analyzer for range data rather than using global data.
5508 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5509 gimple-ssa-evrp-analyze.h
5510 (class sprintf_dom_walker): Add after_dom_children member function.
5511 Add evrp_range_analyzer member.
5512 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5513 range analyzer as needed.
5514 (sprintf_dom_walker::after_dom_children): New member function.
5515 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5517 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5518 (evrp_range_analyzer::pop_to_marker): Likewise.
5520 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5522 PR tree-optimization/84419
5523 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5524 with the required type if its current type is compatible but
5527 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5530 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5531 after vectorization.
5533 2018-02-20 Martin Liska <mliska@suse.cz>
5536 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5537 possible values if we don't have a hint.
5539 2018-02-20 Martin Liska <mliska@suse.cz>
5543 * final.c (shorten_branches): Build align_tab array with one
5545 * opts.c (finish_options): Add alignment option limit check.
5546 (MAX_CODE_ALIGN): Likewise.
5547 (MAX_CODE_ALIGN_VALUE): Likewise.
5548 * doc/invoke.texi: Document maximum allowed option value for
5549 all -falign-* options.
5551 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5554 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5555 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5556 * var-tracking.c (emit_note_insn_var_location): Remove all references
5557 to NOTE_INSN_CALL_ARG_LOCATION.
5558 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5559 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5560 Use copy_rtx_if_shared.
5561 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5562 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5563 (dwarf2out_var_location): Remove handling of
5564 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5566 * final.c (final_scan_insn): Remove all references to
5567 NOTE_INSN_CALL_ARG_LOCATION.
5568 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5569 before dumping final insns.
5570 * except.c (emit_note_eh_region_end): Remove all references to
5571 NOTE_INSN_CALL_ARG_LOCATION.
5572 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5573 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5574 * config/arc/arc.c (hwloop_optimize): Likewise.
5575 * config/arm/arm.c (create_fix_barrier): Likewise.
5576 * config/s390/s390.c (s390_chunkify_start): Likewise.
5577 * config/sh/sh.c (find_barrier): Likewise.
5578 * config/i386/i386.c (rest_of_insert_endbranch,
5579 ix86_seh_fixup_eh_fallthru): Likewise.
5580 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5581 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5582 * config/frv/frv.c (frv_function_prologue): Likewise.
5583 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5585 (note_outside_basic_block_p): Remove all references to
5586 NOTE_INSN_CALL_ARG_LOCATION.
5587 * gengtype.c (adjust_field_rtx_def): Likewise.
5588 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5590 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5591 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5594 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5597 PR tree-optimization/84452
5598 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5599 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5602 2018-02-19 Martin Liska <mliska@suse.cz>
5605 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5607 2018-02-19 Martin Liska <mliska@suse.cz>
5608 Richard Sandiford <richard.sandiford@linaro.org>
5610 PR tree-optimization/82491
5611 * gimple-fold.c (get_base_constructor): Make earlier bail out
5614 2018-02-19 Carl Love <cel@us.ibm.com>
5616 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5617 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5619 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5620 P8V_BUILTIN_VEC_NEG.
5622 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5624 * config/rl78/rl78.md (movdf): New define expand.
5626 2018-02-19 Martin Liska <mliska@suse.cz>
5629 * doc/invoke.texi: Fix typo.
5630 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5632 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5634 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5635 handle rs6000_single_float and rs6000_double_float specially for
5638 2018-02-16 Jeff Law <law@redhat.com>
5640 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5643 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5646 * ipa-inline.c (inline_small_functions): Fix a typo.
5648 2018-02-16 Nathan Sidwell <nathan@acm.org>
5650 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5652 2018-02-16 Carl Love <cel@us.ibm.com>
5654 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5655 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5656 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5657 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5658 expansion to P8V_BUILTIN_VEC_FLOAT2.
5660 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5662 PR rtl-optimization/70023
5663 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5664 src_regno into account.
5666 2018-02-16 Carl Love <cel@us.ibm.com>
5668 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5669 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5670 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5671 * config/rs6000/rs6000.c: Remove case statements for
5672 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5673 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5674 and P9V_BUILTIN_VEC_VINSERT4B.
5675 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5676 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5677 * config/rs6000/vsx.md:
5678 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5681 2018-02-16 Carl Love <cel@us.ibm.com>
5683 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5685 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5687 * config/rs6000/rs6000-c.c: Add the definitions for
5688 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5689 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5690 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5691 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5692 definition for insert4b and define insn *insert3b_internal.
5693 * doc/extend.texi: Add documentation for vec_extract4b.
5695 2018-02-16 Nathan Sidwell <nathan@acm.org>
5697 * doc/extend.texi (Backwards Compatibility): Mention friend
5698 injection. Note for-scope is deprecated.
5699 * doc/invoke.texi (-ffriend-injection): Deprecate.
5701 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5703 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5704 that moved to I2, also allow destinations that are a paradoxical
5705 subreg (instead of a normal reg).
5707 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5710 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5713 2018-02-16 Richard Biener <rguenther@suse.de>
5715 PR tree-optimization/84037
5716 PR tree-optimization/84016
5718 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5719 Adjust vec_construct for the fact we need additional higher latency
5720 128bit inserts for AVX256 and AVX512 vector builds.
5721 (ix86_add_stmt_cost): Scale vector construction cost for
5724 2018-02-16 Richard Biener <rguenther@suse.de>
5726 PR tree-optimization/84417
5727 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5728 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5729 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5731 2018-02-16 Martin Liska <mliska@suse.cz>
5734 * internal-fn.def (ASAN_CHECK): Set proper flags.
5735 (ASAN_MARK): Likewise.
5737 2018-02-16 Julia Koval <julia.koval@intel.com>
5739 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5740 from PTA_CANNONLAKE.
5742 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5745 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5746 Use ++iter rather than iter++ for std::list iterators.
5747 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5748 defer deleting them until all nodes in the forest are processed. Do
5749 free even leaf nodes. Change to_process into auto_vec.
5752 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5753 * vec.h (vec_default_construct): Use memset instead of placement new
5754 if BROKEN_VALUE_INITIALIZATION is defined.
5755 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5756 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5759 PR rtl-optimization/83723
5760 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5761 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5762 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5764 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5766 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5768 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5770 PR rtl-optimization/81443
5771 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5772 from inner REGs to paradoxical SUBREGs.
5774 2018-02-16 Richard Biener <rguenther@suse.de>
5776 PR tree-optimization/84399
5777 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5778 For operands we can analyze at their definition make sure we can
5779 analyze them at each use as well.
5781 2018-02-16 Richard Biener <rguenther@suse.de>
5783 PR tree-optimization/84190
5784 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5785 volatile accesses if the decl isn't volatile.
5787 2018-02-15 Jason Merrill <jason@redhat.com>
5789 PR c++/84314 - ICE with templates and fastcall attribute.
5790 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5791 TYPE_CANONICAL on an existing type.
5793 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5795 PR tree-optimization/84383
5796 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5797 dstoff nor call operand_equal_p if dstbase is NULL.
5799 PR tree-optimization/84334
5800 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5801 also a CONSTANT_CLASS_P, punt.
5803 2018-02-14 Jim Wilson <jimw@sifive.com>
5805 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5806 first SMALL_OPERAND check. New local min_second_step. Move assert
5807 to where locals are set. Add TARGET_RVC support.
5808 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5810 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5812 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5814 2018-02-14 Martin Sebor <msebor@redhat.com>
5816 PR tree-optimization/83698
5817 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5818 arrays constrain the offset range to their bounds.
5819 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5820 (builtin_access::overlap): Avoid setting the size of overlap if it's
5822 (maybe_diag_overlap): Also consider arrays when deciding what values
5823 of offsets to include in diagnostics.
5825 2018-02-14 Martin Sebor <msebor@redhat.com>
5828 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5829 that correspond to the kind of a declaration.
5831 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5834 * config/pa/pa.md: Load address of PIC label using the linkage table
5835 if the label is nonlocal.
5837 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5839 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5840 warning message if user requests -maltivec=be.
5841 * doc/invoke.texi: Document deprecation of -maltivec=be.
5843 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5846 * config/rs6000/rs6000-c.c: Update definitions for
5847 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5848 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5850 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5853 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5854 add _get_ssp intrinsics. Remove argument from
5855 __builtin_ia32_rdssp[d|q].
5856 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5857 * config/i386/i386-builtin.def: Remove argument from
5858 __builtin_ia32_rdssp[d|q].
5859 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5860 ix86_expand_special_args_builtin for _rdssp[d|q].
5861 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5862 Clear register before usage.
5863 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5864 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5866 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5868 PR tree-optimization/84357
5869 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5870 operand 1 of an ARRAY_REF too.
5872 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5875 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5876 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5878 (set_of_reg): New struct.
5879 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5880 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5881 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5883 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5884 Split into bitclr, bitset, bitinvert patterns if appropriate.
5885 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5886 use rx_fuse_in_memory_bitop.
5887 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5888 to named insn, correct maximum insn length.
5890 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5893 * machmode.def: Define a complex mode for PARTIAL_INT.
5894 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5896 * doc/rtl.texi: Document CSPImode.
5897 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5899 (msp430_hard_regno_nregs_with_padding): Likewise.
5901 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5904 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5906 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5908 PR rtl-optimization/84169
5909 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5910 we generated a parallel as new i3 and we split that to new i2 and i3
5911 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5912 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5913 those to i2, not i1. Partially rewrite this scan code.
5915 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5918 * stor-layout.c (place_field): For variable length fields, adjust
5919 offset_align afterwards not just based on the field's alignment,
5920 but also on the size.
5923 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5924 of exps and logs in the use_exp2 case.
5926 2018-02-13 Jeff Law <law@redhat.com>
5928 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5931 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5934 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5938 * common.opt (gas-loc-support, gas-locview-support): New.
5939 (ginline-points, ginternal-reset-location-views): New.
5940 * doc/invoke.texi: Document them. Use @itemx where intended.
5941 (gvariable-location-views): Adjust.
5942 * target.def (reset_location_view): New.
5943 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5944 (TARGET_RESET_LOCATION_VIEW): New.
5945 * doc/tm.texi: Rebuilt.
5946 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5947 (dwarf2out_default_as_locview_support): New.
5948 (output_asm_line_debug_info): Use option variables.
5949 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5950 (output_loc_list): Likewise.
5951 (add_high_low_attributes): Check option variables.
5952 Don't output entry view attribute in strict mode.
5953 (gen_inlined_subroutine_die): Check option variables.
5954 (dwarf2out_inline_entry): Likewise.
5955 (init_sections_and_labels): Likewise.
5956 (dwarf2out_early_finish): Likewise.
5957 (maybe_reset_location_view): New, from...
5958 (dwarf2out_var_location): ... here. Call it.
5959 * debug.h (dwarf2out_default_as_loc_support): Declare.
5960 (dwarf2out_default_as_locview_support): Declare.
5961 * hooks.c (hook_int_rtx_insn_0): New.
5962 * hooks.h (hook_int_rtx_insn_0): Declare.
5963 * toplev.c (process_options): Take -gas-loc-support and
5964 -gas-locview-support from dwarf2out. Enable
5965 -gvariable-location-views by default only with locview
5966 assembler support. Enable -ginternal-reset-location-views by
5967 default only if the target defines the corresponding hook.
5968 Enable -ginline-points by default if location views are
5969 enabled; force it disabled if statement frontiers are
5971 * tree-inline.c (expand_call_inline): Check option variables.
5972 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5974 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5976 PR tree-optimization/84321
5977 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5978 handling. Also check whether the anti-range contains any values
5979 that satisfy the mask; switch to a VR_RANGE if not.
5981 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5984 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5986 2018-02-13 Martin Jambor <mjambor@suse.cz>
5989 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5990 of call statements, also set location of a load to a temporary.
5992 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5994 * config/rl78/rl78.c (add_vector_labels): New function.
5995 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5996 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5997 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5998 which checks that no arguments are passed.
5999 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6000 * doc/extend.texi: Documentation for the new attribute.
6002 2018-02-13 Andreas Schwab <schwab@suse.de>
6004 * config/riscv/linux.h (CPP_SPEC): Define.
6006 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6009 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6010 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6011 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6012 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6013 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6014 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6015 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6016 temporarily for AES and PCLMUL builtins.
6018 PR tree-optimization/84339
6019 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6020 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6024 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6025 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6026 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6028 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6030 * omp-simd-clone.h: New file.
6031 * omp-simd-clone.c: Include omp-simd-clone.h.
6032 (expand_simd_clones): No longer static.
6033 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6034 cgraph.h and omp-simd-clone.h.
6035 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6036 (vect_recog_widen_shift_pattern): Formatting fix.
6037 (vect_pattern_recog_1): Don't check optab for calls.
6040 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6041 operands[2] into a REG before using gen_lowpart on it.
6043 2018-02-12 Jeff Law <law@redhat.com>
6046 * config/sh/sh.c (find_barrier): Consider a sibling call
6049 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6050 successfully back substituting a reg.
6052 2018-02-12 Richard Biener <rguenther@suse.de>
6054 PR tree-optimization/84037
6055 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6056 parameter, move visited init to caller.
6057 (vect_slp_analyze_operations): Separate cost from validity
6058 check, initialize visited once for all instances.
6059 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6061 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6063 (vect_model_promotion_demotion_cost): Likewise.
6064 (vectorizable_bswap): Guard cost modeling with !slp_node
6065 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6067 (vectorizable_call): Likewise.
6068 (vectorizable_conversion): Likewise.
6069 (vectorizable_assignment): Likewise.
6070 (vectorizable_shift): Likewise.
6071 (vectorizable_operation): Likewise.
6072 (vectorizable_store): Likewise.
6073 (vectorizable_load): Likewise.
6074 (vectorizable_condition): Likewise.
6075 (vectorizable_comparison): Likewise.
6077 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6080 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6081 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6082 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6084 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6087 * params.def (inline-min-speedup): Increase from 8 to 15.
6088 (max-inline-insns-auto): Decrease from 40 to 30.
6089 * ipa-split.c (consider_split): Add some buffer for function to
6090 be considered inlining candidate.
6091 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6094 2018-02-12 Richard Biener <rguenther@suse.de>
6096 PR tree-optimization/84037
6097 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6098 matched stmts if we cannot swap the non-matched ones.
6100 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6102 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6103 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6104 _mm_maskz_scalef_round_ss): New intrinsics.
6105 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6106 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6107 __builtin_ia32_scalefss_round): Remove.
6108 (__builtin_ia32_scalefsd_mask_round,
6109 __builtin_ia32_scalefss_mask_round): New intrinsics.
6110 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6111 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6112 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6113 "<round_constraint>")): Changed to ...
6114 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6115 "<round_scalar_constraint>")): ... this.
6116 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6117 %0, %1, %2<round_op3>}"): Changed to ...
6118 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6119 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6120 %2<round_scalar_mask_op3>}"): ... this.
6121 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6123 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6125 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6126 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6127 (_mm_maskz_sqrt_round_ss): New intrinsics.
6128 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6129 (__builtin_ia32_sqrtsd_mask_round)
6130 (__builtin_ia32_sqrtss_mask_round): New builtins.
6131 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6132 (__builtin_ia32_sqrtss_round): Remove.
6133 (__builtin_ia32_sqrtsd_mask_round)
6134 (__builtin_ia32_sqrtss_mask_round): New builtins.
6135 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6136 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6137 ((match_operand:VF_128 1 "vector_operand"
6138 "xBm,<round_constraint>")): Changed to ...
6139 ((match_operand:VF_128 1 "vector_operand"
6140 "xBm,<round_scalar_constraint>")): ... this.
6141 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6142 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6143 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6144 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6145 %<iptr>1<round_scalar_mask_op3>}): ... this.
6146 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6147 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6149 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6152 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6153 Cast vec_cmpeq result to correct type.
6154 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6155 Cast vec_cmpgt result to correct type.
6157 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6159 * final.c (final_scan_insn_1): Renamed from...
6160 (final_scan_insn): ... this. New wrapper, to recover
6161 seen from the outermost call in recursive ones.
6162 * config/sparc/sparc.c (output_return): Drop seen from call.
6163 (output_sibcall): Likewise.
6164 * config/visium/visium.c (output_branch): Likewise.
6166 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6168 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6171 2018-02-10 Alan Modra <amodra@gmail.com>
6174 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6175 Specify LR as an input.
6177 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6180 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6181 remove_member_access_dummy_vars): New functions.
6182 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6183 lower_omp_1, execute_lower_omp): Use them.
6185 PR rtl-optimization/84308
6186 * shrink-wrap.c (spread_components): Release todo vector.
6188 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6190 PR rtl-optimization/57193
6191 * ira-color.c (struct allocno_color_data): Add member
6192 conflict_allocno_hard_prefs.
6193 (update_conflict_allocno_hard_prefs): New.
6194 (bucket_allocno_compare_func): Add a preference based on
6195 conflict_allocno_hard_prefs.
6196 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6197 (color_allocnos): Remove a dead code. Initiate
6198 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6200 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6203 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6204 constraint from =wa to wa. Avoid a subreg on the output operand,
6205 instead use a pseudo and subreg it in a move.
6206 (p9_xxbrd_<mode>): Changed to ...
6207 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6208 (p9_xxbrd_v2df): New expander.
6209 (p9_xxbrw_<mode>): Changed to ...
6210 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6211 (p9_xxbrw_v4sf): New expander.
6213 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6215 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6217 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6220 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6221 multiply in 32-bit mode.
6222 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6223 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6226 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6228 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6229 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6230 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6231 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6233 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6236 * dwarf2out.c (is_trivial_indirect_ref): New function.
6237 (dwarf2out_late_global_decl): Do not generate a location
6238 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6239 and that are not defined in the current unit.
6241 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6243 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6244 instead of a libcall for UNORDERED.
6246 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6249 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6250 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6252 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6255 * config/s390/s390.c (s390_set_current_function): Invoke
6256 s390_indirect_branch_settings also if fndecl didn't change.
6258 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6260 * config/rs6000/rs6000.md (blockage): Set length to zero.
6262 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6264 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6266 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6269 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6270 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6274 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6275 PARALLEL incoming that failed vt_get_decl_and_offset check.
6278 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6279 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6281 (get_variable_section): For decls in named .bss* sections pass true as
6282 second argument to bss_initializer_p.
6284 2018-02-09 Marek Polacek <polacek@redhat.com>
6285 Jakub Jelinek <jakub@redhat.com>
6288 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6289 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6290 Sync some changes from cxx_fold_indirect_ref.
6292 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6294 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6296 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6297 (BLOCK_INLINE_ENTRY_LABEL): New.
6298 (dwarf2out_var_location): Disregard inline entry markers.
6299 (inline_entry_data): New struct.
6300 (inline_entry_data_hasher): New hashtable type.
6301 (inline_entry_data_hasher::hash): New.
6302 (inline_entry_data_hasher::equal): New.
6303 (inline_entry_data_table): New variable.
6304 (add_high_low_attributes): Add DW_AT_entry_pc and
6305 DW_AT_GNU_entry_view attributes if a pending entry is found
6306 in inline_entry_data_table. Add old entry_pc attribute only
6307 if debug nonbinding markers are disabled.
6308 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6309 markers are enabled.
6310 (block_within_block_p, dwarf2out_inline_entry): New.
6311 (dwarf2out_finish): Check that no entries remained in
6312 inline_entry_data_table.
6313 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6314 (final_scan_insn, notice_source_line): Likewise.
6315 (rest_of_clean_state): Skip inline entry markers.
6316 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6318 * gimple.c (gimple_build_debug_inline_entry): New.
6319 * gimple.h (enum gimple_debug_subcode): Add
6320 GIMPLE_DEBUG_INLINE_ENTRY.
6321 (gimple_build_debug_inline_entry): Declare.
6322 (gimple_debug_inline_entry_p): New.
6323 (gimple_debug_nonbind_marker_p): Adjust.
6324 * insn-notes.def (INLINE_ENTRY): New.
6325 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6326 inline entry marker notes.
6327 (print_insn): Likewise.
6328 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6329 (INSN_DEBUG_MARKER_KIND): Likewise.
6330 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6331 * tree-inline.c (expand_call_inline): Build and insert
6332 debug_inline_entry stmt.
6333 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6334 inline entry blocks early, if nonbind markers are enabled.
6335 (dump_scope_block): Dump fragment info.
6336 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6337 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6338 (gimple_build_debug_inline_entry): New.
6339 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6340 Enable/disable inline entry points too.
6341 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6342 (DEBUG_INSN): Describe inline entry markers.
6344 * common.opt (gvariable-location-views): New.
6345 (gvariable-location-views=incompat5): New.
6346 * config.in: Rebuilt.
6347 * configure: Rebuilt.
6348 * configure.ac: Test assembler for view support.
6349 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6350 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6351 * dwarf2out.c (var_loc_view): New typedef.
6352 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6353 (dwarf2out_locviews_in_attribute): New.
6354 (dwarf2out_locviews_in_loclist): New.
6355 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6356 (enum dw_line_info_opcode): Add LI_adv_address.
6357 (struct dw_line_info_table): Add view.
6358 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6359 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6360 (zero_view_p): New variable.
6361 (ZERO_VIEW_P): New macro.
6362 (output_asm_line_debug_info): New.
6363 (struct var_loc_node): Add view.
6364 (add_AT_view_list, AT_loc_list): New.
6365 (add_var_loc_to_decl): Add view param. Test it against last.
6366 (new_loc_list): Add view params. Record them.
6367 (AT_loc_list_ptr): Handle loc and view lists.
6368 (view_list_to_loc_list_val_node): New.
6369 (print_dw_val): Handle dw_val_class_view_list.
6370 (size_of_die): Likewise.
6371 (value_format): Likewise.
6372 (loc_list_has_views): New.
6373 (gen_llsym): Set vl_symbol too.
6374 (maybe_gen_llsym, skip_loc_list_entry): New.
6375 (dwarf2out_maybe_output_loclist_view_pair): New.
6376 (output_loc_list): Output view list or entries too.
6377 (output_view_list_offset): New.
6378 (output_die): Handle dw_val_class_view_list.
6379 (output_dwarf_version): New.
6380 (output_compilation_unit_header): Use it.
6381 (output_skeleton_debug_sections): Likewise.
6382 (output_rnglists, output_line_info): Likewise.
6383 (output_pubnames, output_aranges): Update version comments.
6384 (output_one_line_info_table): Output view numbers in asm comments.
6385 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6386 Call maybe_gen_llsym.
6387 (loc_list_from_tree_1): Adjust.
6388 (add_AT_location_description): Create view list attribute if
6389 needed, check it's absent otherwise.
6390 (convert_cfa_to_fb_loc_list): Adjust.
6391 (maybe_emit_file): Call output_asm_line_debug_info for test.
6392 (dwarf2out_var_location): Reset views as needed. Precompute
6393 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6394 attribute. Set view.
6395 (new_line_info_table): Reset next view.
6396 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6397 (dwarf2out_source_line): Likewise. Output view resets and labels to
6398 the assembler, or select appropriate line info opcodes.
6399 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6400 (optimize_string_length): Catch it. Adjust.
6401 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6402 dw_val_class_view_list, and remove it if no longer needed.
6403 (hash_loc_list): Hash view numbers.
6404 (loc_list_hasher::equal): Compare them.
6405 (optimize_location_lists): Check whether a view list symbol is
6406 needed, and whether the locview attribute is present, and
6407 whether they match. Remove the locview attribute if no longer
6409 (index_location_lists): Call skip_loc_list_entry for test.
6410 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6411 Use output_dwarf_version.
6412 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6413 (struct dw_val_node): Add val_view_list.
6414 * final.c (SEEN_NEXT_VIEW): New.
6415 (set_next_view_needed): New.
6416 (clear_next_view_needed): New.
6417 (maybe_output_next_view): New.
6418 (final_start_function): Rename to...
6419 (final_start_function_1): ... this. Take pointer to FIRST,
6420 add SEEN parameter. Emit param bindings in the initial view.
6421 (final_start_function): Reintroduce SEEN-less interface.
6422 (final): Rename to...
6423 (final_1): ... this. Take SEEN parameter. Output final pending
6424 next view at the end.
6425 (final): Reintroduce seen-less interface.
6426 (final_scan_insn): Output pending next view before switching
6427 sections or ending a block. Mark the next view as needed when
6428 outputting variable locations. Notify debug backend of section
6429 changes, and of location view changes.
6430 (rest_of_handle_final): Adjust.
6431 * toplev.c (process_options): Autodetect value for debug variable
6432 location views option. Warn on incompat5 without -gdwarf-5.
6433 * doc/invoke.texi (gvariable-location-views): New.
6434 (gvariable-location-views=incompat5): New.
6435 (gno-variable-location-views): New.
6437 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6439 PR tree-optimization/84136
6440 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6441 that the result of find_edge is non-NULL.
6443 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6446 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6447 storing integer register in SImode. Fix cost of 256 and 512
6448 byte aligned SSE register store.
6450 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6452 * config/i386/i386.c (ix86_multiplication_cost): Fix
6453 multiplication cost for TARGET_AVX512DQ.
6455 2018-02-08 Marek Polacek <polacek@redhat.com>
6457 PR tree-optimization/84238
6458 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6461 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6463 PR tree-optimization/84265
6464 * tree-vect-stmts.c (vectorizable_store): Don't treat
6465 VMAT_CONTIGUOUS accesses as grouped.
6466 (vectorizable_load): Likewise.
6468 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6470 PR tree-optimization/81635
6471 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6472 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6473 (test_round_for_mask): New functions.
6474 (wide_int_cc_tests): Call test_round_for_mask.
6475 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6476 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6477 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6478 range returned by get_range_info.
6480 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6483 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6484 * symtab.c: Include builtins.h
6485 (symtab_node::output_to_lto_symbol_table_p): Move here
6486 from lto-streamer-out.c:output_symbol_p.
6487 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6488 (output_symbol_p): Move all logic to symtab.c
6489 (produce_symtab): Update.
6491 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6493 * config/s390/s390-opts.h (enum indirect_branch): Define.
6494 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6495 (s390_indirect_branch_via_thunk)
6496 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6497 (enum s390_indirect_branch_type): Define.
6498 * config/s390/s390.c (struct s390_frame_layout, struct
6499 machine_function): Remove.
6500 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6501 (indirect_branch_table_label_no, indirect_branch_table_name):
6503 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6504 (enum s390_indirect_branch_option): Define.
6505 (s390_return_addr_from_memory): New function.
6506 (s390_handle_string_attribute): New function.
6507 (s390_attribute_table): Add new attribute handler.
6508 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6509 (s390_indirect_branch_via_thunk): New function.
6510 (s390_indirect_branch_via_inline_thunk): New function.
6511 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6512 sibling call optimization for non z10 compiles.
6513 (s390_emit_call): Force indirect branch target to be a single
6514 register. Add r1 clobber for non-z10 compiles.
6515 (s390_emit_epilogue): Emit return jump via return_use expander.
6516 (s390_reorg): Handle JUMP_INSNs as execute targets.
6517 (s390_option_override_internal): Perform validity checks for the
6518 new command line options.
6519 (s390_indirect_branch_attrvalue): New function.
6520 (s390_indirect_branch_settings): New function.
6521 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6522 (s390_output_indirect_thunk_function): New function.
6523 (s390_code_end): Implement target hook.
6524 (s390_case_values_threshold): Implement target hook.
6525 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6527 * config/s390/s390.h (struct s390_frame_layout)
6528 (struct machine_function): Move here from s390.c.
6529 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6530 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6531 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6532 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6533 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6534 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6535 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6536 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6537 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6538 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6539 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6540 (mnemonic attribute): Add values which aren't recognized
6542 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6543 pattern for branch conversion. Fix mnemonic attribute.
6544 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6545 indirect branch via thunk if requested.
6546 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6547 ("*indirect_jump"): Disable for branch conversion using out of
6549 ("indirect_jump_via_thunk<mode>_z10")
6550 ("indirect_jump_via_thunk<mode>")
6551 ("indirect_jump_via_inlinethunk<mode>_z10")
6552 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6553 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6554 ("casesi_jump_via_inlinethunk<mode>_z10")
6555 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6556 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6557 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6558 ("*indirect2_jump"): Disable for branch conversion.
6559 ("casesi_jump"): Turn into expander and expand patterns for branch
6561 ("return_use"): New expander.
6562 ("*return"): Emit return via thunk and rename it to ...
6563 ("*return<mode>"): ... this one.
6564 * config/s390/s390.opt: Add new options and and enum for the
6567 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6569 * lra-constraints.c (match_reload): Unconditionally use
6570 gen_lowpart_SUBREG, rather than selecting between that
6571 and equivalent gen_rtx_SUBREG code.
6573 2018-02-08 Richard Biener <rguenther@suse.de>
6575 PR tree-optimization/84233
6576 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6577 changed flag instead of boguously re-using phi_inserted.
6579 2018-02-08 Martin Jambor <mjambor@suse.cz>
6581 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6582 static local variables.
6584 2018-02-08 Richard Biener <rguenther@suse.de>
6586 PR tree-optimization/84278
6587 * tree-vect-stmts.c (vectorizable_store): When looking for
6588 smaller vector types to perform grouped strided loads/stores
6589 make sure the mode is supported by the target.
6590 (vectorizable_load): Likewise.
6592 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6594 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6595 Increase LDP/STP opportunities by adding adjacent callee-saves.
6597 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6599 PR rtl-optimization/84068
6600 PR rtl-optimization/83459
6601 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6603 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6605 PR tree-optimization/84224
6606 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6607 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6610 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6613 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6614 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6615 count, remove one USE.
6617 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6619 * doc/install.texi (Configuration): Document the
6620 --with-long-double-format={ibm,ieee} PowerPC configuration
6624 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6625 Convert from define_expand to be define_insn_and_split. Rework
6626 float/double/_Float128 conversions to QI/HI/SImode to work with
6627 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6628 conversions to QI/HImode types did a store and then a load to
6629 truncate the value. For conversions to VSX registers, don't split
6630 the insn, instead emit the code directly. Use the code iterator
6631 any_fix to combine signed and unsigned conversions.
6632 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6633 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6634 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6635 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6636 (fix_<mode>di2_hw): Likewise.
6637 (fixuns_<mode>di2_hw): Likewise.
6638 (fix_<mode>si2_hw): Likewise.
6639 (fixuns_<mode>si2_hw): Likewise.
6640 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6641 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6642 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6643 fix<uns>_trunc<SFDF:mode>si2_p8.
6644 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6646 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6647 (fix<uns>_<mode>_mem): Likewise.
6648 (fctiw<u>z_<mode>_mem): Likewise.
6649 (fix<uns>_<mode>_mem): Likewise.
6650 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6651 the register allocator from doing a direct move to the GPRs to do
6652 a store, and instead use the ISA 3.0 store byte/half-word from
6653 vector register instruction. For IEEE 128-bit floating point,
6654 also optimize stores of 32-bit ints.
6655 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6657 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6659 * genextract.c (push_pathstr_operand): New function to support
6661 (walk_rtx): Call push_pathstr_operand.
6662 (print_path): Support [a-zA-Z].
6664 2018-02-07 Richard Biener <rguenther@suse.de>
6666 PR tree-optimization/84037
6667 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6668 (cse_and_gimplify_to_preheader): Declare.
6669 (vect_get_place_in_interleaving_chain): Likewise.
6670 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6672 (_loop_vec_info::~_loop_vec_info): Delete it.
6673 (cse_and_gimplify_to_preheader): New function.
6674 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6675 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6676 (vectorizable_load): Likewise. For grouped stores always base
6677 the IV on the first element.
6678 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6679 condition before gimplifying.
6681 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6683 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6684 *DIV_EXPR and *MOD_EXPR.
6686 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6689 * config/i386/i386.c (ix86_option_override_internal): Mask out
6690 the CF_SET bit when checking -fcf-protection.
6692 2018-02-07 Tom de Vries <tom@codesourcery.com>
6695 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6698 2018-02-07 Richard Biener <rguenther@suse.de>
6700 PR tree-optimization/84204
6701 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6704 PR tree-optimization/84205
6705 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6706 special-case isl_ast_op_zdiv_r.
6708 PR tree-optimization/84223
6709 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6710 Only add conditions from within the region.
6711 (gather_bbs::after_dom_children): Adjust.
6713 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6716 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6717 * config/avr/avr.md: Only post-reload split REG-REG moves if
6718 either register is GENERAL_REG_P.
6720 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6722 PR tree-optimization/84235
6723 * tree-ssa-scopedtables.c
6724 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6725 if the subtraction is performed in floating point type where NaNs are
6726 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6727 build 1. Formatting fix.
6729 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6732 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6733 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6734 and skip it regardless of bb boundaries. Use CALL_P macro,
6735 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6736 unnecessarily, formatting fix.
6738 2018-02-06 Michael Collison <michael.collison@arm.com>
6740 * config/arm/thumb2.md:
6741 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6742 (*thumb_mov_notscc): Ditto.
6744 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6747 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6748 unsigned_fix, not "s".
6750 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6752 * configure.ac (gcc_fn_eh_frame_ro): New function.
6753 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6754 correct .eh_frame permissions.
6755 * configure: Regenerate.
6757 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6759 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6762 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6764 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6765 Display warning message for -mno-speculate-indirect-jumps.
6767 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6769 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6771 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6773 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6775 PR tree-optimization/84225
6776 * tree-eh.c (find_trapping_overflow): Only call
6777 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6779 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6782 * config/i386/i386.c: Reimplement the check of possible options
6783 -mibt/-mshstk conbination. Change error messages.
6784 * doc/invoke.texi: Fix a typo: remove extra '='.
6786 2018-02-06 Marek Polacek <polacek@redhat.com>
6788 PR tree-optimization/84228
6789 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6791 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6794 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6795 emitted arch directives.
6796 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6797 __ARM_FEATURE_COPROC before changing architectures.
6799 2018-02-06 Richard Biener <rguenther@suse.de>
6801 * config/i386/i386.c (print_reg): Fix typo.
6802 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6804 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6806 * configure: Regenerate.
6808 2018-02-05 Martin Sebor <msebor@redhat.com>
6810 PR tree-optimization/83369
6811 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6814 2018-02-05 Martin Liska <mliska@suse.cz>
6816 * doc/invoke.texi: Cherry-pick upstream r323995.
6818 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6820 * ira.c (ira_init_register_move_cost): Adjust comment.
6822 2018-02-05 Martin Liska <mliska@suse.cz>
6824 PR gcov-profile/84137
6825 * doc/gcov.texi: Fix typo in documentation.
6827 2018-02-05 Martin Liska <mliska@suse.cz>
6829 PR gcov-profile/83879
6830 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6831 using dlopen functionality.
6833 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6835 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6836 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6837 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6838 _mm_maskz_range_round_ss): New intrinsics.
6839 (__builtin_ia32_rangesd128_round)
6840 (__builtin_ia32_rangess128_round): Remove.
6841 (__builtin_ia32_rangesd128_mask_round,
6842 __builtin_ia32_rangess128_mask_round): New builtins.
6843 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6844 __builtin_ia32_rangess128_round): Remove.
6845 (__builtin_ia32_rangesd128_mask_round,
6846 __builtin_ia32_rangess128_mask_round): New builtins.
6847 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6848 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6849 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6850 "<round_saeonly_constraint>")): Changed to ...
6851 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6852 "<round_saeonly_scalar_constraint>")): ... this.
6853 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6854 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6855 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6856 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6857 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6859 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6861 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6863 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6864 Remove all values except native, 8540 and 8548.
6866 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6868 * config/i386/i386.c (ix86_output_function_return): Pass
6869 INVALID_REGNUM, instead of -1, as invalid register number to
6870 indirect_thunk_name and output_indirect_thunk.
6872 2018-02-02 Julia Koval <julia.koval@intel.com>
6874 * config.gcc: Add -march=icelake.
6875 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6876 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6877 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6878 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6879 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6880 (processor_target_table): Add icelake.
6881 (ix86_option_override_internal): Handle new PTAs.
6882 (get_builtin_code_for_version): Handle icelake.
6883 (M_INTEL_COREI7_ICELAKE): New.
6884 (fold_builtin_cpu): Handle icelake.
6885 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6886 * doc/invoke.texi: Add -march=icelake.
6888 2018-02-02 Julia Koval <julia.koval@intel.com>
6890 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6891 to wide_int_bitmask.
6892 * wide-int-bitmask.h: New.
6894 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6897 * config/i386/i386.md: Replace Pmode with word_mode in
6898 builtin_setjmp_setup and builtin_longjmp to support x32.
6900 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6904 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6906 (rs6000_supported_cpu_names): New static variable.
6907 (linux_cpu_translation_table): Likewise.
6908 (elf_platform) <cpu>: Define new static variable and use it.
6909 Translate kernel AT_PLATFORM name to canonical name if needed.
6910 Error if platform name is unknown.
6912 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6915 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6917 2018-02-01 Jeff Law <law@redhat.com>
6920 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6921 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6922 the scratch if RELEASE_VIA_POP is false.
6923 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6924 If we have to save a temporary register, decrement SIZE appropriately.
6925 Pass new arguments to release_scratch_register_on_entry.
6926 (ix86_adjust_stack_and_probe): Likewise.
6927 (ix86_emit_probe_stack_range): Pass new arguments to
6928 release_scratch_register_on_entry.
6930 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6932 PR rtl-optimization/84157
6933 * combine.c (change_zero_ext): Use REG_P predicate in
6934 front of HARD_REGISTER_P predicate.
6936 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6938 * config/avr/avr.c (avr_option_override): Move disabling of
6939 -fdelete-null-pointer-checks to...
6940 * common/config/avr/avr-common.c (avr_option_optimization_table):
6943 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6945 PR tree-optimization/81635
6946 * tree-data-ref.c (split_constant_offset_1): For types that
6947 wrap on overflow, try to use range info to prove that wrapping
6950 2018-02-01 Renlin Li <renlin.li@arm.com>
6953 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6955 (aarch64_register_move_cost): Likewise.
6956 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6958 (REG_CLASS_NAMES): Likewise.
6959 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6960 TAILCALL_ADDR_REGS. Remove IP registers.
6961 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6963 2018-02-01 Richard Biener <rguenther@suse.de>
6965 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6966 for specifying RPO order and allow NULL for that.
6967 * domwalk.c (dom_walker::dom_walker): Likewise.
6968 (dom_walker::walk): Handle NULL RPO order.
6969 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6971 (rewrite_update_dom_walker): Likewise.
6972 (mark_def_dom_walker): Likewise.
6974 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6976 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6977 (aarch64_maybe_expand_sve_subreg_move): Declare.
6978 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6979 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6981 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6982 that are semantically a reverse operation.
6983 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6984 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6985 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6987 (aarch64_can_change_mode_class): For big-endian, forbid changes
6988 between two SVE modes if they have different element sizes.
6990 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6992 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6993 the TImode handling for big-endian targets.
6995 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6997 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6998 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7000 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7001 Remove BSWAP handing for big-endian targets and use the form of
7002 LD1RQ appropariate for the mode.
7004 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7006 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7007 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7010 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7013 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7014 check for operands that need to go through aarch64_sve_reload_be.
7016 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7018 PR tree-optimization/81661
7019 PR tree-optimization/84117
7020 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7021 * tree-eh.c: Include gimplify.h.
7022 (find_trapping_overflow, replace_trapping_overflow,
7023 rewrite_to_non_trapping_overflow): New functions.
7024 * tree-vect-loop.c: Include tree-eh.h.
7025 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7026 * tree-data-ref.c: Include tree-eh.h.
7027 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7029 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7031 PR rtl-optimization/84123
7032 * combine.c (change_zero_ext): Check if hard register satisfies
7033 can_change_dest_mode before calling gen_lowpart_SUBREG.
7035 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7038 * ira.c (ira_init_register_move_cost): Remove assert.
7040 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7042 PR rtl-optimization/84071
7043 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7044 * doc/tm.texi: Regenerate.
7046 2018-01-31 Richard Biener <rguenther@suse.de>
7048 PR tree-optimization/84132
7049 * tree-data-ref.c (analyze_miv_subscript): Properly
7050 check whether evolution_function_is_affine_multivariate_p
7051 before calling gcd_of_steps_may_divide_p.
7053 2018-01-31 Julia Koval <julia.koval@intel.com>
7056 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7057 * config/i386/i386.md (rdpid_rex64) New.
7058 (rdpid): Make 32bit only.
7060 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7063 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7064 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7066 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7069 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7071 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7073 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7075 PR rtl-optimization/84071
7076 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7077 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7079 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7081 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7082 (arc_attribute_table): Add 'aux' attribute.
7083 (arc_in_small_data_p): Consider aux like variables.
7084 (arc_is_aux_reg_p): New function.
7085 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7086 (arc_get_aux_arg): New function.
7087 (prepare_move_operands): Handle aux-register access.
7088 (arc_handle_aux_attribute): New function.
7089 * doc/extend.texi (ARC Variable attributes): Add subsection.
7091 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7093 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7094 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7095 (arc_attribute_table): Add 'uncached' attribute.
7096 (arc_print_operand): Print '.di' flag for uncached memory
7098 (arc_in_small_data_p): Do not consider for small data the uncached
7100 (arc_is_uncached_mem_p): New function.
7101 * config/arc/predicates.md (compact_store_memory_operand): Check
7102 for uncached memory accesses.
7103 (nonvol_nonimm_operand): Likewise.
7104 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7106 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7109 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7110 falign-loops=): Add Optimization flag.
7112 2018-01-30 Jeff Law <law@redhat.com>
7115 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7116 INT_REGISTERS_SAVED. Check it prior to calling
7117 get_scratch_register_on_entry.
7118 (ix86_adjust_stack_and_probe): Similarly.
7119 (ix86_emit_probe_stack_range): Similarly.
7120 (ix86_expand_prologue): Corresponding changes.
7122 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7125 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7126 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7128 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7131 * lra-constraints.c (curr_insn_transform): Process AND in the
7134 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7136 PR rtl-optimization/83986
7137 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7138 dependence against last_pending_memory_flush in addition to
7141 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7143 PR tree-optimization/81611
7144 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7147 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7150 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7153 2018-01-30 Richard Biener <rguenther@suse.de>
7154 Jakub Jelinek <jakub@redhat.com>
7156 PR tree-optimization/84111
7157 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7158 inner loops added during recursion, as they don't have up-to-date
7161 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7164 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7165 (can_inline_edge_by_limits_p): ... here.
7166 (can_early_inline_edge_p, check_callers,
7167 update_caller_keys, update_callee_keys, recursive_inlining,
7168 add_new_edges_to_heap, speculation_useful_p,
7169 inline_small_functions,
7170 inline_small_functions, flatten_function,
7171 inline_to_all_callers_1): Update.
7173 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7175 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7178 2018-01-30 Richard Biener <rguenther@suse.de>
7180 PR tree-optimization/83008
7181 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7182 invariant and constant vector uses in stmts when they need
7185 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7188 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7189 * configure: Regenerate.
7191 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7193 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7195 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7196 Use gen_rtx_REG rather than gen_lowpart.
7198 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7200 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7201 rather than 0 when creating partial subregs.
7203 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7205 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7208 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7211 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7212 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7213 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7214 flags. This restores the settings used before the 2017-07-24.
7215 Turning off pre increment/decrement/modify allows IVOPTS to
7216 optimize DF/SF loops where the index is an int.
7218 2018-01-29 Richard Biener <rguenther@suse.de>
7219 Kelvin Nilsen <kelvin@gcc.gnu.org>
7222 * tree-vect-stmts.c (vectorizable_call): Don't call
7223 targetm.vectorize_builtin_md_vectorized_function if callee is
7226 2018-01-22 Carl Love <cel@us.ibm.com>
7228 * doc/extend.tex: Fix typo in second arg in
7229 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7231 2018-01-29 Richard Biener <rguenther@suse.de>
7233 PR tree-optimization/84086
7234 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7235 (flush_ssaname_freelist): When SSA names were released reset
7236 the SCEV hash table.
7238 2018-01-29 Richard Biener <rguenther@suse.de>
7240 PR tree-optimization/84057
7241 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7242 removed paths when removing edges.
7244 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7246 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7247 -mfunction-return=@var{choice}.
7249 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7252 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7254 (layout::print_source_line): Likewise.
7255 (test_get_line_width_without_trailing_whitespace): Add test cases.
7257 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7260 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7263 2018-01-26 Jim Wilson <jimw@sifive.com>
7265 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7267 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7270 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7272 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7273 and CMP + SUB-immediate -> SUBS.
7275 2018-01-26 Martin Sebor <msebor@redhat.com>
7277 PR tree-optimization/83896
7278 * tree-ssa-strlen.c (get_string_len): Rename...
7279 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7280 Avoid assuming length is constant.
7281 (handle_char_store): Use HOST_WIDE_INT for string length.
7283 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7286 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7287 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7289 2018-01-26 Richard Biener <rguenther@suse.de>
7291 PR rtl-optimization/84003
7292 * dse.c (record_store): Only record redundant stores when
7293 the earlier store aliases at least all accesses the later one does.
7295 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7297 PR rtl-optimization/83985
7298 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7299 REG_CFA_RESTORE insns.
7300 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7301 REG_CFA_RESTORE insns here.
7304 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7305 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7307 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7309 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7310 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7311 (arc_init): Likewise.
7312 (arc_override_options): Likewise.
7313 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7315 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7317 * config/arc/arc.h (TARGET_DBNZ): Define.
7318 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7319 properly set the tune attribute.
7320 (dbnz): Use TARGET_DBNZ guard.
7321 * config/arc/arc.opt (mtune): Add core3 option.
7323 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7325 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7326 recognize new pic like addresses.
7327 (arc_delegitimize_address): Clean up.
7329 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7331 * config/arc/arc-arches.def: Option mrf16 valid for all
7333 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7334 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7335 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7336 * config/arc/arc-tables.opt: Regenerate.
7337 * config/arc/arc.c (arc_conditional_register_usage): Handle
7338 reduced register file case.
7339 (arc_file_start): Set must have build attributes.
7340 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7342 * config/arc/arc.opt (mrf16): Add new option.
7343 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7344 * config/arc/genmultilib.awk: Handle new mrf16 option.
7345 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7346 * config/arc/t-multilib: Regenerate.
7347 * doc/invoke.texi (ARC Options): Document mrf16 option.
7349 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7351 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7352 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7353 (arc_attribute_table): Add 'secure_call' attribute.
7354 (arc_print_operand): Print secure call operand.
7355 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7357 (arc_is_secure_call_p): New function. * config/arc/arc.md
7358 (call_i): Add support for sjli instruction.
7359 (call_value_i): Likewise.
7360 * config/arc/constraints.md (Csc): New constraint.
7362 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7363 John Eric Martin <John.Martin@emmicro-us.com>
7365 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7366 * config/arc/arc.c (_arc_jli_section): New struct.
7367 (arc_jli_section): New type.
7368 (rc_jli_sections): New static variable.
7369 (arc_handle_jli_attribute): New function.
7370 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7371 (arc_file_end): New function.
7372 (TARGET_ASM_FILE_END): Define.
7373 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7374 (arc_add_jli_section): New function.
7375 (jli_call_scan): Likewise.
7376 (arc_reorg): Call jli_call_scan.
7377 (arc_output_addsi): Remove 'S' from printing asm operand.
7378 (arc_is_jli_call_p): New function.
7379 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7381 (movhi_insn): Likewise.
7382 (movsi_insn): Likewise.
7383 (movsi_set_cc_insn): Likewise.
7384 (loadqi_update): Likewise.
7385 (load_zeroextendqisi_update): Likewise.
7386 (load_signextendqisi_update): Likewise.
7387 (loadhi_update): Likewise.
7388 (load_zeroextendhisi_update): Likewise.
7389 (load_signextendhisi_update): Likewise.
7390 (loadsi_update): Likewise.
7391 (loadsf_update): Likewise.
7392 (movsicc_insn): Likewise.
7393 (bset_insn): Likewise.
7394 (bxor_insn): Likewise.
7395 (bclr_insn): Likewise.
7396 (bmsk_insn): Likewise.
7397 (bicsi3_insn): Likewise.
7398 (cmpsi_cc_c_insn): Likewise.
7399 (movsi_ne): Likewise.
7400 (movsi_cond_exec): Likewise.
7401 (clrsbsi2): Likewise.
7412 (call_i): Remove 'S' asm letter, add jli instruction.
7413 (call_value_i): Likewise.
7414 * config/arc/arc.op (mjli-always): New option.
7415 * config/arc/constraints.md (Cji): New constraint.
7416 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7418 (subsf3_fpx): Likewise.
7419 (mulsf3_fpx): Likewise.
7420 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7422 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7424 * doc/invoke.texi (ARC): Document mjli-always option.
7426 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7428 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7429 avoid addition with 0 and use incw and decw where possible.
7431 2018-01-26 Richard Biener <rguenther@suse.de>
7433 PR tree-optimization/81082
7434 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7435 association if it requires casting to unsigned.
7436 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7437 from fold_plusminus_mult_expr to catch important cases late when
7438 range info is available.
7440 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7442 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7443 * configure.ac (hidden_linkonce): New test.
7444 * configure: Regenerate.
7445 * config.in: Regenerate.
7447 2018-01-26 Julia Koval <julia.koval@intel.com>
7449 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7450 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7451 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7452 _mm_mask_bitshuffle_epi64_mask): Fix type.
7453 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7454 USI_FTYPE_V4DI_V4DI_USI): Remove.
7455 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7456 __builtin_ia32_vpshufbitqmb256_mask,
7457 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7458 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7459 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7461 2018-01-26 Alan Modra <amodra@gmail.com>
7464 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7465 UNSPEC_VBPERMQ. Sort other unspecs.
7467 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7469 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7471 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7474 * predict.c (drop_profile): Do not push/pop cfun; update also
7476 (handle_missing_profiles): Fix logic looking for zero profiles.
7478 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7481 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7482 on functions with #pragma omp declare simd or functions with simd
7484 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7485 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7486 Remove trailing \n from warning_at calls.
7488 2018-01-25 Tom de Vries <tom@codesourcery.com>
7491 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7492 for neutered workers.
7494 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7497 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7498 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7500 2018-01-24 Jeff Law <law@redhat.com>
7503 * i386.c (get_probe_interval): Move to earlier point.
7504 (ix86_compute_frame_layout): If -fstack-clash-protection and
7505 the frame is larger than the probe interval, then use pushes
7506 to save registers rather than reg->mem moves.
7507 (ix86_expand_prologue): Remove conditional for int_registers_saved
7510 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7513 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7514 min/max for never referenced object.
7516 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7519 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7521 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7522 attributes from DECL_ATTRIBUTES (decl) without affecting
7523 DECL_ATTRIBUTES (current_function_decl).
7524 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7525 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7527 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7529 PR tree-optimization/83979
7530 * fold-const.c (fold_comparison): Use constant_boolean_node
7531 instead of boolean_{true,false}_node.
7533 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7535 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7538 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7540 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7541 Simplify the clause that sets the length attribute.
7542 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7543 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7544 clause that sets the length attribute.
7545 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7547 2018-01-24 Tom de Vries <tom@codesourcery.com>
7550 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7551 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7552 Add strict parameter.
7553 (prevent_branch_around_nothing): Insert dummy insn between branch to
7554 label and label with no ptx insn inbetween.
7555 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7557 2018-01-24 Tom de Vries <tom@codesourcery.com>
7560 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7561 for neutered threads in warp.
7562 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7564 2018-01-24 Richard Biener <rguenther@suse.de>
7566 PR tree-optimization/83176
7567 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7570 2018-01-24 Richard Biener <rguenther@suse.de>
7572 PR tree-optimization/82819
7573 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7574 code generating pluses that are no-ops in the target precision.
7576 2018-01-24 Richard Biener <rguenther@suse.de>
7579 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7581 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7583 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7584 to merge probabilities.
7585 * predict.c (probably_never_executed): Also mark as cold functions
7586 with global 0 profile and guessed local profile.
7587 * profile-count.c (profile_probability::combine_with_count): New
7589 * profile-count.h (profile_probability::operator*,
7590 profile_probability::operator*=, profile_probability::operator/,
7591 profile_probability::operator/=): Reduce precision to adjusted
7592 and set value to guessed on contradictory divisions.
7593 (profile_probability::combine_with_freq): Remove.
7594 (profile_probability::combine_wiht_count): Declare.
7595 (profile_count::force_nonzero):: Set to adjusted.
7596 (profile_count::probability_in):: Set quality to adjusted.
7597 * tree-ssa-tail-merge.c (replace_block_by): Use
7600 2018-01-23 Andrew Waterman <andrew@sifive.com>
7601 Jim Wilson <jimw@sifive.com>
7603 * config/riscv/riscv.c (riscv_stack_boundary): New.
7604 (riscv_option_override): Set riscv_stack_boundary. Handle
7605 riscv_preferred_stack_boundary_arg.
7606 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7607 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7608 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7609 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7610 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7611 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7613 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7616 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7617 of struct ix86_frame.
7618 (ix86_expand_epilogue): Likewise. Add a local variable for
7619 the reg_save_offset field in struct ix86_frame.
7621 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7623 PR tree-optimization/82604
7624 * tree-loop-distribution.c (enum partition_kind): New enum item
7625 PKIND_PARTIAL_MEMSET.
7626 (partition_builtin_p): Support above new enum item.
7627 (generate_code_for_partition): Ditto.
7628 (compute_access_range): Differentiate cases that equality can be
7629 proven at all loops, the innermost loops or no loops.
7630 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7631 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7632 (finalize_partitions, distribute_loop): Don't fuse partition of
7633 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7634 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7637 2018-01-23 Martin Liska <mliska@suse.cz>
7639 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7640 order to ignore the predictor.
7641 (PRED_POLYMORPHIC_CALL): Likewise.
7642 (PRED_RECURSIVE_CALL): Likewise.
7644 2018-01-23 Martin Liska <mliska@suse.cz>
7646 * tree-profile.c (tree_profiling): Print function header to
7647 aware reader which function we are working on.
7648 * value-prof.c (gimple_find_values_to_profile): Do not print
7649 not interesting value histograms.
7651 2018-01-23 Martin Liska <mliska@suse.cz>
7653 * profile-count.h (enum profile_quality): Add
7654 profile_uninitialized as the first value. Do not number values
7655 as they are zero based.
7656 (profile_count::verify): Update sanity check.
7657 (profile_probability::verify): Likewise.
7659 2018-01-23 Nathan Sidwell <nathan@acm.org>
7661 * doc/invoke.texi (ffor-scope): Deprecate.
7663 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7665 PR tree-optimization/83510
7666 * domwalk.c (set_all_edges_as_executable): New function.
7667 (dom_walker::dom_walker): Convert bool param
7668 "skip_unreachable_blocks" to enum reachability. Move setup of
7669 edge flags to set_all_edges_as_executable and only do it when
7670 reachability is REACHABLE_BLOCKS.
7671 * domwalk.h (enum dom_walker::reachability): New enum.
7672 (dom_walker::dom_walker): Convert bool param
7673 "skip_unreachable_blocks" to enum reachability.
7674 (set_all_edges_as_executable): New decl.
7675 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7676 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7678 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7679 but converting true to REACHABLE_BLOCKS.
7680 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7682 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7683 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7684 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7686 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7687 if check_all_array_refs will be called.
7689 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7691 * tree.c (selftest::test_location_wrappers): Add more test
7694 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7696 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7697 (selftest::test_bit_in_range): Likewise.
7699 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7702 * doc/sourcebuild.texi (vect_float): Say that the selector
7703 only describes the situation when -funsafe-math-optimizations is on.
7704 (vect_float_strict): Document.
7706 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7708 PR tree-optimization/83965
7709 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7710 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7711 instead of checking only for a reduction.
7712 (vect_recog_widen_sum_pattern): Likewise.
7714 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7716 * predict.c (probably_never_executed): Only use precise profile info.
7717 (compute_function_frequency): Skip after inlining hack since we now
7718 have quality checking.
7720 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7722 * profile-count.h (profile_probability::very_unlikely,
7723 profile_probability::unlikely, profile_probability::even): Set
7724 precision to guessed.
7726 2018-01-23 Richard Biener <rguenther@suse.de>
7728 PR tree-optimization/83963
7729 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7730 Properly terminate dominator walk when crossing the exit edge not
7731 when visiting its source block.
7733 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7736 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7737 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7739 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7741 PR tree-optimization/83957
7742 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7743 semicolon after for body surrounded by braces.
7745 PR tree-optimization/83081
7746 * profile-count.h (profile_probability::split): New method.
7747 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7748 Use profile_probability::split.
7749 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7750 when splitting a single conditional jump into 2.
7752 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7754 PR tree-optimization/69452
7755 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7758 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7760 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7761 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7762 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7764 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7766 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7767 * config/rl78/rl78.md (movdi): New define_expand.
7768 * config/rl78/rl78.c (rl78_split_movdi): New function.
7770 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7773 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7775 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7776 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7777 128-bit to produce an UNSPEC move to get the double word with the
7778 signbit and then a shift directly to do signbit.
7779 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7780 implementation with a new version that just does either a direct
7781 move or a regular move. Move memory interface to separate insns.
7782 Move insns so they are next to the expander.
7783 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7784 with signbit move. Split big and little endian case.
7785 (signbit<mode>2_dm_mem_le): Likewise.
7786 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7787 (signbit<mode>2_dm2): Likewise.
7789 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7791 * config/rl78/rl78.md (anddi3): New define_expand.
7793 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7795 * config/rl78/rl78.md (umindi3): New define_expand.
7797 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7799 * config/rl78/rl78.md (smindi3): New define_expand.
7801 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7803 * config/rl78/rl78.md (smaxdi3): New define_expand.
7805 2018-01-22 Carl Love <cel@us.ibm.com>
7807 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7808 LVX_V1TI): Add macro expansion.
7809 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7810 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7811 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7812 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7813 Change check to determine if the instruction is a byte reversing
7814 entry. Fix typo in comment.
7815 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7816 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7817 Add def_builtin calls for new builtins.
7818 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7819 Add define_insn expansion.
7821 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7823 * config/rl78/rl78.md (umaxdi3): New define_expand.
7825 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7827 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7828 for non-QImode registers.
7830 2018-01-22 Richard Biener <rguenther@suse.de>
7832 PR tree-optimization/83963
7833 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7834 including the loop exit block.
7835 (scop_detection::merge_sese): Likewise.
7836 (scop_detection::add_scop): Do it here instead.
7838 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7840 * doc/sourcebuild.texi (arm_softfloat): Document.
7842 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7845 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7846 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7847 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7849 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7850 David Edelsohn <dje.gcc@gmail.com>
7853 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7854 Change "crset eq" to "crset 2".
7855 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7856 (*call_indirect_aix<mode>_nospec): Likewise.
7857 (*call_value_indirect_aix<mode>_nospec): Likewise.
7858 (*call_indirect_elfv2<mode>_nospec): Likewise.
7859 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7860 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7861 change assembly output from . to $.
7862 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7863 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7864 (*tablejump<mode>_internal1_nospec): Likewise.
7866 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7869 * config/sh/sh_optimize_sett_clrt.cc:
7870 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7872 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7874 PR tree-optimization/83940
7875 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7876 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7877 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7878 use it to pass back the definition type.
7879 (vect_check_store_rhs): Likewise rhs_dt_out.
7880 (vect_build_gather_load_calls): Add a mask_dt argument and use
7881 it instead of a call to vect_is_simple_use.
7882 (vectorizable_store): Update calls to vect_check_load_store_mask
7883 and vect_check_store_rhs. Use the dt returned by the latter instead
7884 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7885 instead of calls to vect_is_simple_use. Pass the scalar rather
7886 than the vector operand to vect_is_simple_use when handling
7887 second and subsequent copies of an rhs value.
7888 (vectorizable_load): Update calls to vect_check_load_store_mask
7889 and vect_build_gather_load_calls. Use the cached mask_dt and
7890 gs_info.offset_dt instead of calls to vect_is_simple_use.
7892 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7895 * tree-emutls.c: Include gimplify.h.
7896 (lower_emutls_2): New function.
7897 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7898 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7899 it before further processing.
7902 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7903 UINTVAL (trueop1) instead of INTVAL (op1).
7905 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7909 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7910 INCOMING_FRAME_SP_OFFSET if not defined.
7911 (scan_trace): Add ENTRY argument. If true and
7912 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7913 emit a note to adjust the CFA offset.
7914 (create_cfi_notes): Adjust scan_trace callers.
7915 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7916 INCOMING_FRAME_SP_OFFSET in the CIE.
7917 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7918 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7920 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7921 * doc/tm.texi: Regenerated.
7923 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7925 PR rtl-optimization/83147
7926 * lra-constraints.c (remove_inheritance_pseudos): Use
7927 lra_substitute_pseudo_within_insn.
7929 2018-01-19 Tom de Vries <tom@codesourcery.com>
7930 Cesar Philippidis <cesar@codesourcery.com>
7933 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7935 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7938 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7939 spaces for function labels.
7941 2018-01-19 Martin Liska <mliska@suse.cz>
7943 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7944 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7945 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7946 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7947 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7948 (PRED_CONST_RETURN): Change from 69 to 65.
7949 (PRED_NULL_RETURN): Change from 91 to 71.
7950 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7951 (PRED_LOOP_GUARD): Change from 66 to 73.
7953 2018-01-19 Martin Liska <mliska@suse.cz>
7955 * predict.c (predict_insn_def): Add new assert.
7956 (struct branch_predictor): Change type to signed integer.
7957 (test_prediction_value_range): Amend test to cover
7959 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7960 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7961 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7962 (PRED_LOOP_IV_COMPARE): Likewise.
7963 * predict.h (PROB_UNINITIALIZED): Define new constant.
7965 2018-01-19 Martin Liska <mliska@suse.cz>
7967 * predict.c (dump_prediction): Add new format for
7968 analyze_brprob.py script which is enabled with -details
7970 * profile-count.h (precise_p): New function.
7972 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7974 PR tree-optimization/83922
7975 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7976 there are no statements that need masking.
7977 (vect_active_double_reduction_p): New function.
7978 (vect_analyze_loop_operations): Use it when handling phis that
7979 are not in the loop header.
7981 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7983 PR tree-optimization/83914
7984 * tree-vect-loop.c (vectorizable_induction): Don't convert
7985 init_expr or apply the peeling adjustment for inductions
7986 that are nested within the vectorized loop.
7988 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7990 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7993 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7997 * function.h (gimplify_parameters): Add gimple_seq * argument.
7998 * function.c: Include gimple.h and options.h.
7999 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8000 for the added local temporaries if needed.
8001 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8002 if there are any parameter cleanups, wrap whole body into a
8003 try/finally with the cleanups.
8005 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8008 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8009 Use GET_MODE_CLASS for scalar floating point.
8011 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8015 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8016 Fix call of call_cgraph_insertion_hooks.
8018 2018-01-18 Martin Sebor <msebor@redhat.com>
8020 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8022 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8025 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8028 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8031 * common.opt: (-ffile-prefix-map): New option.
8032 * opts.c (common_handle_option): Defer it.
8033 * opts-global.c (handle_common_deferred_options): Handle it.
8034 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8035 * file-prefix-map.h: New file.
8036 (remap_debug_filename, add_debug_prefix_map): ...here.
8037 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8038 * final.c (debug_prefix_map, add_debug_prefix_map
8039 remap_debug_filename): Move to...
8040 * file-prefix-map.c: New file.
8041 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8042 generalize, get rid of alloca(), use strrchr() instead of strchr().
8043 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8044 Implement in terms of add_prefix_map().
8045 (remap_macro_filename, remap_debug_filename): Implement in term of
8047 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8048 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8049 * dbxout.c: Include file-prefix-map.h.
8050 * varasm.c: Likewise.
8051 * vmsdbgout.c: Likewise.
8052 * xcoffout.c: Likewise.
8053 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8054 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8055 * doc/invoke.texi (-ffile-prefix-map): Document.
8056 (-fdebug-prefix-map): Update description.
8058 2018-01-18 Martin Liska <mliska@suse.cz>
8060 * config/i386/i386.c (indirect_thunk_name): Document that also
8062 (output_indirect_thunk): Document why both instructions
8063 (pause and lfence) are generated.
8065 2018-01-18 Richard Biener <rguenther@suse.de>
8067 PR tree-optimization/83887
8068 * graphite-scop-detection.c
8069 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8070 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8071 (scop_detection::merge_sese): Re-implement with a flood-fill
8072 algorithm that properly finds a SESE region if it exists.
8074 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8077 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8078 pointer_diff optimizations use view_convert instead of convert.
8080 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8082 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8083 Generate different code for -mno-speculate-indirect-jumps.
8084 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8085 (*call_indirect_aix<mode>): Disable for
8086 -mno-speculate-indirect-jumps.
8087 (*call_indirect_aix<mode>_nospec): New define_insn.
8088 (*call_value_indirect_aix<mode>): Disable for
8089 -mno-speculate-indirect-jumps.
8090 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8091 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8092 -mno-speculate-indirect-jumps.
8093 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8095 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8097 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8098 long double type, set the flags for noting the default long double
8099 type, even if we don't pass or return a long double type.
8101 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8104 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8107 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8109 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8110 support for merge[hl].
8111 (fold_mergehl_helper): New helper function.
8112 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8113 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8114 (altivec_vmrglw_direct): Add xxmrglw insn.
8116 2018-01-17 Andrew Waterman <andrew@sifive.com>
8118 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8119 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8121 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8124 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8125 call the lto_location_cache before reading the
8126 DECL_SOURCE_LOCATION of the types.
8128 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8129 Richard Sandiford <richard.sandiford@linaro.org>
8131 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8132 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8133 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8134 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8135 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8137 * config/aarch64/constraints.md (aarch64_movti_operand):
8139 * config/aarch64/predicates.md (Uti): Add new constraint.
8141 2018-01-17 Carl Love <cel@us.ibm.com>
8143 * config/rs6000/vsx.md (define_expand xl_len_r,
8144 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8145 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8147 (define_expand, define_insn): Move the shift left from the
8148 define_insn to the define_expand for lxvl and stxvl instructions.
8149 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8150 and XL_LEN_R definitions to PURE.
8152 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8154 * config/i386/i386.c (indirect_thunk_name): Declare regno
8155 as unsigned int. Compare regno with INVALID_REGNUM.
8156 (output_indirect_thunk): Ditto.
8157 (output_indirect_thunk_function): Ditto.
8158 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8159 in the call to output_indirect_thunk_function.
8161 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8164 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8165 rather than the size of inner_type to determine the stack slot size
8166 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8168 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8171 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8174 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8176 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8177 endian Linux systems to optionally enable multilibs for selecting
8178 the long double type if the user configured an explicit type.
8179 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8180 have no long double multilibs if not defined.
8181 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8182 warn if the user used -mabi={ieee,ibm}longdouble and we built
8183 multilibs for long double.
8184 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8185 appropriate multilib option.
8186 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8188 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8189 for building long double multilibs.
8190 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8192 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8194 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8197 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8199 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8202 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8205 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8208 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8210 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8211 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8213 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8215 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8216 different rtl trees depending on TARGET_64BIT.
8217 (rs6000_gen_lvx): Likewise.
8219 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8221 * config/visium/visium.md (nop): Tweak comment.
8222 (hazard_nop): Likewise.
8224 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8226 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8227 -mspeculate-indirect-jumps.
8228 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8229 for -mno-speculate-indirect-jumps.
8230 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8231 (*call_value_indirect_elfv2<mode>): Disable for
8232 -mno-speculate-indirect-jumps.
8233 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8234 (indirect_jump): Emit different RTL for
8235 -mno-speculate-indirect-jumps.
8236 (*indirect_jump<mode>): Disable for
8237 -mno-speculate-indirect-jumps.
8238 (*indirect_jump<mode>_nospec): New define_insn.
8239 (tablejump): Emit different RTL for
8240 -mno-speculate-indirect-jumps.
8241 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8242 (tablejumpsi_nospec): New define_expand.
8243 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8244 (tablejumpdi_nospec): New define_expand.
8245 (*tablejump<mode>_internal1): Disable for
8246 -mno-speculate-indirect-jumps.
8247 (*tablejump<mode>_internal1_nospec): New define_insn.
8248 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8251 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8253 * caller-save.c (insert_save): Drop unnecessary parameter. All
8256 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8257 Richard Biener <rguenth@suse.de>
8260 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8261 return early, inline manually is_gimple_sizepos. Make sure if we
8262 call gimplify_expr we don't end up with a gimple constant.
8263 * tree.c (variably_modified_type_p): Don't return true for
8264 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8265 * gimplify.h (is_gimple_sizepos): Remove.
8267 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8269 PR tree-optimization/83857
8270 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8271 vectorizable_live_operation for pure SLP statements.
8272 (vectorizable_live_operation): Handle PHIs.
8274 2018-01-16 Richard Biener <rguenther@suse.de>
8276 PR tree-optimization/83867
8277 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8278 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8280 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8283 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8284 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8285 If off is not INTEGER_CST, issue a may not be aligned warning
8286 rather than isn't aligned. Use isn%'t rather than isn't.
8287 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8289 <case MULT_EXPR>: Improve the case when bottom and one of the
8290 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8291 operand, in that case check if the other operand is multiple of
8292 bottom divided by the INTEGER_CST operand.
8294 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8297 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8298 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8299 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8300 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8301 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8302 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8303 (pa_function_arg_size): New function.
8305 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8307 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8308 in a separate statement.
8310 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8312 PR tree-optimization/83847
8313 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8314 group gathers and scatters.
8316 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8318 PR rtl-optimization/86620
8319 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8321 PR rtl-optimization/83213
8322 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8323 to last if both are JUMP_INSNs.
8325 PR tree-optimization/83843
8326 * gimple-ssa-store-merging.c
8327 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8328 store_immediate_info for bswap/nop orig_stores.
8330 2018-01-15 Andrew Waterman <andrew@sifive.com>
8332 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8334 <UDIV>: Increase cost if !TARGET_DIV.
8336 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8338 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8339 (define_attr "cr_logical_3op"): New.
8340 (cceq_ior_compare): Adjust.
8341 (cceq_ior_compare_complement): Adjust.
8342 (*cceq_rev_compare): Adjust.
8343 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8344 (is_cracked_insn): Adjust.
8345 (insn_must_be_first_in_group): Adjust.
8346 * config/rs6000/40x.md: Adjust.
8347 * config/rs6000/440.md: Adjust.
8348 * config/rs6000/476.md: Adjust.
8349 * config/rs6000/601.md: Adjust.
8350 * config/rs6000/603.md: Adjust.
8351 * config/rs6000/6xx.md: Adjust.
8352 * config/rs6000/7450.md: Adjust.
8353 * config/rs6000/7xx.md: Adjust.
8354 * config/rs6000/8540.md: Adjust.
8355 * config/rs6000/cell.md: Adjust.
8356 * config/rs6000/e300c2c3.md: Adjust.
8357 * config/rs6000/e500mc.md: Adjust.
8358 * config/rs6000/e500mc64.md: Adjust.
8359 * config/rs6000/e5500.md: Adjust.
8360 * config/rs6000/e6500.md: Adjust.
8361 * config/rs6000/mpc.md: Adjust.
8362 * config/rs6000/power4.md: Adjust.
8363 * config/rs6000/power5.md: Adjust.
8364 * config/rs6000/power6.md: Adjust.
8365 * config/rs6000/power7.md: Adjust.
8366 * config/rs6000/power8.md: Adjust.
8367 * config/rs6000/power9.md: Adjust.
8368 * config/rs6000/rs64.md: Adjust.
8369 * config/rs6000/titan.md: Adjust.
8371 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8373 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8374 ix86_indirect_branch_register logic.
8376 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8378 * config/i386/constraints.md (Bs): Update
8379 ix86_indirect_branch_register check. Don't check
8380 ix86_indirect_branch_register with GOT_memory_operand.
8382 * config/i386/predicates.md (GOT_memory_operand): Don't check
8383 ix86_indirect_branch_register here.
8384 (GOT32_symbol_operand): Likewise.
8386 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8388 * config/i386/predicates.md (constant_call_address_operand):
8389 Rewrite ix86_indirect_branch_register logic.
8390 (sibcall_insn_operand): Likewise.
8392 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8394 * config/i386/constraints.md (Bs): Replace
8395 ix86_indirect_branch_thunk_register with
8396 ix86_indirect_branch_register.
8398 * config/i386/i386.md (indirect_jump): Likewise.
8399 (tablejump): Likewise.
8400 (*sibcall_memory): Likewise.
8401 (*sibcall_value_memory): Likewise.
8402 Peepholes of indirect call and jump via memory: Likewise.
8403 * config/i386/i386.opt: Likewise.
8404 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8405 (GOT_memory_operand): Likewise.
8406 (call_insn_operand): Likewise.
8407 (sibcall_insn_operand): Likewise.
8408 (GOT32_symbol_operand): Likewise.
8410 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8413 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8414 type rather than type addr's type points to.
8415 (expand_omp_atomic_mutex): Likewise.
8416 (expand_omp_atomic): Likewise.
8418 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8421 * config/i386/i386.c (output_indirect_thunk_function): Use
8422 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8423 for __x86_return_thunk.
8425 2018-01-15 Richard Biener <rguenther@suse.de>
8428 * expmed.c (extract_bit_field_1): Fix typo.
8430 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8433 * config/arm/iterators.md (VF): New mode iterator.
8434 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8435 Remove integer-related logic from pattern.
8436 (neon_vabd<mode>_3): Likewise.
8438 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8441 * common.opt (fstrict-overflow): No longer an alias.
8442 (fwrapv-pointer): New option.
8443 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8444 also for pointer types based on flag_wrapv_pointer.
8445 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8446 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8447 opts->x_flag_wrapv got set.
8448 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8449 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8450 POINTER_TYPE_OVERFLOW_UNDEFINED.
8451 * match.pd: Likewise in address comparison pattern.
8452 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8454 2018-01-15 Richard Biener <rguenther@suse.de>
8457 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8458 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8459 Reset type names to their identifier if their TYPE_DECL doesn't
8460 have linkage (and thus is used for ODR and devirt).
8461 (save_debug_info_for_decl): Remove.
8462 (save_debug_info_for_type): Likewise.
8463 (add_tree_to_fld_list): Adjust.
8464 * tree-pretty-print.c (dump_generic_node): Make dumping of
8465 type names more robust.
8467 2018-01-15 Richard Biener <rguenther@suse.de>
8469 * BASE-VER: Bump to 8.0.1.
8471 2018-01-14 Martin Sebor <msebor@redhat.com>
8474 * builtins.c (check_access): Avoid warning when the no-warning bit
8477 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8479 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8480 * ira-color (allocno_hard_regs_compare): Likewise.
8482 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8485 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8486 Use .pushsection/.popsection.
8488 2018-01-14 Martin Sebor <msebor@redhat.com>
8491 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8493 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8495 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8496 entry from extra_headers.
8497 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8498 extra_headers, make the list bitwise identical to the i?86-*-* one.
8500 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8502 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8503 -mcmodel=large with -mindirect-branch=thunk,
8504 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8505 -mfunction-return=thunk-extern.
8506 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8507 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8508 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8510 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8512 * config/i386/i386.c (print_reg): Print the name of the full
8513 integer register without '%'.
8514 (ix86_print_operand): Handle 'V'.
8515 * doc/extend.texi: Document 'V' modifier.
8517 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8519 * config/i386/constraints.md (Bs): Disallow memory operand for
8520 -mindirect-branch-register.
8522 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8523 (GOT_memory_operand): Likewise.
8524 (call_insn_operand): Likewise.
8525 (sibcall_insn_operand): Likewise.
8526 (GOT32_symbol_operand): Likewise.
8527 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8528 for -mindirect-branch-register.
8529 (tablejump): Likewise.
8530 (*sibcall_memory): Likewise.
8531 (*sibcall_value_memory): Likewise.
8532 Disallow peepholes of indirect call and jump via memory for
8533 -mindirect-branch-register.
8534 (*call_pop): Replace m with Bw.
8535 (*call_value_pop): Likewise.
8536 (*sibcall_pop_memory): Replace m with Bs.
8537 * config/i386/i386.opt (mindirect-branch-register): New option.
8538 * doc/invoke.texi: Document -mindirect-branch-register option.
8540 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8542 * config/i386/i386-protos.h (ix86_output_function_return): New.
8543 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8544 set function_return_type.
8545 (indirect_thunk_name): Add ret_p to indicate thunk for function
8547 (output_indirect_thunk_function): Pass false to
8548 indirect_thunk_name.
8549 (ix86_output_indirect_branch_via_reg): Likewise.
8550 (ix86_output_indirect_branch_via_push): Likewise.
8551 (output_indirect_thunk_function): Create alias for function
8552 return thunk if regno < 0.
8553 (ix86_output_function_return): New function.
8554 (ix86_handle_fndecl_attribute): Handle function_return.
8555 (ix86_attribute_table): Add function_return.
8556 * config/i386/i386.h (machine_function): Add
8557 function_return_type.
8558 * config/i386/i386.md (simple_return_internal): Use
8559 ix86_output_function_return.
8560 (simple_return_internal_long): Likewise.
8561 * config/i386/i386.opt (mfunction-return=): New option.
8562 (indirect_branch): Mention -mfunction-return=.
8563 * doc/extend.texi: Document function_return function attribute.
8564 * doc/invoke.texi: Document -mfunction-return= option.
8566 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8568 * config/i386/i386-opts.h (indirect_branch): New.
8569 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8570 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8571 with local indirect jump when converting indirect call and jump.
8572 (ix86_set_indirect_branch_type): New.
8573 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8574 (indirectlabelno): New.
8575 (indirect_thunk_needed): Likewise.
8576 (indirect_thunk_bnd_needed): Likewise.
8577 (indirect_thunks_used): Likewise.
8578 (indirect_thunks_bnd_used): Likewise.
8579 (INDIRECT_LABEL): Likewise.
8580 (indirect_thunk_name): Likewise.
8581 (output_indirect_thunk): Likewise.
8582 (output_indirect_thunk_function): Likewise.
8583 (ix86_output_indirect_branch_via_reg): Likewise.
8584 (ix86_output_indirect_branch_via_push): Likewise.
8585 (ix86_output_indirect_branch): Likewise.
8586 (ix86_output_indirect_jmp): Likewise.
8587 (ix86_code_end): Call output_indirect_thunk_function if needed.
8588 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8590 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8591 (ix86_attribute_table): Add indirect_branch.
8592 * config/i386/i386.h (machine_function): Add indirect_branch_type
8593 and has_local_indirect_jump.
8594 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8596 (tablejump): Likewise.
8597 (*indirect_jump): Use ix86_output_indirect_jmp.
8598 (*tablejump_1): Likewise.
8599 (simple_return_indirect_internal): Likewise.
8600 * config/i386/i386.opt (mindirect-branch=): New option.
8601 (indirect_branch): New.
8604 (thunk-inline): Likewise.
8605 (thunk-extern): Likewise.
8606 * doc/extend.texi: Document indirect_branch function attribute.
8607 * doc/invoke.texi: Document -mindirect-branch= option.
8609 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8612 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8614 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8616 * ipa-inline.c (want_inline_small_function_p): Return false if
8617 inlining has already failed with CIF_FINAL_ERROR.
8618 (update_caller_keys): Call want_inline_small_function_p before
8620 (update_callee_keys): Likewise.
8622 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8624 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8626 (rs6000_quadword_masked_address_p): Likewise.
8627 (quad_aligned_load_p): Likewise.
8628 (quad_aligned_store_p): Likewise.
8629 (const_load_sequence_p): Add comment to describe the outer-most loop.
8630 (mimic_memory_attributes_and_flags): New function.
8631 (rs6000_gen_stvx): Likewise.
8632 (replace_swapped_aligned_store): Likewise.
8633 (rs6000_gen_lvx): Likewise.
8634 (replace_swapped_aligned_load): Likewise.
8635 (replace_swapped_load_constant): Capitalize argument name in
8636 comment describing this function.
8637 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8638 and stores that access quad-word aligned addresses and replace
8639 with stvx or lvx instructions when appropriate.
8640 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8641 New function prototype.
8642 (rs6000_quadword_masked_address_p): Likewise.
8643 (rs6000_gen_lvx): Likewise.
8644 (rs6000_gen_stvx): Likewise.
8645 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8646 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8647 when memory address is aligned.
8648 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8649 this split to select lvx instruction when memory address is aligned.
8650 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8651 instruction when memory address is aligned.
8652 (*vsx_le_perm_load_v16qi): Likewise.
8653 (four unnamed splitters): Modify to select the stvx instruction
8654 when memory is aligned.
8656 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8658 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8659 which appears in the queue multiple times.
8661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8662 Alan Hayward <alan.hayward@arm.com>
8663 David Sherwood <david.sherwood@arm.com>
8665 * tree-vectorizer.h (vec_lower_bound): New structure.
8666 (_loop_vec_info): Add check_nonzero and lower_bounds.
8667 (LOOP_VINFO_CHECK_NONZERO): New macro.
8668 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8669 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8670 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8671 fields. Make seg_len the distance travelled, not including the
8673 (dr_direction_indicator): Declare.
8674 (dr_zero_step_indicator): Likewise.
8675 (dr_known_forward_stride_p): Likewise.
8676 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8678 (runtime_alias_check_p): Allow runtime alias checks with
8680 (operator ==): Compare access_size and align.
8681 (prune_runtime_alias_test_list): Rework for new distinction between
8682 the access_size and seg_len.
8683 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8685 (get_segment_min_max): New function.
8686 (create_intersect_range_checks): Use it.
8687 (dr_step_indicator): New function.
8688 (dr_direction_indicator): Likewise.
8689 (dr_zero_step_indicator): Likewise.
8690 (dr_known_forward_stride_p): Likewise.
8691 * tree-loop-distribution.c (data_ref_segment_size): Return
8692 DR_STEP * (niters - 1).
8693 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8695 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8696 (vect_preserves_scalar_order_p): New function, split out from...
8697 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8698 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8699 (vect_vfa_access_size): New function.
8700 (vect_vfa_align): Likewise.
8701 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8702 (dump_lower_bound): New function.
8703 (vect_check_lower_bound): Likewise.
8704 (vect_small_gap_p): Likewise.
8705 (vectorizable_with_step_bound_p): Likewise.
8706 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8707 depencies if the vectorization factor is 1. Convert the checks
8708 for nonzero steps into checks on the bounds of DR_STEP. Try using
8709 a bunds check for variable steps if the minimum required step is
8710 relatively small. Update calls to the dr_with_seg_len
8711 constructor and to vect_compile_time_alias.
8712 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8714 (vect_loop_versioning): Call it.
8715 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8717 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8720 Alan Hayward <alan.hayward@arm.com>
8721 David Sherwood <david.sherwood@arm.com>
8723 * doc/sourcebuild.texi (vect_scatter_store): Document.
8724 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8726 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8728 * genopinit.c (main): Add supports_vec_scatter_store and
8729 supports_vec_scatter_store_cached to target_optabs.
8730 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8731 IFN_MASK_SCATTER_STORE.
8732 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8734 * internal-fn.h (internal_store_fn_p): Declare.
8735 (internal_fn_stored_value_index): Likewise.
8736 * internal-fn.c (scatter_store_direct): New macro.
8737 (expand_scatter_store_optab_fn): New function.
8738 (direct_scatter_store_optab_supported_p): New macro.
8739 (internal_store_fn_p): New function.
8740 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8741 IFN_MASK_SCATTER_STORE.
8742 (internal_fn_mask_index): Likewise.
8743 (internal_fn_stored_value_index): New function.
8744 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8746 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8747 * optabs-query.c (supports_vec_scatter_store_p): New function.
8748 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8749 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8750 true for scatter stores.
8751 (vect_gather_scatter_fn_p): Handle scatter stores too.
8752 (vect_check_gather_scatter): Consider using scatter stores if
8753 supports_vec_scatter_store_p.
8754 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8756 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8757 internal_fn_stored_value_index.
8758 (check_load_store_masking): Handle scatter stores too.
8759 (vect_get_store_rhs): Make public.
8760 (vectorizable_call): Use internal_store_fn_p.
8761 (vectorizable_store): Handle scatter store internal functions.
8762 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8763 when deciding whether the end of the group has been reached.
8764 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8765 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8766 (mask_scatter_store<mode>): New insns.
8768 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8769 Alan Hayward <alan.hayward@arm.com>
8770 David Sherwood <david.sherwood@arm.com>
8772 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8773 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8774 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8776 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8777 Use vect_truncate_gather_scatter_offset if we can't treat the
8778 operation as a normal gather load or scatter store.
8779 (get_group_load_store_type): Take the gather_scatter_info
8780 as argument. Try using a gather load or scatter store for
8781 single-element groups.
8782 (get_load_store_type): Update calls to get_group_load_store_type
8783 and vect_use_strided_gather_scatters_p.
8785 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8786 Alan Hayward <alan.hayward@arm.com>
8787 David Sherwood <david.sherwood@arm.com>
8789 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8790 optional tree argument.
8791 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8793 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8794 but continue to use the current value as a fallback.
8795 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8796 to compare the updates.
8797 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8798 (get_load_store_type): Use it when handling a strided access.
8799 (vect_get_strided_load_store_ops): New function.
8800 (vect_get_data_ptr_increment): Likewise.
8801 (vectorizable_load): Handle strided gather loads. Always pass
8802 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8804 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8805 Alan Hayward <alan.hayward@arm.com>
8806 David Sherwood <david.sherwood@arm.com>
8808 * doc/md.texi (gather_load@var{m}): Document.
8809 (mask_gather_load@var{m}): Likewise.
8810 * genopinit.c (main): Add supports_vec_gather_load and
8811 supports_vec_gather_load_cached to target_optabs.
8812 * optabs-tree.c (init_tree_optimization_optabs): Use
8813 ggc_cleared_alloc to allocate target_optabs.
8814 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8815 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8817 * internal-fn.h (internal_load_fn_p): Declare.
8818 (internal_gather_scatter_fn_p): Likewise.
8819 (internal_fn_mask_index): Likewise.
8820 (internal_gather_scatter_fn_supported_p): Likewise.
8821 * internal-fn.c (gather_load_direct): New macro.
8822 (expand_gather_load_optab_fn): New function.
8823 (direct_gather_load_optab_supported_p): New macro.
8824 (direct_internal_fn_optab): New function.
8825 (internal_load_fn_p): Likewise.
8826 (internal_gather_scatter_fn_p): Likewise.
8827 (internal_fn_mask_index): Likewise.
8828 (internal_gather_scatter_fn_supported_p): Likewise.
8829 * optabs-query.c (supports_at_least_one_mode_p): New function.
8830 (supports_vec_gather_load_p): Likewise.
8831 * optabs-query.h (supports_vec_gather_load_p): Declare.
8832 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8833 and memory_type field.
8834 (NUM_PATTERNS): Bump to 15.
8835 * tree-vect-data-refs.c: Include internal-fn.h.
8836 (vect_gather_scatter_fn_p): New function.
8837 (vect_describe_gather_scatter_call): Likewise.
8838 (vect_check_gather_scatter): Try using internal functions for
8839 gather loads. Recognize existing calls to a gather load function.
8840 (vect_analyze_data_refs): Consider using gather loads if
8841 supports_vec_gather_load_p.
8842 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8843 (vect_get_gather_scatter_offset_type): Likewise.
8844 (vect_convert_mask_for_vectype): Likewise.
8845 (vect_add_conversion_to_patterm): Likewise.
8846 (vect_try_gather_scatter_pattern): Likewise.
8847 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8848 (vect_vect_recog_func_ptrs): Add it.
8849 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8850 internal_fn_mask_index and internal_gather_scatter_fn_p.
8851 (check_load_store_masking): Take the gather_scatter_info as an
8852 argument and handle gather loads.
8853 (vect_get_gather_scatter_ops): New function.
8854 (vectorizable_call): Check internal_load_fn_p.
8855 (vectorizable_load): Likewise. Handle gather load internal
8857 (vectorizable_store): Update call to check_load_store_masking.
8858 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8859 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8860 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8861 (aarch64_gather_scale_operand_d): New predicates.
8862 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8863 (mask_gather_load<mode>): New insns.
8865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8866 Alan Hayward <alan.hayward@arm.com>
8867 David Sherwood <david.sherwood@arm.com>
8869 * optabs.def (fold_left_plus_optab): New optab.
8870 * doc/md.texi (fold_left_plus_@var{m}): Document.
8871 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8872 * internal-fn.c (fold_left_direct): Define.
8873 (expand_fold_left_optab_fn): Likewise.
8874 (direct_fold_left_optab_supported_p): Likewise.
8875 * fold-const-call.c (fold_const_fold_left): New function.
8876 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8877 * tree-parloops.c (valid_reduction_p): New function.
8878 (gather_scalar_reductions): Use it.
8879 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8880 (vect_finish_replace_stmt): Declare.
8881 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8882 (needs_fold_left_reduction_p): New function, split out from...
8883 (vect_is_simple_reduction): ...here. Accept reductions that
8884 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8885 (vect_force_simple_reduction): Also store the reduction type in
8886 the assignment's STMT_VINFO_REDUC_TYPE.
8887 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8888 (merge_with_identity): New function.
8889 (vect_expand_fold_left): Likewise.
8890 (vectorize_fold_left_reduction): Likewise.
8891 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8892 scalar phi in place for it. Check for target support and reject
8893 cases that would reassociate the operation. Defer the transform
8894 phase to vectorize_fold_left_reduction.
8895 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8896 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8897 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8899 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8901 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8902 call to ifc_temp_var.
8904 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8905 Alan Hayward <alan.hayward@arm.com>
8906 David Sherwood <david.sherwood@arm.com>
8908 * target.def (legitimize_address_displacement): Take the original
8909 offset as a poly_int.
8910 * targhooks.h (default_legitimize_address_displacement): Update
8912 * targhooks.c (default_legitimize_address_displacement): Likewise.
8913 * doc/tm.texi: Regenerate.
8914 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8915 as an argument, moving assert of ad->disp == ad->disp_term to...
8916 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8917 Try calling targetm.legitimize_address_displacement before expanding
8918 the address rather than afterwards, and adjust for the new interface.
8919 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8920 Match the new hook interface. Handle SVE addresses.
8921 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8926 * Makefile.in (OBJS): Add early-remat.o.
8927 * target.def (select_early_remat_modes): New hook.
8928 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8929 * doc/tm.texi: Regenerate.
8930 * targhooks.h (default_select_early_remat_modes): Declare.
8931 * targhooks.c (default_select_early_remat_modes): New function.
8932 * timevar.def (TV_EARLY_REMAT): New timevar.
8933 * passes.def (pass_early_remat): New pass.
8934 * tree-pass.h (make_pass_early_remat): Declare.
8935 * early-remat.c: New file.
8936 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8938 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8945 vfm1 with a bound_epilog parameter.
8946 (vect_do_peeling): Update calls accordingly, and move the prologue
8947 call earlier in the function. Treat the base bound_epilog as 0 for
8948 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8949 this base when peeling for gaps.
8950 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8951 with fully-masked loops.
8952 (vect_estimate_min_profitable_iters): Handle the single peeled
8953 iteration in that case.
8955 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8956 Alan Hayward <alan.hayward@arm.com>
8957 David Sherwood <david.sherwood@arm.com>
8959 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8960 single-element interleaving even if the size is not a power of 2.
8961 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8962 accesses for single-element interleaving if the group size is
8965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8966 Alan Hayward <alan.hayward@arm.com>
8967 David Sherwood <david.sherwood@arm.com>
8969 * doc/md.texi (fold_extract_last_@var{m}): Document.
8970 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8971 * optabs.def (fold_extract_last_optab): New optab.
8972 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8973 * internal-fn.c (fold_extract_direct): New macro.
8974 (expand_fold_extract_optab_fn): Likewise.
8975 (direct_fold_extract_optab_supported_p): Likewise.
8976 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8977 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8978 EXTRACT_LAST_REDUCTION.
8979 (get_initial_def_for_reduction): Do not create an initial vector
8980 for EXTRACT_LAST_REDUCTION reductions.
8981 (vectorizable_reduction): Leave the scalar phi in place for
8982 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8983 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8984 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8985 transform phase to vectorizable_condition.
8986 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8988 (vect_finish_stmt_generation): ...here.
8989 (vect_finish_replace_stmt): New function.
8990 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8991 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8993 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8996 Alan Hayward <alan.hayward@arm.com>
8997 David Sherwood <david.sherwood@arm.com>
8999 * doc/md.texi (extract_last_@var{m}): Document.
9000 * optabs.def (extract_last_optab): New optab.
9001 * internal-fn.def (EXTRACT_LAST): New internal function.
9002 * internal-fn.c (cond_unary_direct): New macro.
9003 (expand_cond_unary_optab_fn): Likewise.
9004 (direct_cond_unary_optab_supported_p): Likewise.
9005 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9006 loops using EXTRACT_LAST.
9007 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9008 (extract_last_<mode>): ...this optab.
9009 (vec_extract<mode><Vel>): Update accordingly.
9011 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9012 Alan Hayward <alan.hayward@arm.com>
9013 David Sherwood <david.sherwood@arm.com>
9015 * target.def (empty_mask_is_expensive): New hook.
9016 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9017 * doc/tm.texi: Regenerate.
9018 * targhooks.h (default_empty_mask_is_expensive): Declare.
9019 * targhooks.c (default_empty_mask_is_expensive): New function.
9020 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9021 if the target says that empty masks are expensive.
9022 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9024 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9027 Alan Hayward <alan.hayward@arm.com>
9028 David Sherwood <david.sherwood@arm.com>
9030 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9031 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9032 (vect_use_loop_mask_for_alignment_p): New function.
9033 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9034 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9035 niters_skip argument. Make sure that the first niters_skip elements
9036 of the first iteration are inactive.
9037 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9038 Update call to vect_set_loop_masks_directly.
9039 (get_misalign_in_elems): New function, split out from...
9040 (vect_gen_prolog_loop_niters): ...here.
9041 (vect_update_init_of_dr): Take a code argument that specifies whether
9042 the adjustment should be added or subtracted.
9043 (vect_update_init_of_drs): Likewise.
9044 (vect_prepare_for_masked_peels): New function.
9045 (vect_do_peeling): Skip prologue peeling if we're using a mask
9046 instead. Update call to vect_update_inits_of_drs.
9047 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9049 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9050 alignment. Do not include the number of peeled iterations in
9051 the minimum threshold in that case.
9052 (vectorizable_induction): Adjust the start value down by
9053 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9054 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9055 Take the number of skipped iterations into account when calculating
9057 * tree-vect-stmts.c (vect_gen_while_not): New function.
9059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9060 Alan Hayward <alan.hayward@arm.com>
9061 David Sherwood <david.sherwood@arm.com>
9063 * doc/sourcebuild.texi (vect_fully_masked): Document.
9064 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9066 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9068 (vect_analyze_loop_2): ...here. Don't check the vectorization
9069 factor against the number of loop iterations if the loop is
9072 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9073 Alan Hayward <alan.hayward@arm.com>
9074 David Sherwood <david.sherwood@arm.com>
9076 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9077 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9078 (dump_groups): Update accordingly.
9079 (iv_use::mem_type): New member variable.
9080 (address_p): New function.
9081 (record_use): Add a mem_type argument and initialize the new
9083 (record_group_use): Add a mem_type argument. Use address_p.
9084 Remove obsolete null checks of base_object. Update call to record_use.
9085 (find_interesting_uses_op): Update call to record_group_use.
9086 (find_interesting_uses_cond): Likewise.
9087 (find_interesting_uses_address): Likewise.
9088 (get_mem_type_for_internal_fn): New function.
9089 (find_address_like_use): Likewise.
9090 (find_interesting_uses_stmt): Try find_address_like_use before
9091 calling find_interesting_uses_op.
9092 (addr_offset_valid_p): Use the iv mem_type field as the type
9093 of the addressed memory.
9094 (add_autoinc_candidates): Likewise.
9095 (get_address_cost): Likewise.
9096 (split_small_address_groups_p): Use address_p.
9097 (split_address_groups): Likewise.
9098 (add_iv_candidate_for_use): Likewise.
9099 (autoinc_possible_for_pair): Likewise.
9100 (rewrite_groups): Likewise.
9101 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9102 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9103 (get_alias_ptr_type_for_ptr_address): New function.
9104 (rewrite_use_address): Rewrite address uses in calls that were
9105 identified by find_address_like_use.
9107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9108 Alan Hayward <alan.hayward@arm.com>
9109 David Sherwood <david.sherwood@arm.com>
9111 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9113 * gimple-expr.h (is_gimple_addressable: Likewise.
9114 * gimple-expr.c (is_gimple_address): Likewise.
9115 * internal-fn.c (expand_call_mem_ref): New function.
9116 (expand_mask_load_optab_fn): Use it.
9117 (expand_mask_store_optab_fn): Likewise.
9119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9120 Alan Hayward <alan.hayward@arm.com>
9121 David Sherwood <david.sherwood@arm.com>
9123 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9124 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9125 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9126 (cond_umax@var{mode}): Document.
9127 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9128 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9129 (cond_umin_optab, cond_umax_optab): New optabs.
9130 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9131 (COND_IOR, COND_XOR): New internal functions.
9132 * internal-fn.h (get_conditional_internal_fn): Declare.
9133 * internal-fn.c (cond_binary_direct): New macro.
9134 (expand_cond_binary_optab_fn): Likewise.
9135 (direct_cond_binary_optab_supported_p): Likewise.
9136 (get_conditional_internal_fn): New function.
9137 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9138 Cope with reduction statements that are vectorized as calls rather
9140 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9141 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9142 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9143 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9144 (UNSPEC_COND_EOR): New unspecs.
9145 (optab): Add mappings for them.
9146 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9147 (sve_int_op, sve_fp_op): New int attributes.
9149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9150 Alan Hayward <alan.hayward@arm.com>
9151 David Sherwood <david.sherwood@arm.com>
9153 * optabs.def (while_ult_optab): New optab.
9154 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9155 * internal-fn.def (WHILE_ULT): New internal function.
9156 * internal-fn.h (direct_internal_fn_supported_p): New override
9157 that takes two types as argument.
9158 * internal-fn.c (while_direct): New macro.
9159 (expand_while_optab_fn): New function.
9160 (convert_optab_supported_p): Likewise.
9161 (direct_while_optab_supported_p): New macro.
9162 * wide-int.h (wi::udiv_ceil): New function.
9163 * tree-vectorizer.h (rgroup_masks): New structure.
9164 (vec_loop_masks): New typedef.
9165 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9167 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9168 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9169 (vect_max_vf): New function.
9170 (slpeel_make_loop_iterate_ntimes): Delete.
9171 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9172 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9173 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9174 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9175 internal-fn.h, stor-layout.h and optabs-query.h.
9176 (vect_set_loop_mask): New function.
9177 (add_preheader_seq): Likewise.
9178 (add_header_seq): Likewise.
9179 (interleave_supported_p): Likewise.
9180 (vect_maybe_permute_loop_masks): Likewise.
9181 (vect_set_loop_masks_directly): Likewise.
9182 (vect_set_loop_condition_masked): Likewise.
9183 (vect_set_loop_condition_unmasked): New function, split out from
9184 slpeel_make_loop_iterate_ntimes.
9185 (slpeel_make_loop_iterate_ntimes): Rename to..
9186 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9187 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9188 (vect_do_peeling): Update call accordingly.
9189 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9191 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9192 mask_compare_type, can_fully_mask_p and fully_masked_p.
9193 (release_vec_loop_masks): New function.
9194 (_loop_vec_info): Use it to free the loop masks.
9195 (can_produce_all_loop_masks_p): New function.
9196 (vect_get_max_nscalars_per_iter): Likewise.
9197 (vect_verify_full_masking): Likewise.
9198 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9199 retries, and free the mask rgroups before retrying. Check loop-wide
9200 reasons for disallowing fully-masked loops. Make the final decision
9201 about whether use a fully-masked loop or not.
9202 (vect_estimate_min_profitable_iters): Do not assume that peeling
9203 for the number of iterations will be needed for fully-masked loops.
9204 (vectorizable_reduction): Disable fully-masked loops.
9205 (vectorizable_live_operation): Likewise.
9206 (vect_halve_mask_nunits): New function.
9207 (vect_double_mask_nunits): Likewise.
9208 (vect_record_loop_mask): Likewise.
9209 (vect_get_loop_mask): Likewise.
9210 (vect_transform_loop): Handle the case in which the final loop
9211 iteration might handle a partial vector. Call vect_set_loop_condition
9212 instead of slpeel_make_loop_iterate_ntimes.
9213 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9214 (check_load_store_masking): New function.
9215 (prepare_load_store_mask): Likewise.
9216 (vectorizable_store): Handle fully-masked loops.
9217 (vectorizable_load): Likewise.
9218 (supportable_widening_operation): Use vect_halve_mask_nunits for
9220 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9221 (vect_gen_while): New function.
9222 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9223 (aarch64_uqdec<mode>): New insn.
9225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9226 Alan Hayward <alan.hayward@arm.com>
9227 David Sherwood <david.sherwood@arm.com>
9229 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9230 (reduc_xor_scal_optab): New optabs.
9231 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9232 (reduc_xor_scal_@var{m}): Document.
9233 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9234 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9236 * fold-const-call.c (fold_const_call): Handle them.
9237 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9238 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9239 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9240 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9241 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9242 (UNSPEC_XORV): New unspecs.
9243 (optab): Add entries for them.
9244 (BITWISEV): New int iterator.
9245 (bit_reduc_op): New int attributes.
9247 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9248 Alan Hayward <alan.hayward@arm.com>
9249 David Sherwood <david.sherwood@arm.com>
9251 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9252 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9253 * optabs.def (vec_shl_insert_optab): New optab.
9254 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9255 (duplicate_and_interleave): Likewise.
9256 * tree-vect-loop.c: Include internal-fn.h.
9257 (neutral_op_for_slp_reduction): New function, split out from
9258 get_initial_defs_for_reduction.
9259 (get_initial_def_for_reduction): Handle option 2 for variable-length
9260 vectors by loading the neutral value into a vector and then shifting
9261 the initial value into element 0.
9262 (get_initial_defs_for_reduction): Replace the code argument with
9263 the neutral value calculated by neutral_op_for_slp_reduction.
9264 Use gimple_build_vector for constant-length vectors.
9265 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9266 but the first group_size elements have a neutral value.
9267 Use duplicate_and_interleave otherwise.
9268 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9269 Update call to get_initial_defs_for_reduction. Handle SLP
9270 reductions for variable-length vectors by creating one vector
9271 result for each scalar result, with the elements associated
9272 with other scalar results stubbed out with the neutral value.
9273 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9274 Require IFN_VEC_SHL_INSERT for double reductions on
9275 variable-length vectors, or SLP reductions that have
9276 a neutral value. Require can_duplicate_and_interleave_p
9277 support for variable-length unchained SLP reductions if there
9278 is no neutral value, such as for MIN/MAX reductions. Also require
9279 the number of vector elements to be a multiple of the number of
9280 SLP statements when doing variable-length unchained SLP reductions.
9281 Update call to vect_create_epilog_for_reduction.
9282 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9283 and remove initial values.
9284 (duplicate_and_interleave): Make public.
9285 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9286 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9289 Alan Hayward <alan.hayward@arm.com>
9290 David Sherwood <david.sherwood@arm.com>
9292 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9293 (can_duplicate_and_interleave_p): New function.
9294 (vect_get_and_check_slp_defs): Take the vector of statements
9295 rather than just the current one. Remove excess parentheses.
9296 Restriction rejectinon of vect_constant_def and vect_external_def
9297 for variable-length vectors to boolean types, or types for which
9298 can_duplicate_and_interleave_p is false.
9299 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9300 (duplicate_and_interleave): New function.
9301 (vect_get_constant_vectors): Use gimple_build_vector for
9302 constant-length vectors and suitable variable-length constant
9303 vectors. Use duplicate_and_interleave for other variable-length
9304 vectors. Don't defer the update when inserting new statements.
9306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9307 Alan Hayward <alan.hayward@arm.com>
9308 David Sherwood <david.sherwood@arm.com>
9310 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9311 min_profitable_iters doesn't go negative.
9313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9314 Alan Hayward <alan.hayward@arm.com>
9315 David Sherwood <david.sherwood@arm.com>
9317 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9318 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9319 * optabs.def (vec_mask_load_lanes_optab): New optab.
9320 (vec_mask_store_lanes_optab): Likewise.
9321 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9322 (MASK_STORE_LANES): Likewise.
9323 * internal-fn.c (mask_load_lanes_direct): New macro.
9324 (mask_store_lanes_direct): Likewise.
9325 (expand_mask_load_optab_fn): Handle masked operations.
9326 (expand_mask_load_lanes_optab_fn): New macro.
9327 (expand_mask_store_optab_fn): Handle masked operations.
9328 (expand_mask_store_lanes_optab_fn): New macro.
9329 (direct_mask_load_lanes_optab_supported_p): Likewise.
9330 (direct_mask_store_lanes_optab_supported_p): Likewise.
9331 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9333 (vect_load_lanes_supported): Likewise.
9334 * tree-vect-data-refs.c (strip_conversion): New function.
9335 (can_group_stmts_p): Likewise.
9336 (vect_analyze_data_ref_accesses): Use it instead of checking
9337 for a pair of assignments.
9338 (vect_store_lanes_supported): Take a masked_p parameter.
9339 (vect_load_lanes_supported): Likewise.
9340 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9341 vect_store_lanes_supported and vect_load_lanes_supported.
9342 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9343 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9344 parameter. Don't allow gaps for masked accesses.
9345 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9346 and vect_load_lanes_supported.
9347 (get_load_store_type): Take a masked_p parameter and update
9348 call to get_group_load_store_type.
9349 (vectorizable_store): Update call to get_load_store_type.
9350 Handle IFN_MASK_STORE_LANES.
9351 (vectorizable_load): Update call to get_load_store_type.
9352 Handle IFN_MASK_LOAD_LANES.
9354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9355 Alan Hayward <alan.hayward@arm.com>
9356 David Sherwood <david.sherwood@arm.com>
9358 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9360 * config/aarch64/aarch64-protos.h
9361 (aarch64_sve_struct_memory_operand_p): Declare.
9362 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9363 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9364 (VPRED, vpred): Handle SVE structure modes.
9365 * config/aarch64/constraints.md (Utx): New constraint.
9366 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9367 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9368 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9369 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9370 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9371 structure modes. Split into pieces after RA.
9372 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9373 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9375 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9376 SVE structure modes.
9377 (aarch64_classify_address): Likewise.
9378 (sizetochar): Move earlier in file.
9379 (aarch64_print_operand): Handle SVE register lists.
9380 (aarch64_array_mode): New function.
9381 (aarch64_sve_struct_memory_operand_p): Likewise.
9382 (TARGET_ARRAY_MODE): Redefine.
9384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9385 Alan Hayward <alan.hayward@arm.com>
9386 David Sherwood <david.sherwood@arm.com>
9388 * target.def (array_mode): New target hook.
9389 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9390 * doc/tm.texi: Regenerate.
9391 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9392 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9393 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9395 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9398 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9399 Alan Hayward <alan.hayward@arm.com>
9400 David Sherwood <david.sherwood@arm.com>
9402 * fold-const.c (fold_binary_loc): Check the argument types
9403 rather than the result type when testing for a vector operation.
9405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9407 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9408 * doc/tm.texi: Regenerate.
9410 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9411 Alan Hayward <alan.hayward@arm.com>
9412 David Sherwood <david.sherwood@arm.com>
9414 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9415 (sve): Document new AArch64 extension.
9416 * doc/md.texi (w): Extend the description of the AArch64
9417 constraint to include SVE vectors.
9418 (Upl, Upa): Document new AArch64 predicate constraints.
9419 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9421 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9422 (msve-vector-bits=): New option.
9423 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9424 SVE when these are disabled.
9425 (sve): New extension.
9426 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9427 modes. Adjust their number of units based on aarch64_sve_vg.
9428 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9429 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9430 aarch64_addr_query_type.
9431 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9432 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9433 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9434 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9435 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9436 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9437 (aarch64_simd_imm_zero_p): Delete.
9438 (aarch64_check_zero_based_sve_index_immediate): Declare.
9439 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9440 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9441 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9442 (aarch64_sve_float_mul_immediate_p): Likewise.
9443 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9445 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9446 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9447 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9448 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9449 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9450 (aarch64_regmode_natural_size): Likewise.
9451 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9452 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9454 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9455 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9456 for VG and the SVE predicate registers.
9457 (V_ALIASES): Add a "z"-prefixed alias.
9458 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9459 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9460 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9461 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9462 (REG_CLASS_NAMES): Add entries for them.
9463 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9464 and the predicate registers.
9465 (aarch64_sve_vg): Declare.
9466 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9467 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9468 (REGMODE_NATURAL_SIZE): Define.
9469 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9471 * config/aarch64/aarch64.c: Include cfgrtl.h.
9472 (simd_immediate_info): Add a constructor for series vectors,
9473 and an associated step field.
9474 (aarch64_sve_vg): New variable.
9475 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9476 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9477 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9478 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9479 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9480 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9481 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9482 (aarch64_get_mask_mode): New functions.
9483 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9484 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9485 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9486 predicate modes and predicate registers. Explicitly restrict
9487 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9488 to store a vector mode if it is recognized by
9489 aarch64_classify_vector_mode.
9490 (aarch64_regmode_natural_size): New function.
9491 (aarch64_hard_regno_caller_save_mode): Return the original mode
9493 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9494 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9495 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9496 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9498 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9499 does not overlap dest if the function is frame-related. Handle
9501 (aarch64_split_add_offset): New function.
9502 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9503 them aarch64_add_offset.
9504 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9505 and update call to aarch64_sub_sp.
9506 (aarch64_add_cfa_expression): New function.
9507 (aarch64_expand_prologue): Pass extra temporary registers to the
9508 functions above. Handle the case in which we need to emit new
9509 DW_CFA_expressions for registers that were originally saved
9510 relative to the stack pointer, but now have to be expressed
9511 relative to the frame pointer.
9512 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9514 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9515 IP0 and IP1 values for SVE frames.
9516 (aarch64_expand_vec_series): New function.
9517 (aarch64_expand_sve_widened_duplicate): Likewise.
9518 (aarch64_expand_sve_const_vector): Likewise.
9519 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9520 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9521 into the register, rather than emitting a SET directly.
9522 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9523 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9524 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9525 (offset_9bit_signed_scaled_p): New functions.
9526 (aarch64_replicate_bitmask_imm): New function.
9527 (aarch64_bitmask_imm): Use it.
9528 (aarch64_cannot_force_const_mem): Reject expressions involving
9529 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9530 (aarch64_classify_index): Handle SVE indices, by requiring
9531 a plain register index with a scale that matches the element size.
9532 (aarch64_classify_address): Handle SVE addresses. Assert that
9533 the mode of the address is VOIDmode or an integer mode.
9534 Update call to aarch64_classify_symbol.
9535 (aarch64_classify_symbolic_expression): Update call to
9536 aarch64_classify_symbol.
9537 (aarch64_const_vec_all_in_range_p): New function.
9538 (aarch64_print_vector_float_operand): Likewise.
9539 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9540 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9541 and the FP immediates 1.0 and 0.5.
9542 (aarch64_print_address_internal): Handle SVE addresses.
9543 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9544 (aarch64_regno_regclass): Handle predicate registers.
9545 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9547 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9548 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9549 (aarch64_convert_sve_vector_bits): New function.
9550 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9551 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9553 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9554 Handle SVE vector and predicate modes. Accept VL-based constants
9555 that need only one temporary register, and VL offsets that require
9556 no temporary registers.
9557 (aarch64_conditional_register_usage): Mark the predicate registers
9558 as fixed if SVE isn't available.
9559 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9560 Return true for SVE vector and predicate modes.
9561 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9562 rather than an unsigned int. Handle SVE modes.
9563 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9565 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9567 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9568 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9569 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9570 (aarch64_sve_float_mul_immediate_p): New functions.
9571 (aarch64_sve_valid_immediate): New function.
9572 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9573 Explicitly reject structure modes. Check for INDEX constants.
9574 Handle PTRUE and PFALSE constants.
9575 (aarch64_check_zero_based_sve_index_immediate): New function.
9576 (aarch64_simd_imm_zero_p): Delete.
9577 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9578 vector modes. Accept constants in the range of CNT[BHWD].
9579 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9580 ask for an Advanced SIMD mode.
9581 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9582 (aarch64_simd_vector_alignment): Handle SVE predicates.
9583 (aarch64_vectorize_preferred_vector_alignment): New function.
9584 (aarch64_simd_vector_alignment_reachable): Use it instead of
9586 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9587 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9589 (MAX_VECT_LEN): Delete.
9590 (expand_vec_perm_d): Add a vec_flags field.
9591 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9592 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9593 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9595 (aarch64_evpc_rev): Rename to...
9596 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9597 (aarch64_evpc_rev_global): New function.
9598 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9599 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9601 (aarch64_evpc_sve_tbl): New function.
9602 (aarch64_expand_vec_perm_const_1): Update after rename of
9603 aarch64_evpc_rev. Handle SVE permutes too, trying
9604 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9605 than aarch64_evpc_tbl.
9606 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9607 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9608 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9609 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9610 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9611 (aarch64_expand_sve_vcond): New functions.
9612 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9613 of aarch64_vector_mode_p.
9614 (aarch64_dwarf_poly_indeterminate_value): New function.
9615 (aarch64_compute_pressure_classes): Likewise.
9616 (aarch64_can_change_mode_class): Likewise.
9617 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9618 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9619 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9620 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9621 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9622 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9623 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9624 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9626 (Dn, Dl, Dr): Accept const as well as const_vector.
9627 (Dz): Likewise. Compare against CONST0_RTX.
9628 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9629 of "vector" where appropriate.
9630 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9631 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9632 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9633 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9634 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9635 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9636 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9637 (v_int_equiv): Extend to SVE modes.
9638 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9640 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9641 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9642 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9643 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9644 (SVE_COND_FP_CMP): New int iterators.
9645 (perm_hilo): Handle the new unpack unspecs.
9646 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9648 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9649 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9650 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9651 (aarch64_equality_operator, aarch64_constant_vector_operand)
9652 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9653 (aarch64_sve_nonimmediate_operand): Likewise.
9654 (aarch64_sve_general_operand): Likewise.
9655 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9656 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9657 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9658 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9659 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9660 (aarch64_sve_float_arith_immediate): Likewise.
9661 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9662 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9663 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9664 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9665 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9666 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9667 (aarch64_sve_float_arith_operand): Likewise.
9668 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9669 (aarch64_sve_float_mul_operand): Likewise.
9670 (aarch64_sve_vec_perm_operand): Likewise.
9671 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9672 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9673 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9674 as well as const_vector.
9675 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9676 in file. Use CONST0_RTX and CONSTM1_RTX.
9677 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9678 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9679 Use aarch64_simd_imm_zero.
9680 * config/aarch64/aarch64-sve.md: New file.
9681 * config/aarch64/aarch64.md: Include it.
9682 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9683 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9684 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9685 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9686 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9687 (sve): New attribute.
9688 (enabled): Disable instructions with the sve attribute unless
9690 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9691 aarch64_expand_mov_immediate.
9692 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9693 CNT[BHSD] immediates.
9694 (movti): Split CONST_POLY_INT moves into two halves.
9695 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9696 Split additions that need a temporary here if the destination
9697 is the stack pointer.
9698 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9699 (*add<mode>3_poly_1): New instruction.
9700 (set_clobber_cc): New expander.
9702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9704 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9705 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9706 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9707 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9708 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9709 Change innermode from fixed_mode_size to machine_mode.
9710 (simplify_subreg): Update call accordingly. Handle a constant-sized
9711 subreg of a variable-length CONST_VECTOR.
9713 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9714 Alan Hayward <alan.hayward@arm.com>
9715 David Sherwood <david.sherwood@arm.com>
9717 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9718 (add_offset_to_base): New function, split out from...
9719 (create_mem_ref): ...here. When handling a scale other than 1,
9720 check first whether the address is valid without the offset.
9721 Add it into the base if so, leaving the index and scale as-is.
9723 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9726 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9727 fold_for_warn before checking if arg2 is INTEGER_CST.
9729 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9731 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9732 (store_multiple_operation): Delete.
9733 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9734 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9735 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9736 guarded by TARGET_STRING.
9737 (rs6000_output_load_multiple): Delete.
9738 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9739 OPTION_MASK_STRING / TARGET_STRING handling.
9740 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9741 (const rs6000_opt_masks) <"string">: Change mask to 0.
9742 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9743 (MASK_STRING): Delete.
9744 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9746 (load_multiple): Delete.
9753 (store_multiple): Delete.
9760 (movmemsi_8reg): Delete.
9761 (corresponding unnamed define_insn): Delete.
9762 (movmemsi_6reg): Delete.
9763 (corresponding unnamed define_insn): Delete.
9764 (movmemsi_4reg): Delete.
9765 (corresponding unnamed define_insn): Delete.
9766 (movmemsi_2reg): Delete.
9767 (corresponding unnamed define_insn): Delete.
9768 (movmemsi_1reg): Delete.
9769 (corresponding unnamed define_insn): Delete.
9770 * config/rs6000/rs6000.opt (mno-string): New.
9771 (mstring): Replace by deprecation warning stub.
9772 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9774 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9776 * regrename.c (regrename_do_replace): If replacing the same
9777 reg multiple times, try to reuse last created gen_raw_REG.
9780 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9781 main to workaround a bug in GDB.
9783 2018-01-12 Tom de Vries <tom@codesourcery.com>
9786 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9788 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9790 PR rtl-optimization/80481
9791 * ira-color.c (get_cap_member): New function.
9792 (allocnos_conflict_by_live_ranges_p): Use it.
9793 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9794 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9796 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9799 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9800 (*saddl_se_1): Ditto.
9802 (*ssubl_se_1): Ditto.
9804 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9806 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9807 rather than wi::to_widest for DR_INITs.
9808 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9809 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9810 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9812 (vect_analyze_group_access_1): Note that here.
9814 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9816 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9817 polynomial type sizes.
9819 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9821 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9822 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9823 (gimple_add_tmp_var): Likewise.
9825 2018-01-12 Martin Liska <mliska@suse.cz>
9827 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9828 (gimple_alloc_sizes): Likewise.
9829 (dump_gimple_statistics): Use PRIu64 in printf format.
9830 * gimple.h: Change uint64_t to int.
9832 2018-01-12 Martin Liska <mliska@suse.cz>
9834 * tree-core.h: Use uint64_t instead of int.
9835 * tree.c (tree_node_counts): Likewise.
9836 (tree_node_sizes): Likewise.
9837 (dump_tree_statistics): Use PRIu64 in printf format.
9839 2018-01-12 Martin Liska <mliska@suse.cz>
9841 * Makefile.in: As qsort_chk is implemented in vec.c, add
9842 vec.o to linkage of gencfn-macros.
9843 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9844 passing the info to record_node_allocation_statistics.
9845 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9847 * ggc-common.c (struct ggc_usage): Add operator== and use
9848 it in operator< and compare function.
9849 * mem-stats.h (struct mem_usage): Likewise.
9850 * vec.c (struct vec_usage): Remove operator< and compare
9851 function. Can be simply inherited.
9853 2018-01-12 Martin Jambor <mjambor@suse.cz>
9856 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9857 * tree-ssa-math-opts.c: Include domwalk.h.
9858 (convert_mult_to_fma_1): New function.
9859 (fma_transformation_info): New type.
9860 (fma_deferring_state): Likewise.
9861 (cancel_fma_deferring): New function.
9862 (result_of_phi): Likewise.
9863 (last_fma_candidate_feeds_initial_phi): Likewise.
9864 (convert_mult_to_fma): Added deferring logic, split actual
9865 transformation to convert_mult_to_fma_1.
9866 (math_opts_dom_walker): New type.
9867 (math_opts_dom_walker::after_dom_children): New method, body moved
9868 here from pass_optimize_widening_mul::execute, added deferring logic
9870 (pass_optimize_widening_mul::execute): Moved most of code to
9871 math_opts_dom_walker::after_dom_children.
9872 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9873 * config/i386/i386.c (ix86_option_override_internal): Added
9874 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9876 2018-01-12 Richard Biener <rguenther@suse.de>
9879 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9880 inline instance vars.
9882 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9885 * config/rx/rx.c (rx_is_restricted_memory_address):
9888 2018-01-12 Richard Biener <rguenther@suse.de>
9890 PR tree-optimization/80846
9891 * target.def (split_reduction): New target hook.
9892 * targhooks.c (default_split_reduction): New function.
9893 * targhooks.h (default_split_reduction): Declare.
9894 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9895 target requests first reduce vectors by combining low and high
9897 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9898 (get_vectype_for_scalar_type_and_size): Export.
9899 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9900 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9901 * doc/tm.texi: Regenerate.
9902 * config/i386/i386.c (ix86_split_reduction): Implement
9903 TARGET_VECTORIZE_SPLIT_REDUCTION.
9905 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9908 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9909 in PIC mode except for TARGET_VXWORKS_RTP.
9910 * config/sparc/sparc.c: Include cfgrtl.h.
9911 (TARGET_INIT_PIC_REG): Define.
9912 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9913 (sparc_pic_register_p): New predicate.
9914 (sparc_legitimate_address_p): Use it.
9915 (sparc_legitimize_pic_address): Likewise.
9916 (sparc_delegitimize_address): Likewise.
9917 (sparc_mode_dependent_address_p): Likewise.
9918 (gen_load_pcrel_sym): Remove 4th parameter.
9919 (load_got_register): Adjust call to above. Remove obsolete stuff.
9920 (sparc_expand_prologue): Do not call load_got_register here.
9921 (sparc_flat_expand_prologue): Likewise.
9922 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9923 (sparc_use_pseudo_pic_reg): New function.
9924 (sparc_init_pic_reg): Likewise.
9925 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9926 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9928 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9930 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9931 Add item for branch_cost.
9933 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9935 PR rtl-optimization/83565
9936 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9937 not extend the result to a larger mode for rotate operations.
9938 (num_sign_bit_copies1): Likewise.
9940 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9943 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9945 Use values-Xc.o for -pedantic.
9946 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9948 2018-01-12 Martin Liska <mliska@suse.cz>
9951 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9953 (possible_polymorphic_call_targets): Use it.
9954 (ipa_devirt): Likewise.
9956 2018-01-12 Martin Liska <mliska@suse.cz>
9958 * profile-count.h (enum profile_quality): Use 0 as invalid
9959 enum value of profile_quality.
9961 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9963 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9964 -mext-string options.
9966 2018-01-12 Richard Biener <rguenther@suse.de>
9968 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9969 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9970 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9972 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9974 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9976 * configure.ac (--with-long-double-format): Add support for the
9977 configuration option to change the default long double format on
9979 * config.gcc (powerpc*-linux*-*): Likewise.
9980 * configure: Regenerate.
9981 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9982 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9983 used without modification.
9985 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9987 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9988 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9989 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9990 MISC_BUILTIN_SPEC_BARRIER.
9991 (rs6000_init_builtins): Likewise.
9992 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9994 (speculation_barrier): New define_insn.
9995 * doc/extend.texi: Document __builtin_speculation_barrier.
9997 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10000 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10001 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10002 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10004 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10005 integral modes instead of "ss" and "sd".
10006 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10007 vectors with 32-bit and 64-bit elements.
10008 (vecdupssescalarmodesuffix): New mode attribute.
10009 (vec_dup<mode>): Use it.
10011 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10014 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10015 frame if argument is passed on stack.
10017 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10020 * ree.c (combine_reaching_defs): Optimize also
10021 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10022 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10024 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10026 PR middle-end/83189
10027 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10029 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10031 PR middle-end/83718
10032 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10033 after they are computed.
10035 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10037 PR tree-optimization/83695
10038 * gimple-loop-linterchange.cc
10039 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10040 reset cached scev information after interchange.
10041 (pass_linterchange::execute): Remove call to scev_reset_htab.
10043 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10045 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10046 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10047 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10048 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10049 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10050 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10051 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10052 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10053 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10054 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10055 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10056 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10057 (V_lane_reg): Likewise.
10058 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10060 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10061 (vfmal_lane_low<mode>_intrinsic,
10062 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10063 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10064 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10065 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10066 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10067 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10069 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10071 * config/arm/arm-cpus.in (fp16fml): New feature.
10072 (ALL_SIMD): Add fp16fml.
10073 (armv8.2-a): Add fp16fml as an option.
10074 (armv8.3-a): Likewise.
10075 (armv8.4-a): Add fp16fml as part of fp16.
10076 * config/arm/arm.h (TARGET_FP16FML): Define.
10077 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10079 * config/arm/arm-modes.def (V2HF): Define.
10080 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10081 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10082 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10083 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10084 vfmsl_low, vfmsl_high): New set of builtins.
10085 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10086 (vfml_op): New code attribute.
10087 (VFMLHALVES): New int iterator.
10088 (VFML, VFMLSEL): New mode attributes.
10089 (V_reg): Define mapping for V2HF.
10090 (V_hi, V_lo): New mode attributes.
10091 (VF_constraint): Likewise.
10092 (vfml_half, vfml_half_selector): New int attributes.
10093 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10095 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10096 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10098 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10099 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10100 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10101 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10103 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10104 Document new effective target and option set.
10106 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10108 * config/arm/arm-cpus.in (armv8_4): New feature.
10109 (ARMv8_4a): New fgroup.
10110 (armv8.4-a): New arch.
10111 * config/arm/arm-tables.opt: Regenerate.
10112 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10113 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10114 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10115 Add matching rules for -march=armv8.4-a and extensions.
10116 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10118 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10121 * config/rx/rx.md (BW): New mode attribute.
10122 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10124 2018-01-11 Richard Biener <rguenther@suse.de>
10126 PR tree-optimization/83435
10127 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10128 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10129 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10131 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10132 Alan Hayward <alan.hayward@arm.com>
10133 David Sherwood <david.sherwood@arm.com>
10135 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10137 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10138 (aarch64_print_address_internal): Use it to check for a zero offset.
10140 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10141 Alan Hayward <alan.hayward@arm.com>
10142 David Sherwood <david.sherwood@arm.com>
10144 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10145 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10146 Return a poly_int64 rather than a HOST_WIDE_INT.
10147 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10148 rather than a HOST_WIDE_INT.
10149 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10150 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10151 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10152 final_offset from HOST_WIDE_INT to poly_int64.
10153 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10154 to_constant when getting the number of units in an Advanced SIMD
10156 (aarch64_builtin_vectorized_function): Check for a constant number
10158 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10160 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10161 attribute instead of GET_MODE_NUNITS.
10162 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10163 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10164 GET_MODE_SIZE for fixed-size registers.
10165 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10166 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10167 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10168 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10169 (aarch64_print_operand, aarch64_print_address_internal)
10170 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10171 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10172 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10173 Handle polynomial GET_MODE_SIZE.
10174 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10175 wider than SImode without modification.
10176 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10177 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10178 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10179 passing and returning SVE modes.
10180 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10181 rather than GEN_INT.
10182 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10183 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10184 (aarch64_allocate_and_probe_stack_space): Likewise.
10185 (aarch64_layout_frame): Cope with polynomial offsets.
10186 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10187 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10188 polynomial offsets.
10189 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10190 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10191 poly_int64 rather than a HOST_WIDE_INT.
10192 (aarch64_get_separate_components, aarch64_process_components)
10193 (aarch64_expand_prologue, aarch64_expand_epilogue)
10194 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10195 (aarch64_anchor_offset): New function, split out from...
10196 (aarch64_legitimize_address): ...here.
10197 (aarch64_builtin_vectorization_cost): Handle polynomial
10198 TYPE_VECTOR_SUBPARTS.
10199 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10201 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10202 number of elements from the PARALLEL rather than the mode.
10203 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10204 rather than GET_MODE_BITSIZE.
10205 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10206 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10207 (aarch64_expand_vec_perm_const_1): Handle polynomial
10208 d->perm.length () and d->perm elements.
10209 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10210 Apply to_constant to d->perm elements.
10211 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10212 polynomial CONST_VECTOR_NUNITS.
10213 (aarch64_move_pointer): Take amount as a poly_int64 rather
10215 (aarch64_progress_pointer): Avoid temporary variable.
10216 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10217 the mode attribute instead of GET_MODE.
10219 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10220 Alan Hayward <alan.hayward@arm.com>
10221 David Sherwood <david.sherwood@arm.com>
10223 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10224 x exists before using it.
10225 (aarch64_add_constant_internal): Rename to...
10226 (aarch64_add_offset_1): ...this. Replace regnum with separate
10227 src and dest rtxes. Handle the case in which they're different,
10228 including when the offset is zero. Replace scratchreg with an rtx.
10229 Use 2 additions if there is no spare register into which we can
10230 move a 16-bit constant.
10231 (aarch64_add_constant): Delete.
10232 (aarch64_add_offset): Replace reg with separate src and dest
10233 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10234 Use aarch64_add_offset_1.
10235 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10236 an rtx rather than an int. Take the delta as a poly_int64
10237 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10238 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10239 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10240 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10241 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10242 and aarch64_add_sp.
10243 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10244 aarch64_add_constant.
10246 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10248 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10249 Use scalar_float_mode.
10251 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10253 * config/aarch64/aarch64-simd.md
10254 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10255 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10256 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10257 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10258 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10259 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10260 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10261 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10262 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10263 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10265 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10268 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10269 targ_options->x_arm_arch_string is non NULL.
10271 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10273 * config/aarch64/aarch64.h
10274 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10276 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10279 * expmed.c (emit_store_flag_force): Swap if const op0
10280 and change VOIDmode to mode of op0.
10282 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10284 PR rtl-optimization/83761
10285 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10286 than bytes to mode_for_size.
10288 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10290 PR middle-end/83189
10291 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10292 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10295 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10297 PR middle-end/83575
10298 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10299 when in layout mode.
10300 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10301 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10304 2018-01-10 Michael Collison <michael.collison@arm.com>
10306 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10307 * config/aarch64/aarch64-option-extension.def: Add
10308 AARCH64_OPT_EXTENSION of 'fp16fml'.
10309 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10310 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10311 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10312 * config/aarch64/constraints.md (Ui7): New constraint.
10313 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10314 (VFMLA_SEL_W): Ditto.
10317 (VFMLA16_LOW): New int iterator.
10318 (VFMLA16_HIGH): Ditto.
10319 (UNSPEC_FMLAL): New unspec.
10320 (UNSPEC_FMLSL): Ditto.
10321 (UNSPEC_FMLAL2): Ditto.
10322 (UNSPEC_FMLSL2): Ditto.
10323 (f16mac): New code attribute.
10324 * config/aarch64/aarch64-simd-builtins.def
10325 (aarch64_fmlal_lowv2sf): Ditto.
10326 (aarch64_fmlsl_lowv2sf): Ditto.
10327 (aarch64_fmlalq_lowv4sf): Ditto.
10328 (aarch64_fmlslq_lowv4sf): Ditto.
10329 (aarch64_fmlal_highv2sf): Ditto.
10330 (aarch64_fmlsl_highv2sf): Ditto.
10331 (aarch64_fmlalq_highv4sf): Ditto.
10332 (aarch64_fmlslq_highv4sf): Ditto.
10333 (aarch64_fmlal_lane_lowv2sf): Ditto.
10334 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10335 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10336 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10337 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10338 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10339 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10340 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10341 (aarch64_fmlal_lane_highv2sf): Ditto.
10342 (aarch64_fmlsl_lane_highv2sf): Ditto.
10343 (aarch64_fmlal_laneq_highv2sf): Ditto.
10344 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10345 (aarch64_fmlalq_lane_highv4sf): Ditto.
10346 (aarch64_fmlsl_lane_highv4sf): Ditto.
10347 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10348 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10349 * config/aarch64/aarch64-simd.md:
10350 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10351 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10352 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10353 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10354 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10355 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10356 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10357 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10358 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10359 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10360 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10361 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10362 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10363 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10364 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10365 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10366 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10367 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10368 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10369 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10370 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10371 (vfmlsl_low_u32): Ditto.
10372 (vfmlalq_low_u32): Ditto.
10373 (vfmlslq_low_u32): Ditto.
10374 (vfmlal_high_u32): Ditto.
10375 (vfmlsl_high_u32): Ditto.
10376 (vfmlalq_high_u32): Ditto.
10377 (vfmlslq_high_u32): Ditto.
10378 (vfmlal_lane_low_u32): Ditto.
10379 (vfmlsl_lane_low_u32): Ditto.
10380 (vfmlal_laneq_low_u32): Ditto.
10381 (vfmlsl_laneq_low_u32): Ditto.
10382 (vfmlalq_lane_low_u32): Ditto.
10383 (vfmlslq_lane_low_u32): Ditto.
10384 (vfmlalq_laneq_low_u32): Ditto.
10385 (vfmlslq_laneq_low_u32): Ditto.
10386 (vfmlal_lane_high_u32): Ditto.
10387 (vfmlsl_lane_high_u32): Ditto.
10388 (vfmlal_laneq_high_u32): Ditto.
10389 (vfmlsl_laneq_high_u32): Ditto.
10390 (vfmlalq_lane_high_u32): Ditto.
10391 (vfmlslq_lane_high_u32): Ditto.
10392 (vfmlalq_laneq_high_u32): Ditto.
10393 (vfmlslq_laneq_high_u32): Ditto.
10394 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10395 (AARCH64_FL_FOR_ARCH8_4): New.
10396 (AARCH64_ISA_F16FML): New ISA flag.
10397 (TARGET_F16FML): New feature flag for fp16fml.
10398 (doc/invoke.texi): Document new fp16fml option.
10400 2018-01-10 Michael Collison <michael.collison@arm.com>
10402 * config/aarch64/aarch64-builtins.c:
10403 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10404 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10405 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10406 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10407 (AARCH64_ISA_SHA3): New ISA flag.
10408 (TARGET_SHA3): New feature flag for sha3.
10409 * config/aarch64/iterators.md (sha512_op): New int attribute.
10410 (CRYPTO_SHA512): New int iterator.
10411 (UNSPEC_SHA512H): New unspec.
10412 (UNSPEC_SHA512H2): Ditto.
10413 (UNSPEC_SHA512SU0): Ditto.
10414 (UNSPEC_SHA512SU1): Ditto.
10415 * config/aarch64/aarch64-simd-builtins.def
10416 (aarch64_crypto_sha512hqv2di): New builtin.
10417 (aarch64_crypto_sha512h2qv2di): Ditto.
10418 (aarch64_crypto_sha512su0qv2di): Ditto.
10419 (aarch64_crypto_sha512su1qv2di): Ditto.
10420 (aarch64_eor3qv8hi): Ditto.
10421 (aarch64_rax1qv2di): Ditto.
10422 (aarch64_xarqv2di): Ditto.
10423 (aarch64_bcaxqv8hi): Ditto.
10424 * config/aarch64/aarch64-simd.md:
10425 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10426 (aarch64_crypto_sha512su0qv2di): Ditto.
10427 (aarch64_crypto_sha512su1qv2di): Ditto.
10428 (aarch64_eor3qv8hi): Ditto.
10429 (aarch64_rax1qv2di): Ditto.
10430 (aarch64_xarqv2di): Ditto.
10431 (aarch64_bcaxqv8hi): Ditto.
10432 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10433 (vsha512h2q_u64): Ditto.
10434 (vsha512su0q_u64): Ditto.
10435 (vsha512su1q_u64): Ditto.
10436 (veor3q_u16): Ditto.
10437 (vrax1q_u64): Ditto.
10438 (vxarq_u64): Ditto.
10439 (vbcaxq_u16): Ditto.
10440 * config/arm/types.md (crypto_sha512): New type attribute.
10441 (crypto_sha3): Ditto.
10442 (doc/invoke.texi): Document new sha3 option.
10444 2018-01-10 Michael Collison <michael.collison@arm.com>
10446 * config/aarch64/aarch64-builtins.c:
10447 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10448 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10449 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10450 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10451 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10452 (AARCH64_ISA_SM4): New ISA flag.
10453 (TARGET_SM4): New feature flag for sm4.
10454 * config/aarch64/aarch64-simd-builtins.def
10455 (aarch64_sm3ss1qv4si): Ditto.
10456 (aarch64_sm3tt1aq4si): Ditto.
10457 (aarch64_sm3tt1bq4si): Ditto.
10458 (aarch64_sm3tt2aq4si): Ditto.
10459 (aarch64_sm3tt2bq4si): Ditto.
10460 (aarch64_sm3partw1qv4si): Ditto.
10461 (aarch64_sm3partw2qv4si): Ditto.
10462 (aarch64_sm4eqv4si): Ditto.
10463 (aarch64_sm4ekeyqv4si): Ditto.
10464 * config/aarch64/aarch64-simd.md:
10465 (aarch64_sm3ss1qv4si): Ditto.
10466 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10467 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10468 (aarch64_sm4eqv4si): Ditto.
10469 (aarch64_sm4ekeyqv4si): Ditto.
10470 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10471 (sm3part_op): Ditto.
10472 (CRYPTO_SM3TT): Ditto.
10473 (CRYPTO_SM3PART): Ditto.
10474 (UNSPEC_SM3SS1): New unspec.
10475 (UNSPEC_SM3TT1A): Ditto.
10476 (UNSPEC_SM3TT1B): Ditto.
10477 (UNSPEC_SM3TT2A): Ditto.
10478 (UNSPEC_SM3TT2B): Ditto.
10479 (UNSPEC_SM3PARTW1): Ditto.
10480 (UNSPEC_SM3PARTW2): Ditto.
10481 (UNSPEC_SM4E): Ditto.
10482 (UNSPEC_SM4EKEY): Ditto.
10483 * config/aarch64/constraints.md (Ui2): New constraint.
10484 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10485 * config/arm/types.md (crypto_sm3): New type attribute.
10486 (crypto_sm4): Ditto.
10487 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10488 (vsm3tt1aq_u32): Ditto.
10489 (vsm3tt1bq_u32): Ditto.
10490 (vsm3tt2aq_u32): Ditto.
10491 (vsm3tt2bq_u32): Ditto.
10492 (vsm3partw1q_u32): Ditto.
10493 (vsm3partw2q_u32): Ditto.
10494 (vsm4eq_u32): Ditto.
10495 (vsm4ekeyq_u32): Ditto.
10496 (doc/invoke.texi): Document new sm4 option.
10498 2018-01-10 Michael Collison <michael.collison@arm.com>
10500 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10501 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10502 (AARCH64_FL_FOR_ARCH8_4): New.
10503 (AARCH64_FL_V8_4): New flag.
10504 (doc/invoke.texi): Document new armv8.4-a option.
10506 2018-01-10 Michael Collison <michael.collison@arm.com>
10508 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10509 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10510 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10511 * config/aarch64/aarch64-option-extension.def: Add
10512 AARCH64_OPT_EXTENSION of 'sha2'.
10513 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10514 (crypto): Disable sha2 and aes if crypto disabled.
10515 (crypto): Enable aes and sha2 if enabled.
10516 (simd): Disable sha2 and aes if simd disabled.
10517 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10519 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10520 (TARGET_SHA2): New feature flag for sha2.
10521 (TARGET_AES): New feature flag for aes.
10522 * config/aarch64/aarch64-simd.md:
10523 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10524 conditional on TARGET_AES.
10525 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10526 (aarch64_crypto_sha1hsi): Make pattern conditional
10528 (aarch64_crypto_sha1hv4si): Ditto.
10529 (aarch64_be_crypto_sha1hv4si): Ditto.
10530 (aarch64_crypto_sha1su1v4si): Ditto.
10531 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10532 (aarch64_crypto_sha1su0v4si): Ditto.
10533 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10534 (aarch64_crypto_sha256su0v4si): Ditto.
10535 (aarch64_crypto_sha256su1v4si): Ditto.
10536 (doc/invoke.texi): Document new aes and sha2 options.
10538 2018-01-10 Martin Sebor <msebor@redhat.com>
10540 PR tree-optimization/83781
10541 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10544 2018-01-11 Martin Sebor <msebor@gmail.com>
10545 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10547 PR tree-optimization/83501
10548 PR tree-optimization/81703
10550 * tree-ssa-strlen.c (get_string_cst): Rename...
10551 (get_string_len): ...to this. Handle global constants.
10552 (handle_char_store): Adjust.
10554 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10555 Jim Wilson <jimw@sifive.com>
10557 * config/riscv/riscv-protos.h (riscv_output_return): New.
10558 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10559 (riscv_attribute_table, riscv_output_return),
10560 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10561 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10562 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10564 (riscv_expand_prologue): Add early return for naked function.
10565 (riscv_expand_epilogue): Likewise.
10566 (riscv_function_ok_for_sibcall): Return false for naked function.
10567 (riscv_set_current_function): New.
10568 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10569 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10570 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10571 * doc/extend.texi (RISC-V Function Attributes): New.
10573 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10575 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10576 check for 128-bit long double before checking TCmode.
10577 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10578 128-bit long doubles before checking TFmode or TCmode.
10579 (FLOAT128_IBM_P): Likewise.
10581 2018-01-10 Martin Sebor <msebor@redhat.com>
10583 PR tree-optimization/83671
10584 * builtins.c (c_strlen): Unconditionally return zero for the empty
10586 Use -Warray-bounds for warnings.
10587 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10588 for non-constant array indices with COMPONENT_REF, arrays of
10589 arrays, and pointers to arrays.
10590 (gimple_fold_builtin_strlen): Determine and set length range for
10591 non-constant character arrays.
10593 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10595 PR middle-end/81897
10596 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10599 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10601 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10603 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10606 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10607 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10608 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10609 indexed_or_indirect_operand predicate.
10610 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10611 (*vsx_le_perm_load_v8hi): Likewise.
10612 (*vsx_le_perm_load_v16qi): Likewise.
10613 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10614 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10615 (*vsx_le_perm_store_v8hi): Likewise.
10616 (*vsx_le_perm_store_v16qi): Likewise.
10617 (eight unnamed splitters): Likewise.
10619 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10621 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10622 * config/rs6000/emmintrin.h: Likewise.
10623 * config/rs6000/mmintrin.h: Likewise.
10624 * config/rs6000/xmmintrin.h: Likewise.
10626 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10629 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10631 * tree.c (tree_nop_conversion): Return true for location wrapper
10633 (maybe_wrap_with_location): New function.
10634 (selftest::check_strip_nops): New function.
10635 (selftest::test_location_wrappers): New function.
10636 (selftest::tree_c_tests): Call it.
10637 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10638 (maybe_wrap_with_location): New decl.
10639 (EXPR_LOCATION_WRAPPER_P): New macro.
10640 (location_wrapper_p): New inline function.
10641 (tree_strip_any_location_wrapper): New inline function.
10643 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10646 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10647 stack_realign_offset for the largest alignment of stack slot
10649 (ix86_find_max_used_stack_alignment): New function.
10650 (ix86_finalize_stack_frame_flags): Use it. Set
10651 max_used_stack_alignment if we don't realign stack.
10652 * config/i386/i386.h (machine_function): Add
10653 max_used_stack_alignment.
10655 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10657 * config/arm/arm.opt (-mbranch-cost): New option.
10658 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10661 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10664 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10665 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10667 2018-01-10 Richard Biener <rguenther@suse.de>
10670 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10671 early out so it also covers the case where we have a non-NULL
10674 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10676 PR tree-optimization/83753
10677 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10678 for non-strided grouped accesses if the number of elements is 1.
10680 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10683 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10684 * i386.h (TARGET_USE_GATHER): Define.
10685 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10687 2018-01-10 Martin Liska <mliska@suse.cz>
10690 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10691 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10693 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10694 CLEANUP_NO_PARTITIONING is not set.
10696 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10698 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10699 for vectors, as a partial revert of r254296.
10700 * rtl.h (const_vec_p): Delete.
10701 (const_vec_duplicate_p): Don't test for vector CONSTs.
10702 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10703 * expmed.c (make_tree): Likewise.
10706 * common.md (E, F): Use CONSTANT_P instead of checking for
10708 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10709 checking for CONST_VECTOR.
10711 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10713 PR middle-end/83575
10714 * predict.c (force_edge_cold): Handle in more sane way edges
10715 with no prediction.
10717 2018-01-09 Carl Love <cel@us.ibm.com>
10719 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10721 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10722 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10723 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10724 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10725 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10726 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10727 * config/rs6000/rs6000-protos.h: Add extern defition for
10728 rs6000_generate_float2_double_code.
10729 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10731 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10732 (float2_v2df): Add define_expand.
10734 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10737 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10738 op_mode in the force_to_mode call.
10740 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10742 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10743 instead of checking each element individually.
10744 (aarch64_evpc_uzp): Likewise.
10745 (aarch64_evpc_zip): Likewise.
10746 (aarch64_evpc_ext): Likewise.
10747 (aarch64_evpc_rev): Likewise.
10748 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10749 instead of checking each element individually. Return true without
10751 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10752 whether all selected elements come from the same input, instead of
10753 checking each element individually. Remove calls to gen_rtx_REG,
10754 start_sequence and end_sequence and instead assert that no rtl is
10757 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10759 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10760 order of HIGH and CONST checks.
10762 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10764 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10765 if the destination isn't an SSA_NAME.
10767 2018-01-09 Richard Biener <rguenther@suse.de>
10769 PR tree-optimization/83668
10770 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10772 (canonicalize_loop_form): ... here, renamed from ...
10773 (canonicalize_loop_closed_ssa_form): ... this and amended to
10774 swap successor edges for loop exit blocks to make us use
10775 the RPO order we need for initial schedule generation.
10777 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10779 PR tree-optimization/64811
10780 * match.pd: When optimizing comparisons with Inf, avoid
10781 introducing or losing exceptions from comparisons with NaN.
10783 2018-01-09 Martin Liska <mliska@suse.cz>
10786 * asan.c (shadow_mem_size): Add gcc_assert.
10788 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10790 Don't save registers in main().
10793 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10794 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10795 * config/avr/avr.c (avr_set_current_function): Don't error if
10796 naked, OS_task or OS_main are specified at the same time.
10797 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10799 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10801 * common/config/avr/avr-common.c (avr_option_optimization_table):
10802 Switch on -mmain-is-OS_task for optimizing compilations.
10804 2018-01-09 Richard Biener <rguenther@suse.de>
10806 PR tree-optimization/83572
10807 * graphite.c: Include cfganal.h.
10808 (graphite_transform_loops): Connect infinite loops to exit
10809 and remove fake edges at the end.
10811 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10813 * ipa-inline.c (edge_badness): Revert accidental checkin.
10815 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10818 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10819 symbols; not inline clones.
10821 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10824 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10825 hard registers. Formatting fixes.
10827 PR preprocessor/83722
10828 * gcc.c (try_generate_repro): Pass
10829 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10830 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10833 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10834 Kito Cheng <kito.cheng@gmail.com>
10836 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10837 (riscv_leaf_function_p): Delete.
10838 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10840 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10842 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10844 (do_ifelse): New function.
10845 (do_isel): New function.
10846 (do_sub3): New function.
10847 (do_add3): New function.
10848 (do_load_mask_compare): New function.
10849 (do_overlap_load_compare): New function.
10850 (expand_compare_loop): New function.
10851 (expand_block_compare): Call expand_compare_loop() when appropriate.
10852 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10853 option description.
10854 (-mblock-compare-inline-loop-limit): New option.
10856 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10859 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10860 Reverse order of second and third operands in first alternative.
10861 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10862 of first and second elements in UNSPEC_VPERMR vector.
10863 (altivec_expand_vec_perm_le): Likewise.
10865 2018-01-08 Jeff Law <law@redhat.com>
10867 PR rtl-optimizatin/81308
10868 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10869 (process_switch): If group_case_labels makes a change, then set
10871 (pass_convert_switch::execute): If a switch is converted, then
10872 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10874 PR rtl-optimization/81308
10875 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10878 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10880 PR target/83663 - Revert r255946
10881 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10882 generation for cases where splatting a value is not useful.
10883 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10884 across a vec_duplicate and a paradoxical subreg forming a vector
10885 mode to a vec_concat.
10887 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10889 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10890 -march=armv8.3-a variants.
10891 * config/arm/t-multilib: Likewise.
10892 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10894 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10896 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10898 (cceq_ior_compare_complement): Give it a name so I can use it, and
10899 change boolean_or_operator predicate to boolean_operator so it can
10900 be used to generate a crand.
10901 (eqne): New code iterator.
10902 (bd/bd_neg): New code_attrs.
10903 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10904 a single define_insn.
10905 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10906 decrement (bdnzt/bdnzf/bdzt/bdzf).
10907 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10908 with the new names of the branch decrement patterns, and added the
10909 names of the branch decrement conditional patterns.
10911 2018-01-08 Richard Biener <rguenther@suse.de>
10913 PR tree-optimization/83563
10914 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10917 2018-01-08 Richard Biener <rguenther@suse.de>
10919 PR middle-end/83713
10920 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10922 2018-01-08 Richard Biener <rguenther@suse.de>
10924 PR tree-optimization/83685
10925 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10926 references to abnormals.
10928 2018-01-08 Richard Biener <rguenther@suse.de>
10931 * dwarf2out.c (output_indirect_strings): Handle empty
10932 skeleton_debug_str_hash.
10933 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10935 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10937 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10938 (emit_store_direct): Likewise.
10939 (arc_trampoline_adjust_address): Likewise.
10940 (arc_asm_trampoline_template): New function.
10941 (arc_initialize_trampoline): Use asm_trampoline_template.
10942 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10943 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10944 * config/arc/arc.md (flush_icache): Delete pattern.
10946 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10948 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10949 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10952 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10955 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10956 by not USED_FOR_TARGET.
10957 (make_pass_resolve_sw_modes): Likewise.
10959 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10961 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10964 2018-01-08 Richard Biener <rguenther@suse.de>
10966 PR middle-end/83580
10967 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10969 2018-01-08 Richard Biener <rguenther@suse.de>
10971 PR middle-end/83517
10972 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10974 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10976 PR middle-end/81897
10977 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10978 basic blocks with a small number of successors.
10979 (convert_control_dep_chain_into_preds): Improve handling of
10981 (dump_predicates): Split apart into...
10982 (dump_pred_chain): ...here...
10983 (dump_pred_info): ...and here.
10984 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10985 (can_chain_union_be_invalidated_p): Improve check for invalidation
10987 (uninit_uses_cannot_happen): Avoid unnecessary if
10988 convert_control_dep_chain_into_preds yielded nothing.
10990 2018-01-06 Martin Sebor <msebor@redhat.com>
10992 PR tree-optimization/83640
10993 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10994 subtracting negative offset from size.
10995 (builtin_access::overlap): Adjust offset bounds of the access to fall
10996 within the size of the object if possible.
10998 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11000 PR rtl-optimization/83699
11001 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11002 extract_bit_field_as_subreg to cases in which the extracted
11003 value is also a vector.
11005 * lra-constraints.c (process_alt_operands): Test for the equivalence
11006 substitutions when detecting a possible reload cycle.
11008 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11011 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11012 by default if flag_selective_schedling{,2}. Formatting fixes.
11014 PR rtl-optimization/83682
11015 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11016 if it has non-VECTOR_MODE element mode.
11017 (vec_duplicate_p): Likewise.
11019 PR middle-end/83694
11020 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11021 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11023 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11026 * config/i386/i386-builtin.def
11027 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11028 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11029 Require also OPTION_MASK_ISA_AVX512F in addition to
11030 OPTION_MASK_ISA_GFNI.
11031 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11032 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11033 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11034 to OPTION_MASK_ISA_GFNI.
11035 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11036 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11037 OPTION_MASK_ISA_AVX512BW.
11038 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11039 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11040 addition to OPTION_MASK_ISA_GFNI.
11041 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11042 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11043 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11044 to OPTION_MASK_ISA_GFNI.
11045 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11046 a requirement for all ISAs rather than any of them with a few
11048 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11050 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11051 bitmasks to be enabled with 3 exceptions, instead of requiring any
11052 enabled ISA with lots of exceptions.
11053 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11054 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11055 Change avx512bw in isa attribute to avx512f.
11056 * config/i386/sgxintrin.h: Add license boilerplate.
11057 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11058 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11059 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11060 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11062 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11063 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11064 temporarily sse2 rather than sse if not enabled already.
11067 * config/i386/sse.md (VI248_VLBW): Rename to ...
11068 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11069 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11070 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11071 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11072 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11073 mode iterator instead of VI248_VLBW.
11075 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11077 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11078 (record_modified): Skip clobbers; add debug output.
11079 (param_change_prob): Use sreal frequencies.
11081 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11083 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11084 punt for user-aligned variables.
11086 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11088 * tree-chrec.c (chrec_contains_symbols): Return true for
11091 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11094 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11095 of (x|y) == x for BICS pattern.
11097 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11099 PR tree-optimization/83605
11100 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11101 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11104 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11106 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11107 * config/epiphany/rtems.h: New file.
11109 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11110 Uros Bizjak <ubizjak@gmail.com>
11113 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11114 QIreg_operand instead of register_operand predicate.
11115 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11116 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11117 comments instead of -fmitigate[-_]rop.
11119 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11122 * cgraphunit.c (symbol_table::compile): Switch to text_section
11123 before calling assembly_start debug hook.
11124 * run-rtl-passes.c (run_rtl_passes): Likewise.
11127 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11129 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11130 range_int_cst_p rather than !symbolic_range_p before calling
11131 extract_range_from_multiplicative_op_1.
11133 2018-01-04 Jeff Law <law@redhat.com>
11135 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11136 redundant test in assertion.
11138 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11140 * doc/rtl.texi: Document machine_mode wrapper classes.
11142 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11144 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11145 using tree_to_uhwi.
11147 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11149 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11150 the VEC_PERM_EXPR fold to fail.
11152 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11155 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11156 to switched_sections.
11158 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11161 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11162 test for d.testing.
11164 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11167 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11168 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11170 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11173 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11174 is BLKmode and bitpos not zero or mode change is needed.
11176 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11179 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11182 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11185 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11186 instead of MULT rtx. Update all corresponding splitters.
11187 (*saddl_se): Ditto.
11188 (*ssub<modesuffix>): Ditto.
11189 (*ssubl_se): Ditto.
11190 (*cmp_sadd_di): Update split patterns.
11191 (*cmp_sadd_si): Ditto.
11192 (*cmp_sadd_sidi): Ditto.
11193 (*cmp_ssub_di): Ditto.
11194 (*cmp_ssub_si): Ditto.
11195 (*cmp_ssub_sidi): Ditto.
11196 * config/alpha/predicates.md (const23_operand): New predicate.
11197 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11198 Look for ASHIFT, not MULT inner operand.
11199 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11201 2018-01-04 Martin Liska <mliska@suse.cz>
11203 PR gcov-profile/83669
11204 * gcov.c (output_intermediate_file): Add version to intermediate
11206 * doc/gcov.texi: Document new field 'version' in intermediate
11207 file format. Fix location of '-k' option of gcov command.
11209 2018-01-04 Martin Liska <mliska@suse.cz>
11212 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11214 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11216 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11218 2018-01-03 Martin Sebor <msebor@redhat.com>
11220 PR tree-optimization/83655
11221 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11222 checking calls with invalid arguments.
11224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11226 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11227 (vectorizable_mask_load_store): Delete.
11228 (vectorizable_call): Return false for masked loads and stores.
11229 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11230 instead of gimple_assign_rhs1.
11231 (vectorizable_load): Handle IFN_MASK_LOAD.
11232 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11236 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11238 (vectorizable_mask_load_store): ...here.
11239 (vectorizable_load): ...and here.
11241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11243 * tree-vect-stmts.c (vect_build_all_ones_mask)
11244 (vect_build_zero_merge_argument): New functions, split out from...
11245 (vectorizable_load): ...here.
11247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11249 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11251 (vectorizable_mask_load_store): ...here.
11252 (vectorizable_store): ...and here.
11254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11256 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11258 (vectorizable_mask_load_store): ...here.
11260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11262 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11263 (vect_model_store_cost): Take a vec_load_store_type instead of a
11265 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11266 (vect_model_store_cost): Take a vec_load_store_type instead of a
11268 (vectorizable_mask_load_store): Update accordingly.
11269 (vectorizable_store): Likewise.
11270 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11274 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11275 IFN_MASK_LOAD calls here rather than...
11276 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11279 Alan Hayward <alan.hayward@arm.com>
11280 David Sherwood <david.sherwood@arm.com>
11282 * expmed.c (extract_bit_field_1): For vector extracts,
11283 fall back to extract_bit_field_as_subreg if vec_extract
11286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11287 Alan Hayward <alan.hayward@arm.com>
11288 David Sherwood <david.sherwood@arm.com>
11290 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11291 they are variable or constant sized.
11292 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11293 slots for constant-sized data.
11295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11296 Alan Hayward <alan.hayward@arm.com>
11297 David Sherwood <david.sherwood@arm.com>
11299 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11300 handling COND_EXPRs with boolean comparisons, try to find a better
11301 basis for the mask type than the boolean itself.
11303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11305 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11306 is calculated and how it can be overridden.
11307 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11308 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11310 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11314 Alan Hayward <alan.hayward@arm.com>
11315 David Sherwood <david.sherwood@arm.com>
11317 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11318 Remove the mode argument.
11319 (aarch64_simd_valid_immediate): Remove the mode and inverse
11321 * config/aarch64/iterators.md (bitsize): New iterator.
11322 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11323 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11324 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11325 aarch64_simd_valid_immediate.
11326 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11327 (aarch64_reg_or_bic_imm): Likewise.
11328 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11329 with an insn_type enum and msl with a modifier_type enum.
11330 Replace element_width with a scalar_mode. Change the shift
11331 to unsigned int. Add constructors for scalar_float_mode and
11332 scalar_int_mode elements.
11333 (aarch64_vect_float_const_representable_p): Delete.
11334 (aarch64_can_const_movi_rtx_p)
11335 (aarch64_simd_scalar_immediate_valid_for_move)
11336 (aarch64_simd_make_constant): Update call to
11337 aarch64_simd_valid_immediate.
11338 (aarch64_advsimd_valid_immediate_hs): New function.
11339 (aarch64_advsimd_valid_immediate): Likewise.
11340 (aarch64_simd_valid_immediate): Remove mode and inverse
11341 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11342 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11343 and aarch64_float_const_representable_p on the result.
11344 (aarch64_output_simd_mov_immediate): Remove mode argument.
11345 Update call to aarch64_simd_valid_immediate and use of
11346 simd_immediate_info.
11347 (aarch64_output_scalar_simd_mov_immediate): Update call
11350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11351 Alan Hayward <alan.hayward@arm.com>
11352 David Sherwood <david.sherwood@arm.com>
11354 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11355 (mode_nunits): Likewise CONST_MODE_NUNITS.
11356 * machmode.def (ADJUST_NUNITS): Document.
11357 * genmodes.c (mode_data::need_nunits_adj): New field.
11358 (blank_mode): Update accordingly.
11359 (adj_nunits): New variable.
11360 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11362 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11363 listed in adj_nunits.
11364 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11365 listed in adj_nunits. Don't emit case statements for such modes.
11366 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11367 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11368 nothing if adj_nunits is nonnull.
11369 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11370 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11371 (emit_mode_fbit): Update use of print_maybe_const_decl.
11372 (emit_move_size): Likewise. Treat the array as non-const
11374 (emit_mode_adjustments): Handle adj_nunits.
11376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11378 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11379 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11380 (VECTOR_MODES): Use it.
11381 (make_vector_modes): Take the prefix as an argument.
11383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11384 Alan Hayward <alan.hayward@arm.com>
11385 David Sherwood <david.sherwood@arm.com>
11387 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11388 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11389 for MODE_VECTOR_BOOL.
11390 * machmode.def (VECTOR_BOOL_MODE): Document.
11391 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11392 (make_vector_bool_mode): New function.
11393 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11395 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11396 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11398 * stor-layout.c (int_mode_for_mode): Likewise.
11399 * tree.c (build_vector_type_for_mode): Likewise.
11400 * varasm.c (output_constant_pool_2): Likewise.
11401 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11402 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11403 for MODE_VECTOR_BOOL.
11404 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11405 of mode class checks.
11406 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11407 instead of a list of mode class checks.
11408 (expand_vector_scalar_condition): Likewise.
11409 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11412 Alan Hayward <alan.hayward@arm.com>
11413 David Sherwood <david.sherwood@arm.com>
11415 * machmode.h (mode_size): Change from unsigned short to
11417 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11418 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11419 or if measurement_type is not polynomial.
11420 (fixed_size_mode::includes_p): Check for constant-sized modes.
11421 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11422 return a poly_uint16 rather than an unsigned short.
11423 (emit_mode_size): Change the type of mode_size from unsigned short
11424 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11425 (emit_mode_adjustments): Cope with polynomial vector sizes.
11426 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11428 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11430 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11431 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11432 * caller-save.c (setup_save_areas): Likewise.
11433 (replace_reg_with_saved_mem): Likewise.
11434 * calls.c (emit_library_call_value_1): Likewise.
11435 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11436 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11437 (gen_lowpart_for_combine): Likewise.
11438 * convert.c (convert_to_integer_1): Likewise.
11439 * cse.c (equiv_constant, cse_insn): Likewise.
11440 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11441 (cselib_subst_to_values): Likewise.
11442 * dce.c (word_dce_process_block): Likewise.
11443 * df-problems.c (df_word_lr_mark_ref): Likewise.
11444 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11445 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11446 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11447 (rtl_for_decl_location): Likewise.
11448 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11449 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11450 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11451 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11452 (expand_expr_real_1): Likewise.
11453 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11454 (pad_below): Likewise.
11455 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11456 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11457 * ira.c (get_subreg_tracking_sizes): Likewise.
11458 * ira-build.c (ira_create_allocno_objects): Likewise.
11459 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11460 (ira_sort_regnos_for_alter_reg): Likewise.
11461 * ira-costs.c (record_operand_costs): Likewise.
11462 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11463 (resolve_simple_move): Likewise.
11464 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11465 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11466 (lra_constraints): Likewise.
11467 (CONST_POOL_OK_P): Reject variable-sized modes.
11468 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11469 (add_pseudo_to_slot, lra_spill): Likewise.
11470 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11471 * optabs-query.c (get_best_extraction_insn): Likewise.
11472 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11473 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11474 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11475 * recog.c (offsettable_address_addr_space_p): Likewise.
11476 * regcprop.c (maybe_mode_change): Likewise.
11477 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11478 * regrename.c (build_def_use): Likewise.
11479 * regstat.c (dump_reg_info): Likewise.
11480 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11481 (find_reloads, find_reloads_subreg_address): Likewise.
11482 * reload1.c (eliminate_regs_1): Likewise.
11483 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11484 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11485 (simplify_binary_operation_1, simplify_subreg): Likewise.
11486 * targhooks.c (default_function_arg_padding): Likewise.
11487 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11488 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11489 (verify_gimple_assign_ternary): Likewise.
11490 * tree-inline.c (estimate_move_cost): Likewise.
11491 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11492 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11493 (get_address_cost_ainc): Likewise.
11494 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11495 (vect_supportable_dr_alignment): Likewise.
11496 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11497 (vectorizable_reduction): Likewise.
11498 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11499 (vectorizable_operation, vectorizable_load): Likewise.
11500 * tree.c (build_same_sized_truth_vector_type): Likewise.
11501 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11502 * var-tracking.c (emit_note_insn_var_location): Likewise.
11503 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11504 (ADDR_VEC_ALIGN): Likewise.
11506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11507 Alan Hayward <alan.hayward@arm.com>
11508 David Sherwood <david.sherwood@arm.com>
11510 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11512 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11513 or if measurement_type is polynomial.
11514 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11515 * combine.c (make_extraction): Likewise.
11516 * dse.c (find_shift_sequence): Likewise.
11517 * dwarf2out.c (mem_loc_descriptor): Likewise.
11518 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11519 (extract_bit_field, extract_low_bits): Likewise.
11520 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11521 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11522 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11523 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11524 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11525 * reload.c (find_reloads): Likewise.
11526 * reload1.c (alter_reg): Likewise.
11527 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11528 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11529 * tree-if-conv.c (predicate_mem_writes): Likewise.
11530 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11531 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11532 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11533 * valtrack.c (dead_debug_insert_temp): Likewise.
11534 * varasm.c (mergeable_constant_section): Likewise.
11535 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11538 Alan Hayward <alan.hayward@arm.com>
11539 David Sherwood <david.sherwood@arm.com>
11541 * expr.c (expand_assignment): Cope with polynomial mode sizes
11542 when assigning to a CONCAT.
11544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11545 Alan Hayward <alan.hayward@arm.com>
11546 David Sherwood <david.sherwood@arm.com>
11548 * machmode.h (mode_precision): Change from unsigned short to
11550 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11552 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11553 or if measurement_type is not polynomial.
11554 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11555 in which the mode is already known to be a scalar_int_mode.
11556 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11557 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11559 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11560 for GET_MODE_PRECISION.
11561 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11562 for GET_MODE_PRECISION.
11563 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11565 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11566 (expand_field_assignment, make_extraction): Likewise.
11567 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11568 (get_last_value): Likewise.
11569 * convert.c (convert_to_integer_1): Likewise.
11570 * cse.c (cse_insn): Likewise.
11571 * expr.c (expand_expr_real_1): Likewise.
11572 * lra-constraints.c (simplify_operand_subreg): Likewise.
11573 * optabs-query.c (can_atomic_load_p): Likewise.
11574 * optabs.c (expand_atomic_load): Likewise.
11575 (expand_atomic_store): Likewise.
11576 * ree.c (combine_reaching_defs): Likewise.
11577 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11578 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11579 * tree.h (type_has_mode_precision_p): Likewise.
11580 * ubsan.c (instrument_si_overflow): Likewise.
11582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11583 Alan Hayward <alan.hayward@arm.com>
11584 David Sherwood <david.sherwood@arm.com>
11586 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11587 polynomial numbers of units.
11588 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11589 (valid_vector_subparts_p): New function.
11590 (build_vector_type): Remove temporary shim and take the number
11591 of units as a poly_uint64 rather than an int.
11592 (build_opaque_vector_type): Take the number of units as a
11593 poly_uint64 rather than an int.
11594 * tree.c (build_vector_from_ctor): Handle polynomial
11595 TYPE_VECTOR_SUBPARTS.
11596 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11597 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11598 (build_vector_from_val): If the number of units is variable,
11599 use build_vec_duplicate_cst for constant operands and
11600 VEC_DUPLICATE_EXPR otherwise.
11601 (make_vector_type): Remove temporary is_constant ().
11602 (build_vector_type, build_opaque_vector_type): Take the number of
11603 units as a poly_uint64 rather than an int.
11604 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11606 * cfgexpand.c (expand_debug_expr): Likewise.
11607 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11608 (store_constructor, expand_expr_real_1): Likewise.
11609 (const_scalar_mask_from_tree): Likewise.
11610 * fold-const-call.c (fold_const_reduction): Likewise.
11611 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11612 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11613 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11614 (fold_relational_const): Likewise.
11615 (native_interpret_vector): Likewise. Change the size from an
11616 int to an unsigned int.
11617 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11618 TYPE_VECTOR_SUBPARTS.
11619 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11620 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11621 duplicating a non-constant operand into a variable-length vector.
11622 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11623 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11624 * ipa-icf.c (sem_variable::equals): Likewise.
11625 * match.pd: Likewise.
11626 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11627 * print-tree.c (print_node): Likewise.
11628 * stor-layout.c (layout_type): Likewise.
11629 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11630 * tree-cfg.c (verify_gimple_comparison): Likewise.
11631 (verify_gimple_assign_binary): Likewise.
11632 (verify_gimple_assign_ternary): Likewise.
11633 (verify_gimple_assign_single): Likewise.
11634 * tree-pretty-print.c (dump_generic_node): Likewise.
11635 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11636 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11637 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11638 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11639 (vect_shift_permute_load_chain): Likewise.
11640 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11641 (expand_vector_condition, optimize_vector_constructor): Likewise.
11642 (lower_vec_perm, get_compute_type): Likewise.
11643 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11644 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11645 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11646 (vect_recog_mask_conversion_pattern): Likewise.
11647 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11648 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11649 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11650 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11651 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11652 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11653 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11654 (supportable_widening_operation): Likewise.
11655 (supportable_narrowing_operation): Likewise.
11656 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11658 * varasm.c (output_constant): Likewise.
11660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11661 Alan Hayward <alan.hayward@arm.com>
11662 David Sherwood <david.sherwood@arm.com>
11664 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11665 so that both the length == 3 and length != 3 cases set up their
11666 own permute vectors. Add comments explaining why we know the
11667 number of elements is constant.
11668 (vect_permute_load_chain): Likewise.
11670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11671 Alan Hayward <alan.hayward@arm.com>
11672 David Sherwood <david.sherwood@arm.com>
11674 * machmode.h (mode_nunits): Change from unsigned char to
11676 (ONLY_FIXED_SIZE_MODES): New macro.
11677 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11678 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11679 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11681 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11682 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11683 or if measurement_type is not polynomial.
11684 * genmodes.c (ZERO_COEFFS): New macro.
11685 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11687 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11688 Use ZERO_COEFFS when emitting initializers.
11689 * data-streamer.h (bp_pack_poly_value): New function.
11690 (bp_unpack_poly_value): Likewise.
11691 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11692 for GET_MODE_NUNITS.
11693 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11694 for GET_MODE_NUNITS.
11695 * tree.c (make_vector_type): Remove temporary shim and make
11696 the real function take the number of units as a poly_uint64
11697 rather than an int.
11698 (build_vector_type_for_mode): Handle polynomial nunits.
11699 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11700 * emit-rtl.c (const_vec_series_p_1): Likewise.
11701 (gen_rtx_CONST_VECTOR): Likewise.
11702 * fold-const.c (test_vec_duplicate_folding): Likewise.
11703 * genrecog.c (validate_pattern): Likewise.
11704 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11705 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11706 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11707 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11708 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11709 * rtlanal.c (subreg_get_info): Likewise.
11710 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11711 (vect_grouped_load_supported): Likewise.
11712 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11713 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11714 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11715 (simplify_const_unary_operation, simplify_binary_operation_1)
11716 (simplify_const_binary_operation, simplify_ternary_operation)
11717 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11718 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11719 instead of CONST_VECTOR_NUNITS.
11720 * varasm.c (output_constant_pool_2): Likewise.
11721 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11722 explicit-encoded elements in the XVEC for variable-length vectors.
11724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11726 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11729 Alan Hayward <alan.hayward@arm.com>
11730 David Sherwood <david.sherwood@arm.com>
11732 * coretypes.h (fixed_size_mode): Declare.
11733 (fixed_size_mode_pod): New typedef.
11734 * builtins.h (target_builtins::x_apply_args_mode)
11735 (target_builtins::x_apply_result_mode): Change type to
11736 fixed_size_mode_pod.
11737 * builtins.c (apply_args_size, apply_result_size, result_vector)
11738 (expand_builtin_apply_args_1, expand_builtin_apply)
11739 (expand_builtin_return): Update accordingly.
11741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11743 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11744 * cselib.c (cselib_hash_rtx): Likewise.
11745 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11746 CONST_VECTOR encoding.
11748 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11749 Jeff Law <law@redhat.com>
11752 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11753 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11754 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11755 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11758 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11759 explicitly probe *sp in a noreturn function if there were any callee
11760 register saves or frame pointer is needed.
11762 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11765 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11766 BLKmode for ternary, binary or unary expressions.
11769 * var-tracking.c (delete_vta_debug_insn): New inline function.
11770 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11771 insns from get_insns () to NULL instead of each bb separately.
11772 Use delete_vta_debug_insn. No longer static.
11773 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11774 delete_vta_debug_insns callers.
11775 * rtl.h (delete_vta_debug_insns): Declare.
11776 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11777 instead of variable_tracking_main.
11779 2018-01-03 Martin Sebor <msebor@redhat.com>
11781 PR tree-optimization/83603
11782 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11783 arguments past the endof the argument list in functions declared
11784 without a prototype.
11785 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11786 Avoid checking when arguments are null.
11788 2018-01-03 Martin Sebor <msebor@redhat.com>
11791 * doc/extend.texi (attribute const): Fix a typo.
11792 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11793 issuing -Wsuggest-attribute for void functions.
11795 2018-01-03 Martin Sebor <msebor@redhat.com>
11797 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11798 offset_int::from instead of wide_int::to_shwi.
11799 (maybe_diag_overlap): Remove assertion.
11800 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11801 * gimple-ssa-sprintf.c (format_directive): Same.
11802 (parse_directive): Same.
11803 (sprintf_dom_walker::compute_format_length): Same.
11804 (try_substitute_return_value): Same.
11806 2018-01-03 Jeff Law <law@redhat.com>
11808 PR middle-end/83654
11809 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11810 non-constant residual for zero at runtime and avoid probing in
11811 that case. Reorganize code for trailing problem to mirror handling
11814 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11816 PR tree-optimization/83501
11817 * tree-ssa-strlen.c (get_string_cst): New.
11818 (handle_char_store): Call get_string_cst.
11820 2018-01-03 Martin Liska <mliska@suse.cz>
11822 PR tree-optimization/83593
11823 * tree-ssa-strlen.c: Include tree-cfg.h.
11824 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11825 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11826 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11828 (strlen_dom_walker::before_dom_children): Call
11829 gimple_purge_dead_eh_edges. Dump tranformation with details
11831 (strlen_dom_walker::before_dom_children): Update call by adding
11832 new argument cleanup_eh.
11833 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11835 2018-01-03 Martin Liska <mliska@suse.cz>
11838 * cif-code.def (VARIADIC_THUNK): New enum value.
11839 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11842 2018-01-03 Jan Beulich <jbeulich@suse.com>
11844 * sse.md (mov<mode>_internal): Tighten condition for when to use
11845 vmovdqu<ssescalarsize> for TI and OI modes.
11847 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11849 Update copyright years.
11851 2018-01-03 Martin Liska <mliska@suse.cz>
11854 * ipa-visibility.c (function_and_variable_visibility): Skip
11855 functions with noipa attribure.
11857 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11859 * gcc.c (process_command): Update copyright notice dates.
11860 * gcov-dump.c (print_version): Ditto.
11861 * gcov.c (print_version): Ditto.
11862 * gcov-tool.c (print_version): Ditto.
11863 * gengtype.c (create_file): Ditto.
11864 * doc/cpp.texi: Bump @copying's copyright year.
11865 * doc/cppinternals.texi: Ditto.
11866 * doc/gcc.texi: Ditto.
11867 * doc/gccint.texi: Ditto.
11868 * doc/gcov.texi: Ditto.
11869 * doc/install.texi: Ditto.
11870 * doc/invoke.texi: Ditto.
11872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11874 * vector-builder.h (vector_builder::m_full_nelts): Change from
11875 unsigned int to poly_uint64.
11876 (vector_builder::full_nelts): Update prototype accordingly.
11877 (vector_builder::new_vector): Likewise.
11878 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11879 (vector_builder::operator ==): Likewise.
11880 (vector_builder::finalize): Likewise.
11881 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11882 Take the number of elements as a poly_uint64 rather than an
11884 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11885 from unsigned int to poly_uint64.
11886 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11887 (vec_perm_indices::new_vector): Likewise.
11888 (vec_perm_indices::length): Likewise.
11889 (vec_perm_indices::nelts_per_input): Likewise.
11890 (vec_perm_indices::input_nelts): Likewise.
11891 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11892 number of elements per input as a poly_uint64 rather than an
11893 unsigned int. Use the original encoding for variable-length
11894 vectors, rather than clamping each individual element.
11895 For the second and subsequent elements in each pattern,
11896 clamp the step and base before clamping their sum.
11897 (vec_perm_indices::series_p): Handle polynomial element counts.
11898 (vec_perm_indices::all_in_range_p): Likewise.
11899 (vec_perm_indices_to_tree): Likewise.
11900 (vec_perm_indices_to_rtx): Likewise.
11901 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11902 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11903 (tree_vector_builder::new_binary_operation): Handle polynomial
11904 element counts. Return false if we need to know the number
11905 of elements at compile time.
11906 * fold-const.c (fold_vec_perm): Punt if the number of elements
11907 isn't known at compile time.
11909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11911 * vec-perm-indices.h (vec_perm_builder): Change element type
11912 from HOST_WIDE_INT to poly_int64.
11913 (vec_perm_indices::element_type): Update accordingly.
11914 (vec_perm_indices::clamp): Handle polynomial element_types.
11915 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11916 (vec_perm_indices::all_in_range_p): Likewise.
11917 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11919 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11920 polynomial vec_perm_indices element types.
11921 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11922 * fold-const.c (fold_vec_perm): Likewise.
11923 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11924 * tree-vect-generic.c (lower_vec_perm): Likewise.
11925 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11926 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11927 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11934 rather than an int. Use plus_constant.
11935 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11936 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11939 Alan Hayward <alan.hayward@arm.com>
11940 David Sherwood <david.sherwood@arm.com>
11942 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11943 a HOST_WIDE_INT to a poly_int64.
11945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11946 Alan Hayward <alan.hayward@arm.com>
11947 David Sherwood <david.sherwood@arm.com>
11949 * calls.c (load_register_parameters): Cope with polynomial
11950 mode sizes. Require a constant size for BLKmode parameters
11951 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11952 forces a parameter to be padded at the lsb end in order to
11953 fill a complete number of words, require the parameter size
11954 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
11961 from unsigned int to poly_uint64_pod.
11962 (alter_reg): Treat mode sizes as polynomial.
11964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11965 Alan Hayward <alan.hayward@arm.com>
11966 David Sherwood <david.sherwood@arm.com>
11968 * reload.c (complex_word_subreg_p): New function.
11969 (reload_inner_reg_of_subreg, push_reload): Use it.
11971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11972 Alan Hayward <alan.hayward@arm.com>
11973 David Sherwood <david.sherwood@arm.com>
11975 * lra-constraints.c (process_alt_operands): Reject matched
11976 operands whose sizes aren't ordered.
11977 (match_reload): Refer to this check here.
11979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11980 Alan Hayward <alan.hayward@arm.com>
11981 David Sherwood <david.sherwood@arm.com>
11983 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11984 that the mode size is in the set {1, 2, 4, 8, 16}.
11986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11987 Alan Hayward <alan.hayward@arm.com>
11988 David Sherwood <david.sherwood@arm.com>
11990 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11991 Use plus_constant instead of gen_rtx_PLUS.
11993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11994 Alan Hayward <alan.hayward@arm.com>
11995 David Sherwood <david.sherwood@arm.com>
11997 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11998 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11999 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12000 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12001 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12002 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12003 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12004 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12005 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12006 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12008 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12009 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12010 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12011 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12012 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12013 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12014 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12015 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12016 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12017 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12019 * expr.c (emit_move_resolve_push): Treat the input and result
12020 of PUSH_ROUNDING as a poly_int64.
12021 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12022 (emit_push_insn): Likewise.
12023 * lra-eliminations.c (mark_not_eliminable): Likewise.
12024 * recog.c (push_operand): Likewise.
12025 * reload1.c (elimination_effects): Likewise.
12026 * rtlanal.c (nonzero_bits1): Likewise.
12027 * calls.c (store_one_arg): Likewise. Require the padding to be
12028 known at compile time.
12030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12031 Alan Hayward <alan.hayward@arm.com>
12032 David Sherwood <david.sherwood@arm.com>
12034 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12035 Use plus_constant instead of gen_rtx_PLUS.
12037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12038 Alan Hayward <alan.hayward@arm.com>
12039 David Sherwood <david.sherwood@arm.com>
12041 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12042 rather than an int.
12044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12045 Alan Hayward <alan.hayward@arm.com>
12046 David Sherwood <david.sherwood@arm.com>
12048 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12049 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12050 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
12057 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12058 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12059 as a poly_uint64 too.
12061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12062 Alan Hayward <alan.hayward@arm.com>
12063 David Sherwood <david.sherwood@arm.com>
12065 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12068 Alan Hayward <alan.hayward@arm.com>
12069 David Sherwood <david.sherwood@arm.com>
12071 * combine.c (can_change_dest_mode): Handle polynomial
12072 REGMODE_NATURAL_SIZE.
12073 * expmed.c (store_bit_field_1): Likewise.
12074 * expr.c (store_constructor): Likewise.
12075 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12076 and polynomial REGMODE_NATURAL_SIZE.
12077 (gen_lowpart_common): Likewise.
12078 * reginfo.c (record_subregs_of_mode): Likewise.
12079 * rtlanal.c (read_modify_subreg_p): Likewise.
12081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12082 Alan Hayward <alan.hayward@arm.com>
12083 David Sherwood <david.sherwood@arm.com>
12085 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12086 numbers of elements.
12088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12089 Alan Hayward <alan.hayward@arm.com>
12090 David Sherwood <david.sherwood@arm.com>
12092 * match.pd: Cope with polynomial numbers of vector elements.
12094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12095 Alan Hayward <alan.hayward@arm.com>
12096 David Sherwood <david.sherwood@arm.com>
12098 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12099 in a POINTER_PLUS_EXPR.
12101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12102 Alan Hayward <alan.hayward@arm.com>
12103 David Sherwood <david.sherwood@arm.com>
12105 * omp-simd-clone.c (simd_clone_subparts): New function.
12106 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12107 (ipa_simd_modify_function_body): Likewise.
12109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12110 Alan Hayward <alan.hayward@arm.com>
12111 David Sherwood <david.sherwood@arm.com>
12113 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12114 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12115 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12116 (expand_vector_condition, vector_element): Likewise.
12117 (subparts_gt): New function.
12118 (get_compute_type): Use subparts_gt.
12119 (count_type_subparts): Delete.
12120 (expand_vector_operations_1): Use subparts_gt instead of
12121 count_type_subparts.
12123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12124 Alan Hayward <alan.hayward@arm.com>
12125 David Sherwood <david.sherwood@arm.com>
12127 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12128 (vect_compile_time_alias): ...this new function. Do the calculation
12129 on poly_ints rather than trees.
12130 (vect_prune_runtime_alias_test_list): Update call accordingly.
12132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12133 Alan Hayward <alan.hayward@arm.com>
12134 David Sherwood <david.sherwood@arm.com>
12136 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12138 (vect_schedule_slp_instance): Likewise.
12140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12141 Alan Hayward <alan.hayward@arm.com>
12142 David Sherwood <david.sherwood@arm.com>
12144 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12145 constant and extern definitions for variable-length vectors.
12146 (vect_get_constant_vectors): Note that the number of units
12147 is known to be constant.
12149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12150 Alan Hayward <alan.hayward@arm.com>
12151 David Sherwood <david.sherwood@arm.com>
12153 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12154 of units as polynomial. Choose between WIDE and NARROW based
12157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12158 Alan Hayward <alan.hayward@arm.com>
12159 David Sherwood <david.sherwood@arm.com>
12161 * tree-vect-stmts.c (simd_clone_subparts): New function.
12162 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12165 Alan Hayward <alan.hayward@arm.com>
12166 David Sherwood <david.sherwood@arm.com>
12168 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12169 vectors as polynomial. Use build_index_vector for
12170 IFN_GOMP_SIMD_LANE.
12172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12173 Alan Hayward <alan.hayward@arm.com>
12174 David Sherwood <david.sherwood@arm.com>
12176 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12177 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12178 for variable-length vectors.
12179 (vectorizable_mask_load_store): Treat the number of units as
12180 polynomial, asserting that it is constant if the condition has
12181 already been enforced.
12182 (vectorizable_store, vectorizable_load): Likewise.
12184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12185 Alan Hayward <alan.hayward@arm.com>
12186 David Sherwood <david.sherwood@arm.com>
12188 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12189 of units as polynomial. Punt if we can't tell at compile time
12190 which vector contains the final result.
12192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12193 Alan Hayward <alan.hayward@arm.com>
12194 David Sherwood <david.sherwood@arm.com>
12196 * tree-vect-loop.c (vectorizable_induction): Treat the number
12197 of units as polynomial. Punt on SLP inductions. Use an integer
12198 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12199 cast of such a series for variable-length floating-point
12202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12203 Alan Hayward <alan.hayward@arm.com>
12204 David Sherwood <david.sherwood@arm.com>
12206 * tree.h (build_index_vector): Declare.
12207 * tree.c (build_index_vector): New function.
12208 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12209 of units as polynomial, forcibly converting it to a constant if
12210 vectorizable_reduction has already enforced the condition.
12211 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12212 to create a {1,2,3,...} vector.
12213 (vectorizable_reduction): Treat the number of units as polynomial.
12214 Choose vectype_in based on the largest scalar element size rather
12215 than the smallest number of units. Enforce the restrictions
12218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12219 Alan Hayward <alan.hayward@arm.com>
12220 David Sherwood <david.sherwood@arm.com>
12222 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12223 number of units as polynomial.
12225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12226 Alan Hayward <alan.hayward@arm.com>
12227 David Sherwood <david.sherwood@arm.com>
12229 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12230 * target.def (autovectorize_vector_sizes): Return the vector sizes
12231 by pointer, using vector_sizes rather than a bitmask.
12232 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12233 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12234 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12236 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12237 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12238 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12239 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12240 * omp-general.c (omp_max_vf): Likewise.
12241 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12242 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12243 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12244 * tree-vect-slp.c (vect_slp_bb): Likewise.
12245 * doc/tm.texi: Regenerate.
12246 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12248 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12249 the vector size as a poly_uint64 rather than an unsigned int.
12250 (current_vector_size): Change from an unsigned int to a poly_uint64.
12251 (get_vectype_for_scalar_type): Update accordingly.
12252 * tree.h (build_truth_vector_type): Take the size and number of
12253 units as a poly_uint64 rather than an unsigned int.
12254 (build_vector_type): Add a temporary overload that takes
12255 the number of units as a poly_uint64 rather than an unsigned int.
12256 * tree.c (make_vector_type): Likewise.
12257 (build_truth_vector_type): Take the number of units as a poly_uint64
12258 rather than an unsigned int.
12260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12261 Alan Hayward <alan.hayward@arm.com>
12262 David Sherwood <david.sherwood@arm.com>
12264 * target.def (get_mask_mode): Take the number of units and length
12265 as poly_uint64s rather than unsigned ints.
12266 * targhooks.h (default_get_mask_mode): Update accordingly.
12267 * targhooks.c (default_get_mask_mode): Likewise.
12268 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12269 * doc/tm.texi: Regenerate.
12271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12272 Alan Hayward <alan.hayward@arm.com>
12273 David Sherwood <david.sherwood@arm.com>
12275 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12276 * omp-general.c (omp_max_vf): Likewise.
12277 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12278 (expand_omp_simd): Handle polynomial safelen.
12279 * omp-low.c (omplow_simd_context): Add a default constructor.
12280 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12281 (lower_rec_simd_input_clauses): Update accordingly.
12282 (lower_rec_input_clauses): Likewise.
12284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12285 Alan Hayward <alan.hayward@arm.com>
12286 David Sherwood <david.sherwood@arm.com>
12288 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12289 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12290 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12291 (vect_analyze_slp_cost): Likewise.
12292 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12293 (vect_model_load_cost): Likewise.
12295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12296 Alan Hayward <alan.hayward@arm.com>
12297 David Sherwood <david.sherwood@arm.com>
12299 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12300 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12301 from an unsigned int * to a poly_uint64_pod *.
12302 (calculate_unrolling_factor): New function.
12303 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12306 Alan Hayward <alan.hayward@arm.com>
12307 David Sherwood <david.sherwood@arm.com>
12309 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12310 from an unsigned int to a poly_uint64.
12311 (_loop_vec_info::slp_unrolling_factor): Likewise.
12312 (_loop_vec_info::vectorization_factor): Change from an int
12314 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12315 (vect_get_num_vectors): New function.
12316 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12317 (vect_get_num_copies): Use vect_get_num_vectors.
12318 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12319 to an unsigned int *.
12320 (vect_analyze_data_refs): Change min_vf from an int * to a
12322 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12323 than an unsigned HOST_WIDE_INT.
12324 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12325 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12326 to an unsigned int *.
12327 (vect_analyze_data_ref_dependences): Likewise.
12328 (vect_compute_data_ref_alignment): Handle polynomial vf.
12329 (vect_enhance_data_refs_alignment): Likewise.
12330 (vect_prune_runtime_alias_test_list): Likewise.
12331 (vect_shift_permute_load_chain): Likewise.
12332 (vect_supportable_dr_alignment): Likewise.
12333 (dependence_distance_ge_vf): Take the vectorization factor as a
12334 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12335 (vect_analyze_data_refs): Change min_vf from an int * to a
12337 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12338 vfm1 as a poly_uint64 rather than an int. Make the same change
12339 for the returned bound_scalar.
12340 (vect_gen_vector_loop_niters): Handle polynomial vf.
12341 (vect_do_peeling): Likewise. Update call to
12342 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12343 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12345 * tree-vect-loop.c (vect_determine_vectorization_factor)
12346 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12347 (vect_get_known_peeling_cost): Likewise.
12348 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12349 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12350 (vect_transform_loop): Likewise. Use the lowest possible VF when
12351 updating the upper bounds of the loop.
12352 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12353 rather than an int.
12354 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12355 polynomial unroll factors.
12356 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12357 (vect_make_slp_decision): Likewise.
12358 (vect_supported_load_permutation_p): Likewise, and polynomial
12360 (vect_analyze_slp_cost): Handle polynomial vf.
12361 (vect_slp_analyze_node_operations): Likewise.
12362 (vect_slp_analyze_bb_1): Likewise.
12363 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12364 than an unsigned HOST_WIDE_INT.
12365 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12366 (vectorizable_load): Handle polynomial vf.
12367 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12369 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12372 Alan Hayward <alan.hayward@arm.com>
12373 David Sherwood <david.sherwood@arm.com>
12375 * match.pd: Handle bit operations involving three constants
12376 and try to fold one pair.
12378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12380 * tree-vect-loop-manip.c: Include gimple-fold.h.
12381 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12382 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12383 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12384 Add a path that uses a step of VF instead of 1, but disable it
12386 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12387 and niters_no_overflow parameters. Update calls to
12388 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12389 Create a new SSA name if the latter choses to use a ste other
12390 than zero, and return it via niters_vector_mult_vf_var.
12391 * tree-vect-loop.c (vect_transform_loop): Update calls to
12392 vect_do_peeling, vect_gen_vector_loop_niters and
12393 slpeel_make_loop_iterate_ntimes.
12394 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12395 (vect_gen_vector_loop_niters): Update declarations after above changes.
12397 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12399 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12400 128-bit round to integer instructions.
12401 (ceil<mode>2): Likewise.
12402 (btrunc<mode>2): Likewise.
12403 (round<mode>2): Likewise.
12405 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12407 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12408 unaligned VSX load/store on P8/P9.
12409 (expand_block_clear): Allow the use of unaligned VSX
12410 load/store on P8/P9.
12412 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12414 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12416 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12417 swap associated with both a load and a store.
12419 2018-01-02 Andrew Waterman <andrew@sifive.com>
12421 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12422 * config/riscv/riscv.md (clear_cache): Use it.
12424 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12426 * web.c: Remove out-of-date comment.
12428 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12430 * expr.c (fixup_args_size_notes): Check that any existing
12431 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12432 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12433 (emit_single_push_insn): ...here.
12435 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12437 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12438 (const_vector_encoded_nelts): New function.
12439 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12440 (const_vector_int_elt, const_vector_elt): Declare.
12441 * emit-rtl.c (const_vector_int_elt_1): New function.
12442 (const_vector_elt): Likewise.
12443 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12444 of CONST_VECTOR_ELT.
12446 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12448 * expr.c: Include rtx-vector-builder.h.
12449 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12450 directly on the tree encoding.
12451 (const_vector_from_tree): Likewise.
12452 * optabs.c: Include rtx-vector-builder.h.
12453 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12454 sequence of "u" values.
12455 * vec-perm-indices.c: Include rtx-vector-builder.h.
12456 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12457 directly on the vec_perm_indices encoding.
12459 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12461 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12462 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12463 * rtx-vector-builder.h: New file.
12464 * rtx-vector-builder.c: Likewise.
12465 * rtl.h (rtx_def::u2): Add a const_vector field.
12466 (CONST_VECTOR_NPATTERNS): New macro.
12467 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12468 (CONST_VECTOR_DUPLICATE_P): Likewise.
12469 (CONST_VECTOR_STEPPED_P): Likewise.
12470 (CONST_VECTOR_ENCODED_ELT): Likewise.
12471 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12472 (unwrap_const_vec_duplicate): Likewise.
12473 (const_vec_series_p): Check for a non-duplicated vector encoding.
12474 Say that the function only returns true for integer vectors.
12475 * emit-rtl.c: Include rtx-vector-builder.h.
12476 (gen_const_vec_duplicate_1): Delete.
12477 (gen_const_vector): Call gen_const_vec_duplicate instead of
12478 gen_const_vec_duplicate_1.
12479 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12480 (gen_const_vec_duplicate): Use rtx_vector_builder.
12481 (gen_const_vec_series): Likewise.
12482 (gen_rtx_CONST_VECTOR): Likewise.
12483 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12484 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12485 Build a new vector rather than modifying a CONST_VECTOR in-place.
12486 (handle_special_swappables): Update call accordingly.
12487 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12488 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12489 Build a new vector rather than modifying a CONST_VECTOR in-place.
12490 (handle_special_swappables): Update call accordingly.
12492 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12494 * simplify-rtx.c (simplify_const_binary_operation): Use
12495 CONST_VECTOR_ELT instead of XVECEXP.
12497 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12499 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12500 the selector elements to be different from the data elements
12501 if the selector is a VECTOR_CST.
12502 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12503 ssizetype for the selector.
12505 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12507 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12508 before testing each element individually.
12509 * tree-vect-generic.c (lower_vec_perm): Likewise.
12511 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12513 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12514 * selftest-run-tests.c (selftest::run_tests): Call it.
12515 * vector-builder.h (vector_builder::operator ==): New function.
12516 (vector_builder::operator !=): Likewise.
12517 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12518 (vec_perm_indices::all_from_input_p): New function.
12519 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12520 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12521 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12522 instead of reading the VECTOR_CST directly. Detect whether both
12523 vector inputs are the same before constructing the vec_perm_indices,
12524 and update the number of inputs argument accordingly. Use the
12525 utility functions added above. Only construct sel2 if we need to.
12527 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12529 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12530 the broadcast of the low byte.
12531 (expand_mult_highpart): Use an explicit encoding for the permutes.
12532 * optabs-query.c (can_mult_highpart_p): Likewise.
12533 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12534 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12535 (vectorizable_bswap): Likewise.
12536 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12537 explicit encoding for the power-of-2 permutes.
12538 (vect_permute_store_chain): Likewise.
12539 (vect_grouped_load_supported): Likewise.
12540 (vect_permute_load_chain): Likewise.
12542 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12544 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12545 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12546 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12547 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12548 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12549 (vect_gen_perm_mask_any): Likewise.
12551 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12553 * int-vector-builder.h: New file.
12554 * vec-perm-indices.h: Include int-vector-builder.h.
12555 (vec_perm_indices): Redefine as an int_vector_builder.
12556 (auto_vec_perm_indices): Delete.
12557 (vec_perm_builder): Redefine as a stand-alone class.
12558 (vec_perm_indices::vec_perm_indices): New function.
12559 (vec_perm_indices::clamp): Likewise.
12560 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12561 (vec_perm_indices::new_vector): New function.
12562 (vec_perm_indices::new_expanded_vector): Update for new
12563 vec_perm_indices class.
12564 (vec_perm_indices::rotate_inputs): New function.
12565 (vec_perm_indices::all_in_range_p): Operate directly on the
12566 encoded form, without computing elided elements.
12567 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12568 encoding. Update for new vec_perm_indices class.
12569 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12570 the given vec_perm_builder.
12571 (expand_vec_perm_var): Update vec_perm_builder constructor.
12572 (expand_mult_highpart): Use vec_perm_builder instead of
12573 auto_vec_perm_indices.
12574 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12575 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12576 or double series encoding as appropriate.
12577 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12578 vec_perm_indices instead of auto_vec_perm_indices.
12579 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12580 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12581 (vect_permute_store_chain): Likewise.
12582 (vect_grouped_load_supported): Likewise.
12583 (vect_permute_load_chain): Likewise.
12584 (vect_shift_permute_load_chain): Likewise.
12585 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12586 (vect_transform_slp_perm_load): Likewise.
12587 (vect_schedule_slp_instance): Likewise.
12588 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12589 (vectorizable_mask_load_store): Likewise.
12590 (vectorizable_bswap): Likewise.
12591 (vectorizable_store): Likewise.
12592 (vectorizable_load): Likewise.
12593 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12594 vec_perm_indices instead of auto_vec_perm_indices. Use
12595 tree_to_vec_perm_builder to read the vector from a tree.
12596 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12597 vec_perm_builder instead of a vec_perm_indices.
12598 (have_whole_vector_shift): Use vec_perm_builder and
12599 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12600 truncation to calc_vec_perm_mask_for_shift.
12601 (vect_create_epilog_for_reduction): Likewise.
12602 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12603 from auto_vec_perm_indices to vec_perm_indices.
12604 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12605 instead of changing individual elements.
12606 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12607 the vector in d.perm.
12608 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12609 from auto_vec_perm_indices to vec_perm_indices.
12610 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12611 instead of changing individual elements.
12612 (arm_vectorize_vec_perm_const): Use new_vector to install
12613 the vector in d.perm.
12614 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12615 Update vec_perm_builder constructor.
12616 (rs6000_expand_interleave): Likewise.
12617 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12618 (rs6000_expand_interleave): Likewise.
12620 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12622 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12623 to qimode could truncate the indices.
12624 * optabs.c (expand_vec_perm_var): Likewise.
12626 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12628 * Makefile.in (OBJS): Add vec-perm-indices.o.
12629 * vec-perm-indices.h: New file.
12630 * vec-perm-indices.c: Likewise.
12631 * target.h (vec_perm_indices): Replace with a forward class
12633 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12634 * optabs.h: Include vec-perm-indices.h.
12635 (expand_vec_perm): Delete.
12636 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12637 (expand_vec_perm_const): Declare.
12638 * target.def (vec_perm_const_ok): Replace with...
12639 (vec_perm_const): ...this new hook.
12640 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12641 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12642 * doc/tm.texi: Regenerate.
12643 * optabs.def (vec_perm_const): Delete.
12644 * doc/md.texi (vec_perm_const): Likewise.
12645 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12646 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12647 expand_vec_perm for constant permutation vectors. Assert that
12648 the mode of variable permutation vectors is the integer equivalent
12649 of the mode that is being permuted.
12650 * optabs-query.h (selector_fits_mode_p): Declare.
12651 * optabs-query.c: Include vec-perm-indices.h.
12652 (selector_fits_mode_p): New function.
12653 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12654 is defined, instead of checking whether the vec_perm_const_optab
12655 exists. Use targetm.vectorize.vec_perm_const instead of
12656 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12657 fit in the vector mode before using a variable permute.
12658 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12659 vec_perm_indices instead of an rtx.
12660 (expand_vec_perm): Replace with...
12661 (expand_vec_perm_const): ...this new function. Take the selector
12662 as a vec_perm_indices rather than an rtx. Also take the mode of
12663 the selector. Update call to shift_amt_for_vec_perm_mask.
12664 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12665 Use vec_perm_indices::new_expanded_vector to expand the original
12666 selector into bytes. Check whether the indices fit in the vector
12667 mode before using a variable permute.
12668 (expand_vec_perm_var): Make global.
12669 (expand_mult_highpart): Use expand_vec_perm_const.
12670 * fold-const.c: Includes vec-perm-indices.h.
12671 * tree-ssa-forwprop.c: Likewise.
12672 * tree-vect-data-refs.c: Likewise.
12673 * tree-vect-generic.c: Likewise.
12674 * tree-vect-loop.c: Likewise.
12675 * tree-vect-slp.c: Likewise.
12676 * tree-vect-stmts.c: Likewise.
12677 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12679 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12680 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12681 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12682 (aarch64_vectorize_vec_perm_const): ...this new function.
12683 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12684 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12685 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12686 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12687 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12688 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12689 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12691 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12692 check for NEON modes.
12693 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12694 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12695 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12696 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12698 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12699 the old VEC_PERM_CONST conditions.
12700 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12701 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12702 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12703 (ia64_vectorize_vec_perm_const_ok): Merge into...
12704 (ia64_vectorize_vec_perm_const): ...this new function.
12705 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12706 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12707 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12708 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12709 * config/mips/mips.c (mips_expand_vec_perm_const)
12710 (mips_vectorize_vec_perm_const_ok): Merge into...
12711 (mips_vectorize_vec_perm_const): ...this new function.
12712 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12713 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12714 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12715 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12716 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12717 (rs6000_expand_vec_perm_const): Delete.
12718 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12720 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12721 (altivec_expand_vec_perm_const_le): Take each operand individually.
12722 Operate on constant selectors rather than rtxes.
12723 (altivec_expand_vec_perm_const): Likewise. Update call to
12724 altivec_expand_vec_perm_const_le.
12725 (rs6000_expand_vec_perm_const): Delete.
12726 (rs6000_vectorize_vec_perm_const_ok): Delete.
12727 (rs6000_vectorize_vec_perm_const): New function.
12728 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12729 an element count and rtx array.
12730 (rs6000_expand_extract_even): Update call accordingly.
12731 (rs6000_expand_interleave): Likewise.
12732 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12733 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12734 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12735 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12736 (rs6000_expand_vec_perm_const): Delete.
12737 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12738 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12739 (altivec_expand_vec_perm_const_le): Take each operand individually.
12740 Operate on constant selectors rather than rtxes.
12741 (altivec_expand_vec_perm_const): Likewise. Update call to
12742 altivec_expand_vec_perm_const_le.
12743 (rs6000_expand_vec_perm_const): Delete.
12744 (rs6000_vectorize_vec_perm_const_ok): Delete.
12745 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12746 reference to the SPE evmerge intructions.
12747 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12748 an element count and rtx array.
12749 (rs6000_expand_extract_even): Update call accordingly.
12750 (rs6000_expand_interleave): Likewise.
12751 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12752 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12754 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12758 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12759 vector mode and that that mode matches the mode of the data
12761 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12762 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12763 directly using expand_vec_perm_1 when forcing selectors into
12765 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12767 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12769 * optabs-query.h (can_vec_perm_p): Delete.
12770 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12771 * optabs-query.c (can_vec_perm_p): Split into...
12772 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12773 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12774 particular selector is valid.
12775 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12776 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12777 (vect_grouped_load_supported): Likewise.
12778 (vect_shift_permute_load_chain): Likewise.
12779 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12780 (vect_transform_slp_perm_load): Likewise.
12781 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12782 (vectorizable_bswap): Likewise.
12783 (vect_gen_perm_mask_checked): Likewise.
12784 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12785 implementations of variable permutation vectors into account
12786 when deciding which selector to use.
12787 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12788 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12789 with a false third argument.
12790 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12791 to test whether the constant selector is valid and can_vec_perm_var_p
12792 to test whether a variable selector is valid.
12794 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12796 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12797 * optabs-query.c (can_vec_perm_p): Likewise.
12798 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12799 instead of vec_perm_indices.
12800 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12801 (vect_gen_perm_mask_checked): Likewise,
12802 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12803 (vect_gen_perm_mask_checked): Likewise,
12805 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12807 * optabs-query.h (qimode_for_vec_perm): Declare.
12808 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12809 (qimode_for_vec_perm): ...this new function.
12810 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12812 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12814 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12815 does not have a conditional at the top.
12817 2018-01-02 Richard Biener <rguenther@suse.de>
12819 * ipa-inline.c (big_speedup_p): Fix expression.
12821 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12824 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12827 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12831 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12832 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12833 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12834 cond_taken_branch_cost 3->4.
12836 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12838 PR tree-optimization/83581
12839 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12840 TODO_cleanup_cfg if any changes have been made.
12842 PR middle-end/83608
12843 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12844 convert_modes if target mode has the right side, but different mode
12847 PR middle-end/83609
12848 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12849 last argument when extracting from CONCAT. If either from_real or
12850 from_imag is NULL, use expansion through memory. If result is not
12851 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12852 the parts directly to inner mode, if even that fails, use expansion
12855 PR middle-end/83623
12856 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12857 check for bswap in mode rather than HImode and use that in expand_unop
12860 Copyright (C) 2018 Free Software Foundation, Inc.
12862 Copying and distribution of this file, with or without modification,
12863 are permitted in any medium without royalty provided the copyright
12864 notice and this notice are preserved.