1 2014-04-23 Richard Biener <rguenther@suse.de>
4 * gimplify.c (gimple_regimplify_operands): Update the
7 2014-04-21 Michael Meissner <meissner@linux.vnet.ibm.com>
9 Back port from the trunk, subversion id 209546.
11 2014-04-21 Michael Meissner <meissner@linux.vnet.ibm.com>
14 * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64 case):
15 If mode is DDmode and TARGET_E500_DOUBLE allow move.
17 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print some
18 more debug information for E500 if -mdebug=reg.
20 2014-04-18 Richard Henderson <rth@redhat.com>
22 * config/aarch64/aarch64.c (aarch64_register_move_cost): Pass a mode
23 to GET_MODE_SIZE, not a reg_class_t.
25 2014-04-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust for
29 (vsx_xxmrglw_<mode>): Likewise.
31 2014-04-15 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
34 Revert the following patch
36 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
38 Back port mainline subversion id 209025.
39 2014-04-02 Michael Meissner <meissner@linux.vnet.ibm.com>
42 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
43 software floating point or no floating point registers, do not
44 allow any type in the FPRs. Eliminate a test for SPE SIMD types
45 in GPRs that occurs after we tested for GPRs that would never be
48 * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
49 Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
50 since the FMOVE64 type is DFmode/DDmode. If TARGET_E500_DOUBLE,
51 specifically allow DDmode, since that does not use the SPE SIMD
54 2014-04-10 Vladimir Makarov <vmakarov@redhat.com>
56 PR rtl-optimization/60769
57 * lra-constraints.c (simplify_operand_subreg): Force reload of
58 paradoxical subreg if it is not in the class contents.
60 2014-04-10 Jakub Jelinek <jakub@redhat.com>
62 Backport from mainline
63 2014-03-12 Jakub Jelinek <jakub@redhat.com>
64 Marc Glisse <marc.glisse@inria.fr>
66 PR tree-optimization/60502
67 * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
68 instead of build_low_bits_mask.
70 2013-06-13 Marc Glisse <marc.glisse@inria.fr>
72 * tree.c (build_all_ones_cst): New function.
73 * tree.h (build_all_ones_cst): Declare it.
75 2013-05-10 Marc Glisse <marc.glisse@inria.fr>
77 * tree.c (build_minus_one_cst): New function.
78 * tree.h (build_minus_one_cst): Declare new function.
80 2014-04-10 Jakub Jelinek <jakub@redhat.com>
82 Backport from mainline
83 2014-03-28 Jakub Jelinek <jakub@redhat.com>
86 * config/i386/i386.c (ix86_copy_addr_to_reg): Call copy_addr_to_reg
87 also if addr has VOIDmode.
89 2014-03-17 Jakub Jelinek <jakub@redhat.com>
92 * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
93 note creation for the 2010-08-31 changes.
95 2014-03-06 Jakub Jelinek <jakub@redhat.com>
96 Meador Inge <meadori@codesourcery.com>
99 * config/arm/arm.c (arm_tls_symbol_p): Remove.
100 (arm_legitimize_address): Call legitimize_tls_address for any
101 arm_tls_referenced_p expression, handle constant addend. Call it
102 before testing for !TARGET_ARM.
103 (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
105 2014-04-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
107 Backport from mainline r208750
108 2014-03-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
110 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
111 pattern for vector nor instead of subtract from splat(-1).
112 (altivec_expand_vec_perm_const_le): Likewise.
114 Backport from mainline r209235
115 2014-04-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
117 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Use vnand
118 instead of vnor to exploit possible fusion opportunity in the
120 (altivec_expand_vec_perm_const_le): Likewise.
122 2014-04-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
124 Revert following patch
125 2014-04-08 Pat Haugen <pthaugen@us.ibm.com>
127 Backport from mainline
128 2014-04-08 Pat Haugen <pthaugen@us.ibm.com>
130 * config/rs6000/sync.md (AINT mode_iterator): Move definition.
131 (loadsync_<mode>): Change mode.
132 (load_quadpti, store_quadpti): New.
133 (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
134 * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
136 2014-04-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
138 Backport from mainline r202642
139 2013-09-17 Alan Modra <amodra@gmail.com>
142 * config/rs6000/driver-rs6000.c (elf_platform): Revert 2013-06-11
145 2014-04-08 Pat Haugen <pthaugen@us.ibm.com>
147 Backport from mainline
148 2014-04-08 Pat Haugen <pthaugen@us.ibm.com>
150 * config/rs6000/sync.md (AINT mode_iterator): Move definition.
151 (loadsync_<mode>): Change mode.
152 (load_quadpti, store_quadpti): New.
153 (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
154 * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
156 2014-04-07 Martin Jambor <mjambor@suse.cz>
159 * ipa-cp.c (propagate_constants_accross_call): Do not propagate
162 2014-04-07 Dominique d'Humieres <dominiq@lps.ens.fr>
164 Backport from mainline
165 2013-09-14 Iain Sandoe <iains@gcc.gnu.org>
168 * config/darwin.c (darwin_objc2_section): Note if ObjC Metadata
170 (darwin_objc1_section): Likewise.
171 (darwin_file_end): Emit Image Info section when required.
173 2014-04-05 Alan Modra <amodra@gmail.com>
176 2014-01-28 Alan Modra <amodra@gmail.com>
177 * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
178 * configure.ac <recursive call for build != host>: Define
179 GENERATOR_FILE. Comment. Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
180 and LD_FOR_BUILD too.
181 * configure: Regenerate.
183 2014-04-04 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
185 Backport from mainline r208895:
186 2014-03-28 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
188 * config/rs6000/rs6000.c (fusion_gpr_load_p): Refuse optimization
189 if it would clobber the stack pointer, even temporarily.
191 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
193 Back port from main line:
194 2014-04-01 Michael Meissner <meissner@linux.vnet.ibm.com>
196 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
199 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
201 Back port mainline subversion id 209025.
202 2014-04-02 Michael Meissner <meissner@linux.vnet.ibm.com>
205 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
206 software floating point or no floating point registers, do not
207 allow any type in the FPRs. Eliminate a test for SPE SIMD types
208 in GPRs that occurs after we tested for GPRs that would never be
211 * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
212 Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
213 since the FMOVE64 type is DFmode/DDmode. If TARGET_E500_DOUBLE,
214 specifically allow DDmode, since that does not use the SPE SIMD
217 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
219 Backport from mainline r205308
220 2013-11-23 David Edelsohn <dje.gcc@gmail.com>
222 * config/rs6000/rs6000.c (IN_NAMED_SECTION): New macro.
223 (rs6000_xcoff_select_section): Place decls with stricter alignment
225 (rs6000_xcoff_unique_section): Allow unique sections for
226 uninitialized data with strict alignment.
228 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
230 Backport from mainline
231 2013-11-15 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
233 * gcc/configure: Regenerate.
235 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
238 2013-04-25 Alan Modra <amodra@gmail.com>
241 * config/rs6000/rs6000.md (rotlsi3_internal7): Rename to
242 rotlsi3_internal7le and condition on !BYTES_BIG_ENDIAN.
243 (rotlsi3_internal8be): New BYTES_BIG_ENDIAN insn.
244 Repeat for many other rotate/shift and mask patterns using subregs.
246 (ashrdisi3_noppc64): Rename to ashrdisi3_noppc64be and condition
249 2013-06-07 Alan Modra <amodra@gmail.com>
251 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
252 override user -mfp-in-toc.
253 (offsettable_ok_by_alignment): Consider just the current access
254 rather than the whole object, unless BLKmode. Handle
255 CONSTANT_POOL_ADDRESS_P constants that lack a decl too.
256 (use_toc_relative_ref): Allow CONSTANT_POOL_ADDRESS_P constants
258 * config/rs6000/linux64.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Don't
259 override user -mfp-in-toc or -msum-in-toc. Default to
260 -mno-fp-in-toc for -mcmodel=medium.
262 2013-06-18 Alan Modra <amodra@gmail.com>
264 * config/rs6000/rs6000.h (enum data_align): New.
265 (LOCAL_ALIGNMENT, DATA_ALIGNMENT): Use rs6000_data_alignment.
266 (DATA_ABI_ALIGNMENT): Define.
267 (CONSTANT_ALIGNMENT): Correct comment.
268 * config/rs6000/rs6000-protos.h (rs6000_data_alignment): Declare.
269 * config/rs6000/rs6000.c (rs6000_data_alignment): New function.
271 2013-07-11 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
273 * config/rs6000/rs6000.md (""*tls_gd_low<TLSmode:tls_abi_suffix>"):
274 Require GOT register as additional operand in UNSPEC.
275 ("*tls_ld_low<TLSmode:tls_abi_suffix>"): Likewise.
276 ("*tls_got_dtprel_low<TLSmode:tls_abi_suffix>"): Likewise.
277 ("*tls_got_tprel_low<TLSmode:tls_abi_suffix>"): Likewise.
278 ("*tls_gd<TLSmode:tls_abi_suffix>"): Update splitter.
279 ("*tls_ld<TLSmode:tls_abi_suffix>"): Likewise.
280 ("tls_got_dtprel_<TLSmode:tls_abi_suffix>"): Likewise.
281 ("tls_got_tprel_<TLSmode:tls_abi_suffix>"): Likewise.
283 2014-01-23 Pat Haugen <pthaugen@us.ibm.com>
285 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
286 force flag_ira_loop_pressure if set via command line.
288 2014-02-06 Alan Modra <amodra@gmail.com>
291 * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
292 change SDmode to DDmode when lra_in_progress.
294 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
298 2014-03-12 Michael Meissner <meissner@linux.vnet.ibm.com>
300 * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
304 (VEC_base): Likewise.
305 (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
306 registers, we need to swap double words in little endian mode.
308 * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
309 to be a container mode for 128-bit integer operations added in ISA
310 2.07. Unlike TImode and PTImode, the preferred register set is
311 the Altivec/VMX registers for the 128-bit operations.
313 * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
315 (rs6000_split_128bit_ok_p): Likewise.
317 * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
318 macros for creating ISA 2.07 normal and overloaded builtin
319 functions with 3 arguments.
320 (BU_P8V_OVERLOAD_3): Likewise.
321 (VPERM_1T): Add support for V1TImode in 128-bit vector operations
322 for use as overloaded functions.
323 (VPERM_1TI_UNS): Likewise.
324 (VSEL_1TI): Likewise.
325 (VSEL_1TI_UNS): Likewise.
326 (ST_INTERNAL_1ti): Likewise.
327 (LD_INTERNAL_1ti): Likewise.
328 (XXSEL_1TI): Likewise.
329 (XXSEL_1TI_UNS): Likewise.
330 (VPERM_1TI): Likewise.
331 (VPERM_1TI_UNS): Likewise.
332 (XXPERMDI_1TI): Likewise.
334 (LXVD2X_V1TI): Likewise.
335 (STXVD2X_V1TI): Likewise.
336 (VEC_INIT_V1TI): Likewise.
337 (VEC_SET_V1TI): Likewise.
338 (VEC_EXT_V1TI): Likewise.
339 (EQV_V1TI): Likewise.
340 (NAND_V1TI): Likewise.
341 (ORC_V1TI): Likewise.
342 (VADDCUQ): Add support for 128-bit integer arithmetic instructions
343 added in ISA 2.07. Add both normal 'altivec' builtins, and the
347 (VADDEUQM): Likewise.
348 (VADDECUQ): Likewise.
349 (VSUBEUQM): Likewise.
350 (VSUBECUQ): Likewise.
352 * config/rs6000/rs6000-c.c (__int128_type): New static to hold
353 __int128_t and __uint128_t types.
354 (__uint128_type): Likewise.
355 (altivec_categorize_keyword): Add support for vector __int128_t,
356 vector __uint128_t, vector __int128, and vector unsigned __int128
357 as a container type for TImode operations that need to be done in
358 VSX/Altivec registers.
359 (rs6000_macro_to_expand): Likewise.
360 (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
361 to support 128-bit integer instructions vaddcuq, vadduqm,
362 vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
363 (altivec_resolve_overloaded_builtin): Add support for V1TImode.
365 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
366 for V1TImode, and set up preferences to use VSX/Altivec
367 registers. Setup VSX reload handlers.
368 (rs6000_debug_reg_global): Likewise.
369 (rs6000_init_hard_regno_mode_ok): Likewise.
370 (rs6000_preferred_simd_mode): Likewise.
371 (vspltis_constant): Do not allow V1TImode as easy altivec
373 (easy_altivec_constant): Likewise.
374 (output_vec_const_move): Likewise.
375 (rs6000_expand_vector_set): Convert V1TImode set and extract to
377 (rs6000_expand_vector_extract): Likewise.
378 (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
380 (rs6000_const_vec): Add support for V1TImode.
381 (rs6000_emit_le_vsx_load): Swap double words when loading or
382 storing TImode/V1TImode.
383 (rs6000_emit_le_vsx_store): Likewise.
384 (rs6000_emit_le_vsx_move): Likewise.
385 (rs6000_emit_move): Add support for V1TImode.
386 (altivec_expand_ld_builtin): Likewise.
387 (altivec_expand_st_builtin): Likewise.
388 (altivec_expand_vec_init_builtin): Likewise.
389 (altivec_expand_builtin): Likewise.
390 (rs6000_init_builtins): Add support for V1TImode type. Add
391 support for ISA 2.07 128-bit integer builtins. Define type names
392 for the VSX/Altivec vector types.
393 (altivec_init_builtins): Add support for overloaded vector
394 functions with V1TImode type.
395 (rs6000_preferred_reload_class): Prefer Altivec registers for
397 (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
399 (rs6000_split_128bit_ok_p): Likewise.
400 (rs6000_handle_altivec_attribute): Create V1TImode from vector
401 __int128_t and vector __uint128_t.
403 * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
411 (VS_scalar): Likewise.
412 (VS_double): Likewise.
413 (vsx_set_v1ti): New builtin function to create V1TImode from
416 * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say
417 whether we support the ISA 2.07 128-bit integer arithmetic
419 (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
420 (enum rs6000_builtin_type_index): Add fields to hold V1TImode
421 and TImode types for use with the builtin functions.
422 (V1TI_type_node): Likewise.
423 (unsigned_V1TI_type_node): Likewise.
424 (intTI_type_internal_node): Likewise.
425 (uintTI_type_internal_node): Likewise.
427 * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA
428 2.07 128-bit builtin functions.
429 (UNSPEC_VADDEUQM): Likewise.
430 (UNSPEC_VADDECUQ): Likewise.
431 (UNSPEC_VSUBCUQ): Likewise.
432 (UNSPEC_VSUBEUQM): Likewise.
433 (UNSPEC_VSUBECUQ): Likewise.
434 (VM): Add V1TImode to vector mode iterators.
437 (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
438 (altivec_vaddcuq): Likewise.
439 (altivec_vsubuqm): Likewise.
440 (altivec_vsubcuq): Likewise.
441 (altivec_vaddeuqm): Likewise.
442 (altivec_vaddecuq): Likewise.
443 (altivec_vsubeuqm): Likewise.
444 (altivec_vsubecuq): Likewise.
446 * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
448 (BOOL_128): Likewise.
449 (BOOL_REGS_OUTPUT): Likewise.
450 (BOOL_REGS_OP1): Likewise.
451 (BOOL_REGS_OP2): Likewise.
452 (BOOL_REGS_UNARY): Likewise.
453 (BOOL_REGS_AND_CR0): Likewise.
455 * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
456 128-bit integer builtin support.
457 (vec_vadduqm): Likewise.
458 (vec_vaddecuq): Likewise.
459 (vec_vaddeuqm): Likewise.
460 (vec_vsubecuq): Likewise.
461 (vec_vsubeuqm): Likewise.
462 (vec_vsubcuq): Likewise.
463 (vec_vsubuqm): Likewise.
465 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
466 Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
467 vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
468 128-bit integer add/subtract to ISA 2.07.
470 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
472 Apply mainline r207798
473 2014-02-26 Alan Modra <amodra@gmail.com>
476 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
477 find_replacement on parts of insn rtl that might be reloaded.
479 Backport from mainline r208287
480 2014-03-03 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
482 * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
483 reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
484 constraint on constants to permit them being loaded into
485 GENERAL_REGS or BASE_REGS.
487 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
489 Backport from mainline r207699.
490 2014-02-11 Michael Meissner <meissner@linux.vnet.ibm.com>
493 * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
494 for VSX/Altivec vectors that land in GPR registers.
496 Backport from mainline r207808.
497 2014-02-15 Michael Meissner <meissner@linux.vnet.ibm.com>
500 * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
501 (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
502 into 64-bit and 32-bit moves. On 64-bit moves, add support for
503 using direct move instructions on ISA 2.07. Also adjust
504 instruction length for 64-bit.
505 (mov<mode>_64bit, TFmode/TDmode): Likewise.
506 (mov<mode>_32bit, TFmode/TDmode): Likewise.
508 Backport from mainline r207868.
509 2014-02-18 Michael Meissner <meissner@linux.vnet.ibm.com>
512 * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
513 Split 64-bit moves into 2 patterns. Do not allow the use of
514 direct move for TDmode in little endian, since the decimal value
515 has little endian bytes within a word, but the 64-bit pieces are
516 ordered in a big endian fashion, and normal subreg's of TDmode are
518 (mov<mode>_64bit_dm): Likewise.
519 (movtd_64bit_nodm): Likewise.
521 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
523 Backport from mainline r207658
524 2014-02-06 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
526 * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
527 big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
528 -mcall-openbsd, or -mcall-linux.
529 (CC1_ENDIAN_BIG_SPEC): Remove.
530 (CC1_ENDIAN_LITTLE_SPEC): Remove.
531 (CC1_ENDIAN_DEFAULT_SPEC): Remove.
532 (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
533 (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
534 and %cc1_endian_default.
535 * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
537 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
539 Little Endian Vector API Support
540 Backport from mainline r206443
541 2014-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
543 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
544 two duplicate entries.
546 Backport from mainline r206494
547 2014-01-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
549 * doc/invoke.texi: Add -maltivec={be,le} options, and document
550 default element-order behavior for -maltivec.
551 * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
552 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
553 that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
554 when targeting big endian, at least for now.
555 * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
557 Backport from mainline r206541
558 2014-01-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
560 * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
562 Backport from mainline r206590
563 2014-01-13 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
565 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
566 Implement -maltivec=be for vec_insert and vec_extract.
568 Backport from mainline r206641
569 2014-01-15 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
571 * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
572 and vmulosh rather than call gen_vec_widen_smult_*.
573 (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
574 than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
575 (vec_widen_smult_even_v16qi): Likewise.
576 (vec_widen_umult_even_v8hi): Likewise.
577 (vec_widen_smult_even_v8hi): Likewise.
578 (vec_widen_umult_odd_v16qi): Likewise.
579 (vec_widen_smult_odd_v16qi): Likewise.
580 (vec_widen_umult_odd_v8hi): Likewise.
581 (vec_widen_smult_odd_v8hi): Likewise.
582 (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
583 vmuloub rather than call gen_vec_widen_umult_*.
584 (vec_widen_umult_lo_v16qi): Likewise.
585 (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
586 vmulosb rather than call gen_vec_widen_smult_*.
587 (vec_widen_smult_lo_v16qi): Likewise.
588 (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
589 rather than call gen_vec_widen_umult_*.
590 (vec_widen_umult_lo_v8hi): Likewise.
591 (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
592 rather than call gen_vec_widen_smult_*.
593 (vec_widen_smult_lo_v8hi): Likewise.
595 Backport from mainline r207062
596 2014-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
598 * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
599 correction for little endian...
600 * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
603 Backport from mainline r207262
604 2014-01-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
606 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Use
607 CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
608 * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
609 -maltivec=be with LE targets.
610 (vsx_mergeh_<mode>): Likewise.
611 * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New
613 (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
614 (altivec_vmrghb): Replace with define_expand and new
615 *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE
617 (altivec_vmrghb_direct): New define_insn.
618 (altivec_vmrghh): Replace with define_expand and new
619 *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE
621 (altivec_vmrghh_direct): New define_insn.
622 (altivec_vmrghw): Replace with define_expand and new
623 *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE
625 (altivec_vmrghw_direct): New define_insn.
626 (*altivec_vmrghsf): Adjust for endianness.
627 (altivec_vmrglb): Replace with define_expand and new
628 *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE
630 (altivec_vmrglb_direct): New define_insn.
631 (altivec_vmrglh): Replace with define_expand and new
632 *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE
634 (altivec_vmrglh_direct): New define_insn.
635 (altivec_vmrglw): Replace with define_expand and new
636 *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE
638 (altivec_vmrglw_direct): New define_insn.
639 (*altivec_vmrglsf): Adjust for endianness.
640 (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
641 (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
642 (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
643 (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
644 (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
645 (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
646 (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
647 (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
649 Backport from mainline r207318
650 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
652 * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
653 gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
654 remove element index adjustment for endian (now handled in vsx.md
656 (altivec_expand_vec_perm_const): Use
657 gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
658 * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
659 (vsx_xxspltw_<mode>): Adjust element index for little endian.
660 * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
661 define_expand and a new define_insn *altivec_vspltb_internal;
662 adjust for -maltivec=be on a little endian target.
663 (altivec_vspltb_direct): New.
664 (altivec_vsplth): Divide into a define_expand and a new
665 define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
666 little endian target.
667 (altivec_vsplth_direct): New.
668 (altivec_vspltw): Divide into a define_expand and a new
669 define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
670 little endian target.
671 (altivec_vspltw_direct): New.
672 (altivec_vspltsf): Divide into a define_expand and a new
673 define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
674 a little endian target.
676 Backport from mainline r207326
677 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
679 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
680 unused variable "field".
681 * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
682 (vsx_mergeh_<mode>): Likewise.
683 * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
684 (altivec_vmrghh): Likewise.
685 (altivec_vmrghw): Likewise.
686 (altivec_vmrglb): Likewise.
687 (altivec_vmrglh): Likewise.
688 (altivec_vmrglw): Likewise.
689 (altivec_vspltb): Add missing uses.
690 (altivec_vsplth): Likewise.
691 (altivec_vspltw): Likewise.
692 (altivec_vspltsf): Likewise.
694 Backport from mainline r207414
695 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
697 * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
698 (altivec_vsumsws): Add handling for -maltivec=be with a little
700 (altivec_vsumsws_direct): New.
701 (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
704 Backport from mainline r207415
705 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
707 * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
708 for vector types other than V16QImode.
709 * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
710 define_expand, and call altivec_expand_vec_perm_le when producing
711 code with little endian element order.
712 (*altivec_vperm_<mode>_internal): New insn having previous
713 behavior of altivec_vperm_<mode>.
714 (altivec_vperm_<mode>_uns): Change to a define_expand, and call
715 altivec_expand_vec_perm_le when producing code with little endian
717 (*altivec_vperm_<mode>_uns_internal): New insn having previous
718 behavior of altivec_vperm_<mode>_uns.
720 Backport from mainline r207520
721 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
723 * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
724 (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
725 (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
726 (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
728 (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
730 (altivec_vpks<VI_char>ss): Likewise.
731 (altivec_vpks<VI_char>us): Likewise.
732 (altivec_vpku<VI_char>us): Likewise.
733 (altivec_vpku<VI_char>um): Likewise.
734 (altivec_vpku<VI_char>um_direct): New (copy of
735 altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
737 (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
738 target is little endian and -maltivec=be is not specified.
739 (*altivec_vupkhs<VU_char>_direct): New (copy of
740 altivec_vupkhs<VU_char> that always emits vupkhs*, for internal
742 (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
743 target is little endian and -maltivec=be is not specified.
744 (*altivec_vupkls<VU_char>_direct): New (copy of
745 altivec_vupkls<VU_char> that always emits vupkls*, for internal
747 (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
748 little endian and -maltivec=be is not specified.
749 (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
750 little endian and -maltivec=be is not specified.
752 Backport from mainline r207521
753 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
755 * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
756 generation for -maltivec=be.
757 (altivec_vsumsws): Simplify redundant test.
759 Backport from mainline r207525
760 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
762 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
763 CODE_FOR_altivec_vpku[hw]um to
764 CODE_FOR_altivec_vpku[hw]um_direct.
765 * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
766 UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
767 (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
768 UNSPEC_VUNPACK_LO_SIGN_DIRECT.
770 Backport from mainline r207814.
771 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
773 * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
776 Backport from mainline r207815.
777 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
779 * config/rs6000/altivec.md (p8_vmrgew): Handle little endian
781 (p8_vmrgow): Likewise.
783 Backport from mainline r207919.
784 2014-02-19 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
786 * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
789 Backport from mainline 208019
790 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
792 * config/rs6000/altivec.md (altivec_lvxl): Rename as
793 *altivec_lvxl_<mode>_internal and use VM2 iterator instead of
795 (altivec_lvxl_<mode>): New define_expand incorporating
796 -maltivec=be semantics where needed.
797 (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
798 (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
799 semantics where needed.
800 (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
801 (altivec_stvx_<mode>): New define_expand incorporating
802 -maltivec=be semantics where needed.
803 (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
804 VM2 iterator instead of V4SI.
805 (altivec_stvxl_<mode>): New define_expand incorporating
806 -maltivec=be semantics where needed.
807 * config/rs6000/rs6000-builtin.def: Add new built-in definitions
808 LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
809 LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI,
810 STVX_V2DF, STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI,
811 STVXL_V2DF, STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI,
813 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
814 ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
815 similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
816 ALTIVEC_BUILTIN_STVXL.
817 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New
819 (altivec_expand_stvx_be): Likewise.
820 * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
821 (altivec_expand_lvx_be): Likewise.
822 (altivec_expand_stvx_be): Likewise.
823 (altivec_expand_builtin): Add cases for
824 ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
825 ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
826 (altivec_init_builtins): Add definitions for
827 __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
828 __builtin_altivec_stvx_<mode>, and
829 __builtin_altivec_stvxl_<mode>.
831 Backport from mainline 208021
832 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
834 * config/rs6000/altivec.md (altivec_vsumsws): Replace second
836 (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
839 Backport from mainline 208049
840 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
842 * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
843 define_insn with define_expand and new define_insn
844 *altivec_lve<VI_char>x_internal.
845 (altivec_stve<VI_char>x): Replace define_insn with define_expand
846 and new define_insn *altivec_stve<VI_char>x_internal.
847 * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
849 * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
851 (altivec_expand_stvex_be): New function.
853 Backport from mainline
854 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
855 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
858 Backport from mainline
859 2014-02-25 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
860 * config/rs6000/vector.md (*vector_unordered<mode>): Change split
861 to use canonical form for nor<mode>3.
863 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
865 Backport from mainline
866 2014-02-04 Michael Meissner <meissner@linux.vnet.ibm.com>
868 * config/rs6000/rs6000.opt (-mlra): Add switch to enable the LRA
871 * config/rs6000/rs6000.c (TARGET_LRA_P): Add support for -mlra to
872 enable the LRA register allocator. Back port the changes from the
874 (rs6000_legitimate_offset_address_p): Likewise.
875 (legitimate_lo_sum_address_p): Likewise.
876 (use_toc_relative_ref): Likewise.
877 (rs6000_legitimate_address_p): Likewise.
878 (rs6000_emit_move): Likewise.
879 (rs6000_secondary_memory_needed_mode): Likewise.
880 (rs6000_alloc_sdmode_stack_slot): Likewise.
881 (rs6000_lra_p): Likewise.
883 * config/rs6000/sync.md (load_lockedti): Copy TI/PTI variables by
884 64-bit parts to force the register allocator to allocate even/odd
885 register pairs for the quad word atomic instructions.
886 (store_conditionalti): Likewise.
888 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
890 Back port from mainline
891 2014-01-23 Michael Meissner <meissner@linux.vnet.ibm.com>
894 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
895 -mquad-memory-atomic. Update -mquad-memory documentation to say
896 it is only used for non-atomic loads/stores.
898 * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
899 -mquad-memory or -mquad-memory-atomic switches.
901 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
902 -mquad-memory-atomic to ISA 2.07 support.
904 * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
905 to separate support of normal quad word memory operations (ldq,
906 stq) from the atomic quad word memory operations.
908 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
909 support to separate non-atomic quad word operations from atomic
910 quad word operations. Disable non-atomic quad word operations in
911 little endian mode so that we don't have to swap words after the
912 load and before the store.
913 (quad_load_store_p): Add comment about atomic quad word support.
914 (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
915 options printed with -mdebug=reg.
917 * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
918 -mquad-memory-atomic as the test for whether we have quad word
920 (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic,
921 -mquad-memory, or -mp8-vector are used, allow byte/half-word
924 * config/rs6000/sync.md (load_lockedti): Insure that the address
925 is a proper indexed or indirect address for the lqarx instruction.
926 On little endian systems, swap the hi/lo registers after the lqarx
928 (load_lockedpti): Use indexed_or_indirect_operand predicate to
929 insure the address is valid for the lqarx instruction.
930 (store_conditionalti): Insure that the address is a proper indexed
931 or indirect address for the stqcrx. instruction. On little endian
932 systems, swap the hi/lo registers before doing the stqcrx.
934 (store_conditionalpti): Use indexed_or_indirect_operand predicate to
935 insure the address is valid for the stqcrx. instruction.
937 * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
938 Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
939 type of quad memory support is available.
941 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
943 Apply mainline r202190, powerpc64le multilibs and multiarch dir
944 2013-09-03 Alan Modra <amodra@gmail.com>
946 * config.gcc (powerpc*-*-linux*): Add support for little-endian
947 multilibs to big-endian target and vice versa.
948 * config/rs6000/t-linux64: Use := assignment on all vars.
949 (MULTILIB_EXTRA_OPTS): Remove fPIC.
950 (MULTILIB_OSDIRNAMES): Specify using mapping from multilib_options.
951 * config/rs6000/t-linux64le: New file.
952 * config/rs6000/t-linux64bele: New file.
953 * config/rs6000/t-linux64lebe: New file.
955 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
957 Back port from mainline
958 2014-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
961 * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
962 endian support, remove tests for WORDS_BIG_ENDIAN.
963 (p8_mfvsrd_3_<mode>): Likewise.
964 (reload_gpr_from_vsx<mode>): Likewise.
965 (reload_gpr_from_vsxsf): Likewise.
966 (p8_mfvsrd_4_disf): Likewise.
968 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
970 Backport from mainline
971 2013-04-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
974 * config/rs6000/rs6000.c (rs6000_emit_swdiv_high_precision): Remove.
975 (rs6000_emit_swdiv_low_precision): Remove.
976 (rs6000_emit_swdiv): Rewrite to handle between one and four
977 iterations of Newton-Raphson generally; modify required number of
978 iterations for some cases.
979 * config/rs6000/rs6000.h (RS6000_RECIP_HIGH_PRECISION_P): Remove.
981 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
983 Backport from mainline
984 2013-08-19 Peter Bergner <bergner@vnet.ibm.com>
985 Jakub Jelinek <jakub@redhat.com>
987 * builtins.def (BUILT_IN_FABSD32): New DFP ABS builtin.
988 (BUILT_IN_FABSD64): Likewise.
989 (BUILT_IN_FABSD128): Likewise.
990 * builtins.c (expand_builtin): Add support for
991 new DFP ABS builtins.
992 (fold_builtin_1): Likewise.
993 * config/rs6000/dfp.md
994 (*abstd2_fpr): Handle non-overlapping destination
996 (*nabstd2_fpr): Likewise.
998 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1000 Apply mainline r205060.
1001 2013-11-20 Alan Modra <amodra@gmail.com>
1002 * config/rs6000/sysv4.h (CC1_ENDIAN_LITTLE_SPEC): Define as empty.
1003 * config/rs6000/rs6000.c (rs6000_option_override_internal): Default
1004 to strict alignment on older processors when little-endian.
1005 * config/rs6000/linux64.h (PROCESSOR_DEFAULT64): Default to power8
1008 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1010 POWER ELFv2 ABI Support
1011 Backport from mainline r204842:
1013 2013-11-15 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1015 * doc/invoke.texi (-mabi=elfv1, -mabi=elfv2): Document.
1017 Backport from mainline r204809:
1019 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1021 * config/rs6000/sysv4le.h (LINUX64_DEFAULT_ABI_ELFv2): Define.
1023 Backport from mainline r204808:
1025 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1026 Alan Modra <amodra@gmail.com>
1028 * config/rs6000/rs6000.h (RS6000_SAVE_AREA): Handle ABI_ELFv2.
1029 (RS6000_SAVE_TOC): Remove.
1030 (RS6000_TOC_SAVE_SLOT): New macro.
1031 * config/rs6000/rs6000.c (rs6000_parm_offset): New function.
1032 (rs6000_parm_start): Use it.
1033 (rs6000_function_arg_advance_1): Likewise.
1034 (rs6000_emit_prologue): Use RS6000_TOC_SAVE_SLOT.
1035 (rs6000_emit_epilogue): Likewise.
1036 (rs6000_call_aix): Likewise.
1037 (rs6000_output_function_prologue): Do not save/restore r11
1038 around calling _mcount for ABI_ELFv2.
1040 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1041 Alan Modra <amodra@gmail.com>
1043 * config/rs6000/rs6000-protos.h (rs6000_reg_parm_stack_space):
1045 * config/rs6000/rs6000.h (RS6000_REG_SAVE): Remove.
1046 (REG_PARM_STACK_SPACE): Call rs6000_reg_parm_stack_space.
1047 * config/rs6000/rs6000.c (rs6000_parm_needs_stack): New function.
1048 (rs6000_function_parms_need_stack): Likewise.
1049 (rs6000_reg_parm_stack_space): Likewise.
1050 (rs6000_function_arg): Do not replace BLKmode by Pmode when
1051 returning a register argument.
1053 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1054 Michael Gschwind <mkg@us.ibm.com>
1056 * config/rs6000/rs6000.h (FP_ARG_MAX_RETURN): New macro.
1057 (ALTIVEC_ARG_MAX_RETURN): Likewise.
1058 (FUNCTION_VALUE_REGNO_P): Use them.
1059 * config/rs6000/rs6000.c (TARGET_RETURN_IN_MSB): Define.
1060 (rs6000_return_in_msb): New function.
1061 (rs6000_return_in_memory): Handle ELFv2 homogeneous aggregates.
1062 Handle aggregates of up to 16 bytes for ELFv2.
1063 (rs6000_function_value): Handle ELFv2 homogeneous aggregates.
1065 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1066 Michael Gschwind <mkg@us.ibm.com>
1068 * config/rs6000/rs6000.h (AGGR_ARG_NUM_REG): Define.
1069 * config/rs6000/rs6000.c (rs6000_aggregate_candidate): New function.
1070 (rs6000_discover_homogeneous_aggregate): Likewise.
1071 (rs6000_function_arg_boundary): Handle homogeneous aggregates.
1072 (rs6000_function_arg_advance_1): Likewise.
1073 (rs6000_function_arg): Likewise.
1074 (rs6000_arg_partial_bytes): Likewise.
1075 (rs6000_psave_function_arg): Handle BLKmode arguments.
1077 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1078 Michael Gschwind <mkg@us.ibm.com>
1080 * config/rs6000/rs6000.h (AGGR_ARG_NUM_REG): Define.
1081 * config/rs6000/rs6000.c (rs6000_aggregate_candidate): New function.
1082 (rs6000_discover_homogeneous_aggregate): Likewise.
1083 (rs6000_function_arg_boundary): Handle homogeneous aggregates.
1084 (rs6000_function_arg_advance_1): Likewise.
1085 (rs6000_function_arg): Likewise.
1086 (rs6000_arg_partial_bytes): Likewise.
1087 (rs6000_psave_function_arg): Handle BLKmode arguments.
1089 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1091 * config/rs6000/rs6000.c (machine_function): New member
1093 (rs6000_emit_prologue): Set r2_setup_needed if necessary.
1094 (rs6000_output_mi_thunk): Set r2_setup_needed.
1095 (rs6000_output_function_prologue): Output global entry point
1096 prologue and local entry point marker if needed for ABI_ELFv2.
1097 Output -mprofile-kernel code here.
1098 (output_function_profiler): Do not output -mprofile-kernel
1099 code here; moved to rs6000_output_function_prologue.
1100 (rs6000_file_start): Output ".abiversion 2" for ABI_ELFv2.
1102 (rs6000_emit_move): Do not handle dot symbols for ABI_ELFv2.
1103 (rs6000_output_function_entry): Likewise.
1104 (rs6000_assemble_integer): Likewise.
1105 (rs6000_elf_encode_section_info): Likewise.
1106 (rs6000_elf_declare_function_name): Do not create dot symbols
1107 or .opd section for ABI_ELFv2.
1109 (rs6000_trampoline_size): Update for ABI_ELFv2 trampolines.
1110 (rs6000_trampoline_init): Likewise.
1111 (rs6000_elf_file_end): Call file_end_indicate_exec_stack
1114 (rs6000_call_aix): Handle ELFv2 indirect calls. Do not check
1115 for function descriptors in ABI_ELFv2.
1117 * config/rs6000/rs6000.md ("*call_indirect_aix<mode>"): Support
1118 on ABI_AIX only, not ABI_ELFv2.
1119 ("*call_value_indirect_aix<mode>"): Likewise.
1120 ("*call_indirect_elfv2<mode>"): New pattern.
1121 ("*call_value_indirect_elfv2<mode>"): Likewise.
1123 * config/rs6000/predicates.md ("symbol_ref_operand"): Do not
1124 check for function descriptors in ABI_ELFv2.
1125 ("current_file_function_operand"): Likewise.
1127 * config/rs6000/ppc-asm.h [__powerpc64__ && _CALL_ELF == 2]:
1129 (FUNC_NAME): Define ELFv2 variant.
1130 (JUMP_TARGET): Likewise.
1131 (FUNC_START): Likewise.
1132 (HIDDEN_FUNC): Likewise.
1133 (FUNC_END): Likeiwse.
1135 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1137 * config.gcc [powerpc*-*-* | rs6000-*-*]: Support --with-abi=elfv1
1138 and --with-abi=elfv2.
1139 * config/rs6000/option-defaults.h (OPTION_DEFAULT_SPECS): Add "abi".
1140 * config/rs6000/rs6000.opt (mabi=elfv1): New option.
1141 (mabi=elfv2): Likewise.
1142 * config/rs6000/rs6000-opts.h (enum rs6000_abi): Add ABI_ELFv2.
1143 * config/rs6000/linux64.h (DEFAULT_ABI): Do not hard-code to AIX_ABI
1145 (ELFv2_ABI_CHECK): New macro.
1146 (SUBSUBTARGET_OVERRIDE_OPTIONS): Use it to decide whether to set
1147 rs6000_current_abi to ABI_AIX or ABI_ELFv2.
1148 (GLIBC_DYNAMIC_LINKER64): Support ELFv2 ld.so version.
1149 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Predefine
1150 _CALL_ELF and __STRUCT_PARM_ALIGN__ if appropriate.
1152 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Handle ABI_ELFv2.
1153 (debug_stack_info): Likewise.
1154 (rs6000_file_start): Treat ABI_ELFv2 the same as ABI_AIX.
1155 (rs6000_legitimize_tls_address): Likewise.
1156 (rs6000_conditional_register_usage): Likewise.
1157 (rs6000_emit_move): Likewise.
1158 (init_cumulative_args): Likewise.
1159 (rs6000_function_arg_advance_1): Likewise.
1160 (rs6000_function_arg): Likewise.
1161 (rs6000_arg_partial_bytes): Likewise.
1162 (rs6000_output_function_entry): Likewise.
1163 (rs6000_assemble_integer): Likewise.
1164 (rs6000_savres_strategy): Likewise.
1165 (rs6000_stack_info): Likewise.
1166 (rs6000_function_ok_for_sibcall): Likewise.
1167 (rs6000_emit_load_toc_table): Likewise.
1168 (rs6000_savres_routine_name): Likewise.
1169 (ptr_regno_for_savres): Likewise.
1170 (rs6000_emit_prologue): Likewise.
1171 (rs6000_emit_epilogue): Likewise.
1172 (rs6000_output_function_epilogue): Likewise.
1173 (output_profile_hook): Likewise.
1174 (output_function_profiler): Likewise.
1175 (rs6000_trampoline_size): Likewise.
1176 (rs6000_trampoline_init): Likewise.
1177 (rs6000_elf_output_toc_section_asm_op): Likewise.
1178 (rs6000_elf_encode_section_info): Likewise.
1179 (rs6000_elf_reloc_rw_mask): Likewise.
1180 (rs6000_elf_declare_function_name): Likewise.
1181 (rs6000_function_arg_boundary): Treat ABI_ELFv2 the same as ABI_AIX,
1182 except that rs6000_compat_align_parm is always assumed false.
1183 (rs6000_gimplify_va_arg): Likewise.
1184 (rs6000_call_aix): Update comment.
1185 (rs6000_sibcall_aix): Likewise.
1186 * config/rs6000/rs6000.md ("tls_gd_aix<TLSmode:tls_abi_suffix>"):
1187 Treat ABI_ELFv2 the same as ABI_AIX.
1188 ("*tls_gd_call_aix<TLSmode:tls_abi_suffix>"): Likewise.
1189 ("tls_ld_aix<TLSmode:tls_abi_suffix>"): Likewise.
1190 ("*tls_ld_call_aix<TLSmode:tls_abi_suffix>"): Likewise.
1191 ("load_toc_aix_si"): Likewise.
1192 ("load_toc_aix_di"): Likewise.
1194 ("call_value"): Likewise.
1195 ("*call_local_aix<mode>"): Likewise.
1196 ("*call_value_local_aix<mode>"): Likewise.
1197 ("*call_nonlocal_aix<mode>"): Likewise.
1198 ("*call_value_nonlocal_aix<mode>"): Likewise.
1199 ("*call_indirect_aix<mode>"): Likewise.
1200 ("*call_value_indirect_aix<mode>"): Likewise.
1201 ("sibcall"): Likewise.
1202 ("sibcall_value"): Likewise.
1203 ("*sibcall_aix<mode>"): Likewise.
1204 ("*sibcall_value_aix<mode>"): Likewise.
1205 * config/rs6000/predicates.md ("symbol_ref_operand"): Likewise.
1206 ("current_file_function_operand"): Likewise.
1208 Backport from mainline r204807:
1210 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1212 * config/rs6000/rs6000.c (rs6000_arg_partial_bytes): Simplify logic
1213 by making use of the fact that for vector / floating point arguments
1214 passed both in VRs/FPRs and in the fixed parameter area, the partial
1215 bytes mechanism is in fact not used.
1217 Backport from mainline r204806:
1219 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1221 * config/rs6000/rs6000.c (rs6000_psave_function_arg): New function.
1222 (rs6000_finish_function_arg): Likewise.
1223 (rs6000_function_arg): Use rs6000_psave_function_arg and
1224 rs6000_finish_function_arg to handle both vector and floating
1225 point arguments that are also passed in GPRs / the stack.
1227 Backport from mainline r204805:
1229 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1231 * config/rs6000/rs6000.c (USE_FP_FOR_ARG_P): Remove TYPE argument.
1232 (USE_ALTIVEC_FOR_ARG_P): Likewise.
1233 (rs6000_darwin64_record_arg_advance_recurse): Update uses.
1234 (rs6000_function_arg_advance_1):Likewise.
1235 (rs6000_darwin64_record_arg_recurse): Likewise.
1236 (rs6000_function_arg): Likewise.
1237 (rs6000_arg_partial_bytes): Likewise.
1239 Backport from mainline r204804:
1241 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1243 * config/rs6000/rs6000.c (rs6000_option_override_internal): Replace
1244 "DEFAULT_ABI != ABI_AIX" test by testing for ABI_V4 or ABI_DARWIN.
1245 (rs6000_savres_strategy): Likewise.
1246 (rs6000_return_addr): Likewise.
1247 (rs6000_emit_load_toc_table): Replace "DEFAULT_ABI != ABI_AIX" by
1248 testing for ABI_V4 (since ABI_DARWIN is impossible here).
1249 (rs6000_emit_prologue): Likewise.
1250 (legitimate_lo_sum_address_p): Simplify DEFAULT_ABI test.
1251 (rs6000_elf_declare_function_name): Remove duplicated test.
1252 * config/rs6000/rs6000.md ("load_toc_v4_PIC_1"): Explicitly test
1253 for ABI_V4 (instead of "DEFAULT_ABI != ABI_AIX" test).
1254 ("load_toc_v4_PIC_1_normal"): Likewise.
1255 ("load_toc_v4_PIC_1_476"): Likewise.
1256 ("load_toc_v4_PIC_1b"): Likewise.
1257 ("load_toc_v4_PIC_1b_normal"): Likewise.
1258 ("load_toc_v4_PIC_1b_476"): Likewise.
1259 ("load_toc_v4_PIC_2"): Likewise.
1260 ("load_toc_v4_PIC_3b"): Likewise.
1261 ("load_toc_v4_PIC_3c"): Likewise.
1262 * config/rs6000/rs6000.h (RS6000_REG_SAVE): Simplify DEFAULT_ABI test.
1263 (RS6000_SAVE_AREA): Likewise.
1264 (FP_ARG_MAX_REG): Likewise.
1265 (RETURN_ADDRESS_OFFSET): Likewise.
1266 * config/rs6000/sysv.h (TARGET_TOC): Test for ABI_V4 instead
1268 (SUBTARGET_OVERRIDE_OPTIONS): Likewise.
1269 (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
1271 Backport from mainline r204803:
1273 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1275 * config/rs6000/rs6000.c (rs6000_call_indirect_aix): Rename to ...
1276 (rs6000_call_aix): ... this. Handle both direct and indirect calls.
1277 Create call insn directly instead of via various gen_... routines.
1278 Mention special registers used by the call in CALL_INSN_FUNCTION_USAGE.
1279 (rs6000_sibcall_aix): New function.
1280 * config/rs6000/rs6000.md (TOC_SAVE_OFFSET_32BIT): Remove.
1281 (TOC_SAVE_OFFSET_64BIT): Likewise.
1282 (AIX_FUNC_DESC_TOC_32BIT): Likewise.
1283 (AIX_FUNC_DESC_TOC_64BIT): Likewise.
1284 (AIX_FUNC_DESC_SC_32BIT): Likewise.
1285 (AIX_FUNC_DESC_SC_64BIT): Likewise.
1286 ("call" expander): Call rs6000_call_aix.
1287 ("call_value" expander): Likewise.
1288 ("call_indirect_aix<ptrsize>"): Replace this pattern ...
1289 ("call_indirect_aix<ptrsize>_nor11"): ... and this pattern ...
1290 ("*call_indirect_aix<mode>"): ... by this insn pattern.
1291 ("call_value_indirect_aix<ptrsize>"): Replace this pattern ...
1292 ("call_value_indirect_aix<ptrsize>_nor11"): ... and this pattern ...
1293 ("*call_value_indirect_aix<mode>"): ... by this insn pattern.
1294 ("*call_nonlocal_aix32", "*call_nonlocal_aix64"): Replace by ...
1295 ("*call_nonlocal_aix<mode>"): ... this pattern.
1296 ("*call_value_nonlocal_aix32", "*call_value_nonlocal_aix64"): Replace
1297 ("*call_value_nonlocal_aix<mode>"): ... by this pattern.
1298 ("*call_local_aix<mode>"): New insn pattern.
1299 ("*call_value_local_aix<mode>"): Likewise.
1300 ("sibcall" expander): Call rs6000_sibcall_aix.
1301 ("sibcall_value" expander): Likewise. Move earlier in file.
1302 ("*sibcall_nonlocal_aix<mode>"): Replace by ...
1303 ("*sibcall_aix<mode>"): ... this pattern.
1304 ("*sibcall_value_nonlocal_aix<mode>"): Replace by ...
1305 ("*sibcall_value_aix<mode>"): ... this pattern.
1306 * config/rs6000/rs6000-protos.h (rs6000_call_indirect_aix): Remove.
1307 (rs6000_call_aix): Add prototype.
1308 (rs6000_sibcall_aix): Likewise.
1310 Backport from mainline r204799:
1312 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1314 * config/rs6000/rs6000.c (rs6000_emit_prologue): Do not place a
1315 RTX_FRAME_RELATED_P marker on the UNSPEC_MOVESI_FROM_CR insn.
1316 Instead, add USEs of all modified call-saved CR fields to the
1317 insn storing the result to the stack slot, and provide an
1318 appropriate REG_FRAME_RELATED_EXPR for that insn.
1319 * config/rs6000/rs6000.md ("*crsave"): New insn pattern.
1320 * config/rs6000/predicates.md ("crsave_operation"): New predicate.
1322 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1324 ELFv2 ABI Call Support
1325 Backport from mainline r204798:
1327 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1328 Alan Modra <amodra@gmail.com>
1330 * function.c (assign_parms): Use all.reg_parm_stack_space instead
1331 of re-evaluating REG_PARM_STACK_SPACE target macro.
1332 (locate_and_pad_parm): New parameter REG_PARM_STACK_SPACE. Use it
1333 instead of evaluating target macro REG_PARM_STACK_SPACE every time.
1334 (assign_parm_find_entry_rtl): Update call.
1335 * calls.c (initialize_argument_information): Update call.
1336 (emit_library_call_value_1): Likewise.
1337 * expr.h (locate_and_pad_parm): Update prototype.
1339 Backport from mainline r204797:
1341 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1343 * calls.c (store_unaligned_arguments_into_pseudos): Skip PARALLEL
1346 Backport from mainline r197003:
1348 2013-03-23 Eric Botcazou <ebotcazou@adacore.com>
1350 * calls.c (expand_call): Add missing guard to code handling return
1351 of non-BLKmode structures in MSB.
1352 * function.c (expand_function_end): Likewise.
1354 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1356 Backport from mainline r201750.
1357 Note: Default setting of -mcompat-align-parm inverted!
1359 2013-08-14 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1362 * doc/invoke.texi: Add documentation of mcompat-align-parm
1364 * config/rs6000/rs6000.opt: Add mcompat-align-parm option.
1365 * config/rs6000/rs6000.c (rs6000_function_arg_boundary): For AIX
1366 and Linux, correct BLKmode alignment when 128-bit alignment is
1367 required and compatibility flag is not set.
1368 (rs6000_gimplify_va_arg): For AIX and Linux, honor specified
1369 alignment for zero-size arguments when compatibility flag is not
1372 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1374 Little Endian Vector Support
1375 Backport from mainline r205333
1376 2013-11-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1378 * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Correct
1381 Backport from mainline r205241
1382 2013-11-21 Bill Schmidt <wschmidt@vnet.ibm.com>
1384 * config/rs6000/vector.md (vec_pack_trunc_v2df): Revert previous
1385 little endian change.
1386 (vec_pack_sfix_trunc_v2df): Likewise.
1387 (vec_pack_ufix_trunc_v2df): Likewise.
1388 * config/rs6000/rs6000.c (rs6000_expand_interleave): Correct
1389 double checking of endianness.
1391 Backport from mainline r205146
1392 2013-11-20 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1394 * config/rs6000/vsx.md (vsx_set_<mode>): Adjust for little endian.
1395 (vsx_extract_<mode>): Likewise.
1396 (*vsx_extract_<mode>_one_le): New LE variant on
1397 *vsx_extract_<mode>_zero.
1398 (vsx_extract_v4sf): Adjust for little endian.
1400 Backport from mainline r205080
1401 2013-11-19 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1403 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Adjust
1404 V16QI vector splat case for little endian.
1406 Backport from mainline r205045:
1408 2013-11-19 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1410 * config/rs6000/vector.md ("mov<mode>"): Do not call
1411 rs6000_emit_le_vsx_move to move into or out of GPRs.
1412 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Assert
1413 source and destination are not GPR hard regs.
1415 Backport from mainline r204920
1416 2011-11-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1418 * config/rs6000/rs6000.c (rs6000_frame_related): Add split_reg
1419 parameter and use it in REG_FRAME_RELATED_EXPR note.
1420 (emit_frame_save): Call rs6000_frame_related with extra NULL_RTX
1422 (rs6000_emit_prologue): Likewise, but for little endian VSX
1423 stores, pass the source register of the store instead.
1425 Backport from mainline r204862
1426 2013-11-15 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1428 * config/rs6000/altivec.md (UNSPEC_VPERM_X, UNSPEC_VPERM_UNS_X):
1430 (altivec_vperm_<mode>): Revert earlier little endian change.
1431 (*altivec_vperm_<mode>_internal): Remove.
1432 (altivec_vperm_<mode>_uns): Revert earlier little endian change.
1433 (*altivec_vperm_<mode>_uns_internal): Remove.
1434 * config/rs6000/vector.md (vec_realign_load_<mode>): Revise
1437 Backport from mainline r204441
1438 2013-11-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1440 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1441 Remove restriction against use of VSX instructions when generating
1442 code for little endian mode.
1444 Backport from mainline r204440
1445 2013-11-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1447 * config/rs6000/altivec.md (mulv4si3): Ensure we generate vmulouh
1448 for both big and little endian.
1449 (mulv8hi3): Swap input operands for merge high and merge low
1450 instructions for little endian.
1452 Backport from mainline r204439
1453 2013-11-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1455 * config/rs6000/altivec.md (vec_widen_umult_even_v16qi): Change
1456 define_insn to define_expand that uses even patterns for big
1457 endian and odd patterns for little endian.
1458 (vec_widen_smult_even_v16qi): Likewise.
1459 (vec_widen_umult_even_v8hi): Likewise.
1460 (vec_widen_smult_even_v8hi): Likewise.
1461 (vec_widen_umult_odd_v16qi): Likewise.
1462 (vec_widen_smult_odd_v16qi): Likewise.
1463 (vec_widen_umult_odd_v8hi): Likewise.
1464 (vec_widen_smult_odd_v8hi): Likewise.
1465 (altivec_vmuleub): New define_insn.
1466 (altivec_vmuloub): Likewise.
1467 (altivec_vmulesb): Likewise.
1468 (altivec_vmulosb): Likewise.
1469 (altivec_vmuleuh): Likewise.
1470 (altivec_vmulouh): Likewise.
1471 (altivec_vmulesh): Likewise.
1472 (altivec_vmulosh): Likewise.
1474 Backport from mainline r204395
1475 2013-11-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1477 * config/rs6000/vector.md (vec_pack_sfix_trunc_v2df): Adjust for
1479 (vec_pack_ufix_trunc_v2df): Likewise.
1481 Backport from mainline r204363
1482 2013-11-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1484 * config/rs6000/altivec.md (vec_widen_umult_hi_v16qi): Swap
1485 arguments to merge instruction for little endian.
1486 (vec_widen_umult_lo_v16qi): Likewise.
1487 (vec_widen_smult_hi_v16qi): Likewise.
1488 (vec_widen_smult_lo_v16qi): Likewise.
1489 (vec_widen_umult_hi_v8hi): Likewise.
1490 (vec_widen_umult_lo_v8hi): Likewise.
1491 (vec_widen_smult_hi_v8hi): Likewise.
1492 (vec_widen_smult_lo_v8hi): Likewise.
1494 Backport from mainline r204350
1495 2013-11-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1497 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode> for VSX_D):
1498 Replace the define_insn_and_split with a define_insn and two
1499 define_splits, with the split after reload re-permuting the source
1500 register to its original value.
1501 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
1502 (*vsx_le_perm_store_v8hi): Likewise.
1503 (*vsx_le_perm_store_v16qi): Likewise.
1505 Backport from mainline r204321
1506 2013-11-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1508 * config/rs6000/vector.md (vec_pack_trunc_v2df): Adjust for
1511 Backport from mainline r204321
1512 2013-11-02 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
1514 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Adjust for
1517 Backport from mainline r203980
1518 2013-10-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1520 * config/rs6000/altivec.md (mulv8hi3): Adjust for little endian.
1522 Backport from mainline r203930
1523 2013-10-22 Bill Schmidt <wschmidt@vnet.ibm.com>
1525 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Reverse
1526 meaning of merge-high and merge-low masks for little endian; avoid
1527 use of vector-pack masks for little endian for mismatched modes.
1529 Backport from mainline r203877
1530 2013-10-20 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1532 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Adjust for
1534 (vec_unpacku_hi_v8hi): Likewise.
1535 (vec_unpacku_lo_v16qi): Likewise.
1536 (vec_unpacku_lo_v8hi): Likewise.
1538 Backport from mainline r203863
1539 2013-10-19 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1541 * config/rs6000/rs6000.c (vspltis_constant): Make sure we check
1542 all elements for both endian flavors.
1544 Backport from mainline r203714
1545 2013-10-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1547 * gcc/config/rs6000/vector.md (vec_unpacks_hi_v4sf): Correct for
1549 (vec_unpacks_lo_v4sf): Likewise.
1550 (vec_unpacks_float_hi_v4si): Likewise.
1551 (vec_unpacks_float_lo_v4si): Likewise.
1552 (vec_unpacku_float_hi_v4si): Likewise.
1553 (vec_unpacku_float_lo_v4si): Likewise.
1555 Backport from mainline r203713
1556 2013-10-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1558 * config/rs6000/vsx.md (vsx_concat_<mode>): Adjust output for LE.
1559 (vsx_concat_v2sf): Likewise.
1561 Backport from mainline r203458
1562 2013-10-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1564 * config/rs6000/vsx.md (*vsx_le_perm_load_v2di): Generalize to
1565 handle vector float as well.
1566 (*vsx_le_perm_load_v4si): Likewise.
1567 (*vsx_le_perm_store_v2di): Likewise.
1568 (*vsx_le_perm_store_v4si): Likewise.
1570 Backport from mainline r203457
1571 2013-10-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1573 * config/rs6000/vector.md (vec_realign_load<mode>): Generate vperm
1574 directly to circumvent subtract from splat{31} workaround.
1575 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_le): New
1577 * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): New.
1578 * config/rs6000/altivec.md (define_c_enum "unspec"): Add
1579 UNSPEC_VPERM_X and UNSPEC_VPERM_UNS_X.
1580 (altivec_vperm_<mode>): Convert to define_insn_and_split to
1581 separate big and little endian logic.
1582 (*altivec_vperm_<mode>_internal): New define_insn.
1583 (altivec_vperm_<mode>_uns): Convert to define_insn_and_split to
1584 separate big and little endian logic.
1585 (*altivec_vperm_<mode>_uns_internal): New define_insn.
1586 (vec_permv16qi): Add little endian logic.
1588 Backport from mainline r203247
1589 2013-10-07 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1591 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const_le): New.
1592 (altivec_expand_vec_perm_const): Call it.
1594 Backport from mainline r203246
1595 2013-10-07 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1597 * config/rs6000/vector.md (mov<mode>): Emit permuted move
1598 sequences for LE VSX loads and stores at expand time.
1599 * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_move): New
1601 * config/rs6000/rs6000.c (rs6000_const_vec): New.
1602 (rs6000_gen_le_vsx_permute): New.
1603 (rs6000_gen_le_vsx_load): New.
1604 (rs6000_gen_le_vsx_store): New.
1605 (rs6000_gen_le_vsx_move): New.
1606 * config/rs6000/vsx.md (*vsx_le_perm_load_v2di): New.
1607 (*vsx_le_perm_load_v4si): New.
1608 (*vsx_le_perm_load_v8hi): New.
1609 (*vsx_le_perm_load_v16qi): New.
1610 (*vsx_le_perm_store_v2di): New.
1611 (*vsx_le_perm_store_v4si): New.
1612 (*vsx_le_perm_store_v8hi): New.
1613 (*vsx_le_perm_store_v16qi): New.
1614 (*vsx_xxpermdi2_le_<mode>): New.
1615 (*vsx_xxpermdi4_le_<mode>): New.
1616 (*vsx_xxpermdi8_le_V8HI): New.
1617 (*vsx_xxpermdi16_le_V16QI): New.
1618 (*vsx_lxvd2x2_le_<mode>): New.
1619 (*vsx_lxvd2x4_le_<mode>): New.
1620 (*vsx_lxvd2x8_le_V8HI): New.
1621 (*vsx_lxvd2x16_le_V16QI): New.
1622 (*vsx_stxvd2x2_le_<mode>): New.
1623 (*vsx_stxvd2x4_le_<mode>): New.
1624 (*vsx_stxvd2x8_le_V8HI): New.
1625 (*vsx_stxvd2x16_le_V16QI): New.
1627 Backport from mainline r201235
1628 2013-07-24 Bill Schmidt <wschmidt@linux.ibm.com>
1629 Anton Blanchard <anton@au1.ibm.com>
1631 * config/rs6000/altivec.md (altivec_vpkpx): Handle little endian.
1632 (altivec_vpks<VI_char>ss): Likewise.
1633 (altivec_vpks<VI_char>us): Likewise.
1634 (altivec_vpku<VI_char>us): Likewise.
1635 (altivec_vpku<VI_char>um): Likewise.
1637 Backport from mainline r201208
1638 2013-07-24 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
1639 Anton Blanchard <anton@au1.ibm.com>
1641 * config/rs6000/vector.md (vec_realign_load_<mode>): Reorder input
1642 operands to vperm for little endian.
1643 * config/rs6000/rs6000.c (rs6000_expand_builtin): Use lvsr instead
1644 of lvsl to create the control mask for a vperm for little endian.
1646 Backport from mainline r201195
1647 2013-07-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1648 Anton Blanchard <anton@au1.ibm.com>
1650 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Reverse
1651 two operands for little-endian.
1653 Backport from mainline r201193
1654 2013-07-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1655 Anton Blanchard <anton@au1.ibm.com>
1657 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Correct
1658 selection of field for vector splat in little endian mode.
1660 Backport from mainline r201149
1661 2013-07-22 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
1662 Anton Blanchard <anton@au1.ibm.com>
1664 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Fix
1665 endianness when selecting field to splat.
1667 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1669 Backport from mainline r205123:
1671 2013-11-20 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1673 * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Do not
1674 allow subregs of TDmode in FPRs of smaller size in little-endian.
1675 (rs6000_split_multireg_move): When splitting an access to TDmode
1676 in FPRs, do not use simplify_gen_subreg.
1678 Backport from mainline r204927:
1680 2013-11-17 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1682 * config/rs6000/rs6000.c (rs6000_emit_move): Use low word of
1683 sdmode_stack_slot also in little-endian mode.
1685 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1688 Backport from mainline
1689 2013-12-03 Peter Bergner <bergner@vnet.ibm.com>
1691 * config/rs6000/htmintrin.h (_TEXASR_INSTRUCTION_FETCH_CONFLICT): Fix
1693 (_TEXASRU_INSTRUCTION_FETCH_CONFLICT): Likewise.
1695 Backport from mainline r205233.
1696 2013-11-21 Peter Bergner <bergner@vnet.ibm.com>
1698 * doc/extend.texi: Document htm builtins.
1700 Backport from mainline
1701 2013-07-17 Iain Sandoe <iain@codesourcery.com>
1703 * config/rs6000/darwin.h (REGISTER_NAMES): Add HTM registers.
1705 Backport from mainline
1706 2013-07-16 Peter Bergner <bergner@vnet.ibm.com>
1708 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
1709 enable extra ISA flags with TARGET_HTM.
1711 2013-07-16 Jakub Jelinek <jakub@redhat.com>
1712 Peter Bergner <bergner@vnet.ibm.com>
1714 * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTERS): Mention HTM
1715 registers in the comment.
1716 (DWARF_FRAME_REGISTERS): Subtract also the 3 HTM registers.
1717 (DWARF_REG_TO_UNWIND_COLUMN): Use DWARF_FRAME_REGISTERS
1718 rather than FIRST_PSEUDO_REGISTERS.
1720 * config.gcc (powerpc*-*-*): Install htmintrin.h and htmxlintrin.h.
1721 * config/rs6000/t-rs6000 (MD_INCLUDES): Add htm.md.
1722 * config/rs6000/rs6000.opt: Add -mhtm option.
1723 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Add OPTION_MASK_HTM.
1724 (ISA_2_7_MASKS_SERVER): Add OPTION_MASK_HTM.
1725 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
1726 __HTM__ if the HTM instructions are available.
1727 * config/rs6000/predicates.md (u3bit_cint_operand, u10bit_cint_operand)
1728 (htm_spr_reg_operand): New define_predicates.
1729 * config/rs6000/rs6000.md (define_attr "type"): Add htm.
1730 (TFHAR_REGNO, TFIAR_REGNO, TEXASR_REGNO): New define_constants.
1732 * config/rs6000/rs6000-builtin.def (BU_HTM_0, BU_HTM_1, BU_HTM_2)
1733 (BU_HTM_3, BU_HTM_SPR0, BU_HTM_SPR1): Add support macros for defining
1734 HTM builtin functions.
1735 * config/rs6000/rs6000.c (RS6000_BUILTIN_H): New macro.
1736 (rs6000_reg_names, alt_reg_names): Add HTM SPR register names.
1737 (rs6000_init_hard_regno_mode_ok): Add support for HTM instructions.
1738 (rs6000_builtin_mask_calculate): Likewise.
1739 (rs6000_option_override_internal): Likewise.
1740 (bdesc_htm): Add new HTM builtin support.
1741 (htm_spr_num): New function.
1742 (htm_spr_regno): Likewise.
1743 (rs6000_htm_spr_icode): Likewise.
1744 (htm_expand_builtin): Likewise.
1745 (htm_init_builtins): Likewise.
1746 (rs6000_expand_builtin): Add support for HTM builtin functions.
1747 (rs6000_init_builtins): Likewise.
1748 (rs6000_invalid_builtin, rs6000_opt_mask): Add support for -mhtm option.
1749 * config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mhtm.
1750 (TARGET_HTM, MASK_HTM): Define macros.
1751 (FIRST_PSEUDO_REGISTER): Adjust for new HTM SPR registers.
1752 (FIXED_REGISTERS): Likewise.
1753 (CALL_USED_REGISTERS): Likewise.
1754 (CALL_REALLY_USED_REGISTERS): Likewise.
1755 (REG_ALLOC_ORDER): Likewise.
1756 (enum reg_class): Likewise.
1757 (REG_CLASS_NAMES): Likewise.
1758 (REG_CLASS_CONTENTS): Likewise.
1759 (REGISTER_NAMES): Likewise.
1760 (ADDITIONAL_REGISTER_NAMES): Likewise.
1761 (RS6000_BTC_SPR, RS6000_BTC_VOID, RS6000_BTC_32BIT, RS6000_BTC_64BIT)
1762 (RS6000_BTC_MISC_MASK, RS6000_BTM_HTM): New macros.
1763 (RS6000_BTM_COMMON): Add RS6000_BTM_HTM.
1764 * config/rs6000/htm.md: New file.
1765 * config/rs6000/htmintrin.h: New file.
1766 * config/rs6000/htmxlintrin.h: New file.
1768 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1772 2013-11-23 Alan Modra <amodra@gmail.com>
1773 * config/rs6000/vsx.md (fusion peepholes): Disable when !TARGET_VSX.
1775 Backport from mainline
1776 2013-11-12 Michael Meissner <meissner@linux.vnet.ibm.com>
1779 * config/rs6000/rs6000.md (movdi_internal32): Eliminate
1780 constraints that would allow DImode into the traditional Altivec
1781 registers, but cause undesirable code generation when loading 0 as
1783 (movdi_internal64): Likewise.
1784 (cmp<mode>_fpr): Do not use %x for CR register output.
1785 (extendsfdf2_fpr): Fix constraints when -mallow-upper-df and
1786 -mallow-upper-sf debug switches are used.
1788 Backport from mainline
1789 2013-10-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1791 * config/rs6000/rs6000.c (enum rs6000_reload_reg_type): Add new
1792 fields to the reg_addr array that describes the valid addressing
1793 mode for any register, general purpose registers, floating point
1794 registers, and Altivec registers.
1795 (FIRST_RELOAD_REG_CLASS): Likewise.
1796 (LAST_RELOAD_REG_CLASS): Likewise.
1797 (struct reload_reg_map_type): Likewise.
1798 (reload_reg_map_type): Likewise.
1799 (RELOAD_REG_VALID): Likewise.
1800 (RELOAD_REG_MULTIPLE): Likewise.
1801 (RELOAD_REG_INDEXED): Likewise.
1802 (RELOAD_REG_OFFSET): Likewise.
1803 (RELOAD_REG_PRE_INCDEC): Likewise.
1804 (RELOAD_REG_PRE_MODIFY): Likewise.
1805 (reg_addr): Likewise.
1806 (mode_supports_pre_incdec_p): New helper functions to say whether
1807 a given mode supports PRE_INC, PRE_DEC, and PRE_MODIFY.
1808 (mode_supports_pre_modify_p): Likewise.
1809 (rs6000_debug_vector_unit): Rearrange the -mdebug=reg output to
1810 print the valid address mode bits for each mode.
1811 (rs6000_debug_print_mode): Likewise.
1812 (rs6000_debug_reg_global): Likewise.
1813 (rs6000_setup_reg_addr_masks): New function to set up the address
1814 mask bits for each type.
1815 (rs6000_init_hard_regno_mode_ok): Use memset to clear arrays.
1816 Call rs6000_setup_reg_addr_masks to set up the address mask bits.
1817 (rs6000_legitimate_address_p): Use mode_supports_pre_incdec_p and
1818 mode_supports_pre_modify_p to determine if PRE_INC, PRE_DEC, and
1819 PRE_MODIFY are supported.
1820 (rs6000_output_move_128bit): Change to use {src,dest}_vmx_p for altivec
1821 registers, instead of {src,dest}_av_p.
1822 (rs6000_print_options_internal): Tweak the debug output slightly.
1824 Backport from mainline
1825 2013-10-03 Michael Meissner <meissner@linux.vnet.ibm.com>
1827 * config/rs6000/rs6000-builtin.def (XSRDPIM): Use floatdf2,
1828 ceildf2, btruncdf2, instead of vsx_* name.
1830 * config/rs6000/vsx.md (vsx_add<mode>3): Change arithmetic
1831 iterators to only do V2DF and V4SF here. Move the DF code to
1832 rs6000.md where it is combined with SF mode. Replace <VSv> with
1833 just 'v' since only vector operations are handled with these insns
1834 after moving the DF support to rs6000.md.
1835 (vsx_sub<mode>3): Likewise.
1836 (vsx_mul<mode>3): Likewise.
1837 (vsx_div<mode>3): Likewise.
1838 (vsx_fre<mode>2): Likewise.
1839 (vsx_neg<mode>2): Likewise.
1840 (vsx_abs<mode>2): Likewise.
1841 (vsx_nabs<mode>2): Likewise.
1842 (vsx_smax<mode>3): Likewise.
1843 (vsx_smin<mode>3): Likewise.
1844 (vsx_sqrt<mode>2): Likewise.
1845 (vsx_rsqrte<mode>2): Likewise.
1846 (vsx_fms<mode>4): Likewise.
1847 (vsx_nfma<mode>4): Likewise.
1848 (vsx_copysign<mode>3): Likewise.
1849 (vsx_btrunc<mode>2): Likewise.
1850 (vsx_floor<mode>2): Likewise.
1851 (vsx_ceil<mode>2): Likewise.
1852 (vsx_smaxsf3): Delete scalar ops that were moved to rs6000.md.
1853 (vsx_sminsf3): Likewise.
1854 (vsx_fmadf4): Likewise.
1855 (vsx_fmsdf4): Likewise.
1856 (vsx_nfmadf4): Likewise.
1857 (vsx_nfmsdf4): Likewise.
1858 (vsx_cmpdf_internal1): Likewise.
1860 * config/rs6000/rs6000.h (TARGET_SF_SPE): Define macros to make it
1861 simpler to select whether a target has SPE or traditional floating
1862 point support in iterators.
1863 (TARGET_DF_SPE): Likewise.
1864 (TARGET_SF_FPR): Likewise.
1865 (TARGET_DF_FPR): Likewise.
1866 (TARGET_SF_INSN): Macros to say whether floating point support
1867 exists for a given operation for expanders.
1868 (TARGET_DF_INSN): Likewise.
1870 * config/rs6000/rs6000.c (Ftrad): New mode attributes to allow
1871 combining of SF/DF mode operations, using both traditional and VSX
1879 (abs<mode>2): Combine SF/DF modes using traditional floating point
1880 instructions. Add support for using the upper DF registers with
1881 VSX support, and SF registers with power8-vector support. Update
1882 expanders for operations supported by both the SPE and traditional
1883 floating point units.
1884 (abs<mode>2_fpr): Likewise.
1885 (nabs<mode>2): Likewise.
1886 (nabs<mode>2_fpr): Likewise.
1887 (neg<mode>2): Likewise.
1888 (neg<mode>2_fpr): Likewise.
1889 (add<mode>3): Likewise.
1890 (add<mode>3_fpr): Likewise.
1891 (sub<mode>3): Likewise.
1892 (sub<mode>3_fpr): Likewise.
1893 (mul<mode>3): Likewise.
1894 (mul<mode>3_fpr): Likewise.
1895 (div<mode>3): Likewise.
1896 (div<mode>3_fpr): Likewise.
1897 (sqrt<mode>3): Likewise.
1898 (sqrt<mode>3_fpr): Likewise.
1899 (fre<Fs>): Likewise.
1900 (rsqrt<mode>2): Likewise.
1901 (cmp<mode>_fpr): Likewise.
1902 (smax<mode>3): Likewise.
1903 (smin<mode>3): Likewise.
1904 (smax<mode>3_vsx): Likewise.
1905 (smin<mode>3_vsx): Likewise.
1906 (negsf2): Delete SF operations that are merged with DF.
1913 (fmasf4_fpr): Likewise.
1914 (fmssf4_fpr): Likewise.
1915 (nfmasf4_fpr): Likewise.
1916 (nfmssf4_fpr): Likewise.
1917 (sqrtsf2): Likewise.
1918 (rsqrtsf_internal1): Likewise.
1919 (smaxsf3): Likewise.
1920 (sminsf3): Likewise.
1921 (cmpsf_internal1): Likewise.
1922 (copysign<mode>3_fcpsgn): Add VSX/power8-vector support.
1923 (negdf2): Delete DF operations that are merged with SF.
1925 (nabsdf2): Likewise.
1931 (rsqrtdf_internal1): Likewise.
1932 (fmadf4_fpr): Likewise.
1933 (fmsdf4_fpr): Likewise.
1934 (nfmadf4_fpr): Likewise.
1935 (nfmsdf4_fpr): Likewise.
1936 (sqrtdf2): Likewise.
1937 (smaxdf3): Likewise.
1938 (smindf3): Likewise.
1939 (cmpdf_internal1): Likewise.
1940 (lrint<mode>di2): Use TARGET_<MODE>_FPR macro.
1941 (btrunc<mode>2): Delete separate expander, and combine with the
1942 insn and add VSX instruction support. Use TARGET_<MODE>_FPR.
1943 (btrunc<mode>2_fpr): Likewise.
1944 (ceil<mode>2): Likewise.
1945 (ceil<mode>2_fpr): Likewise.
1946 (floor<mode>2): Likewise.
1947 (floor<mode>2_fpr): Likewise.
1948 (fma<mode>4_fpr): Combine SF and DF fused multiply/add support.
1949 Add support for using the upper registers with VSX and
1950 power8-vector. Move insns to be closer to the define_expands. On
1951 VSX systems, prefer the traditional form of FMA over the VSX
1952 version, since the traditional form allows the target not to
1953 overlap with the inputs.
1954 (fms<mode>4_fpr): Likewise.
1955 (nfma<mode>4_fpr): Likewise.
1956 (nfms<mode>4_fpr): Likewise.
1958 Backport from mainline
1959 2013-09-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1961 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Allow
1962 DFmode, DImode, and SFmode in the upper VSX registers based on the
1963 -mupper-regs-{df,sf} flags. Fix wu constraint to be ALTIVEC_REGS
1964 if -mpower8-vector. Combine -mvsx-timode handling with the rest
1965 of the VSX register handling.
1967 * config/rs6000/rs6000.md (f32_lv): Use %x0 for VSX regsters.
1969 (zero_extendsidi2_lfiwzx): Add support for loading into the
1970 Altivec registers with -mpower8-vector. Use wu/wv constraints to
1971 only do VSX memory options on Altivec registers.
1972 (extendsidi2_lfiwax): Likewise.
1973 (extendsfdf2_fpr): Likewise.
1974 (mov<mode>_hardfloat, SF/SD modes): Likewise.
1975 (mov<mode>_hardfloat32, DF/DD modes): Likewise.
1976 (mov<mode>_hardfloat64, DF/DD modes): Likewise.
1977 (movdi_internal64): Likewise.
1979 Backport from mainline
1980 2013-09-23 Michael Meissner <meissner@linux.vnet.ibm.com>
1982 * config/rs6000/rs6000.c (rs6000_vector_reload): Delete, combine
1983 reload helper function arrays into a single array reg_addr.
1984 (reload_fpr_gpr): Likewise.
1985 (reload_gpr_vsx): Likewise.
1986 (reload_vsx_gpr): Likewise.
1987 (struct rs6000_reg_addr): Likewise.
1988 (reg_addr): Likewise.
1989 (rs6000_debug_reg_global): Change rs6000_vector_reload,
1990 reload_fpr_gpr, reload_gpr_vsx, reload_vsx_gpr uses to reg_addr.
1991 (rs6000_init_hard_regno_mode_ok): Likewise.
1992 (rs6000_secondary_reload_direct_move): Likewise.
1993 (rs6000_secondary_reload): Likewise.
1995 * config/rs6000/rs6000.h (enum r6000_reg_class_enum): Add new
1996 constraints: wu, ww, and wy. Repurpose wv constraint added during
1997 power8 changes. Put wg constraint in alphabetical order.
1999 * config/rs6000/rs6000.opt (-mvsx-scalar-float): New debug switch
2000 for future work to add ISA 2.07 VSX single precision support.
2001 (-mvsx-scalar-double): Change default from -1 to 1, update
2002 documentation comment.
2003 (-mvsx-scalar-memory): Rename debug switch to -mupper-regs-df.
2004 (-mupper-regs-df): New debug switch to control whether DF values
2005 can go in the traditional Altivec registers.
2006 (-mupper-regs-sf): New debug switch to control whether SF values
2007 can go in the traditional Altivec registers.
2009 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print wu, ww,
2011 (rs6000_init_hard_regno_mode_ok): Use ssize_t instead of int for
2012 loop variables. Rename -mvsx-scalar-memory to -mupper-regs-df.
2013 Add new constraints, wu/ww/wy. Repurpose wv constraint.
2014 (rs6000_debug_legitimate_address_p): Print if we are running
2015 before, during, or after reload.
2016 (rs6000_secondary_reload): Add a comment.
2017 (rs6000_opt_masks): Add -mupper-regs-df, -mupper-regs-sf.
2019 * config/rs6000/constraints.md (wa constraint): Sort w<x>
2020 constraints. Update documentation string.
2021 (wd constraint): Likewise.
2022 (wf constraint): Likewise.
2023 (wg constraint): Likewise.
2024 (wn constraint): Likewise.
2025 (ws constraint): Likewise.
2026 (wt constraint): Likewise.
2027 (wx constraint): Likewise.
2028 (wz constraint): Likewise.
2029 (wu constraint): New constraint for ISA 2.07 SFmode scalar
2031 (ww constraint): Likewise.
2032 (wy constraint): Likewise.
2033 (wv constraint): Repurpose ISA 2.07 constraint that did not use in
2034 the previous submissions.
2035 * doc/md.texi (PowerPC and IBM RS6000): Likewise.
2037 Backport from mainline
2038 2013-10-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2041 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Only
2042 restrict TImode addresses to single indirect registers if both
2043 -mquad-memory and -mvsx-timode are used.
2044 (rs6000_output_move_128bit): Use quad_load_store_p to determine if
2045 we should emit load/store quad. Remove using %y for quad memory
2048 * config/rs6000/rs6000.md (mov<mode>_ppc64, TI/PTImode): Add
2049 constraints to allow load/store quad on machines where TImode is
2050 not allowed in VSX registers. Use 'n' instead of 'F' constraint
2051 for TImode to load integer constants.
2053 Backport from mainline
2054 2013-10-02 Michael Meissner <meissner@linux.vnet.ibm.com>
2057 * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Turn off
2058 setting -mvsx-timode by default until the underlying problem is
2060 (RS6000_CPU, power7 defaults): Likewise.
2063 2013-08-16 Michael Meissner <meissner@linux.vnet.ibm.com>
2066 * config/rs6000/predicates.md (fusion_gpr_mem_load): Allow the
2067 memory rtx to contain ZERO_EXTEND and SIGN_EXTEND.
2069 * config/rs6000/rs6000-protos.h (fusion_gpr_load_p): Pass operands
2070 array instead of each individual operand as a separate argument.
2071 (emit_fusion_gpr_load): Likewise.
2072 (expand_fusion_gpr_load): Add new function declaration.
2074 * config/rs6000/rs6000.c (fusion_gpr_load_p): Change the calling
2075 signature to have the operands passed as an array, instead of as
2076 separate arguments. Allow ZERO_EXTEND to be in the memory
2077 address, and also SIGN_EXTEND if -mpower8-fusion-sign. Do not
2078 depend on the register live/dead flags when peepholes are run.
2079 (expand_fusion_gpr_load): New function to be called from the
2080 peephole2 pass, to change the register that addis sets to be the
2082 (emit_fusion_gpr_load): Change the calling signature to have the
2083 operands passed as an array, instead of as separate arguments.
2084 Allow ZERO_EXTEND to be in the memory address, and also
2085 SIGN_EXTEND if -mpower8-fusion-sign.
2087 * config/rs6000/rs6000.md (UNSPEC_FUSION_GPR): Delete unused
2089 (power8 fusion peephole/peephole2): Rework the fusion peepholes to
2090 adjust the register addis loads up in the peephole2 pass. Do not
2091 depend on the register live/dead state when the peephole pass is
2095 2013-07-23 Michael Meissner <meissner@linux.vnet.ibm.com>
2097 * config/rs6000/vector.md (xor<mode>3): Move 128-bit boolean
2098 expanders to rs6000.md.
2099 (ior<mode>3): Likewise.
2100 (and<mode>3): Likewise.
2101 (one_cmpl<mode>2): Likewise.
2102 (nor<mode>3): Likewise.
2103 (andc<mode>3): Likewise.
2104 (eqv<mode>3): Likewise.
2105 (nand<mode>3): Likewise.
2106 (orc<mode>3): Likewise.
2108 * config/rs6000/rs6000-protos.h (rs6000_split_logical): New
2111 * config/rs6000/rs6000.c (rs6000_split_logical_inner): Add support
2112 to split multi-word logical operations.
2113 (rs6000_split_logical_di): Likewise.
2114 (rs6000_split_logical): Likewise.
2116 * config/rs6000/vsx.md (VSX_L2): Delete, no longer used.
2117 (vsx_and<mode>3_32bit): Move 128-bit logical insns to rs6000.md,
2118 and allow TImode operations in 32-bit.
2119 (vsx_and<mode>3_64bit): Likewise.
2120 (vsx_ior<mode>3_32bit): Likewise.
2121 (vsx_ior<mode>3_64bit): Likewise.
2122 (vsx_xor<mode>3_32bit): Likewise.
2123 (vsx_xor<mode>3_64bit): Likewise.
2124 (vsx_one_cmpl<mode>2_32bit): Likewise.
2125 (vsx_one_cmpl<mode>2_64bit): Likewise.
2126 (vsx_nor<mode>3_32bit): Likewise.
2127 (vsx_nor<mode>3_64bit): Likewise.
2128 (vsx_andc<mode>3_32bit): Likewise.
2129 (vsx_andc<mode>3_64bit): Likewise.
2130 (vsx_eqv<mode>3_32bit): Likewise.
2131 (vsx_eqv<mode>3_64bit): Likewise.
2132 (vsx_nand<mode>3_32bit): Likewise.
2133 (vsx_nand<mode>3_64bit): Likewise.
2134 (vsx_orc<mode>3_32bit): Likewise.
2135 (vsx_orc<mode>3_64bit): Likewise.
2137 * config/rs6000/rs6000.h (VLOGICAL_REGNO_P): Always allow vector
2138 logical types in GPRs.
2140 * config/rs6000/altivec.md (altivec_and<mode>3): Move 128-bit
2141 logical insns to rs6000.md, and allow TImode operations in
2143 (altivec_ior<mode>3): Likewise.
2144 (altivec_xor<mode>3): Likewise.
2145 (altivec_one_cmpl<mode>2): Likewise.
2146 (altivec_nor<mode>3): Likewise.
2147 (altivec_andc<mode>3): Likewise.
2149 * config/rs6000/rs6000.md (BOOL_128): New mode iterators and mode
2150 attributes for moving the 128-bit logical operations into
2152 (BOOL_REGS_OUTPUT): Likewise.
2153 (BOOL_REGS_OP1): Likewise.
2154 (BOOL_REGS_OP2): Likewise.
2155 (BOOL_REGS_UNARY): Likewise.
2156 (BOOL_REGS_AND_CR0): Likewise.
2157 (one_cmpl<mode>2): Add support for DI logical operations on
2158 32-bit, splitting the operations to 32-bit.
2162 (and<mode>3, 128-bit types): Rewrite 2013-06-06 logical operator
2163 changes to combine the 32/64-bit code, allow logical operations on
2164 TI mode in 32-bit, and to use similar match_operator patterns like
2165 scalar mode uses. Combine the Altivec and VSX code for logical
2166 operations, and move it here.
2167 (ior<mode>3, 128-bit types): Likewise.
2168 (xor<mode>3, 128-bit types): Likewise.
2169 (one_cmpl<mode>3, 128-bit types): Likewise.
2170 (nor<mode>3, 128-bit types): Likewise.
2171 (andc<mode>3, 128-bit types): Likewise.
2172 (eqv<mode>3, 128-bit types): Likewise.
2173 (nand<mode>3, 128-bit types): Likewise.
2174 (orc<mode>3, 128-bit types): Likewise.
2175 (and<mode>3_internal): Likewise.
2176 (bool<mode>3_internal): Likewise.
2177 (boolc<mode>3_internal1): Likewise.
2178 (boolc<mode>3_internal2): Likewise.
2179 (boolcc<mode>3_internal1): Likewise.
2180 (boolcc<mode>3_internal2): Likewise.
2181 (eqv<mode>3_internal1): Likewise.
2182 (eqv<mode>3_internal2): Likewise.
2183 (one_cmpl1<mode>3_internal): Likewise.
2185 Back port from mainline:
2186 2013-06-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2187 Pat Haugen <pthaugen@us.ibm.com>
2188 Peter Bergner <bergner@vnet.ibm.com>
2190 * lib/target-supports.exp (check_p8vector_hw_available) Add power8
2192 (check_effective_target_powerpc_p8vector_ok): Likewise.
2193 (is-effective-target): Likewise.
2194 (check_vect_support_and_set_flags): Likewise.
2196 Backport from mainline
2197 2013-07-31 Michael Meissner <meissner@linux.vnet.ibm.com>
2199 * config/rs6000/predicates.md (fusion_gpr_addis): New predicates
2200 to support power8 load fusion.
2201 (fusion_gpr_mem_load): Likewise.
2203 * config/rs6000/rs6000-modes.def (PTImode): Update a comment.
2205 * config/rs6000/rs6000-protos.h (fusion_gpr_load_p): New
2206 declarations for power8 load fusion.
2207 (emit_fusion_gpr_load): Likewise.
2209 * config/rs6000/rs6000.c (rs6000_option_override_internal): If
2210 tuning for power8, turn on fusion mode by default. Turn on sign
2211 extending fusion mode if normal fusion mode is on, and we are at
2213 (fusion_gpr_load_p): New function, return true if we can fuse an
2214 addis instruction with a dependent load to a GPR.
2215 (emit_fusion_gpr_load): Emit the instructions for power8 load
2218 * config/rs6000/vsx.md (VSX_M2): New iterator for fusion
2220 (VSX load fusion peepholes): New peepholes to fuse together an
2221 addi instruction with a VSX load instruction.
2223 * config/rs6000/rs6000.md (GPR load fusion peepholes): New
2224 peepholes to fuse an addis instruction with a load to a GPR base
2225 register. If we are supporting sign extending fusions, convert
2226 sign extending loads to zero extending loads and add an explicit
2229 Backport from mainline
2230 2013-07-18 Pat Haugen <pthaugen@us.ibm.com>
2232 * config/rs6000/rs6000.c (rs6000_option_override_internal): Adjust flag
2233 interaction for new Power8 flags and VSX.
2235 Back port from the trunk
2236 2013-06-28 Michael Meissner <meissner@linux.vnet.ibm.com>
2239 * config/rs6000/rs6000.h (MODES_TIEABLE_P): Do not allow PTImode
2240 to tie with any other modes. Eliminate Altivec vector mode tests,
2241 since these are a subset of ALTIVEC or VSX vector modes. Simplify
2242 code, to return 0 if testing MODE2 for a condition, if we've
2243 already tested MODE1 for the same condition.
2245 Backport from mainline
2246 2013-06-28 Pat Haugen <pthaugen@us.ibm.com>
2248 * config/rs6000/rs6000.md (define_insn ""): Fix insn type.
2250 Back port from the trunk
2251 2013-06-26 Michael Meissner <meissner@linux.vnet.ibm.com>
2252 Pat Haugen <pthaugen@us.ibm.com>
2253 Peter Bergner <bergner@vnet.ibm.com>
2255 * config/rs6000/power8.md: New.
2256 * config/rs6000/rs6000-cpus.def (RS6000_CPU table): Adjust processor
2257 setting for power8 entry.
2258 * config/rs6000/t-rs6000 (MD_INCLUDES): Add power8.md.
2259 * config/rs6000/rs6000.c (is_microcoded_insn, is_cracked_insn): Adjust
2260 test for Power4/Power5 only.
2261 (insn_must_be_first_in_group, insn_must_be_last_in_group): Add Power8
2263 (force_new_group): Adjust comment.
2264 * config/rs6000/rs6000.md: Include power8.md.
2266 Back port from the trunk
2267 2013-06-14 Michael Meissner <meissner@linux.vnet.ibm.com>
2270 * config/rs6000/rs6000.md (mov<mode>_ppc64): Call
2271 rs6000_output_move_128bit to handle emitting quad memory
2272 operations. Set attribute length to 8 bytes.
2274 Back port from the trunk
2275 2013-06-13 Michael Meissner <meissner@linux.vnet.ibm.com>
2277 * config/rs6000/rs6000.c (rs6000_option_override_internal): Move
2278 test for clearing quad memory on 32-bit later.
2280 Back port from the trunk
2282 2013-06-12 Michael Meissner <meissner@linux.vnet.ibm.com>
2283 Pat Haugen <pthaugen@us.ibm.com>
2284 Peter Bergner <bergner@vnet.ibm.com>
2286 * config/rs6000/rs6000.c (emit_load_locked): Add support for
2287 power8 byte, half-word, and quad-word atomic instructions.
2288 (emit_store_conditional): Likewise.
2289 (rs6000_expand_atomic_compare_and_swap): Likewise.
2290 (rs6000_expand_atomic_op): Likewise.
2292 * config/rs6000/sync.md (larx): Add new modes for power8.
2294 (AINT): New mode iterator to include TImode as well as normal
2295 integer modes on power8.
2296 (fetchop_pred): Use int_reg_operand instead of gpc_reg_operand so
2297 that VSX registers are not considered. Use AINT mode iterator
2298 instead of INT1 to allow inclusion of quad word atomic operations
2300 (load_locked<mode>): Likewise.
2301 (store_conditional<mode>): Likewise.
2302 (atomic_compare_and_swap<mode>): Likewise.
2303 (atomic_exchange<mode>): Likewise.
2304 (atomic_nand<mode>): Likewise.
2305 (atomic_fetch_<fetchop_name><mode>): Likewise.
2306 (atomic_nand_fetch<mode>): Likewise.
2307 (mem_thread_fence): Use gen_loadsync_<mode> instead of enumerating
2309 (ATOMIC): On power8, add QImode, HImode modes.
2310 (load_locked<QHI:mode>_si): Varients of load_locked for QI/HI
2311 modes that promote to SImode.
2312 (load_lockedti): Convert TImode arguments to PTImode, so that we
2313 get a guaranteed even/odd register pair.
2314 (load_lockedpti): Likewise.
2315 (store_conditionalti): Likewise.
2316 (store_conditionalpti): Likewise.
2318 * config/rs6000/rs6000.md (QHI): New mode iterator for power8
2319 atomic load/store instructions.
2322 Back port from the trunk
2324 2013-06-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2325 Pat Haugen <pthaugen@us.ibm.com>
2326 Peter Bergner <bergner@vnet.ibm.com>
2328 * config/rs6000/vector.md (GPR move splitter): Do not split moves
2329 of vectors in GPRS if they are direct moves or quad word load or
2332 * config/rs6000/rs6000-protos.h (rs6000_output_move_128bit): Add
2334 (direct_move_p): Likewise.
2335 (quad_load_store_p): Likewise.
2337 * config/rs6000/rs6000.c (enum rs6000_reg_type): Simplify register
2338 classes into bins based on the physical register type.
2339 (reg_class_to_reg_type): Likewise.
2340 (IS_STD_REG_TYPE): Likewise.
2341 (IS_FP_VECT_REG_TYPE): Likewise.
2342 (reload_fpr_gpr): Arrays to determine what insn to use if we can
2343 use direct move instructions.
2344 (reload_gpr_vsx): Likewise.
2345 (reload_vsx_gpr): Likewise.
2346 (rs6000_init_hard_regno_mode_ok): Precalculate the register type
2347 information that is a simplification of register classes. Also
2348 precalculate direct move reload helpers.
2349 (direct_move_p): New function to return true if the operation can
2350 be done as a direct move instruciton.
2351 (quad_load_store_p): New function to return true if the operation
2352 is a quad memory operation.
2353 (rs6000_legitimize_address): If quad memory, only allow register
2354 indirect for TImode addresses.
2355 (rs6000_legitimate_address_p): Likewise.
2356 (enum reload_reg_type): Delete, replace with rs6000_reg_type.
2357 (rs6000_reload_register_type): Likewise.
2358 (register_to_reg_type): Return register type.
2359 (rs6000_secondary_reload_simple_move): New helper function for
2360 secondary reload and secondary memory needed to identify anything
2361 that is a simple move, and does not need reloading.
2362 (rs6000_secondary_reload_direct_move): New helper function for
2363 secondary reload to identify cases that can be done with several
2364 instructions via the direct move instructions.
2365 (rs6000_secondary_reload_move): New helper function for secondary
2366 reload to identify moves between register types that can be done.
2367 (rs6000_secondary_reload): Add support for quad memory operations
2368 and for direct move.
2369 (rs6000_secondary_memory_needed): Likewise.
2370 (rs6000_debug_secondary_memory_needed): Change argument names.
2371 (rs6000_output_move_128bit): New function to return the move to
2372 use for 128-bit moves, including knowing about the various
2373 limitations of quad memory operations.
2375 * config/rs6000/vsx.md (vsx_mov<mode>): Add support for quad
2376 memory operations. call rs6000_output_move_128bit for the actual
2377 instruciton(s) to generate.
2378 (vsx_movti_64bit): Likewise.
2380 * config/rs6000/rs6000.md (UNSPEC_P8V_FMRGOW): New unspec values.
2381 (UNSPEC_P8V_MTVSRWZ): Likewise.
2382 (UNSPEC_P8V_RELOAD_FROM_GPR): Likewise.
2383 (UNSPEC_P8V_MTVSRD): Likewise.
2384 (UNSPEC_P8V_XXPERMDI): Likewise.
2385 (UNSPEC_P8V_RELOAD_FROM_VSX): Likewise.
2386 (UNSPEC_FUSION_GPR): Likewise.
2387 (FMOVE128_GPR): New iterator for direct move.
2388 (f32_lv): New mode attribute for load/store of SFmode/SDmode
2392 (zero_extend<mode>di2_internal1): Add support for power8 32-bit
2393 loads and direct move instructions.
2394 (zero_extendsidi2_lfiwzx): Likewise.
2395 (extendsidi2_lfiwax): Likewise.
2396 (extendsidi2_nocell): Likewise.
2397 (floatsi<mode>2_lfiwax): Likewise.
2399 (floatunssi<mode>2_lfiwzx): Likewise.
2401 (fix_trunc<mode>_stfiwx): Likewise.
2402 (fixuns_trunc<mode>_stfiwx): Likewise.
2403 (mov<mode>_hardfloat, 32-bit floating point): Likewise.
2404 (mov<move>_hardfloat64, 64-bit floating point): Likewise.
2405 (parity<mode>2_cmpb): Set length/type attr.
2406 (unnamed shift right patterns, mov<mode>_internal2): Change type attr
2407 for 'mr.' to fast_compare.
2408 (bpermd_<mode>): Change type attr to popcnt.
2409 (p8_fmrgow_<mode>): New insns for power8 direct move support.
2410 (p8_mtvsrwz_1): Likewise.
2411 (p8_mtvsrwz_2): Likewise.
2412 (reload_fpr_from_gpr<mode>): Likewise.
2413 (p8_mtvsrd_1): Likewise.
2414 (p8_mtvsrd_2): Likewise.
2415 (p8_xxpermdi_<mode>): Likewise.
2416 (reload_vsx_from_gpr<mode>): Likewise.
2417 (reload_vsx_from_gprsf): Likewise.
2418 (p8_mfvsrd_3_<mode>): LIkewise.
2419 (reload_gpr_from_vsx<mode>): Likewise.
2420 (reload_gpr_from_vsxsf): Likewise.
2421 (p8_mfvsrd_4_disf): Likewise.
2422 (multi-word GPR splits): Do not split direct moves or quad memory
2425 Backport from the trunk
2427 2013-06-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2428 Pat Haugen <pthaugen@us.ibm.com>
2429 Peter Bergner <bergner@vnet.ibm.com>
2431 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
2432 Document new power8 builtins.
2434 * config/rs6000/vector.md (and<mode>3): Add a clobber/scratch of a
2435 condition code register, to allow 128-bit logical operations to be
2436 done in the VSX or GPR registers.
2437 (nor<mode>3): Use the canonical form for nor.
2438 (eqv<mode>3): Add expanders for power8 xxleqv, xxlnand, xxlorc,
2439 vclz*, and vpopcnt* vector instructions.
2440 (nand<mode>3): Likewise.
2441 (orc<mode>3): Likewise.
2442 (clz<mode>2): LIkewise.
2443 (popcount<mode>2): Likewise.
2445 * config/rs6000/predicates.md (int_reg_operand): Rework tests so
2446 that only the GPRs are recognized.
2448 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2449 support for new power8 builtins.
2451 * config/rs6000/rs6000-builtin.def (xscvspdpn): Add new power8
2453 (xscvdpspn): Likewise.
2459 (vpopcnt): Likewise.
2460 (vpopcntb): Likewise.
2461 (vpopcnth): Likewise.
2462 (vpopcntw): Likewise.
2463 (vpopcntd): Likewise.
2468 (eqv_v16qi3): Likewise.
2469 (eqv_v8hi3): Likewise.
2470 (eqv_v4si3): Likewise.
2471 (eqv_v2di3): Likewise.
2472 (eqv_v4sf3): Likewise.
2473 (eqv_v2df3): Likewise.
2475 (nand_v16qi3): Likewise.
2476 (nand_v8hi3): Likewise.
2477 (nand_v4si3): Likewise.
2478 (nand_v2di3): Likewise.
2479 (nand_v4sf3): Likewise.
2480 (nand_v2df3): Likewise.
2482 (orc_v16qi3): Likewise.
2483 (orc_v8hi3): Likewise.
2484 (orc_v4si3): Likewise.
2485 (orc_v2di3): Likewise.
2486 (orc_v4sf3): Likewise.
2487 (orc_v2df3): Likewise.
2489 * config/rs6000/rs6000.c (rs6000_option_override_internal): Only
2490 allow power8 quad mode in 64-bit.
2491 (rs6000_builtin_vectorized_function): Add support to vectorize
2492 ISA 2.07 count leading zeros, population count builtins.
2493 (rs6000_expand_vector_init): On ISA 2.07 use xscvdpspn to form
2494 V4SF vectors instead of xscvdpsp to avoid IEEE related traps.
2495 (builtin_function_type): Add vgbbd builtin function which takes an
2497 (altivec_expand_vec_perm_const): Add support for new power8 merge
2500 * config/rs6000/vsx.md (VSX_L2): New iterator for 128-bit types,
2501 that does not include TImdoe for use with 32-bit.
2502 (UNSPEC_VSX_CVSPDPN): Support for power8 xscvdpspn and xscvspdpn
2504 (UNSPEC_VSX_CVDPSPN): Likewise.
2505 (vsx_xscvdpspn): Likewise.
2506 (vsx_xscvspdpn): Likewise.
2507 (vsx_xscvdpspn_scalar): Likewise.
2508 (vsx_xscvspdpn_directmove): Likewise.
2509 (vsx_and<mode>3): Split logical operations into 32-bit and
2510 64-bit. Add support to do logical operations on TImode as well as
2511 VSX vector types. Allow logical operations to be done in either
2512 VSX registers or in general purpose registers in 64-bit mode. Add
2513 splitters if GPRs were used. For AND, add clobber of CCmode to
2514 allow use of ANDI on GPRs. Rewrite nor to use the canonical RTL
2516 (vsx_and<mode>3_32bit): Likewise.
2517 (vsx_and<mode>3_64bit): Likewise.
2518 (vsx_ior<mode>3): Likewise.
2519 (vsx_ior<mode>3_32bit): Likewise.
2520 (vsx_ior<mode>3_64bit): Likewise.
2521 (vsx_xor<mode>3): Likewise.
2522 (vsx_xor<mode>3_32bit): Likewise.
2523 (vsx_xor<mode>3_64bit): Likewise.
2524 (vsx_one_cmpl<mode>2): Likewise.
2525 (vsx_one_cmpl<mode>2_32bit): Likewise.
2526 (vsx_one_cmpl<mode>2_64bit): Likewise.
2527 (vsx_nor<mode>3): Likewise.
2528 (vsx_nor<mode>3_32bit): Likewise.
2529 (vsx_nor<mode>3_64bit): Likewise.
2530 (vsx_andc<mode>3): Likewise.
2531 (vsx_andc<mode>3_32bit): Likewise.
2532 (vsx_andc<mode>3_64bit): Likewise.
2533 (vsx_eqv<mode>3_32bit): Add support for power8 xxleqv, xxlnand,
2534 and xxlorc instructions.
2535 (vsx_eqv<mode>3_64bit): Likewise.
2536 (vsx_nand<mode>3_32bit): Likewise.
2537 (vsx_nand<mode>3_64bit): Likewise.
2538 (vsx_orc<mode>3_32bit): Likewise.
2539 (vsx_orc<mode>3_64bit): Likewise.
2541 * config/rs6000/rs6000.h (VLOGICAL_REGNO_P): Update comment.
2543 * config/rs6000/altivec.md (UNSPEC_VGBBD): Add power8 vgbbd
2545 (p8_vmrgew): Add power8 vmrgew and vmrgow instructions.
2546 (p8_vmrgow): Likewise.
2547 (altivec_and<mode>3): Add clobber of CCmode to allow AND using
2548 GPRs to be split under VSX.
2549 (p8v_clz<mode>2): Add power8 count leading zero support.
2550 (p8v_popcount<mode>2): Add power8 population count support.
2551 (p8v_vgbbd): Add power8 gather bits by bytes by doubleword
2554 * config/rs6000/rs6000.md (eqv<mode>3): Add support for powerp eqv
2557 * config/rs6000/altivec.h (vec_eqv): Add defines to export power8
2559 (vec_nand): Likewise.
2560 (vec_vclz): Likewise.
2561 (vec_vclzb): Likewise.
2562 (vec_vclzd): Likewise.
2563 (vec_vclzh): Likewise.
2564 (vec_vclzw): Likewise.
2565 (vec_vgbbd): Likewise.
2566 (vec_vmrgew): Likewise.
2567 (vec_vmrgow): Likewise.
2568 (vec_vpopcnt): Likewise.
2569 (vec_vpopcntb): Likewise.
2570 (vec_vpopcntd): Likewise.
2571 (vec_vpopcnth): Likewise.
2572 (vec_vpopcntw): Likewise.
2576 2013-05-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2577 Pat Haugen <pthaugen@us.ibm.com>
2578 Peter Bergner <bergner@vnet.ibm.com>
2580 * config/rs6000/vector.md (VEC_I): Add support for new power8 V2DI
2584 (vrotl<mode>3): Likewise.
2585 (vashl<mode>3): Likewise.
2586 (vlshr<mode>3): Likewise.
2587 (vashr<mode>3): Likewise.
2589 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2590 support for power8 V2DI builtins.
2592 * config/rs6000/rs6000-builtin.def (abs_v2di): Add support for
2593 power8 V2DI builtins.
2594 (vupkhsw): Likewise.
2595 (vupklsw): Likewise.
2596 (vaddudm): Likewise.
2601 (vpkudum): Likewise.
2602 (vpksdss): Likewise.
2603 (vpkudus): Likewise.
2604 (vpksdus): Likewise.
2609 (vsubudm): Likewise.
2610 (vcmpequd): Likewise.
2611 (vcmpgtsd): Likewise.
2612 (vcmpgtud): Likewise.
2613 (vcmpequd_p): Likewise.
2614 (vcmpgtsd_p): Likewise.
2615 (vcmpgtud_p): Likewise.
2616 (vupkhsw): Likewise.
2617 (vupklsw): Likewise.
2618 (vaddudm): Likewise.
2623 (vpksdss): Likewise.
2624 (vpksdus): Likewise.
2625 (vpkudum): Likewise.
2626 (vpkudus): Likewise.
2631 (vsubudm): Likewise.
2633 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Add
2634 support for power8 V2DI instructions.
2636 * config/rs6000/altivec.md (UNSPEC_VPKUHUM): Add support for
2637 power8 V2DI instructions. Combine pack and unpack insns to use an
2638 iterator for each mode. Check whether a particular mode supports
2639 Altivec instructions instead of just checking TARGET_ALTIVEC.
2640 (UNSPEC_VPKUWUM): Likewise.
2641 (UNSPEC_VPKSHSS): Likewise.
2642 (UNSPEC_VPKSWSS): Likewise.
2643 (UNSPEC_VPKUHUS): Likewise.
2644 (UNSPEC_VPKSHUS): Likewise.
2645 (UNSPEC_VPKUWUS): Likewise.
2646 (UNSPEC_VPKSWUS): Likewise.
2647 (UNSPEC_VPACK_SIGN_SIGN_SAT): Likewise.
2648 (UNSPEC_VPACK_SIGN_UNS_SAT): Likewise.
2649 (UNSPEC_VPACK_UNS_UNS_SAT): Likewise.
2650 (UNSPEC_VPACK_UNS_UNS_MOD): Likewise.
2651 (UNSPEC_VUPKHSB): Likewise.
2652 (UNSPEC_VUNPACK_HI_SIGN): Likewise.
2653 (UNSPEC_VUNPACK_LO_SIGN): Likewise.
2654 (UNSPEC_VUPKHSH): Likewise.
2655 (UNSPEC_VUPKLSB): Likewise.
2656 (UNSPEC_VUPKLSH): Likewise.
2658 (VI_char): Likewise.
2659 (VI_scalar): Likewise.
2660 (VI_unit): Likewise.
2662 (VP_small): Likewise.
2663 (VP_small_lc): Likewise.
2664 (VU_char): Likewise.
2665 (add<mode>3): Likewise.
2666 (altivec_vaddcuw): Likewise.
2667 (altivec_vaddu<VI_char>s): Likewise.
2668 (altivec_vadds<VI_char>s): Likewise.
2669 (sub<mode>3): Likewise.
2670 (altivec_vsubcuw): Likewise.
2671 (altivec_vsubu<VI_char>s): Likewise.
2672 (altivec_vsubs<VI_char>s): Likewise.
2673 (altivec_vavgs<VI_char>): Likewise.
2674 (altivec_vcmpbfp): Likewise.
2675 (altivec_eq<mode>): Likewise.
2676 (altivec_gt<mode>): Likewise.
2677 (altivec_gtu<mode>): Likewise.
2678 (umax<mode>3): Likewise.
2679 (smax<mode>3): Likewise.
2680 (umin<mode>3): Likewise.
2681 (smin<mode>3): Likewise.
2682 (altivec_vpkuhum): Likewise.
2683 (altivec_vpkuwum): Likewise.
2684 (altivec_vpkshss): Likewise.
2685 (altivec_vpkswss): Likewise.
2686 (altivec_vpkuhus): Likewise.
2687 (altivec_vpkshus): Likewise.
2688 (altivec_vpkuwus): Likewise.
2689 (altivec_vpkswus): Likewise.
2690 (altivec_vpks<VI_char>ss): Likewise.
2691 (altivec_vpks<VI_char>us): Likewise.
2692 (altivec_vpku<VI_char>us): Likewise.
2693 (altivec_vpku<VI_char>um): Likewise.
2694 (altivec_vrl<VI_char>): Likewise.
2695 (altivec_vsl<VI_char>): Likewise.
2696 (altivec_vsr<VI_char>): Likewise.
2697 (altivec_vsra<VI_char>): Likewise.
2698 (altivec_vsldoi_<mode>): Likewise.
2699 (altivec_vupkhsb): Likewise.
2700 (altivec_vupkhs<VU_char>): Likewise.
2701 (altivec_vupkls<VU_char>): Likewise.
2702 (altivec_vupkhsh): Likewise.
2703 (altivec_vupklsb): Likewise.
2704 (altivec_vupklsh): Likewise.
2705 (altivec_vcmpequ<VI_char>_p): Likewise.
2706 (altivec_vcmpgts<VI_char>_p): Likewise.
2707 (altivec_vcmpgtu<VI_char>_p): Likewise.
2708 (abs<mode>2): Likewise.
2709 (vec_unpacks_hi_v16qi): Likewise.
2710 (vec_unpacks_hi_v8hi): Likewise.
2711 (vec_unpacks_lo_v16qi): Likewise.
2712 (vec_unpacks_hi_<VP_small_lc>): Likewise.
2713 (vec_unpacks_lo_v8hi): Likewise.
2714 (vec_unpacks_lo_<VP_small_lc>): Likewise.
2715 (vec_pack_trunc_v8h): Likewise.
2716 (vec_pack_trunc_v4si): Likewise.
2717 (vec_pack_trunc_<mode>): Likewise.
2719 * config/rs6000/altivec.h (vec_vaddudm): Add defines for power8
2721 (vec_vmaxsd): Likewise.
2722 (vec_vmaxud): Likewise.
2723 (vec_vminsd): Likewise.
2724 (vec_vminud): Likewise.
2725 (vec_vpksdss): Likewise.
2726 (vec_vpksdus): Likewise.
2727 (vec_vpkudum): Likewise.
2728 (vec_vpkudus): Likewise.
2729 (vec_vrld): Likewise.
2730 (vec_vsld): Likewise.
2731 (vec_vsrad): Likewise.
2732 (vec_vsrd): Likewise.
2733 (vec_vsubudm): Likewise.
2734 (vec_vupkhsw): Likewise.
2735 (vec_vupklsw): Likewise.
2737 2013-05-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2738 Pat Haugen <pthaugen@us.ibm.com>
2739 Peter Bergner <bergner@vnet.ibm.com>
2741 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Add
2742 documentation for the power8 crypto builtins.
2744 * config/rs6000/t-rs6000 (MD_INCLUDES): Add crypto.md.
2746 * config/rs6000/rs6000-builtin.def (BU_P8V_AV_1): Add support
2747 macros for defining power8 builtin functions.
2748 (BU_P8V_AV_2): Likewise.
2749 (BU_P8V_AV_P): Likewise.
2750 (BU_P8V_VSX_1): Likewise.
2751 (BU_P8V_OVERLOAD_1): Likewise.
2752 (BU_P8V_OVERLOAD_2): Likewise.
2753 (BU_CRYPTO_1): Likewise.
2754 (BU_CRYPTO_2): Likewise.
2755 (BU_CRYPTO_3): Likewise.
2756 (BU_CRYPTO_OVERLOAD_1): Likewise.
2757 (BU_CRYPTO_OVERLOAD_2): Likewise.
2758 (XSCVSPDP): Fix typo, point to the correct instruction.
2759 (VCIPHER): Add power8 crypto builtins.
2760 (VCIPHERLAST): Likewise.
2761 (VNCIPHER): Likewise.
2762 (VNCIPHERLAST): Likewise.
2763 (VPMSUMB): Likewise.
2764 (VPMSUMH): Likewise.
2765 (VPMSUMW): Likewise.
2766 (VPERMXOR_V2DI): Likewise.
2767 (VPERMXOR_V4SI: Likewise.
2768 (VPERMXOR_V8HI: Likewise.
2769 (VPERMXOR_V16QI: Likewise.
2770 (VSHASIGMAW): Likewise.
2771 (VSHASIGMAD): Likewise.
2773 (VPERMXOR): Likewise.
2774 (VSHASIGMA): Likewise.
2776 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
2777 __CRYPTO__ if the crypto instructions are available.
2778 (altivec_overloaded_builtins): Add support for overloaded power8
2781 * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
2782 support for power8 crypto builtins.
2783 (builtin_function_type): Likewise.
2784 (altivec_init_builtins): Add support for builtins that take vector
2785 long long (V2DI) arguments.
2787 * config/rs6000/crypto.md: New file, define power8 crypto
2790 2013-05-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2791 Pat Haugen <pthaugen@us.ibm.com>
2792 Peter Bergner <bergner@vnet.ibm.com>
2794 * doc/invoke.texi (Option Summary): Add power8 options.
2795 (RS/6000 and PowerPC Options): Likewise.
2797 * doc/md.texi (PowerPC and IBM RS6000 constraints): Update to use
2798 constraints.md instead of rs6000.h. Reorder w* constraints. Add
2799 wm, wn, wr documentation.
2801 * gcc/config/rs6000/constraints.md (wm): New constraint for VSX
2802 registers if direct move instructions are enabled.
2803 (wn): New constraint for no registers.
2804 (wq): New constraint for quad word even GPR registers.
2805 (wr): New constraint if 64-bit instructions are enabled.
2806 (wv): New constraint if power8 vector instructions are enabled.
2807 (wQ): New constraint for quad word memory locations.
2809 * gcc/config/rs6000/predicates.md (const_0_to_15_operand): New
2810 constraint for 0..15 for crypto instructions.
2811 (gpc_reg_operand): If VSX allow registers in VSX registers as well
2812 as GPR and floating point registers.
2813 (int_reg_operand): New predicate to match only GPR registers.
2814 (base_reg_operand): New predicate to match base registers.
2815 (quad_int_reg_operand): New predicate to match even GPR registers
2816 for quad memory operations.
2817 (vsx_reg_or_cint_operand): New predicate to allow vector logical
2818 operations in both GPR and VSX registers.
2819 (quad_memory_operand): New predicate for quad memory operations.
2820 (reg_or_indexed_operand): New predicate for direct move support.
2822 * gcc/config/rs6000/rs6000-cpus.def (ISA_2_5_MASKS_EMBEDDED):
2823 Inherit from ISA_2_4_MASKS, not ISA_2_2_MASKS.
2824 (ISA_2_7_MASKS_SERVER): New mask for ISA 2.07 (i.e. power8).
2825 (POWERPC_MASKS): Add power8 options.
2826 (power8 cpu): Use ISA_2_7_MASKS_SERVER instead of specifying the
2829 * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
2830 Define _ARCH_PWR8 and __POWER8_VECTOR__ for power8.
2832 * gcc/config/rs6000/rs6000.opt (-mvsx-timode): Add documentation.
2833 (-mpower8-fusion): New power8 options.
2834 (-mpower8-fusion-sign): Likewise.
2835 (-mpower8-vector): Likewise.
2836 (-mcrypto): Likewise.
2837 (-mdirect-move): Likewise.
2838 (-mquad-memory): Likewise.
2840 * gcc/config/rs6000/rs6000.c (power8_cost): Initial definition for
2842 (rs6000_hard_regno_mode_ok): Make PTImode only match even GPR
2844 (rs6000_debug_reg_print): Print the base register class if
2846 (rs6000_debug_vector_unit): Add p8_vector.
2847 (rs6000_debug_reg_global): If -mdebug=reg, print power8 constraint
2848 definitions. Also print fusion state.
2849 (rs6000_init_hard_regno_mode_ok): Set up power8 constraints.
2850 (rs6000_builtin_mask_calculate): Add power8 builtin support.
2851 (rs6000_option_override_internal): Add support for power8.
2852 (rs6000_common_init_builtins): Add debugging for skipped builtins
2854 (rs6000_adjust_cost): Add power8 support.
2855 (rs6000_issue_rate): Likewise.
2856 (insn_must_be_first_in_group): Likewise.
2857 (insn_must_be_last_in_group): Likewise.
2858 (force_new_group): Likewise.
2859 (rs6000_register_move_cost): Likewise.
2860 (rs6000_opt_masks): Likewise.
2862 * config/rs6000/rs6000.h (ASM_CPU_POWER8_SPEC): If we don't have a
2863 power8 capable assembler, default to power7 options.
2864 (TARGET_DIRECT_MOVE): Likewise.
2865 (TARGET_CRYPTO): Likewise.
2866 (TARGET_P8_VECTOR): Likewise.
2867 (VECTOR_UNIT_P8_VECTOR_P): Define power8 vector support.
2868 (VECTOR_UNIT_VSX_OR_P8_VECTOR_P): Likewise.
2869 (VECTOR_MEM_P8_VECTOR_P): Likewise.
2870 (VECTOR_MEM_VSX_OR_P8_VECTOR_P): Likewise.
2871 (VECTOR_MEM_ALTIVEC_OR_VSX_P): Likewise.
2872 (TARGET_XSCVDPSPN): Likewise.
2873 (TARGET_XSCVSPDPN): Likewsie.
2874 (TARGET_SYNC_HI_QI): Likewise.
2875 (TARGET_SYNC_TI): Likewise.
2876 (MASK_CRYPTO): Likewise.
2877 (MASK_DIRECT_MOVE): Likewise.
2878 (MASK_P8_FUSION): Likewise.
2879 (MASK_P8_VECTOR): Likewise.
2880 (REG_ALLOC_ORDER): Move fr13 to be lower in priority so that the
2881 TFmode temporary used by some of the direct move instructions to
2882 get two FP temporary registers does not force creation of a stack
2884 (VLOGICAL_REGNO_P): Allow vector logical operations in GPRs.
2885 (MODES_TIEABLE_P): Move the VSX tests above the Altivec tests so
2886 that any VSX registers are tieable, even if they are also an
2887 Altivec vector mode.
2888 (r6000_reg_class_enum): Add wm, wr, wv constraints.
2889 (RS6000_BTM_P8_VECTOR): Power8 builtin support.
2890 (RS6000_BTM_CRYPTO): Likewise.
2891 (RS6000_BTM_COMMON): Likewise.
2893 * config/rs6000/rs6000.md (cpu attribute): Add power8.
2894 * config/rs6000/rs6000-opts.h (PROCESSOR_POWER8): Likewise.
2895 (enum rs6000_vector): Add power8 vector support.
2898 Backport from mainline
2899 2013-03-20 Pat Haugen <pthaugen@us.ibm.com>
2901 * config/rs6000/predicates.md (indexed_address, update_address_mem
2902 update_indexed_address_mem): New predicates.
2903 * config/rs6000/vsx.md (vsx_extract_<mode>_zero): Set correct "type"
2904 attribute for load/store instructions.
2905 * config/rs6000/dfp.md (movsd_store): Likewise.
2906 (movsd_load): Likewise.
2907 * config/rs6000/rs6000.md (zero_extend<mode>di2_internal1): Likewise.
2908 (unnamed HI->DI extend define_insn): Likewise.
2909 (unnamed SI->DI extend define_insn): Likewise.
2910 (unnamed QI->SI extend define_insn): Likewise.
2911 (unnamed QI->HI extend define_insn): Likewise.
2912 (unnamed HI->SI extend define_insn): Likewise.
2913 (unnamed HI->SI extend define_insn): Likewise.
2914 (extendsfdf2_fpr): Likewise.
2915 (movsi_internal1): Likewise.
2916 (movsi_internal1_single): Likewise.
2917 (movhi_internal): Likewise.
2918 (movqi_internal): Likewise.
2919 (movcc_internal1): Correct mnemonic for stw insn. Set correct "type"
2920 attribute for load/store instructions.
2921 (mov<mode>_hardfloat): Set correct "type" attribute for load/store
2923 (mov<mode>_softfloat): Likewise.
2924 (mov<mode>_hardfloat32): Likewise.
2925 (mov<mode>_hardfloat64): Likewise.
2926 (mov<mode>_softfloat64): Likewise.
2927 (movdi_internal32): Likewise.
2928 (movdi_internal64): Likewise.
2929 (probe_stack_<mode>): Likewise.
2931 Backport from mainline
2932 2013-03-20 Michael Meissner <meissner@linux.vnet.ibm.com>
2934 * config/rs6000/vector.md (VEC_R): Add 32-bit integer, binary
2935 floating point, and decimal floating point to reload iterator.
2937 * config/rs6000/constraints.md (wl constraint): New constraints to
2938 return FLOAT_REGS if certain options are used to reduce the number
2939 of separate patterns that exist in the file.
2940 (wx constraint): Likewise.
2941 (wz constraint): Likewise.
2943 * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
2944 -mdebug=reg, print wg, wl, wx, and wz constraints.
2945 (rs6000_init_hard_regno_mode_ok): Initialize new constraints.
2946 Initialize the reload functions for 64-bit binary/decimal floating
2948 (reg_offset_addressing_ok_p): If we are on a power7 or later, use
2949 LFIWZX and STFIWX to load/store 32-bit decimal types, and don't
2950 create the buffer on the stack to overcome not having a 32-bit
2952 (rs6000_emit_move): Likewise.
2953 (rs6000_secondary_memory_needed_rtx): Likewise.
2954 (rs6000_alloc_sdmode_stack_slot): Likewise.
2955 (rs6000_preferred_reload_class): On VSX, we can create SFmode 0.0f
2956 via xxlxor, just like DFmode 0.0.
2958 * config/rs6000/rs6000.h (TARGET_NO_SDMODE_STACK): New macro)
2959 (define as 1 if we are running on a power7 or newer.
2960 (enum r6000_reg_class_enum): Add new constraints.
2962 * config/rs6000/dfp.md (movsd): Delete, combine with binary
2963 floating point moves in rs6000.md. Combine power6x (mfpgpr) moves
2964 with other moves by using conditional constraits (wg). Use LFIWZX
2965 and STFIWX for loading SDmode on power7. Use xxlxor to create
2967 (movsd splitter): Likewise.
2968 (movsd_hardfloat): Likewise.
2969 (movsd_softfloat): Likewise.
2971 * config/rs6000/rs6000.md (FMOVE32): New iterators to combine
2972 binary and decimal floating point moves.
2973 (fmove_ok): New attributes to combine binary and decimal floating
2974 point moves, and to combine power6x (mfpgpr) moves along normal
2976 (real_value_to_target): Likewise.
2983 (movsf): Combine binary and decimal floating point moves. Combine
2984 power6x (mfpgpr) moves with other moves by using conditional
2985 constraits (wg). Use LFIWZX and STFIWX for loading SDmode on
2987 (mov<mode> for SFmode/SDmode); Likewise.
2988 (SFmode/SDmode splitters): Likewise.
2989 (movsf_hardfloat): Likewise.
2990 (mov<mode>_hardfloat for SFmode/SDmode): Likewise.
2991 (movsf_softfloat): Likewise.
2992 (mov<mode>_softfloat for SFmode/SDmode): Likewise.
2994 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wl)
2995 (wx and wz constraints.
2997 * config/rs6000/constraints.md (wg constraint): New constraint to
2998 return FLOAT_REGS if -mmfpgpr (power6x) was used.
3000 * config/rs6000/rs6000.h (enum r6000_reg_class_enum): Add wg
3003 * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
3004 -mdebug=reg, print wg, wl, wx, and wz constraints.
3005 (rs6000_init_hard_regno_mode_ok): Initialize new constraints.
3006 Initialize the reload functions for 64-bit binary/decimal floating
3008 (reg_offset_addressing_ok_p): If we are on a power7 or later, use
3009 LFIWZX and STFIWX to load/store 32-bit decimal types, and don't
3010 create the buffer on the stack to overcome not having a 32-bit
3012 (rs6000_emit_move): Likewise.
3013 (rs6000_secondary_memory_needed_rtx): Likewise.
3014 (rs6000_alloc_sdmode_stack_slot): Likewise.
3015 (rs6000_preferred_reload_class): On VSX, we can create SFmode 0.0f
3016 via xxlxor, just like DFmode 0.0.
3019 * config/rs6000/dfp.md (movdd): Delete, combine with binary
3020 floating point moves in rs6000.md. Combine power6x (mfpgpr) moves
3021 with other moves by using conditional constraits (wg). Use LFIWZX
3022 and STFIWX for loading SDmode on power7.
3023 (movdd splitters): Likewise.
3024 (movdd_hardfloat32): Likewise.
3025 (movdd_softfloat32): Likewise.
3026 (movdd_hardfloat64_mfpgpr): Likewise.
3027 (movdd_hardfloat64): Likewise.
3028 (movdd_softfloat64): Likewise.
3030 * config/rs6000/rs6000.md (FMOVE64): New iterators to combine
3031 64-bit binary and decimal floating point moves.
3032 (FMOVE64X): Likewise.
3033 (movdf): Combine 64-bit binary and decimal floating point moves.
3034 Combine power6x (mfpgpr) moves with other moves by using
3035 conditional constraits (wg).
3036 (mov<mode> for DFmode/DDmode): Likewise.
3037 (DFmode/DDmode splitters): Likewise.
3038 (movdf_hardfloat32): Likewise.
3039 (mov<mode>_hardfloat32 for DFmode/DDmode): Likewise.
3040 (movdf_softfloat32): Likewise.
3041 (movdf_hardfloat64_mfpgpr): Likewise.
3042 (movdf_hardfloat64): Likewise.
3043 (mov<mode>_hardfloat64 for DFmode/DDmode): Likewise.
3044 (movdf_softfloat64): Likewise.
3045 (mov<mode>_softfloat64 for DFmode/DDmode): Likewise.
3046 (reload_<mode>_load): Move to later in the file so they aren't in
3047 the middle of the floating point move insns.
3048 (reload_<mode>_store): Likewise.
3050 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wg
3053 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print out wg
3054 constraint if -mdebug=reg.
3055 (rs6000_initi_hard_regno_mode_ok): Enable wg constraint if
3056 -mfpgpr. Enable using dd reload support if needed.
3058 * config/rs6000/dfp.md (movtd): Delete, combine with 128-bit
3059 binary and decimal floating point moves in rs6000.md.
3060 (movtd_internal): Likewise.
3062 * config/rs6000/rs6000.md (FMOVE128): Combine 128-bit binary and
3063 decimal floating point moves.
3065 (movtf_internal): Likewise.
3066 (mov<mode>_internal, TDmode/TFmode): Likewise.
3067 (movtf_softfloat): Likewise.
3068 (mov<mode>_softfloat, TDmode/TFmode): Likewise.
3070 * config/rs6000/rs6000.md (movdi_mfpgpr): Delete, combine with
3071 movdi_internal64, using wg constraint for move direct operations.
3072 (movdi_internal64): Likewise.
3074 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print
3075 MODES_TIEABLE_P for selected modes. Print the numerical value of
3076 the various virtual registers. Use GPR/FPR first/last values)
3077 (instead of hard coding the register numbers. Print which modes
3078 have reload functions registered.
3079 (rs6000_option_override_internal): If -mdebug=reg, trace the
3080 options settings before/after setting cpu, target and subtarget
3082 (rs6000_secondary_reload_trace): Improve the RTL dump for
3083 -mdebug=addr and for secondary reload failures in
3084 rs6000_secondary_reload_inner.
3085 (rs6000_secondary_reload_fail): Likewise.
3086 (rs6000_secondary_reload_inner): Likewise.
3088 * config/rs6000/rs6000.md (FIRST_GPR_REGNO): Add convenience
3089 macros for first/last GPR and FPR registers.
3090 (LAST_GPR_REGNO): Likewise.
3091 (FIRST_FPR_REGNO): Likewise.
3092 (LAST_FPR_REGNO): Likewise.
3094 * config/rs6000/vector.md (mul<mode>3): Use the combined macro
3095 VECTOR_UNIT_ALTIVEC_OR_VSX_P instead of separate calls to
3096 VECTOR_UNIT_ALTIVEC_P and VECTOR_UNIT_VSX_P.
3097 (vcond<mode><mode>): Likewise.
3098 (vcondu<mode><mode>): Likewise.
3099 (vector_gtu<mode>): Likewise.
3100 (vector_gte<mode>): Likewise.
3101 (xor<mode>3): Don't allow logical operations on TImode in 32-bit
3102 to prevent the compiler from converting DImode operations to
3104 (ior<mode>3): Likewise.
3105 (and<mode>3): Likewise.
3106 (one_cmpl<mode>2): Likewise.
3107 (nor<mode>3): Likewise.
3108 (andc<mode>3): Likewise.
3110 * config/rs6000/constraints.md (wt constraint): New constraint
3111 that returns VSX_REGS if TImode is allowed in VSX registers.
3113 * config/rs6000/predicates.md (easy_fp_constant): 0.0f is an easy
3116 * config/rs6000/rs6000-modes.def (PTImode): Define, PTImode is
3117 similar to TImode, but it is restricted to being in the GPRs.
3119 * config/rs6000/rs6000.opt (-mvsx-timode): New switch to allow
3120 TImode to occupy a single VSX register.
3122 * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Default to
3123 -mvsx-timode for power7/power8.
3124 (power7 cpu): Likewise.
3125 (power8 cpu): Likewise.
3127 * config/rs6000/rs6000.c (rs6000_hard_regno_nregs_internal): Make
3128 sure that TFmode/TDmode take up two registers if they are ever
3129 allowed in the upper VSX registers.
3130 (rs6000_hard_regno_mode_ok): If -mvsx-timode, allow TImode in VSX
3132 (rs6000_init_hard_regno_mode_ok): Likewise.
3133 (rs6000_debug_reg_global): Add debugging for PTImode and wt
3134 constraint. Print if LRA is turned on.
3135 (rs6000_option_override_internal): Give an error if -mvsx-timode
3136 and VSX is not enabled.
3137 (invalid_e500_subreg): Handle PTImode, restricting it to GPRs. If
3138 -mvsx-timode, restrict TImode to reg+reg addressing, and PTImode
3139 to reg+offset addressing. Use PTImode when checking offset
3140 addresses for validity.
3141 (reg_offset_addressing_ok_p): Likewise.
3142 (rs6000_legitimate_offset_address_p): Likewise.
3143 (rs6000_legitimize_address): Likewise.
3144 (rs6000_legitimize_reload_address): Likewise.
3145 (rs6000_legitimate_address_p): Likewise.
3146 (rs6000_eliminate_indexed_memrefs): Likewise.
3147 (rs6000_emit_move): Likewise.
3148 (rs6000_secondary_reload): Likewise.
3149 (rs6000_secondary_reload_inner): Handle PTImode. Allow 64-bit
3150 reloads to fpr registers to continue to use reg+offset addressing)
3151 (but 64-bit reloads to altivec registers need reg+reg addressing.
3152 Drop test for PRE_MODIFY, since VSX loads/stores no longer support
3153 it. Treat LO_SUM like a PLUS operation.
3154 (rs6000_secondary_reload_class): If type is 64-bit, prefer to use
3155 FLOAT_REGS instead of VSX_RGS to allow use of reg+offset
3157 (rs6000_cannot_change_mode_class): Do not allow TImode in VSX
3158 registers to share a register with a smaller sized type, since VSX
3159 puts scalars in the upper 64-bits.
3160 (print_operand): Add support for PTImode.
3161 (rs6000_register_move_cost): Use VECTOR_MEM_VSX_P instead of
3162 VECTOR_UNIT_VSX_P to catch types that can be loaded in VSX
3163 registers, but don't have arithmetic support.
3164 (rs6000_memory_move_cost): Add test for VSX.
3165 (rs6000_opt_masks): Add -mvsx-timode.
3167 * config/rs6000/vsx.md (VSm): Change to use 64-bit aligned moves
3170 (VSr): Use wt constraint for TImode.
3171 (VSv): Drop TImode support.
3172 (vsx_movti): Delete, replace with versions for 32-bit and 64-bit.
3173 (vsx_movti_64bit): Likewise.
3174 (vsx_movti_32bit): Likewise.
3175 (vec_store_<mode>): Use VSX iterator instead of vector iterator.
3176 (vsx_and<mode>3): Delete use of '?' constraint on inputs, just put
3177 one '?' on the appropriate output constraint. Do not allow TImode
3178 logical operations on 32-bit systems.
3179 (vsx_ior<mode>3): Likewise.
3180 (vsx_xor<mode>3): Likewise.
3181 (vsx_one_cmpl<mode>2): Likewise.
3182 (vsx_nor<mode>3): Likewise.
3183 (vsx_andc<mode>3): Likewise.
3184 (vsx_concat_<mode>): Likewise.
3185 (vsx_xxpermdi_<mode>): Fix thinko for non V2DF/V2DI modes.
3187 * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Map from
3188 OPTION_MASK_VSX_TIMODE.
3189 (enum rs6000_reg_class_enum): Add RS6000_CONSTRAINT_wt.
3190 (STACK_SAVEAREA_MODE): Use PTImode instead of TImode.
3192 * config/rs6000/rs6000.md (INT mode attribute): Add PTImode.
3193 (TI2 iterator): New iterator for TImode, PTImode.
3194 (wd mode attribute): Add values for vector types.
3195 (movti_string): Replace TI move operations with operations for
3196 TImode and PTImode. Add support for TImode being allowed in VSX
3198 (mov<mode>_string, TImode/PTImode): Likewise.
3199 (movti_ppc64): Likewise.
3200 (mov<mode>_ppc64, TImode/PTImode): Likewise.
3201 (TI mode splitters): Likewise.
3203 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wt
3206 2014-04-04 Richard Biener <rguenther@suse.de>
3208 * tree-ssanames.c (make_ssa_name_fn): Fix assert.
3210 2014-04-02 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3212 * config/s390/s390.c (s390_expand_insv): Use GET_MODE_BITSIZE.
3214 2014-04-01 Richard Biener <rguenther@suse.de>
3216 * gimple.h (struct gimple_statement_base): Align subcode to
3219 2014-04-01 Sebastian Huber <sebastian.huber@embedded-brains.de>
3221 * doc/invoke.texi (mapp-regs): Clarify.
3223 2014-03-31 H.J. Lu <hongjiu.lu@intel.com>
3225 PR rtl-optimization/60700
3226 Backport from mainline
3227 2013-07-30 Zhenqiang Chen <zhenqiang.chen@linaro.org>
3229 PR rtl-optimization/57637
3230 * function.c (move_insn_for_shrink_wrap): Also check the
3231 GEN set of the LIVE problem for the liveness analysis
3232 if it exists, otherwise give up.
3234 2014-03-30 Kaz Kojima <kkojima@gcc.gnu.org>
3236 Backport from mainline
3237 2014-03-19 Kaz Kojima <kkojima@gcc.gnu.org>
3240 * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
3242 2014-03-26 Martin Jambor <mjambor@suse.cz>
3245 * ipa.c (symtab_remove_unreachable_nodes): Clear thunk and
3246 alias flags of nodes in the border.
3248 2014-03-26 Eric Botcazou <ebotcazou@adacore.com>
3250 PR rtl-optimization/60452
3251 * rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
3252 <case REG>: Return 1 for invalid offsets from the frame pointer.
3254 2014-03-24 Richard Biener <rguenther@suse.de>
3256 PR tree-optimization/60429
3257 * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Remove
3260 2014-03-23 Eric Botcazou <ebotcazou@adacore.com>
3262 PR rtl-optimization/60601
3263 * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
3265 * gcc.c (eval_spec_function): Initialize save_growing_value.
3267 2014-03-20 Jakub Jelinek <jakub@redhat.com>
3270 * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
3271 into CONST, put pic register as first operand of PLUS. Use
3272 gen_const_mem for both 32-bit and 64-bit PIC got loads.
3274 2014-03-20 Eric Botcazou <ebotcazou@adacore.com>
3276 * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
3277 around for store forwarding issue in the FPU on the UT699.
3278 * config/sparc/sparc.md (in_branch_delay): Return false for single FP
3279 loads and operations if -mfix-ut699 is specified.
3280 (divtf3_hq): Tweak attribute.
3281 (sqrttf2_hq): Likewise.
3283 2014-03-18 Kai Tietz <ktietz@redhat.com>
3285 PR rtl-optimization/56356
3286 * sdbout.c (sdbout_parms): Verify that parms'
3287 incoming argument is valid.
3288 (sdbout_reg_parms): Likewise.
3290 2014-03-18 Eric Botcazou <ebotcazou@adacore.com>
3292 * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
3293 proper constant for the store mode.
3295 2014-03-17 Mikael Pettersson <mikpelinux@gmail.com>
3296 Committed by Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3298 Backport from mainline:
3300 2013-06-20 Joern Rennecke <joern.rennecke@embecosm.com>
3302 PR rtl-optimization/57425
3303 PR rtl-optimization/57569
3304 * alias.c (write_dependence_p): Remove parameters mem_mode and
3305 canon_mem_addr. Add parameters x_mode, x_addr and x_canonicalized.
3306 Changed all callers.
3307 (canon_anti_dependence): Get comments and semantics in sync.
3308 Add parameter mem_canonicalized. Changed all callers.
3309 * rtl.h (canon_anti_dependence): Update prototype.
3311 2013-06-16 Joern Rennecke <joern.rennecke@embecosm.com>
3313 PR rtl-optimization/57425
3314 PR rtl-optimization/57569
3315 * alias.c (write_dependence_p): Add new parameters mem_mode,
3316 canon_mem_addr and mem_canonicalized. Change type of writep to bool.
3317 Changed all callers.
3318 (canon_anti_dependence): New function.
3319 * cse.c (check_dependence): Use canon_anti_dependence.
3320 * cselib.c (cselib_invalidate_mem): Likewise.
3321 * rtl.h (canon_anti_dependence): Declare.
3323 2014-03-17 Richard Biener <rguenther@suse.de>
3325 Backport from mainline
3326 2014-03-11 Richard Biener <rguenther@suse.de>
3328 PR tree-optimization/60429
3329 PR tree-optimization/60485
3330 * tree-ssa-structalias.c (set_union_with_increment): Properly
3331 take into account all fields that overlap the shifted vars.
3332 (do_sd_constraint): Likewise.
3333 (do_ds_constraint): Likewise.
3334 (get_constraint_for_ptr_offset): Likewise.
3336 2014-03-15 Eric Botcazou <ebotcazou@adacore.com>
3338 * config/sparc/sparc-protos.h (tls_call_delay): Delete.
3339 (eligible_for_call_delay): New prototype.
3340 * config/sparc/sparc.c (tls_call_delay): Rename into...
3341 (eligible_for_call_delay): ...this. Return false if the instruction
3342 cannot be put in the delay slot of a branch.
3343 (eligible_for_restore_insn): Simplify.
3344 (eligible_for_return_delay): Return false if the instruction cannot be
3345 put in the delay slot of a branch and simplify.
3346 (eligible_for_sibcall_delay): Return false if the instruction cannot be
3347 put in the delay slot of a branch.
3348 * config/sparc/sparc.md (fix_ut699): New attribute.
3349 (tls_call_delay): Delete.
3350 (in_call_delay): Reimplement.
3351 (eligible_for_sibcall_delay): Rename into...
3352 (in_sibcall_delay): ...this.
3353 (eligible_for_return_delay): Rename into...
3354 (in_return_delay): ...this.
3355 (in_branch_delay): Reimplement.
3356 (in_uncond_branch_delay): Delete.
3357 (in_annul_branch_delay): Delete.
3359 2014-03-14 Georg-Johann Lay <avr@gjlay.de>
3361 Backport from 2014-03-14 trunk r208562.
3364 * config/avr/avr.c (avr_set_current_function): Pass function name
3365 through default_strip_name_encoding before sanity checking instead
3366 of skipping the first char of the assembler name.
3368 2014-03-13 Georg-Johann Lay <avr@gjlay.de>
3370 Backport from 2014-03-13 trunk r208532.
3373 * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
3374 calls of avr_out_plus_1.
3376 2014-03-13 Joey Ye <joey.ye@arm.com>
3378 Backport from mainline
3379 2014-03-12 Thomas Preud'homme <thomas.preudhomme@arm.com>
3381 PR tree-optimization/60454
3382 * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
3384 2014-03-06 Matthias Klose <doko@ubuntu.com>
3386 * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
3387 MULTILIB_OSDIRNAMES is not defined.
3389 2014-03-06 Jakub Jelinek <jakub@redhat.com>
3391 PR tree-optimization/60276
3392 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Avoid
3393 a -Wsign-compare warning.
3395 * Makefile.in (tree-ssa-uninit.o): Depend on $(PARAMS_H).
3397 Backport from mainline
3398 2014-02-21 Jakub Jelinek <jakub@redhat.com>
3400 PR tree-optimization/56490
3401 * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
3402 * tree-ssa-uninit.c: Include params.h.
3403 (compute_control_dep_chain): Add num_calls argument, return false
3404 if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
3405 num_calls to recursive call.
3406 (find_predicates): Change dep_chain into normal array, add num_calls
3407 variable and adjust compute_control_dep_chain caller.
3408 (find_def_preds): Likewise.
3410 2014-02-13 Jakub Jelinek <jakub@redhat.com>
3413 * expr.c (compress_float_constant): If x is a hard register,
3414 extend into a pseudo and then move to x.
3416 2014-02-11 Richard Henderson <rth@redhat.com>
3417 Jakub Jelinek <jakub@redhat.com>
3420 * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
3421 around drhs if type conversion to lacc->type is not useless.
3423 2014-02-08 Jakub Jelinek <jakub@redhat.com>
3426 * ipa-cp.c (determine_versionability): Fail at -O0
3427 or __attribute__((optimize (0))) or -fno-ipa-cp functions.
3428 * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
3430 2014-02-06 Jakub Jelinek <jakub@redhat.com>
3433 * tree.h (opts_for_fn): New inline function.
3434 (opt_for_fn): Define.
3435 * config/i386/i386.c (ix86_function_regparm): Use
3436 opt_for_fn (decl, optimize) instead of optimize.
3438 2014-02-05 Jakub Jelinek <jakub@redhat.com>
3441 * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
3442 bb with no successors.
3444 2014-03-04 Richard Biener <rguenther@suse.de>
3446 PR tree-optimization/60382
3447 * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
3448 dead PHIs a reduction.
3450 2014-02-25 Richard Biener <rguenther@suse.de>
3452 Backport from mainline
3453 2014-02-21 Richard Biener <rguenther@suse.de>
3455 PR tree-optimization/60276
3456 * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
3457 (STMT_VINFO_MIN_NEG_DIST): New macro.
3458 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
3459 STMT_VINFO_MIN_NEG_DIST.
3460 * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
3461 made for negative dependence distances still hold.
3463 2014-02-25 Richard Biener <rguenther@suse.de>
3465 Backport from mainline
3466 2014-02-21 Richard Biener <rguenther@suse.de>
3469 * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
3470 DECL_INITIAL for globals not in the current function context.
3472 2014-02-20 Richard Biener <rguenther@suse.de>
3475 * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
3478 2014-02-14 Richard Biener <rguenther@suse.de>
3480 PR tree-optimization/60183
3481 * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating
3483 (tree_ssa_phiprop): Calculate and free post-dominators.
3485 2014-02-25 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3488 * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
3491 2014-02-24 John David Anglin <danglin@gcc.gnu.org>
3493 * config/pa/pa.c (pa_output_move_double): Don't valididate when
3494 adjusting offsetable addresses.
3496 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
3498 * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg names
3500 2014-02-23 Edgar E. Iglesias <edgar.iglesias@xilinx.com>
3502 * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
3505 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
3507 * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk
3508 and define TARGET_ASM_OUTPUT_MI_THUNK and
3509 TARGET_ASM_CAN_OUTPUT_MI_THUNK.
3511 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
3513 * config/microblaze/predicates.md: Add cmp_op predicate.
3514 * config/microblaze/microblaze.md: Add branch_compare instruction
3515 which uses cmp_op predicate and emits cmp insn before branch.
3516 * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
3517 to microblaze_expand_conditional_branch and consolidate logic.
3518 (microblaze_expand_conditional_branch): emit branch_compare
3519 insn instead of handling cmp op separate from branch insn.
3521 2014-02-21 Martin Jambor <mjambor@suse.cz>
3524 * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
3525 info when checking whether lattices are bottom.
3527 2014-02-21 Jakub Jelinek <jakub@redhat.com>
3529 * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
3530 mode for mask of V8SFmode permutation.
3532 2014-02-20 Richard Henderson <rth@redhat.com>
3535 * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
3536 on failure the store back into EXPECT. Always make a new pseudo for
3539 2014-02-20 Jakub Jelinek <jakub@redhat.com>
3542 * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
3543 gen_reg_rtx if d->testing_p.
3544 (expand_vec_perm_pshufb2, expand_vec_perm_even_odd_1,
3545 expand_vec_perm_broadcast_1): Return early if d->testing_p and
3546 we will certainly return true.
3548 2014-02-20 Richard Biener <rguenther@suse.de>
3550 * tree-cfg.c (replace_uses_by): Mark altered BBs before
3551 doing the substitution.
3553 2014-02-19 H.J. Lu <hongjiu.lu@intel.com>
3555 Backport from mainline
3556 2014-02-19 H.J. Lu <hongjiu.lu@intel.com>
3559 * config/i386/i386.c (construct_container): Remove TFmode check
3560 for X86_64_INTEGER_CLASS.
3562 2014-02-19 Uros Bizjak <ubizjak@gmail.com>
3564 Backport from mainline
3565 2014-02-19 Uros Bizjak <ubizjak@gmail.com>
3568 * config/i386/i386.c (type_natural_mode): Warn for ABI changes
3569 only when -Wpsabi is enabled.
3571 2014-02-19 Terry Guo <terry.guo@arm.com>
3573 Backport from mainline
3574 2014-02-08 Terry Guo <terry.guo@arm.com>
3576 * doc/invoke.texi: Document ARM -march=armv7e-m.
3578 2014-02-18 Kai Tietz <ktietz@redhat.com>
3580 Backport from mainline
3581 2014-02-18 Kai Tietz <ktietz@redhat.com>
3584 * config/i386/i386.c (ix86_expand_prologue): Use
3585 rax register as displacement for restoring %r10, %rax.
3586 Additional fix wrong offset for restoring both-registers.
3588 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
3590 * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
3591 assertion with conditional return.
3593 2014-02-18 Jakub Jelinek <jakub@redhat.com>
3594 Uros Bizjak <ubizjak@gmail.com>
3597 * config/i386/driver-i386.c (host_detect_local_cpu): If
3598 YMM state is not saved by the OS, also clear has_f16c. Move
3599 CPUID 0x80000001 handling before YMM state saving checking.
3601 2014-02-14 Roland McGrath <mcgrathr@google.com>
3603 * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
3604 * configure: Regenerated.
3605 * config.in: Regenerated.
3606 * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
3607 instead of ASM_SHORT.
3609 2014-02-13 Uros Bizjak <ubizjak@gmail.com>
3611 Backport from mainline
3612 2014-02-13 Uros Bizjak <ubizjak@gmail.com>
3614 * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
3615 operands[2], not operands[3].
3617 2014-02-13 Dominik Vogt <vogt@linux.vnet.ibm.com>
3619 * config/s390/s390.c (s390_asm_output_function_label): Fix crash
3620 caused by bad second argument to warning_at() with -mhotpatch and
3621 nested functions (e.g. with gfortran).
3623 2014-02-12 H.J. Lu <hongjiu.lu@intel.com>
3625 Backport from mainline
3626 2014-02-12 H.J. Lu <hongjiu.lu@intel.com>
3627 Uros Bizjak <ubizjak@gmail.com>
3630 * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
3632 2014-02-12 Eric Botcazou <ebotcazou@adacore.com>
3634 PR rtl-optimization/60116
3635 * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
3636 other_insn once the combination has been validated.
3638 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
3640 * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
3641 * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
3643 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
3645 * config/microblaze/microblaze.c: Extend mcpu version format
3647 2014-02-10 David Holsgrove <david.holsgrove@xilinx.com>
3649 * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
3651 2014-02-10 Richard Biener <rguenther@suse.de>
3653 Backport from mainline
3654 2014-01-30 Richard Biener <rguenther@suse.de>
3656 PR tree-optimization/59903
3657 * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
3660 2014-02-10 Richard Biener <rguenther@suse.de>
3662 PR tree-optimization/60115
3663 * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
3664 MEM_REF handling. Properly verify that the accesses are not
3665 out of the objects bound.
3667 2014-02-05 James Greenhalgh <james.greenhalgh@arm.com>
3669 Backport from mainline.
3670 2014-02-05 James Greenhalgh <james.greenhalgh@arm.com>
3673 * doc/invoke.texi (-march): Clarify documentation for ARM.
3677 2014-02-04 John David Anglin <danglin@gcc.gnu.org>
3680 * config/pa/pa.c (legitimize_tls_address): Return original address
3681 if not passed a SYMBOL_REF rtx.
3682 (hppa_legitimize_address): Call legitimize_tls_address for all TLS
3684 (pa_emit_move_sequence): Simplify TLS source operands.
3685 (pa_legitimate_constant_p): Reject all TLS constants.
3686 * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
3687 (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
3689 2014-02-04 Uros Bizjak <ubizjak@gmail.com>
3691 Backport from mainline
3692 2014-02-02 Uros Bizjak <ubizjak@gmail.com>
3695 * config/i386/i386.c (classify_argument): Fix handling of bit_offset
3696 when calculating size of integer atomic types.
3698 2014-02-02 Uros Bizjak <ubizjak@gmail.com>
3700 Backport from mainline
3701 2014-01-30 Jakub Jelinek <jakub@redhat.com>
3703 * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
3705 2014-01-31 Richard Henderson <rth@redhat.com>
3708 * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
3709 until after else_eh is processed.
3711 2014-01-30 David Holsgrove <david.holsgrove@xilinx.com>
3713 Backport from mainline
3714 * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
3715 comparison_operator with ordered_comparison_operator.
3717 2014-01-25 Walter Lee <walt@tilera.com>
3719 Backport from mainline
3720 2014-01-25 Walter Lee <walt@tilera.com>
3722 * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
3723 avoid clobbering a live register.
3725 2014-01-25 Walter Lee <walt@tilera.com>
3727 Backport from mainline
3728 2014-01-25 Walter Lee <walt@tilera.com>
3730 * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
3731 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
3732 * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
3733 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
3735 2014-01-25 Walter Lee <walt@tilera.com>
3737 Backport from mainline
3738 2014-01-25 Walter Lee <walt@tilera.com>
3740 * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
3741 insns before bundling.
3742 * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
3744 2014-01-25 Walter Lee <walt@tilera.com>
3746 Backport from mainline
3747 2014-01-25 Walter Lee <walt@tilera.com>
3749 * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
3750 PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
3751 * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
3753 2014-01-25 Walter Lee <walt@tilera.com>
3755 Backport from mainline
3756 2014-01-25 Walter Lee <walt@tilera.com>
3758 * config/tilepro/tilepro.md (ctzdi2): Use register_operand
3763 2014-01-25 Walter Lee <walt@tilera.com>
3765 Backport from mainline
3766 2014-01-25 Walter Lee <walt@tilera.com>
3768 * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
3769 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3771 2014-01-24 H.J. Lu <hongjiu.lu@intel.com>
3773 Backport from mainline
3774 2014-01-23 H.J. Lu <hongjiu.lu@intel.com>
3777 * config/i386/i386.md (pushsf splitter): Get stack adjustment
3778 from push operand if code of push isn't PRE_DEC.
3780 2014-01-23 David Holsgrove <david.holsgrove@xilinx.com>
3782 Backport from mainline.
3783 * config/microblaze/microblaze.md: Add trap insn and attribute
3785 2014-01-23 Marek Polacek <polacek@redhat.com>
3787 Backport from mainline
3788 2013-10-21 Marek Polacek <polacek@redhat.com>
3791 * fold-const.c (fold_range_test): Return 0 if the type is not
3794 2014-01-22 David Holsgrove <david.holsgrove@xilinx.com>
3796 * config/microblaze/microblaze.md: Correct bswaphi2 insn.
3798 2014-01-22 Uros Bizjak <ubizjak@gmail.com>
3800 Backport from mainline
3801 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
3803 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3804 for SImode_address_operand operands, having only a REG argument.
3806 2014-01-20 Jakub Jelinek <jakub@redhat.com>
3809 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3810 if operands[1] is a REG or ZERO_EXTEND of a REG.
3812 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
3813 H.J. Lu <hongjiu.lu@intel.com>
3816 * config/i386/i386.md (*lea<mode>): Zero-extend return register
3817 to DImode for zero-extended addresses.
3819 2014-01-21 Andrew Pinski <apinski@cavium.com>
3820 Steve Ellcey <sellcey@mips.com>
3823 * config/mips/mips.c (mips_print_operand): Check operand mode instead
3826 2014-01-21 Andrey Belevantsev <abel@ispras.ru>
3828 Backport from mainline
3829 2013-12-23 Andrey Belevantsev <abel@ispras.ru>
3831 PR rtl-optimization/57422
3832 * sel-sched.c (mark_unavailable_hard_regs): Fix typo when calling
3833 add_to_hard_reg_set.
3835 2014-01-20 Jakub Jelinek <jakub@redhat.com>
3838 * tree.h (fold_builtin_strcat): New prototype.
3839 * builtins.c (fold_builtin_strcat): No longer static. Add len
3840 argument, if non-NULL, don't call c_strlen. Optimize
3841 directly into __builtin_memcpy instead of __builtin_strcpy.
3842 (fold_builtin_2): Adjust fold_builtin_strcat caller.
3843 * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
3845 2014-01-20 Richard Biener <rguenther@suse.de>
3848 * builtins.c (fold_builtin_strcat): Remove case better handled
3849 by tree-ssa-strlen.c.
3851 2014-01-19 John David Anglin <danglin@gcc.gnu.org>
3853 * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
3854 long non-pic millicode calls.
3856 2014-01-17 John David Anglin <danglin@gcc.gnu.org>
3858 * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
3859 call to $$dyncall when TARGET_LONG_CALLS is true.
3861 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
3863 Backport from mainline
3864 2014-01-14 H.J. Lu <hongjiu.lu@intel.com>
3867 * config/i386/i386.c (type_natural_mode): Add a bool parameter
3868 to indicate if type is used for function return value. Warn
3869 ABI change if the vector mode isn't available for function
3871 (ix86_function_arg_advance): Pass false to type_natural_mode.
3872 (ix86_function_arg): Likewise.
3873 (ix86_gimplify_va_arg): Likewise.
3874 (function_arg_32): Don't warn ABI change.
3875 (ix86_function_value): Pass true to type_natural_mode.
3876 (ix86_return_in_memory): Likewise.
3877 (ix86_struct_value_rtx): Removed.
3878 (TARGET_STRUCT_VALUE_RTX): Likewise.
3880 2014-01-17 Charles Baylis <charles.baylis@linaro.org>
3882 Backport from mainline
3883 2013-12-19 Charles Baylis <charles.baylis@linaro.org>
3886 * config/arm/arm-ldmstm.ml: Use low_register_operand for Thumb
3888 * config/arm/ldmstm.md: Regenerate.
3890 2013-12-19 Charles Baylis <charles.baylis@linaro.org>
3893 * config/arm/predicates.md (arm_hard_general_register_operand):
3895 (arm_hard_register_operand): Remove.
3896 * config/arm/arm-ldmstm.ml: Use arm_hard_general_register_operand
3898 * config/arm/ldmstm.md: Regenerate.
3900 2013-12-19 Charles Baylis <charles.baylis@linaro.org>
3903 * config/arm/predicates.md (vfp_hard_register_operand): New predicate.
3904 * config/arm/arm.md (vfp_pop_multiple_with_writeback): Use
3905 vfp_hard_register_operand.
3907 2014-01-17 Kugan Vivekanandarajah <kuganv@linaro.org>
3909 Backport from mainline
3910 2014-01-15 Matthew Gretton-Dann <matthew.gretton-dann@linaro.org>
3911 Kugan Vivekanandarajah <kuganv@linaro.org>
3914 * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
3917 2014-01-17 Terry Guo <terry.guo@arm.com>
3920 * config/arm/arm.md (prefetch): Set insn type attribute to load1.
3922 2014-01-16 Jakub Jelinek <jakub@redhat.com>
3925 * config/i386/i386.c (ix86_expand_builtin): If target doesn't
3926 satisfy operand 0 predicate for gathers, use a new pseudo as
3929 2014-01-16 Richard Henderson <rth@redhat.com>
3932 * reginfo.c (global_regs_decl): Globalize.
3933 * rtl.h (global_regs_decl): Declare.
3934 * ira.c (do_reload): Diagnose frame_pointer_needed and it
3935 reserved via global_regs.
3937 2014-01-16 Peter Bergner <bergner@vnet.ibm.com>
3939 Backport from mainline
3940 2014-01-15 Uros Bizjak <ubizjak@gmail.com>
3942 * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
3943 VALID_AVX256_REG_OR_OI_MODE.
3945 2013-09-05 Peter Bergner <bergner@vnet.ibm.com>
3948 * reginfo.c (choose_hard_reg_mode): Scan through all mode classes
3949 looking for widest mode.
3951 2014-01-16 Marek Polacek <polacek@redhat.com>
3953 Backported from mainline
3954 2014-01-16 Marek Polacek <polacek@redhat.com>
3957 * gimple-low.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
3958 it is error_mark_node.
3960 2014-01-15 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3963 * config/s390/s390.c (s390_preferred_reload_class): Don't return
3964 ADDR_REGS for invalid symrefs in non-PIC code.
3966 2014-01-14 Uros Bizjak <ubizjak@gmail.com>
3969 2014-01-08 Uros Bizjak <ubizjak@gmail.com>
3971 * config/i386/i386.c (ix86_data_alignment): Calculate max_align
3972 from prefetch_block tune setting.
3974 2014-01-13 Jakub Jelinek <jakub@redhat.com>
3976 Backported from mainline
3977 2014-01-10 Jakub Jelinek <jakub@redhat.com>
3979 PR tree-optimization/59745
3980 * tree-predcom.c (tree_predictive_commoning_loop): Call
3981 free_affine_expand_cache if giving up because components is NULL.
3983 2014-01-10 Yufeng Zhang <yufeng.zhang@arm.com>
3985 * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
3986 with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
3987 rtx is const0_rtx or not.
3989 2014-01-10 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3991 * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
3992 extraction in good case.
3994 2014-01-10 Huacai Chen <chenhc@lemote.com>
3996 * config/mips/driver-native.c (host_detect_local_cpu): Handle new
3997 kernel strings for Loongson-2E/2F/3A.
3999 2014-01-10 Richard Biener <rguenther@suse.de>
4001 PR tree-optimization/59715
4002 * tree-flow.h (split_critical_edges): Declare.
4003 * tree-cfg.c (split_critical_edges): Export.
4004 * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4006 2014-01-09 Richard Sandiford <rdsandiford@googlemail.com>
4008 * config/mips/mips.h (ISA_HAS_WSBH): Define.
4009 * config/mips/mips.md (UNSPEC_WSBH, UNSPEC_DSBH, UNSPEC_DSHD): New
4011 (bswaphi2, bswapsi2, bswapdi2, wsbh, dsbh, dshd): New patterns.
4013 2014-01-09 Richard Sandiford <rdsandiford@googlemail.com>
4015 PR rtl-optimization/59137
4016 * reorg.c (steal_delay_list_from_target): Call update_block for
4018 (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
4020 2014-01-09 Richard Sandiford <rdsandiford@googlemail.com>
4023 2012-10-07 Richard Sandiford <rdsandiford@googlemail.com>
4025 * config/mips/mips.c (mips_truncated_op_cost): New function.
4026 (mips_rtx_costs): Adjust test for BADDU.
4027 * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
4029 2012-10-02 Richard Sandiford <rdsandiford@googlemail.com>
4031 * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
4032 (*baddu_si): ...this new pattern.
4034 2014-01-09 Richard Biener <rguenther@suse.de>
4036 Backport from mainline
4037 2013-11-18 Richard Biener <rguenther@suse.de>
4039 PR tree-optimization/59125
4040 PR tree-optimization/54570
4041 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): When inlining
4042 is not complete do not treat component-references with offset zero
4043 but different fields as equal.
4044 * tree-object-size.c: Include tree-phinodes.h and ssa-iterators.h.
4045 (compute_object_sizes): Apply TLC. Propagate the constant
4046 results into all uses and fold their stmts.
4047 * passes.def (pass_all_optimizations): Move pass_object_sizes
4048 after the first pass_forwprop and before pass_fre.
4050 2013-12-03 Jakub Jelinek <jakub@redhat.com>
4052 PR tree-optimization/59362
4053 * tree-object-size.c (object_sizes): Change into array of
4054 vec<unsigned HOST_WIDE_INT>.
4055 (compute_builtin_object_size): Check computed bitmap for
4056 non-NULL instead of object_sizes. Call safe_grow on object_sizes
4057 vector if new SSA_NAMEs appeared.
4058 (init_object_sizes): Check computed bitmap for non-NULL.
4059 Call safe_grow on object_sizes elements instead of initializing
4061 (fini_object_sizes): Call release on object_sizes elements, don't
4064 2014-01-09 Richard Earnshaw <rearnsha@arm.com>
4066 PR rtl-optimization/54300
4067 * regcprop.c (copyprop_hardreg_forward_1): Ensure any unused
4068 outputs in a single-set are killed from the value chains.
4070 2014-01-09 Jakub Jelinek <jakub@redhat.com>
4072 PR rtl-optimization/59724
4073 * ifcvt.c (cond_exec_process_if_block): Don't call
4074 flow_find_head_matching_sequence with 0 longest_match.
4075 * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4076 non-active insns if !stop_after.
4077 (try_head_merge_bb): Revert 2014-01-07 changes.
4079 2014-01-09 Hans-Peter Nilsson <hp@axis.com>
4081 Backport from mainline
4082 2013-12-23 Hans-Peter Nilsson <hp@axis.com>
4085 * config/cris/predicates.md (cris_nonsp_register_operand):
4086 New define_predicate.
4087 * config/cris/cris.md: Replace register_operand with
4088 cris_nonsp_register_operand for destinations in all
4089 define_splits where a register is set more than once.
4091 2014-01-08 H.J. Lu <hongjiu.lu@intel.com>
4093 Backport from mainline
4094 2013-12-25 H.J. Lu <hongjiu.lu@intel.com>
4097 * config/i386/i386.c (struct ptt): Add a field for processor name.
4098 (processor_target_table): Sync with processor_type. Add
4100 (cpu_names): Removed.
4101 (ix86_option_override_internal): Default x_ix86_tune_string
4102 to processor_target_table[TARGET_CPU_DEFAULT].name.
4103 (ix86_function_specific_print): Assert arch and tune <
4104 PROCESSOR_max. Use processor_target_table to print arch and
4106 * config/i386/i386.h (TARGET_CPU_DEFAULT): Default to
4107 PROCESSOR_GENERIC32.
4108 (target_cpu_default): Removed.
4109 (processor_type): Reordered.
4111 2014-01-08 Uros Bizjak <ubizjak@gmail.com>
4113 Backport from mainline
4114 2014-01-05 Uros Bizjak <ubizjak@gmail.com>
4116 * config/i386/i386.c (ix86_data_alignment): Calculate max_align
4117 from prefetch_block tune setting.
4118 (nocona_cost): Correct size of prefetch block to 64.
4120 2014-01-08 Martin Jambor <mjambor@suse.cz>
4123 * ipa-prop.c (ipa_compute_jump_functions): Bail out if not optimizing.
4124 (parm_preserved_before_stmt_p): Assume modification present when not
4127 2014-01-07 John David Anglin <danglin@gcc.gnu.org>
4130 * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
4131 for 14-bit register offsets when INT14_OK_STRICT is false.
4133 2014-01-07 Roland Stigge <stigge@antcom.de>
4134 Michael Meissner <meissner@linux.vnet.ibm.com>
4137 * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
4138 Only check TFmode for SPE constants. Don't check TImode or TDmode.
4140 2014-01-07 Jakub Jelinek <jakub@redhat.com>
4142 PR rtl-optimization/58668
4143 * cfgcleanup.c (flow_find_cross_jump): Don't count
4144 any jumps if dir_p is NULL. Remove p1 variable and make USE/CLOBBER
4145 check consistent with other places.
4146 (flow_find_head_matching_sequence): Don't count USE or CLOBBER insns.
4147 (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
4149 * ifcvt.c (count_bb_insns): Don't count USE or CLOBBER insns.
4151 2014-01-07 Mike Stump <mikestump@comcast.net>
4152 Jakub Jelinek <jakub@redhat.com>
4155 * tree.h (struct tree_optimization_option): Change optabs
4156 type from unsigned char * to void *.
4157 * optabs.c (init_tree_optimization_optabs): Adjust
4158 TREE_OPTIMIZATION_OPTABS initialization.
4160 2014-01-07 Jakub Jelinek <jakub@redhat.com>
4162 Backported from mainline
4163 2013-12-16 Jakub Jelinek <jakub@redhat.com>
4167 * gimple.h (walk_stmt_load_store_addr_fn): New typedef.
4168 (walk_stmt_load_store_addr_ops, walk_stmt_load_store_ops): Use it
4169 for callback params.
4170 * gimple.c (walk_stmt_load_store_ops): Likewise.
4171 (walk_stmt_load_store_addr_ops): Likewise. Adjust all callback
4172 calls to supply the gimple operand containing the base tree
4173 as an extra argument.
4174 * tree-ssa-ter.c (find_ssaname, find_ssaname_in_store): New helper
4176 (find_replaceable_in_bb): For calls or GIMPLE_ASM, only set
4177 same_root_var if USE is used somewhere in the stores of the stmt.
4178 * ipa-prop.c (visit_ref_for_mod_analysis): Remove name of the stmt
4179 argument and ATTRIBUTE_UNUSED, add another unnamed tree argument.
4180 * ipa-pure-const.c (check_load, check_store, check_ipa_load,
4181 check_ipa_store): Likewise.
4182 * gimple.c (gimple_ior_addresses_taken_1): Likewise.
4183 * ipa-split.c (test_nonssa_use, mark_nonssa_use): Likewise.
4184 (verify_non_ssa_vars, visit_bb): Adjust their callers.
4185 * cfgexpand.c (add_scope_conflicts_1): Use
4186 walk_stmt_load_store_addr_fn type for visit variable.
4187 (visit_op, visit_conflict): Remove name of the stmt
4188 argument and ATTRIBUTE_UNUSED, add another unnamed tree argument.
4189 * tree-sra.c (asm_visit_addr): Likewise. Remove name of the data
4190 argument and ATTRIBUTE_UNUSED.
4191 * cgraphbuild.c (mark_address, mark_load, mark_store): Add another
4192 unnamed tree argument.
4194 2014-01-03 Andreas Schwab <schwab@linux-m68k.org>
4196 * config/m68k/m68k.c (handle_move_double): Handle pushes with
4197 overlapping registers also for registers other than the stack
4200 2014-01-03 Jakub Jelinek <jakub@redhat.com>
4203 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
4206 2014-01-01 Jakub Jelinek <jakub@redhat.com>
4208 PR rtl-optimization/59647
4209 * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
4210 new_rtx into UNSIGNED_FLOAT rtxes.
4212 2013-12-28 Eric Botcazou <ebotcazou@adacore.com>
4214 * doc/invoke.texi (output file options): Document -fada-spec-parent.
4216 2013-12-26 Uros Bizjak <ubizjak@gmail.com>
4218 * config/i386/driver-i386.c (decode_caches_intel): Add missing entries.
4220 2013-12-20 Jakub Jelinek <jakub@redhat.com>
4223 Backported from mainline
4224 2013-08-19 Dehao Chen <dehao@google.com>
4226 * value-prof.c (gimple_ic): Fix the bug of adding EH edge.
4228 2013-12-19 James Greenhalgh <james.greenhalgh@arm.com>
4230 Backport from Mainline.
4231 2013-05-01 James Greenhalgh <james.greenhalgh@arm.com>
4233 * config/aarch64/aarch64-simd-builtins.def (cmhs): Rename to...
4235 (cmhi): Rename to...
4237 * config/aarch64/aarch64-simd.md
4238 (simd_mode): Add SF.
4239 (aarch64_vcond_internal): Use new names for unsigned comparison insns.
4240 (aarch64_cm<optab><mode>): Rewrite to not use UNSPECs.
4241 * config/aarch64/aarch64.md (*cstore<mode>_neg): Rename to...
4242 (cstore<mode>_neg): ...This.
4243 * config/aarch64/iterators.md
4245 (unspec): Remove UNSPEC_CM<EQ, LE, LT, GE, GT, HS, HI, TST>.
4247 (UCOMPARISONS): Likewise.
4248 (optab): Add missing comparisons.
4256 (V_cmp_result): Add DF, SF modes.
4257 (v_cmp_result): Likewise.
4260 * config/aarch64/predicates.md (aarch64_reg_or_fp_zero): New.
4262 Partial Backport from mainline.
4263 2013-05-01 James Greenhalgh <james.greenhalgh@arm.com>
4265 * config/aarch64/arm_neon.h
4266 (vc<eq, lt, le, gt, ge, tst><qsd>_<u><8,16,32,64>): Remap
4267 to builtins or C as appropriate.
4269 2013-12-19 Dominik Vogt <vogt@linux.vnet.ibm.com>
4270 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
4272 Backport from mainline
4273 2013-12-19 Dominik Vogt <vogt@linux.vnet.ibm.com>
4274 * config/s390/s390.c (s390_hotpatch_trampoline_halfwords_default): New
4276 (s390_hotpatch_trampoline_halfwords_max): New constant
4277 (s390_hotpatch_trampoline_halfwords): New static variable
4278 (get_hotpatch_attribute): New function
4279 (s390_handle_hotpatch_attribute): New function
4280 (s390_attribute_table): New target specific attribute table to implement
4281 the hotpatch attribute
4282 (s390_option_override): Parse hotpatch options
4283 (s390_function_num_hotpatch_trampoline_halfwords): New function
4284 (s390_can_inline_p): Implement target hook to
4285 suppress hotpatching for explicitly inlined functions
4286 (s390_asm_output_function_label): Generate hotpatch prologue
4287 (TARGET_ATTRIBUTE_TABLE): Define to implement target attribute table
4288 (TARGET_CAN_INLINE_P): Define to implement target hook
4289 * config/s390/s390.opt (mhotpatch): New options -mhotpatch, -mhotpatch=
4290 * config/s390/s390-protos.h (s390_asm_output_function_label): Add
4292 * config/s390/s390.h (ASM_OUTPUT_FUNCTION_LABEL): Target specific
4293 function label generation for hotpatching
4294 (FUNCTION_BOUNDARY): Align functions to eight bytes
4295 * doc/extend.texi: Document hotpatch attribute
4296 * doc/invoke.texi: Document -mhotpatch option
4298 2013-12-18 Eric Botcazou <ebotcazou@adacore.com>
4300 * config/arm/arm.c (arm_expand_epilogue_apcs_frame): Fix thinko.
4302 2013-12-12 Vladimir Makarov <vmakarov@redhat.com>
4305 * lra-coalesce.c (lra_coalesce): Invalidate inheritance pseudo
4306 values if necessary.
4308 2013-12-12 Jakub Jelinek <jakub@redhat.com>
4311 * gimplify.c (omp_check_private): Add copyprivate argument, if it
4312 is true, don't check omp_privatize_by_reference.
4313 (gimplify_scan_omp_clauses): For OMP_CLAUSE_COPYPRIVATE verify
4314 decl is private in outer context. Adjust omp_check_private caller.
4316 2013-12-10 Eric Botcazou <ebotcazou@adacore.com>
4318 PR rtl-optimization/58295
4319 * simplify-rtx.c (simplify_truncation): Restrict the distribution for
4320 WORD_REGISTER_OPERATIONS targets.
4322 2013-12-10 Kai Tietz <ktietz@redhat.com>
4325 * config/i386/i386.c (ix86_expand_prologue): Address saved
4326 registers stack-relative, not via frame-pointer.
4328 2013-12-09 Alan Modra <amodra@gmail.com>
4331 2013-12-05 Alan Modra <amodra@gmail.com>
4332 * configure.ac (BUILD_CXXFLAGS) Don't use ALL_CXXFLAGS for
4334 <recursive call for build != host>: Clear GMPINC. Don't bother
4336 * configure: Regenerate.
4338 2013-12-08 Uros Bizjak <ubizjak@gmail.com>
4340 Backport from mainline
4341 2013-12-06 Uros Bizjak <ubizjak@gmail.com>
4344 * config/i386/i386.c (type_natural_mode): Properly handle
4345 size 8 for !TARGET_64BIT.
4347 2013-12-07 Ralf Corsépius <ralf.corsepius@rtems.org>
4349 * config.gcc (microblaze*-*-rtems*): Add TARGET_BIG_ENDIAN_DEFAULT.
4351 2013-12-06 Jakub Jelinek <jakub@redhat.com>
4353 PR tree-optimization/59388
4354 * tree-ssa-reassoc.c (update_range_test): If op == range->exp,
4355 gimplify tem after stmt rather than before it.
4357 2013-12-06 Oleg Endo <olegendo@gcc.gnu.org>
4359 Backport from mainline
4360 2013-11-26 Oleg Endo <olegendo@gcc.gnu.org>
4364 * config/sh/sh.c (max_mov_insn_displacement, disp_addr_displacement):
4365 Prefix function names with 'sh_'. Make them non-static.
4366 * config/sh/sh-protos.h (sh_disp_addr_displacement,
4367 sh_max_mov_insn_displacement): Add declarations.
4368 * config/sh/constraints.md (Q): Reject QImode.
4369 (Sdd): Use match_code "mem".
4370 (Snd): Fix erroneous matching of non-memory operands.
4371 * config/sh/predicates.md (short_displacement_mem_operand): New
4373 (general_movsrc_operand): Disallow PC relative QImode loads.
4374 * config/sh/sh.md (*mov<mode>_reg_reg): Remove it.
4375 (*movqi, *movhi): Merge both insns into...
4376 (*mov<mode>): ... this new insn. Replace generic 'm' constraints with
4377 'Snd' and 'Sdd' constraints. Calculate insn length dynamically based
4378 on the operand types.
4380 2013-12-06 Richard Biener <rguenther@suse.de>
4382 Backport from mainline
4383 2013-11-29 Richard Biener <rguenther@suse.de>
4385 PR tree-optimization/59334
4386 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Fix bug
4389 2013-11-28 Richard Biener <rguenther@suse.de>
4391 PR tree-optimization/59330
4392 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Simplify
4393 and fix delayed marking of free calls not necessary.
4395 2013-12-06 Richard Biener <rguenther@suse.de>
4397 Backport from mainline
4398 2013-11-27 Richard Biener <rguenther@suse.de>
4400 PR tree-optimization/59288
4401 * tree-vect-loop.c (get_initial_def_for_induction): Do not
4402 re-analyze the PHI but use STMT_VINFO_LOOP_PHI_EVOLUTION_PART.
4404 2013-11-19 Richard Biener <rguenther@suse.de>
4406 PR tree-optimization/59164
4407 * tree-vect-loop.c (vect_analyze_loop_operations): Adjust
4408 check whether we can create an epilogue loop to reflect the
4409 cases where we create one.
4411 2013-09-05 Richard Biener <rguenther@suse.de>
4413 PR tree-optimization/58137
4414 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size):
4415 Do not create vectors of pointers.
4416 * tree-vect-loop.c (get_initial_def_for_induction): Use proper
4417 types for the components of the vector initializer.
4418 * tree-cfg.c (verify_gimple_assign_binary): Remove special-casing
4419 allowing pointer vectors with PLUS_EXPR/MINUS_EXPR.
4421 2013-12-06 Oleg Endo <olegendo@gcc.gnu.org>
4425 * config/sh/sh.md (*cbranch_t): Check that there are no labels between
4426 the s1 insn and the testing insn. Remove REG_DEAD note from s1 insn.
4428 2013-12-05 Richard Biener <rguenther@suse.de>
4430 Backport from mainline
4431 2013-11-19 Richard Biener <rguenther@suse.de>
4434 * tree-ssa-ter.c (find_replaceable_in_bb): Avoid forwarding
4435 loads into stmts that may clobber it.
4437 2013-12-04 Jakub Jelinek <jakub@redhat.com>
4439 PR rtl-optimization/58726
4440 * combine.c (force_to_mode): Fix comment typo. Don't destructively
4441 modify x for ROTATE, ROTATERT and IF_THEN_ELSE.
4443 2013-12-04 Jakub Jelinek <jakub@redhat.com>
4444 Uros Bizjak <ubizjak@gmail.com>
4447 * config/i386/i386.c (ix86_legitimate_combined_insn): If for
4448 !TARGET_AVX there is misaligned MEM operand with vector mode
4449 and get_attr_ssememalign is 0, return false.
4450 (ix86_expand_special_args_builtin): Add get_pointer_alignment
4451 computed alignment and for non-temporal loads/stores also
4452 at least GET_MODE_ALIGNMENT as MEM_ALIGN.
4453 * config/i386/sse.md
4454 (<sse>_loadu<ssemodesuffix><avxsizesuffix>,
4455 <sse>_storeu<ssemodesuffix><avxsizesuffix>,
4456 <sse2>_loaddqu<avxsizesuffix>,
4457 <sse2>_storedqu<avxsizesuffix>, <sse3>_lddqu<avxsizesuffix>,
4458 sse_vmrcpv4sf2, sse_vmrsqrtv4sf2, sse2_cvtdq2pd, sse_movhlps,
4459 sse_movlhps, sse_storehps, sse_loadhps, sse_loadlps,
4460 *vec_interleave_highv2df, *vec_interleave_lowv2df,
4461 *vec_extractv2df_1_sse, sse2_loadhpd, sse2_loadlpd, sse2_movsd,
4462 sse4_1_<code>v8qiv8hi2, sse4_1_<code>v4qiv4si2,
4463 sse4_1_<code>v4hiv4si2, sse4_1_<code>v2qiv2di2,
4464 sse4_1_<code>v2hiv2di2, sse4_1_<code>v2siv2di2, sse4_2_pcmpestr,
4465 *sse4_2_pcmpestr_unaligned, sse4_2_pcmpestri, sse4_2_pcmpestrm,
4466 sse4_2_pcmpestr_cconly, sse4_2_pcmpistr, *sse4_2_pcmpistr_unaligned,
4467 sse4_2_pcmpistri, sse4_2_pcmpistrm, sse4_2_pcmpistr_cconly): Add
4468 ssememalign attribute.
4469 * config/i386/i386.md (ssememalign): New define_attr.
4471 2013-12-03 Jakub Jelinek <jakub@redhat.com>
4474 * gimplify.c (nonlocal_vla_vars): New variable.
4475 (gimplify_var_or_parm_decl): Put VAR_DECLs for VLAs into
4476 nonlocal_vla_vars chain.
4477 (gimplify_body): Call declare_vars on nonlocal_vla_vars chain
4478 if outer_bind has DECL_INITIAL (current_function_decl) block.
4481 * optabs.c (emit_conditional_move): Save and restore
4482 pending_stack_adjust and stack_pointer_delta if cmove can't be used.
4484 2013-12-02 Jakub Jelinek <jakub@redhat.com>
4486 PR tree-optimization/59358
4487 * tree-vrp.c (union_ranges): To check for the partially
4488 overlapping ranges or adjacent ranges, also compare *vr0max
4491 2013-12-02 Richard Biener <rguenther@suse.de>
4493 PR tree-optimization/59139
4494 * tree-ssa-loop-niter.c (chain_of_csts_start): Properly match
4495 code in get_val_for.
4496 (get_val_for): Use gcc_checking_asserts.
4498 2013-11-27 Tom de Vries <tom@codesourcery.com>
4499 Marc Glisse <marc.glisse@inria.fr>
4502 * fold-const.c (fold_indirect_ref_1): Don't create out-of-bounds
4504 * gimplify.c (gimple_fold_indirect_ref): Same.
4506 2013-12-01 Eric Botcazou <ebotcazou@adacore.com>
4508 * config/i386/winnt.c (i386_pe_asm_named_section): Be prepared for an
4511 2013-12-01 Bernd Edlinger <bernd.edlinger@hotmail.de>
4513 * expr.c (emit_group_store): Fix off-by-one BITFIELD_END argument.
4515 2013-11-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4517 Backport from mainline
4518 2013-11-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4520 * config/arm/iterators.md (vrint_conds): New int attribute.
4521 * config/arm/vfp.md (<vrint_pattern><SDF:mode>2): Set conds attribute.
4522 (smax<mode>3): Likewise.
4523 (smin<mode>3): Likewise.
4525 2013-11-28 Uros Bizjak <ubizjak@gmail.com>
4527 Backport from mainline
4528 2013-11-27 Uros Bizjak <ubizjak@gmail.com>
4529 Ganesh Gopalasubramanian <Ganesh.Gopalasubramanian@amd.com>
4532 * gcc.target/i386/xop-frczX.c: New test.
4534 2013-11-28 Terry Guo <terry.guo@arm.com>
4536 Backport mainline r205391
4537 2013-11-26 Terry Guo <terry.guo@arm.com>
4539 * config/arm/arm.c (require_pic_register): Handle high pic base
4540 register for thumb-1.
4541 (arm_load_pic_register): Also initialize high pic base register.
4542 * doc/invoke.texi: Update documentation for option -mpic-register.
4544 2013-11-27 Jakub Jelinek <jakub@redhat.com>
4546 Backported from mainline
4547 2013-11-26 Jakub Jelinek <jakub@redhat.com>
4549 PR tree-optimization/59014
4550 * tree-vrp.c (register_edge_assert_for_1): Don't look
4551 through conversions from non-integral types or through
4552 narrowing conversions.
4554 2013-11-27 Eric Botcazou <ebotcazou@adacore.com>
4557 * expr.c (emit_group_store): Don't write past the end of the structure.
4558 (store_bit_field): Fix formatting.
4560 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
4562 Backport from mainline
4563 2013-09-17 Sebastian Huber <sebastian.huber@embedded-brains.de>
4565 * config/sparc/t-rtems: Add leon3 multilibs.
4567 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
4569 Backport from mainline
4570 2013-08-09 Eric Botcazou <ebotcazou@adacore.com>
4572 * configure.ac: Add GAS check for LEON instructions on SPARC.
4573 * configure: Regenerate.
4574 * config.in: Likewise.
4575 * config.gcc (with_cpu): Remove sparc-leon*-* and deal with LEON in the
4577 * config/sparc/sparc.opt (LEON, LEON3): New masks.
4578 * config/sparc/sparc.h (ASM_CPU32_DEFAULT_SPEC): Set to AS_LEON_FLAG
4580 (ASM_CPU_SPEC): Pass AS_LEON_FLAG if -mcpu=leon or -mcpu=leon3.
4581 (AS_LEON_FLAG): New macro.
4582 * config/sparc/sparc.c (sparc_option_override): Set MASK_LEON for leon
4583 and MASK_LEON3 for leon3 and unset them if HAVE_AS_LEON is not defined.
4584 Deal with LEON and LEON3 for the memory model.
4585 * config/sparc/sync.md (atomic_compare_and_swap<mode>): Enable if LEON3
4586 (atomic_compare_and_swap<mode>_1): Likewise.
4587 (*atomic_compare_and_swap<mode>_1): Likewise.
4589 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
4591 Backport from mainline
4592 2013-07-23 Eric Botcazou <ebotcazou@adacore.com>
4594 * doc/invoke.texi (SPARC Options): Document new leon3 processor value.
4596 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
4598 Backport from mainline
4599 2013-07-22 Eric Botcazou <ebotcazou@adacore.com>
4601 * config.gcc (sparc*-*-*): Accept leon3 processor.
4602 (sparc-leon*-*): Merge with sparc*-*-* and add leon3 support.
4603 * doc/invoke.texi (SPARC Options): Adjust -mfix-ut699 entry.
4604 * config/sparc/sparc-opts.h (enum processor_type): Add PROCESSOR_LEON3.
4605 * config/sparc/sparc.opt (enum processor_type): Add leon3.
4606 (mfix-ut699): Adjust comment.
4607 * config/sparc/sparc.h (TARGET_CPU_leon3): New define.
4608 (CPP_CPU32_DEFAULT_SPEC): Add leon3 support.
4609 (CPP_CPU_SPEC): Likewise.
4610 (ASM_CPU_SPEC): Likewise.
4611 * config/sparc/sparc.c (leon3_cost): New constant.
4612 (sparc_option_override): Add leon3 support.
4613 (mem_ref): New function.
4614 (sparc_gate_work_around_errata): Return true if -mfix-ut699 is enabled.
4615 (sparc_do_work_around_errata): Look into the instruction in the delay
4616 slot and adjust accordingly. Add fix for the data cache nullify issues
4617 of the UT699. Change insertion position for the NOP.
4618 * config/sparc/leon.md (leon_fpalu, leon_fpmds, write_buf): Delete.
4619 (leon3_load): New reservation.
4620 (leon_store): Bump latency to 2.
4621 (grfpu): New automaton.
4622 (grfpu_alu): New unit.
4623 (grfpu_ds): Likewise.
4624 (leon_fp_alu): Adjust.
4625 (leon_fp_mult): Delete.
4626 (leon_fp_div): Split into leon_fp_divs and leon_fp_divd.
4627 (leon_fp_sqrt): Split into leon_fp_sqrts and leon_fp_sqrtd.
4628 * config/sparc/sparc.md (cpu): Add leon3.
4629 * config/sparc/sync.md (atomic_exchangesi): Disable if -mfix-ut699.
4631 (atomic_test_and_set): Likewise.
4634 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
4636 Backport from mainline
4637 2013-04-10 Steven Bosscher <steven@gcc.gnu.org>
4639 * config/sparc/sparc.c: Include tree-pass.h.
4640 (TARGET_MACHINE_DEPENDENT_REORG): Do not redefine.
4641 (sparc_reorg): Rename to sparc_do_work_around_errata. Move to
4642 head of file. Change return type. Split off gate function.
4643 (sparc_gate_work_around_errata): New function.
4644 (pass_work_around_errata): New pass definition.
4645 (insert_pass_work_around_errata) New pass insert definition to
4646 insert pass_work_around_errata just after delayed-branch scheduling.
4647 (sparc_option_override): Insert the pass.
4648 * config/sparc/t-sparc (sparc.o): Add TREE_PASS_H dependence.
4650 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
4652 Backport from mainline
4653 2013-05-28 Eric Botcazou <ebotcazou@adacore.com>
4655 * doc/invoke.texi (SPARC Options): Document -mfix-ut699.
4656 * builtins.c (expand_builtin_mathfn) <BUILT_IN_SQRT>: Try to widen the
4657 mode if the instruction isn't available in the original mode.
4658 * config/sparc/sparc.opt (mfix-ut699): New option.
4659 * config/sparc/sparc.md (muldf3_extend): Disable if -mfix-ut699.
4660 (divdf3): Turn into expander.
4661 (divdf3_nofix): New insn.
4662 (divdf3_fix): Likewise.
4663 (divsf3): Disable if -mfix-ut699.
4664 (sqrtdf2): Turn into expander.
4665 (sqrtdf2_nofix): New insn.
4666 (sqrtdf2_fix): Likewise.
4667 (sqrtsf2): Disable if -mfix-ut699.
4669 2013-11-22 Eric Botcazou <ebotcazou@adacore.com>
4671 * print-rtl.c (print_rtx) <case MEM>: Output a space if no MEM_EXPR.
4673 2013-11-21 Zhenqiang Chen <zhenqiang.chen@linaro.org>
4676 Backport from mainline: r197467 and r198999.
4677 2013-04-03 Jeff Law <law@redhat.com>
4679 * Makefile.in (lra-constraints.o): Depend on $(OPTABS_H).
4680 (lra-eliminations.o): Likewise.
4682 2013-05-16 Jeff Law <law@redhat.com>
4684 * Makefile.in (tree-switch-conversion.o): Depend on $(OPTABS_H).
4686 2013-11-20 Eric Botcazou <ebotcazou@adacore.com>
4689 * config/sparc/sparc.c (sparc_fold_builtin) <case CODE_FOR_pdist_vis>:
4690 Make sure neg2_ovf is set before being used.
4692 2013-11-20 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
4693 Dominik Vogt <vogt@linux.vnet.ibm.com>
4695 Backport from mainline
4696 * config/s390/s390.c (s390_canonicalize_comparison): Don't fold
4697 int comparisons with an out of range condition code.
4698 (s390_optimize_nonescaping_tx): Skip empty BBs.
4699 Generate the new tbegin RTX when removing the FPR clobbers (with
4701 (s390_expand_tbegin): Fix the retry loop counter. Copy CC to the
4702 result before doing the retry calculations.
4703 (s390_init_builtins): Make tbegin "returns_twice" and tabort
4705 * config/s390/s390.md (UNSPECV_TBEGIN_TDB): New constant used for
4706 the TDB setting part of an tbegin.
4707 ("tbegin_1", "tbegin_nofloat_1"): Add a set for the TDB.
4708 ("tx_assist"): Set unused argument to an immediate zero instead of
4709 loading zero into a GPR and pass it as argument.
4710 * config/s390/htmxlintrin.h (__TM_simple_begin, __TM_begin):
4711 Remove inline and related attributes.
4712 (__TM_nesting_depth, __TM_is_user_abort, __TM_is_named_user_abort)
4713 (__TM_is_illegal, __TM_is_footprint_exceeded)
4714 (__TM_is_nested_too_deep, __TM_is_conflict): Fix format value
4717 2013-11-19 Uros Bizjak <ubizjak@gmail.com>
4719 Backport from mainline
4720 2013-11-18 Uros Bizjak <ubizjak@gmail.com>
4722 * config/i386/i386.c (ix86_decompose_address): Use REG_P instead of
4723 ix86_address_subreg_operand. Move subreg checks to
4724 ix86_validate_address_register. Move address override check to
4725 ix86_legitimate_address_p.
4726 (ix86_validate_address_register): New function.
4727 (ix86_legitimate_address_p): Call ix86_validate_address_register
4728 to validate base and index registers. Add address override check
4729 from ix86_decompose_address.
4730 (ix86_decompose_address): Remove.
4732 Backport from mainline
4733 2013-11-17 Uros Bizjak <ubizjak@gmail.com>
4736 * config/i386/i386.c (ix86_address_subreg_operand): Do not
4737 reject non-integer subregs.
4738 (ix86_decompose_address): Do not reject invalid CONST_INT RTXes.
4739 Move check for invalid x32 constant addresses ...
4740 (ix86_legitimate_address_p): ... here.
4742 2013-11-19 Richard Biener <rguenther@suse.de>
4744 Backport from mainline
4745 2013-11-07 Richard Biener <rguenther@suse.de>
4747 * tree-dfa.c (get_ref_base_and_extent): Fix casting.
4749 2013-11-19 Richard Biener <rguenther@suse.de>
4751 PR tree-optimization/57517
4752 * tree-predcom.c (combinable_refs_p): Verify the combination
4753 is always executed when the refs are.
4755 2013-11-19 Richard Biener <rguenther@suse.de>
4757 Backport from mainline
4758 2013-11-05 Richard Biener <rguenther@suse.de>
4761 * tree-dfa.c (get_ref_base_and_extent): Merge common code
4762 in MEM_REF and TARGET_MEM_REF handling. Make sure to
4763 process trailing array detection before diving into the
4764 view-converted object (and possibly apply some extra offset).
4766 2013-11-18 Richard Biener <rguenther@suse.de>
4768 Backport from mainline
4769 2013-10-21 Richard Biener <rguenther@suse.de>
4771 PR tree-optimization/58794
4772 * fold-const.c (operand_equal_p): Compare FIELD_DECL operand
4773 of COMPONENT_REFs with OEP_CONSTANT_ADDRESS_OF left in place.
4775 2013-10-21 Richard Biener <rguenther@suse.de>
4778 * fold-const.c (fold_binary_loc): Fold ((T) (X /[ex] C)) * C
4779 to (T) X for sign-changing conversions (or no conversion).
4781 2013-11-06 Richard Biener <rguenther@suse.de>
4783 PR tree-optimization/58653
4784 * tree-predcom.c (ref_at_iteration): Rewrite to generate
4786 (prepare_initializers_chain): Adjust.
4788 PR tree-optimization/59047
4789 * tree-predcom.c (ref_at_iteration): Handle bitfield accesses
4792 2013-10-15 Richard Biener <rguenther@suse.de>
4794 PR tree-optimization/58143
4795 * tree-ssa-loop-im.c (arith_code_with_undefined_signed_overflow):
4797 (rewrite_to_defined_overflow): Likewise.
4798 (move_computations_dom_walker::before_dom): Rewrite stmts
4799 with undefined signed overflow that are not always executed
4800 into unsigned arithmetic.
4802 2013-11-14 Uros Bizjak <ubizjak@gmail.com>
4804 Backport from mainline
4805 2013-11-10 Uros Bizjak <ubizjak@gmail.com>
4807 * mode-switching.c (optimize_mode_switching): Mark block as
4808 nontransparent, if last_mode at block exit is different from no_mode.
4810 Backport from mainline
4811 2013-11-06 Uros Bizjak <ubizjak@gmail.com>
4814 * config/i386/i386.c (ix86_avx_u128_mode_needed): Require
4815 AVX_U128_DIRTY mode for call_insn RTXes that use AVX256 registers.
4816 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY mode for call_insn
4817 RTXes that return in AVX256 register.
4819 2013-11-14 Jakub Jelinek <jakub@redhat.com>
4820 Uros Bizjak <ubizjak@gmail.com>
4823 * config/i386/i386.md (*anddi_2): Only allow CCZmode if
4824 operands[2] satisfies_constraint_Z that might have bit 31 set.
4826 2013-11-12 H.J. Lu <hongjiu.lu@intel.com>
4828 Backported from mainline
4829 2013-11-12 H.J. Lu <hongjiu.lu@intel.com>
4832 * config/i386/i386.c (initial_ix86_tune_features): Set
4833 X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL and
4834 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL for m_HASWELL.
4836 2013-11-11 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4838 Backported from mainline
4839 2013-10-30 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4842 * config/arm/arm.c (arm_expand_epilogue_apcs_frame): Emit blockage
4844 2013-11-11 Jakub Jelinek <jakub@redhat.com>
4846 Backported from mainline
4847 2013-11-06 Jakub Jelinek <jakub@redhat.com>
4850 * expr.c (get_bit_range): Handle *offset == NULL_TREE.
4851 (expand_assignment): If *bitpos is negative, set *offset
4852 and adjust *bitpos, so that it is not negative.
4854 2013-11-05 Jakub Jelinek <jakub@redhat.com>
4856 PR rtl-optimization/58997
4857 * loop-iv.c (iv_subreg): For IV_UNKNOWN_EXTEND, expect
4858 get_iv_value to be in iv->mode rather than iv->extend_mode.
4859 (iv_extend): Likewise. Otherwise, if iv->extend != extend,
4860 use lowpart_subreg on get_iv_value before calling simplify_gen_unary.
4861 * loop-unswitch.c (may_unswitch_on): Make sure op[i] is in the right
4864 2013-11-10 Karlson2k <k2k@narod.ru>
4865 Kai Tietz <ktietz@redhat.com>
4869 * configure.ac: Adding for exported symbols check
4870 and for rdynamic-check executable-extension.
4871 * configure: Regenerated.
4873 2013-11-07 H.J. Lu <hongjiu.lu@intel.com>
4876 * config/i386/i386.md (push peepholer/splitter): Use Pmode
4877 with stack_pointer_rtx.
4879 2013-11-05 Uros Bizjak <ubizjak@gmail.com>
4881 * config/i386/t-rtems (MULTILIB_MATCHES): Fix option typos.
4883 2013-11-05 Jakub Jelinek <jakub@redhat.com>
4885 PR tree-optimization/58984
4886 * ipa-prop.c (ipa_load_from_parm_agg_1): Add SIZE_P argument,
4887 set *SIZE_P if non-NULL on success.
4888 (ipa_load_from_parm_agg, ipa_analyze_indirect_call_uses): Adjust
4890 (ipcp_transform_function): Likewise. Punt if size of access
4891 is different from TYPE_SIZE on v->value's type.
4893 2013-11-03 H.J. Lu <hongjiu.lu@intel.com>
4895 Backport from mainline
4896 2013-10-12 H.J. Lu <hongjiu.lu@intel.com>
4899 * config/i386/i386.c (ix86_copy_addr_to_reg): New function.
4900 (ix86_expand_movmem): Replace copy_addr_to_reg with
4901 ix86_copy_addr_to_reg.
4902 (ix86_expand_setmem): Likewise.
4904 2013-10-29 Uros Bizjak <ubizjak@gmail.com>
4906 Backport from mainline
4907 2013-08-08 Richard Sandiford <rdsandiford@googlemail.com>
4909 PR rtl-optimization/58079
4910 * combine.c (combine_simplify_rtx): Avoid using SUBST if
4911 simplify_comparison has widened a comparison with an integer.
4913 2013-10-29 Martin Jambor <mjambor@suse.cz>
4916 Backport from mainline
4917 2013-05-09 Martin Jambor <mjambor@suse.cz>
4920 * gimple-fold.c (canonicalize_constructor_val): Call
4921 cgraph_get_create_real_symbol_node instead of cgraph_get_create_node.
4923 Backport from mainline
4924 2013-03-16 Jan Hubicka <jh@suse.cz>
4926 * cgraph.h (cgraph_get_create_real_symbol_node): Declare.
4927 * cgraph.c (cgraph_get_create_real_symbol_node): New function.
4928 * cgrpahbuild.c: Use cgraph_get_create_real_symbol_node instead
4929 of cgraph_get_create_node.
4930 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4932 2013-10-28 Tom de Vries <tom@codesourcery.com>
4934 * cfgexpand.c (gimple_expand_cfg): Remove test for parm_birth_insn.
4935 Don't commit insertions after NOTE_INSN_FUNCTION_BEG.
4937 2013-10-26 Uros Bizjak <ubizjak@gmail.com>
4939 Backport from mainline
4940 2013-10-22 Uros Bizjak <ubizjak@gmail.com>
4943 * config/i386/i386.c (put_condition_code) <case GTU, case LEU>:
4944 Remove CCCmode handling.
4945 <case LTU>: Return 'c' suffix for CCCmode.
4946 <case GEU>: Return 'nc' suffix for CCCmode.
4947 (ix86_cc_mode) <case GTU, case LEU>: Do not generate overflow checks.
4948 * config/i386/i386.md (*sub<mode>3_cconly_overflow): Remove.
4949 (*sub<mode>3_cc_overflow): Ditto.
4950 (*subsi3_zext_cc_overflow): Ditto.
4952 2013-10-26 Uros Bizjak <ubizjak@gmail.com>
4954 Backport from mainline
4955 2013-10-19 Uros Bizjak <ubizjak@gmail.com>
4958 * config/i386/i386.c (ix86_function_value_regno): Add DX_REG,
4959 ST1_REG and XMM1_REG for 32bit and 64bit targets. Also add DI_REG
4960 and SI_REG for 64bit SYSV ABI targets.
4962 2013-08-25 Richard Henderson <rth@twiddle.net>
4965 * optabs.c (maybe_emit_atomic_exchange): Use create_input_operand
4966 instead of create_convert_operand_to.
4967 (maybe_emit_sync_lock_test_and_set): Likewise.
4968 (expand_atomic_compare_and_swap): Likewise.
4969 (maybe_emit_compare_and_swap_exchange_loop): Don't convert_modes.
4971 2013-10-25 Eric Botcazou <ebotcazou@adacore.com>
4973 PR rtl-optimization/58831
4974 * alias.c (init_alias_analysis): At the beginning of each iteration, set
4975 the reg_seen[N] bit if static_reg_base_value[N] is non-null.
4977 2013-10-25 Eric Botcazou <ebotcazou@adacore.com>
4979 * recog.c (search_ofs): New static variable moved from...
4980 (peep2_find_free_register): ...here.
4981 (peephole2_optimize): Initialize it.
4983 2013-10-24 David Edelsohn <dje.gcc@gmail.com>
4985 Backport from mainline
4986 2013-10-23 David Edelsohn <dje.gcc@gmail.com>
4989 * config/rs6000/rs6000.md (mulsi3_internal1 and splitter): Add
4990 TARGET_32BIT final condition.
4991 (mulsi3_internal2 and splitter): Same.
4993 2013-10-23 Tom de Vries <tom@codesourcery.com>
4995 PR tree-optimization/58805
4996 * tree-ssa-tail-merge.c (stmt_local_def): Add gimple_vdef check.
4998 2013-10-23 Richard Biener <rguenther@suse.de>
5000 Backport from mainline
5001 2013-06-24 Richard Biener <rguenther@suse.de>
5003 PR tree-optimization/57488
5004 * tree-ssa-pre.c (insert): Clear NEW sets before each iteration.
5006 2013-10-16 Ganesh Gopalasubramanian <Ganesh.Gopalasubramanian@amd.com>
5008 Backport from mainline
5009 2013-10-16 Ganesh Gopalasubramanian
5010 <Ganesh.Gopalasubramanian@amd.com>
5012 * config/i386/i386.c (ix86_option_override_internal): Enable FMA4
5015 2013-10-16 Jakub Jelinek <jakub@redhat.com>
5017 * BASE-VER: Set to 4.8.3.
5018 * DEV-PHASE: Set to prerelease.
5020 2013-10-16 Release Manager
5022 * GCC 4.8.2 released.
5024 2013-10-12 James Greenhalgh <james.greenhalgh@arm.com>
5026 Backport from mainline.
5027 2013-10-12 James Greenhalgh <james.greenhalgh@arm.com>
5029 * config/aarch64/arm_neon.h
5030 (vtbx<1,3>_<psu>8): Fix register constriants.
5032 2013-10-10 Jakub Jelinek <jakub@redhat.com>
5035 * stmt.c (expand_asm_operands): Add FALLTHRU_BB argument,
5036 if any labels are in FALLTHRU_BB, use a special label emitted
5037 immediately after the asm goto insn rather than label_rtx
5039 (expand_asm_stmt): Adjust caller.
5040 * cfgrtl.c (commit_one_edge_insertion): Force splitting of
5041 edge if the last insn in predecessor is a jump with single successor,
5042 but it isn't simplejump_p.
5044 2013-10-09 Jakub Jelinek <jakub@redhat.com>
5046 Backport from mainline
5047 2013-09-26 Richard Biener <rguenther@suse.de>
5049 PR tree-optimization/58539
5050 * tree-vect-loop.c (vect_create_epilog_for_reduction): Honor
5051 the fact that debug statements are not taking part in loop-closed
5054 2013-10-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5056 * config/s390/s390.c (s390_register_info): Make the call-saved FPR
5057 loop to work also for 31bit ABI.
5058 Save the stack pointer for frame_size > 0.
5060 2013-10-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5062 * config/s390/s390.md ("tbegin", "tbegin_nofloat", "tbegin_retry")
5063 ("tbegin_retry_nofloat", "tend", "tabort", "tx_assist"): Remove
5064 constraint letters from expanders.
5065 ("tbegin_retry", "tbegin_retry_nofloat"): Change predicate of the
5066 retry count to general_operand.
5067 ("tabort"): Give operand 0 a mode.
5068 ("tabort_1"): Add mode and constraint letter for operand 0.
5069 * doc/extend.texi: Fix protoype of __builtin_non_tx_store.
5071 2013-10-04 Marcus Shawcroft <marcus.shawcroft@arm.com>
5073 Backport from mainline.
5076 * config/aarch64/aarch64.md (*add_<shift>_<mode>)
5077 (*add_<shift>_si_uxtw,*add_mul_imm_<mode>)
5078 (*sub_<shift>_<mode>)
5079 (*sub_<shift>_si_uxtw,*sub_mul_imm_<mode>, *sub_mul_imm_si_uxtw):
5080 Remove k constraint.
5082 2013-10-02 John David Anglin <danglin@gcc.gnu.org>
5084 * config.gcc (hppa*64*-*-linux*): Don't add pa/t-linux to tmake_file.
5086 2013-10-01 Jakub Jelinek <jakub@redhat.com>
5087 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5090 * config/s390/s390.c (s390_chunkify_start): Handle tablejump_p first,
5091 continue when done, for other jumps look through PARALLEL
5094 2013-09-30 Jakub Jelinek <jakub@redhat.com>
5097 * fold-const.c (fold_ternary_loc): For A < 0 : <sign bit of A> : 0
5098 optimization, punt if sign_bit_p looked through any zero extension.
5100 2013-09-27 Paulo Matos <pmatos@broadcom.com>
5102 Backport from mainline.
5105 2013-03-27 Richard Biener <rguenther@suse.de>
5107 PR tree-optimization/56716
5108 * tree-ssa-structalias.c (perform_var_substitution): Adjust
5109 dumping for ref nodes.
5111 2013-09-27 Paulo Matos <pmatos@broadcom.com>
5113 Backport from mainline.
5115 2013-09-27 Paulo Matos <pmatos@broadcom.com>
5117 * gcc.dg/pr58463.c: New test.
5119 2013-09-23 Eric Botcazou <ebotcazou@adacore.com>
5121 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Recurse on copy
5122 assignment statements.
5124 2013-09-23 Alan Modra <amodra@gmail.com>
5127 * config/rs6000/rs6000.md (bswapdi2_64bit): Disable for volatile mems.
5129 2013-09-23 Alan Modra <amodra@gmail.com>
5131 * config/rs6000/predicates.md (add_cint_operand): New.
5132 (reg_or_add_cint_operand, small_toc_ref): Use add_cint_operand.
5133 * config/rs6000/rs6000.md (largetoc_high_plus): Restrict offset
5134 using add_cint_operand.
5135 (largetoc_high_plus_aix): Likewise.
5136 * config/rs6000/rs6000.c (toc_relative_expr_p): Use add_cint_operand.
5138 2013-09-20 John David Anglin <danglin@gcc.gnu.org>
5141 * config/pa/pa.c (pa_option_override): Disable auto increment and
5142 decrement instructions until reload is completed.
5144 * config/pa/pa.md: In "scc" insn patterns, change output template to
5145 handle const0_rtx in reg_or_0_operand operands.
5147 2013-09-19 Jakub Jelinek <jakub@redhat.com>
5149 * omp-low.c (expand_omp_sections): Always pass len - 1 to
5150 GOMP_sections_start, even if !exit_reachable.
5152 2013-09-18 Richard Earnshaw <rearnsha@arm.com>
5154 * arm.c (arm_expand_prologue): Validate architecture supports
5155 LDRD/STRD before accepting tuning preferences.
5156 (arm_expand_epilogue): Likewise.
5158 2013-09-18 Daniel Morris <danielm@ecoscentric.com>
5159 Paolo Carlini <paolo.carlini@oracle.com>
5162 * doc/implement-cxx.texi: Fix references to the C++ standards.
5164 2013-09-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5166 PR tree-optimization/58088
5167 * fold-const.c (mask_with_trailing_zeros): New function.
5168 (fold_binary_loc): Make sure we don't recurse infinitely
5169 when the X in (X & C1) | C2 is a tree of the form (Y * K1) & K2.
5170 Use mask_with_trailing_zeros where appropriate.
5172 2013-09-14 John David Anglin <danglin@gcc.gnu.org>
5175 * config/pa/pa.c (pa_expand_prologue): Change mode in gen_rtx_POST_INC
5178 2013-09-13 Christian Bruel <christian.bruel@st.com>
5181 * config/sh/sh.md (mov<mode>_reg_reg): Allow memory reloads.
5183 2013-09-11 Andi Kleen <ak@linux.intel.com>
5185 Backport from mainline
5186 * doc/extend.texi: Use __atomic_store_n instead of
5187 __atomic_store in HLE example.
5189 2013-09-11 Andi Kleen <ak@linux.intel.com>
5191 Backport from mainline
5192 * doc/extend.texi: Dont use __atomic_clear in HLE
5195 2013-09-11 Andi Kleen <ak@linux.intel.com>
5197 Backport from mainline
5198 * doc/extend.texi: Document that __atomic_clear and
5199 __atomic_test_and_set should only be used with bool.
5201 2013-09-11 Richard Biener <rguenther@suse.de>
5204 * passes.c (init_optimization_passes): Split critical edges
5205 before late uninit warning pass in the -Og pipeline.
5207 2013-09-11 Jakub Jelinek <jakub@redhat.com>
5209 PR tree-optimization/58385
5210 * fold-const.c (build_range_check): If both low and high are NULL,
5211 use omit_one_operand_loc to preserve exp side-effects.
5213 2013-09-10 Richard Earnshaw <rearnsha@arm.com>
5216 * arm/vfp.md (combine_vcvt_f32_<FCVTI32typename>): Fix pattern to
5217 support conditional execution.
5218 (combine_vcvt_f64_<FCVTI32typename>): Likewise.
5220 2013-09-10 Jakub Jelinek <jakub@redhat.com>
5222 PR rtl-optimization/58365
5223 * cfgcleanup.c (merge_memattrs): Also clear MEM_READONLY_P
5224 resp. MEM_NOTRAP_P if they differ, or set MEM_VOLATILE_P if
5227 2013-09-09 Jakub Jelinek <jakub@redhat.com>
5229 PR tree-optimization/58364
5230 * tree-ssa-reassoc.c (init_range_entry): For BIT_NOT_EXPR on
5231 BOOLEAN_TYPE, only invert in_p and continue with arg0 if
5232 the current range can't be an unconditional true or false.
5234 2013-09-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5237 Backport from mainline
5238 2013-04-30 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
5240 * explow.c (plus_constant): Pass "mode" to immed_double_int_const.
5241 Use gen_int_mode rather than GEN_INT.
5243 2013-09-09 Richard Biener <rguenther@suse.de>
5245 Backport from mainline
5246 2013-08-27 Richard Biener <rguenther@suse.de>
5248 PR tree-optimization/57521
5249 * tree-if-conv.c (if_convertible_bb_p): Verify that at least
5250 one edge is non-critical.
5251 (find_phi_replacement_condition): Make sure to use a non-critical
5252 edge. Cleanup and remove old bug workarounds.
5253 (bb_postdominates_preds): Remove.
5254 (if_convertible_loop_p_1): Do not compute post-dominators.
5255 (combine_blocks): Do not free post-dominators.
5256 (main_tree_if_conversion): Likewise.
5258 2013-09-09 Richard Biener <rguenther@suse.de>
5260 Backport from mainline
5261 2013-09-03 Richard Biener <rguenther@suse.de>
5264 * fold-const.c (negate_expr_p): Fix division case.
5265 (negate_expr): Likewise.
5267 2013-09-09 Richard Biener <rguenther@suse.de>
5269 Backport from mainline
5270 2013-08-29 Richard Biener <rguenther@suse.de>
5272 PR tree-optimization/57685
5273 * tree-vrp.c (register_edge_assert_for_1): Recurse only for
5274 single-use operands to avoid exponential complexity.
5276 2013-09-09 Richard Biener <rguenther@suse.de>
5278 Backport from mainline
5279 2013-08-30 Richard Biener <rguenther@suse.de>
5281 PR tree-optimization/58223
5282 * tree-loop-distribution.c (has_anti_dependence): Rename to ...
5283 (has_anti_or_output_dependence): ... this and adjust to also
5284 look for output dependences.
5285 (mark_nodes_having_upstream_mem_writes): Adjust.
5286 (rdg_flag_uses): Likewise.
5288 2013-09-03 Richard Biener <rguenther@suse.de>
5290 Backport from mainline
5291 2013-08-29 Richard Biener <rguenther@suse.de>
5293 PR tree-optimization/58246
5294 * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Properly
5295 handle the dominance check inside a basic-block.
5297 2013-09-03 Richard Biener <rguenther@suse.de>
5299 Backport from mainline
5300 2013-08-30 Richard Biener <rguenther@suse.de>
5302 PR tree-optimization/58228
5303 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Do not
5304 allow invariant loads in nested loop vectorization.
5306 2013-09-03 Richard Biener <rguenther@suse.de>
5308 Backport from mainline
5309 2013-08-30 Richard Biener <rguenther@suse.de>
5311 PR tree-optimization/58010
5312 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
5313 assert that we have a loop-closed PHI.
5315 2013-09-01 Uros Bizjak <ubizjak@gmail.com>
5317 Backport from mainline
5318 2013-08-31 Uros Bizjak <ubizjak@gmail.com>
5320 * config/alpha/alpha.c (alpha_emit_conditional_move): Update
5321 "cmp" RTX before signed_comparison_operator check to account
5324 2013-09-01 John David Anglin <danglin@gcc.gnu.org>
5326 * config/pa/pa.md: Allow "const 0" operand 1 in "scc" insns.
5328 2013-08-30 Jakub Jelinek <jakub@redhat.com>
5330 PR tree-optimization/58277
5331 * tree-ssa-strlen.c (strlen_enter_block): If do_invalidate gave up
5332 after seeing too many stmts with vdef in between dombb and current
5333 bb, invalidate everything.
5335 2013-08-29 Oleg Endo <olegendo@gcc.gnu.org>
5337 Backport from mainline
5338 2013-08-05 Oleg Endo <olegendo@gcc.gnu.org>
5341 * recog.h (rtx (*insn_gen_fn) (rtx, ...)): Replace typedef with new
5343 * expr.c (move_by_pieces_1, store_by_pieces_2): Replace argument
5344 rtx (*) (rtx, ...) with insn_gen_fn.
5345 * genoutput.c (output_insn_data): Cast gen_? function pointers to
5346 insn_gen_fn::stored_funcptr. Add initializer braces.
5348 Backport from mainline
5349 2013-08-07 Oleg Endo <olegendo@gcc.gnu.org>
5352 * config/rs6000/rs6000.c (gen_2arg_fn_t): Remove typedef.
5353 (rs6000_emit_swdiv_high_precision, rs6000_emit_swdiv_low_precision,
5354 rs6000_emit_swrsqrt): Don't cast result of GEN_FCN to gen_2arg_fn_t.
5356 2013-08-29 Jakub Jelinek <jakub@redhat.com>
5358 Backported from mainline
5359 2013-05-27 Richard Biener <rguenther@suse.de>
5362 PR tree-optimization/57417
5363 * tree-ssa-sccvn.c (vn_reference_fold_indirect): Fix test
5365 (set_ssa_val_to): Compare addresses using
5366 get_addr_base_and_unit_offset.
5368 PR tree-optimization/57396
5369 * tree-affine.c (double_int_constant_multiple_p): Properly
5370 return false for val == 0 and div != 0.
5372 PR tree-optimization/57343
5373 * tree-ssa-loop-niter.c (number_of_iterations_ne_max): Do not
5374 use multiple_of_p if not TYPE_OVERFLOW_UNDEFINED.
5375 (number_of_iterations_cond): Do not build the folded tree.
5377 2013-08-28 Jakub Jelinek <jakub@redhat.com>
5380 * omp-low.c (copy_var_decl): Copy over TREE_NO_WARNING flag.
5382 2013-08-28 Richard Biener <rguenther@suse.de>
5384 Backport from mainline
5385 2013-06-24 Richard Biener <rguenther@suse.de>
5388 * passes.c (init_optimization_passes): Move pass_fold_builtins
5389 and pass_dce earlier with -Og.
5391 2013-08-28 Uros Bizjak <ubizjak@gmail.com>
5393 Backport from mainline
5394 2013-08-27 H.J. Lu <hongjiu.lu@intel.com>
5396 * config/i386/driver-i386.c (host_detect_local_cpu): Update
5397 Haswell processor detection.
5399 Backport from mainline
5400 2013-08-27 Christian Widmer <shadow@umbrox.de>
5403 * config/i386/driver-i386.c (host_detect_local_cpu): Add detection
5404 of Ivy Bridge and Haswell processors. Assume core-avx2 for unknown
5405 AVX2 capable processors.
5407 2013-08-23 Jakub Jelinek <jakub@redhat.com>
5410 * config/i386/x86-64.h (TARGET_SECTION_TYPE_FLAGS): Define.
5411 * config/i386/i386.c (x86_64_elf_section_type_flags): New function.
5413 PR tree-optimization/58209
5414 * tree-tailcall.c (find_tail_calls): Give up for pointer result types
5415 if m or a is non-NULL.
5417 2013-08-21 Richard Earnshaw <rearnsha@arm.com>
5420 * arm.c (aapcs_vfp_allocate): Decompose the argument if the
5421 suggested mode for the assignment isn't compatible with the
5424 2013-08-20 Alan Modra <amodra@gmail.com>
5427 * config/rs6000/rs6000.c (rs6000_emit_prologue): Correct ool_adjust.
5428 (rs6000_emit_epilogue): Likewise.
5430 2013-08-19 Peter Bergner <bergner@vnet.ibm.com>
5431 Jakub Jelinek <jakub@redhat.com>
5433 Backport from mainline
5434 * config/rs6000/dfp.md (*negtd2_fpr): Handle non-overlapping
5435 destination and source operands.
5437 2013-08-18 Jakub Jelinek <jakub@redhat.com>
5439 PR tree-optimization/58006
5440 * tree-parloops.c (take_address_of): Don't ICE if get_name
5442 (eliminate_local_variables_stmt): Remove clobber stmts.
5444 2013-08-16 Jakub Jelinek <jakub@redhat.com>
5446 PR tree-optimization/58164
5447 * gimple.c (walk_stmt_load_store_addr_ops): For visit_addr
5448 walk gimple_goto_dest of GIMPLE_GOTO.
5450 PR tree-optimization/58165
5451 * tree-call-cdce.c (shrink_wrap_one_built_in_call): If
5452 bi_call must be the last stmt in a bb, don't split_block, instead
5453 use fallthru edge from it and give up if there is none.
5454 Release conds vector when returning early.
5456 2013-08-15 David Given <dg@cowlark.com>
5458 Backport from mainline
5459 2013-04-26 Vladimir Makarov <vmakarov@redhat.com>
5461 * lra-constraints.c (process_alt_operands): Use #if HAVE_ATTR_enable
5464 2013-08-14 Jakub Jelinek <jakub@redhat.com>
5466 PR tree-optimization/58145
5467 * tree-sra.c (build_ref_for_offset): If prev_base has
5468 TREE_THIS_VOLATILE or TREE_SIDE_EFFECTS, propagate it to MEM_REF.
5470 2013-08-14 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5472 * config/s390/htmxlintrin.h: Add file missing from last commit.
5473 * config/s390/htmintrin.h: Likewise.
5474 * config/s390/s390intrin.h: Likewise.
5476 2013-08-14 Uros Bizjak <ubizjak@gmail.com>
5478 Backport from mainline
5479 2013-08-13 Uros Bizjak <ubizjak@gmail.com>
5481 * config/i386/sse.md (*sse2_maskmovdqu): Emit addr32 prefix
5482 when Pmode != word_mode. Add length_address attribute.
5483 (sse3_monitor_<mode>): Merge from sse3_monitor and
5484 sse3_monitor64_<mode> insn patterns. Emit addr32 prefix when
5485 Pmode != word_mode. Update insn length attribute.
5486 * config/i386/i386.c (ix86_option_override_internal): Update
5487 ix86_gen_monitor selection for merged sse3_monitor insn.
5489 2013-08-14 Jakub Jelinek <jakub@redhat.com>
5490 Alexandre Oliva <aoliva@redhat.com>
5493 * config/i386/i386.c (ix86_delegitimize_address): For CM_MEDIUM_PIC
5494 and CM_LARGE_PIC ix86_cmodel fall thru into the -m32 code, handle
5495 there also UNSPEC_PLTOFF.
5497 2013-08-13 Jakub Jelinek <jakub@redhat.com>
5500 * asan.c (instrument_strlen_call): Fix typo in comment.
5501 Use char * type even for the lhs of POINTER_PLUS_EXPR.
5503 2013-08-13 Vladimir Makarov <vmakarov@redhat.com>
5505 Backport from mainline
5506 2013-06-06 Vladimir Makarov <vmakarov@redhat.com>
5508 PR rtl-optimization/57459
5509 * lra-constraints.c (update_ebb_live_info): Fix typo for operand
5510 type when setting live regs.
5512 2013-08-13 Marek Polacek <polacek@redhat.com>
5513 Jakub Jelinek <jakub@redhat.com>
5515 PR tree-optimization/57980
5516 * tree-tailcall.c (process_assignment): Return false
5517 when not dealing with integers or floats.
5519 2013-08-12 Andrew Haley <aph@redhat.com>
5521 Backport from mainline:
5522 * 2013-07-11 Andreas Schwab <schwab@suse.de>
5524 * config/aarch64/aarch64-linux.h (CPP_SPEC): Define.
5526 2013-08-13 Uros Bizjak <ubizjak@gmail.com>
5528 Backport from mainline
5529 2013-08-12 Perez Read <netfirewall@gmail.com>
5532 * config/i386/i386.md (*movabs<mode>_1): Add <ptrsize> PTR before
5533 operand 0 for intel asm alternative.
5534 (*movabs<mode>_2): Ditto for operand 1.
5536 2013-08-09 Zhenqiang Chen <zhenqiang.chen@linaro.org>
5538 Backport from mainline:
5539 2013-08-09 Zhenqiang Chen <zhenqiang.chen@linaro.org>
5541 * config/arm/neon.md (vcond): Fix floating-point vector
5542 comparisons against 0.
5544 2013-08-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5546 Backport from mainline:
5547 2013-08-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5549 * config/arm/neon.md (movmisalign<mode>): Disable when we
5550 don't allow unaligned accesses.
5551 (*movmisalign<mode>_neon_store): Likewise.
5552 (*movmisalign<mode>_neon_load): Likewise.
5553 (*movmisalign<mode>_neon_store): Likewise.
5554 (*movmisalign<mode>_neon_load): Likewise.
5556 2013-08-06 Martin Jambor <mjambor@suse.cz>
5559 * gimple-ssa-strength-reduction.c (replace_ref): Make sure built
5560 MEM_REF has proper alignment information.
5562 2013-08-05 Richard Earnshaw <rearnsha@arm.com>
5564 PR rtl-optimization/57708
5565 * recog.c (peep2_find_free_register): Validate all regs in a
5568 2013-08-02 Eric Botcazou <ebotcazou@adacore.com>
5570 * config/sparc/sparc.c (sparc_emit_membar_for_model) <SMM_TSO>: Add
5571 the implied StoreLoad barrier for atomic operations if before.
5573 2013-08-02 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5575 Backports from mainline:
5576 2013-06-27 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5578 * config/s390/s390.c: Rename UNSPEC_CCU_TO_INT to
5579 UNSPEC_STRCMPCC_TO_INT and UNSPEC_CCZ_TO_INT to UNSPEC_CC_TO_INT.
5580 (struct machine_function): Add tbegin_p.
5581 (s390_canonicalize_comparison): Fold CC mode compares to
5582 conditional jump if possible.
5583 (s390_emit_jump): Return the emitted jump.
5584 (s390_branch_condition_mask, s390_branch_condition_mnemonic):
5585 Handle CCRAWmode compares.
5586 (s390_option_override): Default to -mhtm if available.
5587 (s390_reg_clobbered_rtx): Handle floating point regs as well.
5588 (s390_regs_ever_clobbered): Use s390_regs_ever_clobbered also for
5589 FPRs instead of df_regs_ever_live_p.
5590 (s390_optimize_nonescaping_tx): New function.
5591 (s390_init_frame_layout): Extend clobbered_regs array to cover
5593 (s390_emit_prologue): Call s390_optimize_nonescaping_tx.
5594 (s390_expand_tbegin): New function.
5595 (enum s390_builtin): New enum definition.
5596 (code_for_builtin): New array definition.
5597 (s390_init_builtins): New function.
5598 (s390_expand_builtin): New function.
5599 (TARGET_INIT_BUILTINS): Define.
5600 (TARGET_EXPAND_BUILTIN): Define.
5601 * common/config/s390/s390-common.c (processor_flags_table): Add PF_TX.
5602 * config/s390/predicates.md (s390_comparison): Handle CCRAWmode.
5603 (s390_alc_comparison): Likewise.
5604 * config/s390/s390-modes.def: Add CCRAWmode.
5605 * config/s390/s390.h (processor_flags): Add PF_TX.
5606 (TARGET_CPU_HTM): Define macro.
5607 (TARGET_HTM): Define macro.
5608 (TARGET_CPU_CPP_BUILTINS): Define __HTM__ for htm.
5609 * config/s390/s390.md: Rename UNSPEC_CCU_TO_INT to
5610 UNSPEC_STRCMPCC_TO_INT and UNSPEC_CCZ_TO_INT to UNSPEC_CC_TO_INT.
5611 (UNSPECV_TBEGIN, UNSPECV_TBEGINC, UNSPECV_TEND, UNSPECV_TABORT)
5612 (UNSPECV_ETND, UNSPECV_NTSTG, UNSPECV_PPA): New unspecv enum
5614 (TBEGIN_MASK, TBEGINC_MASK): New constants.
5615 ("*cc_to_int"): Move up.
5616 ("*mov<mode>cc", "*cjump_64", "*cjump_31"): Accept integer
5617 constants other than 0.
5618 ("*ccraw_to_int"): New insn and splitter definition.
5619 ("tbegin", "tbegin_nofloat", "tbegin_retry")
5620 ("tbegin_retry_nofloat", "tbeginc", "tend", "tabort")
5621 ("tx_assist"): New expander.
5622 ("tbegin_1", "tbegin_nofloat_1", "*tbeginc_1", "*tend_1")
5623 ("*tabort_1", "etnd", "ntstg", "*ppa"): New insn definition.
5624 * config/s390/s390.opt: Add -mhtm option.
5625 * config/s390/s390-protos.h (s390_emit_jump): Add return type.
5626 * config/s390/htmxlintrin.h: New file.
5627 * config/s390/htmintrin.h: New file.
5628 * config/s390/s390intrin.h: New file.
5629 * doc/extend.texi: Document htm builtins.
5630 * config.gcc: Add the new header files to extra_headers.
5632 2013-07-17 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5634 * config/s390/s390.c: (s390_expand_builtin): Allow -mhtm to be
5635 enabled without -march=zEC12.
5636 * config/s390/s390.h (TARGET_HTM): Do not require EC12 machine
5639 2013-08-01 Ganesh Gopalasubramanian <Ganesh.Gopalasubramanian@amd.com>
5641 Backport from mainline
5642 2013-05-13 Ganesh Gopalasubramanian
5643 <Ganesh.Gopalasubramanian@amd.com>
5645 * config/i386/i386.c (processor_target_table): Modified default
5646 alignment values for AMD BD and BT architectures.
5648 2013-07-31 Sriraman Tallam <tmsriram@google.com>
5650 * config/i386/i386.c (dispatch_function_versions): Fix array
5651 indexing of function_version_info to match actual_versions.
5653 2013-07-31 Sebastian Huber <sebastian.huber@embedded-brains.de>
5655 * config.gcc (*-*-rtems*): Use __cxa_atexit by default.
5656 * config/rs6000/rtems.h (TARGET_LIBGCC_SDATA_SECTION): Define.
5658 2013-07-31 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5660 Backport from mainline
5661 2013-03-27 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5663 * config/s390/s390.h (TARGET_FLT_EVAL_METHOD): Define.
5665 2013-07-23 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5667 * config/s390/linux-unwind.h: Use the proper dwarf to hard reg
5668 mapping for FPRs when creating the fallback framestate.
5670 2013-07-29 Dominik Vogt <vogt@linux.vnet.ibm.com>
5672 * config/s390/s390.md ("movcc"): Swap load and store instructions.
5674 2013-07-25 Terry Guo <terry.guo@arm.com>
5676 Backport from mainline:
5677 2013-07-25 Terry Guo <terry.guo@arm.com>
5679 * config/arm/arm.c (thumb1_size_rtx_costs): Assign proper cost for
5680 shift_add/shift_sub0/shift_sub1 RTXs.
5682 2013-07-22 Iain Sandoe <iain@codesourcery.com>
5684 Backport from mainline:
5685 2013-07-22 Uros Bizjak <ubizjak@gmail.com>
5687 * config/i386/i386.md (nonlocal_goto_receiver): Delete insn if
5688 it is not needed after split.
5690 2013-07-20 Iain Sandoe <iain@codesourcery.com>
5693 * config/i386/i386.c (output_set_got) [TARGET_MACHO]: Adjust to emit a
5694 second label for nonlocal goto receivers. Don't output pic base labels
5695 unless we're producing PIC; mark that action unreachable().
5696 (ix86_save_reg): If the function contains a nonlocal label, save the
5698 * config/darwin-protos.h (machopic_should_output_picbase_label): New.
5699 * gcc/config/darwin.c (emitted_pic_label_num): New GTY.
5700 (update_pic_label_number_if_needed): New.
5701 (machopic_output_function_base_name): Adjust for nonlocal receiver
5703 (machopic_should_output_picbase_label): New.
5704 * config/i386/i386.md (enum unspecv): UNSPECV_NLGR: New.
5705 (nonlocal_goto_receiver): New insn and split.
5707 2013-07-19 Wei Mi <wmi@google.com>
5709 Backport from mainline:
5710 2013-07-18 Vladimir Makarov <vmakarov@redhat.com>
5711 Wei Mi <wmi@google.com>
5713 PR rtl-optimization/57878
5714 * lra-assigns.c (assign_by_spills): Move non_reload_pseudos to the
5715 top. Promote lra_assert to gcc_assert.
5716 (reload_pseudo_compare_func): Check regs first for reload pseudos.
5718 2013-07-11 Georg-Johann Lay <avr@gjlay.de>
5720 Backport from 2013-07-19 trunk r201051.
5723 * config/avr/avr-fixed.md (round<mode>3_const): Turn expander to insn.
5724 * config/avr/avr.md (adjust_len): Add `round'.
5725 * config/avr/avr-protos.h (avr_out_round): New prototype.
5726 (avr_out_plus): Add `out_label' argument.
5727 * config/avr/avr.c (avr_out_plus_1): Add `out_label' argument.
5728 (avr_out_plus): Pass down `out_label' to avr_out_plus_1.
5729 Handle the case where `insn' is just a pattern.
5730 (avr_out_bitop): Handle the case where `insn' is just a pattern.
5731 (avr_out_round): New function.
5732 (avr_adjust_insn_length): Handle ADJUST_LEN_ROUND.
5734 2013-07-19 Kirill Yukhin <kirill.yukhin@intel.com>
5736 * config/i386/bmiintrin.h (_bextr_u32): New.
5737 (_bextr_u64): Ditto.
5742 (_blsmsk_u32): Ditto.
5743 (_blsmsk_u64): Ditto.
5744 (_tzcnt_u32): Ditto.
5745 (_tzcnt_u64): Ditto.
5747 2013-07-17 James Greenhalgh <james.greenhalgh@arm.com>
5749 Backport From mainline:
5750 2013-07-03 James Greenhalgh <james.greenhalgh@arm.com>
5752 * config/aarch64/aarch64-builtins.c
5753 (aarch64_simd_expand_builtin): Handle AARCH64_SIMD_STORE1.
5754 * config/aarch64/aarch64-simd-builtins.def (ld1): New.
5756 * config/aarch64/aarch64-simd.md
5757 (aarch64_ld1<VALL:mode>): New.
5758 (aarch64_st1<VALL:mode>): Likewise.
5759 * config/aarch64/arm_neon.h
5760 (vld1<q>_<fpsu><8, 16, 32, 64>): Convert to RTL builtins.
5762 2013-07-11 Georg-Johann Lay <avr@gjlay.de>
5764 Backport from 2013-07-11 trunk r200901.
5767 * config/avr/avr.c (avr_set_current_function): Sanity-check signal
5768 name seen by assembler/linker if available.
5770 2013-07-10 Georg-Johann Lay <avr@gjlay.de>
5772 Backport from 2013-07-10 trunk r200872.
5775 * config/avr/avr.c (avr_prologue_setup_frame): Trunk -size to mode
5778 2013-07-10 Georg-Johann Lay <avr@gjlay.de>
5780 Backport from 2013-07-10 trunk r200870.
5783 * config/avr/avr-mcus.def (atmega16hva, atmega16hva2, atmega16hvb)
5784 (atmega16m1, atmega16u4, atmega32a, atmega32c1, atmega32hvb)
5785 (atmega32m1, atmega32u4, atmega32u6, atmega64c1, atmega64m1):
5786 Remove duplicate devices.
5787 * config/avr/gen-avr-mmcu-texi.c (print_mcus): Fail on duplicate MCUs.
5788 * config/avr/t-multilib: Regenerate.
5789 * config/avr/avr-tables.opt: Regenerate.
5790 * doc/avr-mmcu.texi: Regenerate.
5792 2013-07-10 Georg-Johann Lay <avr@gjlay.de>
5795 * config/avr/avr.opt (Waddr-space-convert): Fix typo.
5797 2013-07-09 Joseph Myers <joseph@codesourcery.com>
5799 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Only
5800 adjust register size for TDmode and TFmode for VSX registers.
5802 2013-07-08 Kai Tietz <ktietz@redhat.com>
5804 Backport from mainline
5806 * config/i386/i386.c (TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P): Define as
5807 hook_bool_const_tree_true.
5809 2013-07-08 Uros Bizjak <ubizjak@gmail.com>
5811 Backport from mainline
5812 2013-07-07 Uros Bizjak <ubizjak@gmail.com>
5814 * config/i386/driver-i386.c (host_detect_local_cpu): Do not check
5815 signature_TM2_ebx, it interferes with signature_INTEL_ebx.
5817 Backport from mainline
5818 2013-07-06 Uros Bizjak <ubizjak@gmail.com>
5820 * config/i386/sse.md (sse_movlhps): Change alternative 3
5821 of operand 2 to "m".
5823 2013-07-08 Eric Botcazou <ebotcazou@adacore.com>
5825 * Makefile.in (tree-ssa-reassoc.o): Add dependency on $(PARAMS_H).
5827 2013-07-08 Jakub Jelinek <jakub@redhat.com>
5829 PR rtl-optimization/57829
5830 * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Ensure that
5831 mask bits outside of mode are just sign-extension from mode to HWI.
5833 2013-07-03 Jakub Jelinek <jakub@redhat.com>
5836 * config/i386/predicates.md (vsib_address_operand): Disallow
5837 SYMBOL_REF or LABEL_REF in parts.disp if TARGET_64BIT && flag_pic.
5839 2013-06-30 Terry Guo <terry.guo@arm.com>
5841 Backport from mainline
5842 2013-03-27 Bin Cheng <bin.cheng@arm.com>
5845 * config/arm/arm.c (thumb1_rtx_costs, thumb1_size_rtx_costs): Fix
5846 rtx costs for SET/ASHIFT/ASHIFTRT/LSHIFTRT/ROTATERT patterns with
5849 2013-06-28 Jakub Jelinek <jakub@redhat.com>
5852 * config/i386/i386.c (ix86_expand_builtin): If target == NULL
5853 and mode is VOIDmode, don't create a VOIDmode pseudo to copy result
5856 2013-06-27 Jakub Jelinek <jakub@redhat.com>
5859 * config/i386/i386.md (bmi_bextr_<mode>): Swap predicates and
5860 constraints of operand 1 and 2.
5863 * config/i386/i386.md (bmi2_bzhi_<mode>3): Swap AND arguments
5864 to match RTL canonicalization. Swap predicates and
5865 constraints of operand 1 and 2.
5867 * tree-vect-stmts.c (vectorizable_store): Move ptr_incr var
5868 decl before the loop, initialize to NULL.
5869 (vectorizable_load): Initialize ptr_incr to NULL.
5871 2013-06-24 Martin Jambor <mjambor@suse.cz>
5873 PR tree-optimization/57358
5874 * ipa-prop.c (parm_ref_data_preserved_p): Always return true when
5877 2013-06-24 Alan Modra <amodra@gmail.com>
5879 * config/rs6000/rs6000.c (vspltis_constant): Correct for little-endian.
5880 (gen_easy_altivec_constant): Likewise.
5881 * config/rs6000/predicates.md (easy_vector_constant_add_self,
5882 easy_vector_constant_msb): Likewise.
5884 2013-06-21 Uros Bizjak <ubizjak@gmail.com>
5886 Backport from mainline
5887 2013-06-20 Uros Bizjak <ubizjak@gmail.com>
5890 * config/i386/i386.c (construct_container): Report error if
5891 long double is used with disabled x87 float returns.
5893 2013-06-20 Wei Mi <wmi@google.com>
5895 Backport from mainline
5896 2013-06-19 Wei Mi <wmi@google.com>
5898 PR rtl-optimization/57518
5899 * ira.c (set_paradoxical_subreg): Set pdx_subregs[regno]
5900 if regno is used in paradoxical subreg.
5901 (update_equiv_regs): Check pdx_subregs[regno] before
5902 set a reg to be equivalent with a mem.
5905 2013-06-20 David Edelsohn <dje.gcc@gmail.com>
5907 Backport from mainline
5908 2013-06-19 David Edelsohn <dje.gcc@gmail.com>
5911 * collect2.c (collect_atexit): New.
5912 (collect_exit): Delete.
5913 (main): Register collect_atexit with atexit.
5914 (collect_wait): Change collect_exit to exit.
5916 * collect2.h (collect_exit): Delete.
5917 * tlink.c (do_tlink): Rename exit to ret. Change collect_exit to exit.
5919 2013-06-19 Matthias Klose <doko@ubuntu.com>
5922 * file-find.h (find_a_file): Add a mode parameter.
5923 * file-find.c (find_a_file): Likewise.
5924 * gcc-ar.c (main): Call find_a_file with R_OK for the plugin,
5925 with X_OK for the executables.
5926 * collect2.c (main): Call find_a_file with X_OK.
5928 2013-06-19 Igor Zamyatin <igor.zamyatin@intel.com>
5930 * doc/invoke.texi (core-avx2): Document.
5931 (atom): Updated with MOVBE.
5933 2013-06-19 Jakub Jelinek <jakub@redhat.com>
5936 * gcc-ar.c (main): If not CROSS_DIRECTORY_STRUCTURE, look for
5937 PERSONALITY in $PATH derived prefixes.
5939 2013-06-19 Paolo Carlini <paolo.carlini@oracle.com>
5942 * doc/cpp.texi [Standard Predefined Macros, __cplusplus]: Document
5943 that now in C++ the value is correct per the C++ standards.
5945 2013-06-19 Alan Modra <amodra@gmail.com>
5947 Apply mainline patches
5948 2013-06-13 Alan Modra <amodra@gmail.com>
5949 * config/rs6000/rs6000.h (LONG_DOUBLE_LARGE_FIRST): Define.
5950 * config/rs6000/rs6000.md (signbittf2): New insn.
5951 (extenddftf2_internal): Use LONG_DOUBLE_LARGE_FIRST.
5952 (abstf2_internal, cmptf_internal2): Likewise.
5953 * config/rs6000/spe.md (spe_abstf2_cmp, spe_abstf2_tst): Likewise.
5955 2013-06-11 Anton Blanchard <anton@samba.org>
5956 * config/rs6000/rs6000.c (rs6000_adjust_atomic_subword): Calculate
5957 correct shift value in little-endian mode.
5959 2013-06-07 Alan Modra <amodra@gmail.com>
5960 * config/rs6000/rs6000.c (setup_incoming_varargs): Round up
5963 2013-06-04 Alan Modra <amodra@gmail.com>
5964 * config/rs6000/rs6000.c (output_toc): Correct little-endian float
5967 2013-05-10 Alan Modra <amodra@gmail.com>
5968 * configure.ac (HAVE_AS_TLS): Swap powerpc64 and powerpc cases.
5969 (HAVE_LD_LARGE_TOC): Don't mention AIX in help text.
5970 * configure: Regenerate.
5972 2013-05-09 Alan Modra <amodra@gmail.com>
5973 * configure.ac (HAVE_AS_TLS): Enable tests for powerpcle and
5975 * configure: Regenerate.
5977 2013-05-07 Anton Blanchard <anton@samba.org>
5978 * configure.ac (HAVE_LD_LARGE_TOC): Use right linker emulation
5979 for powerpc64 little endian.
5980 * configure: Regenerate.
5982 2013-05-06 Alan Modra <amodra@gmail.com>
5983 * config/rs6000/linux.h (DEFAULT_ASM_ENDIAN): Define.
5984 (LINK_OS_LINUX_EMUL): Use ENDIAN_SELECT.
5985 * config/rs6000/linux64.h (DEFAULT_ASM_ENDIAN): Define.
5986 * config/rs6000/sysv4le.h (DEFAULT_ASM_ENDIAN): Define.
5987 (LINK_TARGET_SPEC): Use ENDIAN_SELECT.
5988 * config/rs6000/sysv4.h (DEFAULT_ASM_ENDIAN): Define as -mbig.
5990 2013-05-06 Alan Modra <amodra@gmail.com>
5991 * config/rs6000/sysv4.h (ENDIAN_SELECT): Define, extracted from
5992 (ASM_SPEC): ..here. Emit DEFAULT_ASM_ENDIAN too.
5993 (DEFAULT_ASM_ENDIAN): Define.
5994 (CC1_SPEC, LINK_TARGET_SPEC): Use ENDIAN_SELECT.
5995 * config/rs6000/linux64.h (ASM_SPEC32): Remove endian options.
5996 Update -K PIC clause from sysv4.h.
5997 (ASM_SPEC_COMMON): Use ENDIAN_SELECT.
5998 (LINK_OS_LINUX_EMUL32, LINK_OS_LINUX_EMUL64): Likewise.
6000 2013-05-06 Alan Modra <amodra@gmail.com>
6001 * config/rs6000/rs6000.md (bswapdi 2nd splitter): Don't swap words
6002 twice for little-endian.
6003 (ashrdi3_no_power, ashrdi3): Support little-endian.
6005 2013-04-25 Alan Modra <amodra@gmail.com>
6006 * config.gcc: Support little-endian powerpc-linux targets.
6007 * config/rs6000/linux.h (LINK_OS_LINUX_EMUL): Define.
6008 (LINK_OS_LINUX_SPEC): Define.
6009 * config/rs6000/linuxspe.h (TARGET_DEFAULT):
6010 Preserve MASK_LITTLE_ENDIAN.
6011 * config/rs6000/default64.h (TARGET_DEFAULT): Likewise.
6012 * config/rs6000/linuxaltivec.h (TARGET_DEFAULT): Likewise.
6013 * config/rs6000/linux64.h (OPTION_LITTLE_ENDIAN): Don't zero.
6014 (LINK_OS_LINUX_EMUL32, LINK_OS_LINUX_EMUL64): Define.
6015 (LINK_OS_LINUX_SPEC32, LINK_OS_LINUX_SPEC64): Use above.
6016 * config/rs6000/rs6000.c (output_toc): Don't use .tc for TARGET_ELF.
6017 Correct fp word order for little-endian. Don't shift toc entries
6018 smaller than a word for little-endian.
6019 * config/rs6000/rs6000.md (bswaphi2, bswapsi2 split): Comment.
6020 (bswapdi2 splits): Correct low-part subreg for little-endian.
6021 Remove wrong BYTES_BIG_ENDIAN tests, and rename vars to remove
6022 low/high where such is correct only for be.
6023 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Allow
6024 little-endian for -mcall-aixdesc.
6026 2013-06-12 Martin Jambor <mjambor@suse.cz>
6028 * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that param_index is
6029 within bounds at the beginning of the function.
6031 2013-06-12 Jakub Jelinek <jakub@redhat.com>
6033 PR tree-optimization/57537
6034 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): If
6035 vect_handle_widen_op_by_const, convert oprnd1 to half_type1.
6037 2013-06-10 Uros Bizjak <ubizjak@gmail.com>
6039 Backport from mainline
6040 2013-06-10 Uros Bizjak <ubizjak@gmail.com>
6042 * config/alpha/alpha.c (alpha_emit_xfloating_compare): Also use
6043 cmp_code to construct REG_EQUAL note.
6045 2013-06-10 Oleg Endo <olegendo@gcc.gnu.org>
6047 Backport from mainline
6048 2013-05-20 Oleg Endo <olegendo@gcc.gnu.org>
6051 * config/sh/sh.md (fmasf4): Remove empty constraints strings.
6052 (*fmasf4, *fmasf4_media): New insns.
6054 2013-06-09 Jakub Jelinek <jakub@redhat.com>
6057 * config/i386/i386.md (TARGET_READ_MODIFY_WRITE peepholes): Ensure
6058 that operands[2] doesn't overlap with operands[0].
6060 2013-06-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
6062 * recog.c (offsettable_address_addr_space_p): Fix calculation of
6063 address mode. Move pointer mode initialization to the same place.
6065 2013-06-07 Sofiane Naci <sofiane.naci@arm.com>
6067 Backport from mainline
6068 * config/aarch64/aarch64.md (*movdi_aarch64): Define "simd" attribute.
6070 2013-06-07 Uros Bizjak <ubizjak@gmail.com>
6072 Backport from mainline
6073 2013-06-05 Uros Bizjak <ubizjak@gmail.com>
6075 * config/alpha/alpha.c (alpha_emit_conditional_move): Swap all
6076 GE, GT, GEU and GTU compares, modulo DImode compares with zero.
6078 Backport from mainline
6079 2013-05-23 Uros Bizjak <ubizjak@gmail.com>
6082 * config/alpha/alpha.md (unspec): Add UNSPEC_XFLT_COMPARE.
6083 * config/alpha/alpha.c (alpha_emit_xfloating_compare): Construct
6084 REG_EQUAL note as UNSPEC_XFLT_COMPARE unspec.
6086 2013-06-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6088 Backport from mainline.
6089 2013-05-22 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6091 * config/rs6000/rs6000.h (MALLOC_ABI_ALIGNMENT): New #define.
6093 2013-06-03 James Greenhalgh <james.greenhalgh@arm.com>
6095 Backport from mainline.
6096 2013-04-25 James Greenhalgh <james.greenhalgh@arm.com>
6098 * config/aarch64/aarch64.c (aarch64_print_operand): Fix asm_fprintf
6099 format specifier in 'X' case.
6101 2013-05-31 Richard Henderson <rth@redhat.com>
6104 * config/i386/i386.c (ix86_seh_fixup_eh_fallthru): New.
6105 (ix86_reorg): Call it.
6107 2012-05-31 Jakub Jelinek <jakub@redhat.com>
6109 * BASE-VER: Set to 4.8.2.
6110 * DEV-PHASE: Set to prerelease.
6112 2013-05-31 Release Manager
6114 * GCC 4.8.1 released.
6116 2013-05-24 Greta Yorsh <Greta.Yorsh@arm.com>
6118 Backport from mainline
6119 2013-05-02 Greta Yorsh <Greta.Yorsh@arm.com>
6122 * config/arm/arm.c (arm_expand_epilogue): Check really_return before
6123 generating simple_return for naked functions.
6125 2013-05-24 Alexander Ivchenko <alexander.ivchenko@intel.com>
6128 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Check
6129 that index is not negative.
6131 2013-05-23 Martin Jambor <mjambor@suse.cz>
6134 * tree.h (contains_bitfld_component_ref_p): Declare.
6135 * tree-sra.c (contains_bitfld_comp_ref_p): Move...
6136 * tree.c (contains_bitfld_component_ref_p): ...here. Adjust its caller.
6137 * ipa-prop.c (determine_known_aggregate_parts): Check that LHS does
6138 not access a bit-field. Assert all final offsets are byte-aligned.
6140 2013-05-23 Richard Biener <rguenther@suse.de>
6142 PR rtl-optimization/57341
6143 * ira.c (validate_equiv_mem_from_store): Use anti_dependence
6144 instead of true_dependence.
6146 2013-05-23 Jakub Jelinek <jakub@redhat.com>
6149 * expmed.c (store_split_bit_field): If op0 is a REG or
6150 SUBREG of a REG, don't lower unit. Handle unit not being
6151 always BITS_PER_WORD.
6153 2013-05-22 Uros Bizjak <ubizjak@gmail.com>
6156 * config/i386/i386.md (*movti_internal_rex64): Emit movaps/movups
6157 for non-sse2 targets.
6158 (*movti_internal): Simplify mode attribute calculation.
6160 2013-05-22 Richard Biener <rguenther@suse.de>
6162 Backport from mainline
6163 2013-05-21 Richard Biener <rguenther@suse.de>
6165 PR tree-optimization/57318
6166 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Do not
6167 estimate stmts with side-effects as likely eliminated.
6169 2013-05-21 Richard Biener <rguenther@suse.de>
6171 PR tree-optimization/57330
6172 * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Properly
6173 preserve the call stmts fntype.
6175 2013-05-21 Richard Biener <rguenther@suse.de>
6177 PR tree-optimization/57303
6178 * tree-ssa-sink.c (statement_sink_location): Properly handle
6181 2013-05-21 Magnus Granberg <baldrick@free.fr>
6184 * Makefile.in (PLUGIN_HEADERS): Add $(TARGET_H).
6186 2013-05-21 Eric Botcazou <ebotcazou@adacore.com>
6188 Backport from mainline
6189 2013-05-14 Eric Botcazou <ebotcazou@adacore.com>
6191 * config/sparc/sp64-elf.h (CPP_SUBTARGET_SPEC): Delete.
6192 * config/sparc/openbsd64.h (CPP_SUBTARGET_SPEC): Likewise.
6194 2013-05-17 Jakub Jelinek <jakub@redhat.com>
6196 PR rtl-optimization/57281
6197 PR rtl-optimization/57300
6198 * config/i386/i386.md (extendsidi2_1 dead reg splitter): Remove.
6199 (extendsidi2_1 peephole2s): Add instead 2 new peephole2s, that undo
6200 what the other splitter did if the registers are dead.
6202 2013-05-17 Uros Bizjak <ubizjak@gmail.com>
6204 Backport from mainline
6205 2013-05-16 Uros Bizjak <ubizjak@gmail.com>
6207 * config/i386/driver-i386.c (host_detect_local_cpu): Determine
6208 cache parameters using detect_caches_amd also for CYRIX,
6209 NSC and TM2 signatures.
6211 2013-05-16 Uros Bizjak <ubizjak@gmail.com>
6212 Dzianis Kahanovich <mahatma@eu.by>
6216 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6217 VIA/Centaur processors and determine their cache parameters
6218 using detect_caches_amd.
6220 2013-05-15 Uros Bizjak <ubizjak@gmail.com>
6222 * config/i386/i386.c (ix86_option_override_internal): Update
6223 processor_alias_table for missing PTA_PRFCHW and PTA_FXSR flags. Add
6224 PTA_POPCNT to corei7 entry. Do not enable SSE prefetch on
6225 non-SSE 3dNow! targets. Enable TARGET_PRFCHW for TARGET_3DNOW targets.
6226 * config/i386/i386.md (prefetch): Enable for TARGET_PRFCHW instead
6228 (*prefetch_3dnow): Enable for TARGET_PRFCHW only.
6230 2013-05-17 Jakub Jelinek <jakub@redhat.com>
6232 * gcc.c (SANITIZER_SPEC): Reject -fsanitize=address -fsanitize=thread
6235 PR tree-optimization/57051
6236 * fold-const.c (const_binop) <case VEC_LSHIFT_EXPR,
6237 case VEC_RSHIFT_EXPR>: Fix BYTES_BIG_ENDIAN handling.
6239 2013-05-16 Jakub Jelinek <jakub@redhat.com>
6241 * omp-low.c (extract_omp_for_data): For collapsed loops,
6242 if at least one of the loops is known at compile time to
6243 iterate zero times, set count to 0.
6244 (expand_omp_regimplify_p): New function.
6245 (expand_omp_for_generic): For collapsed loops, if at least
6246 one of the loops isn't known to iterate at least once,
6247 add runtime check with setting count to 0.
6248 (expand_omp_for_static_nochunk, expand_omp_for_static_chunk):
6249 For unsigned types if it isn't known at compile time that
6250 the loop will iterate at least once, add runtime check to bypass
6251 the whole loop if initial condition isn't true.
6253 2013-05-14 Jakub Jelinek <jakub@redhat.com>
6256 * expr.c (expand_expr_real_2) <case WIDEN_MULT_EXPR>: Handle
6257 the case when both op0 and op1 have VOIDmode.
6259 2013-05-13 Jakub Jelinek <jakub@redhat.com>
6261 PR tree-optimization/57230
6262 * tree-ssa-strlen.c (handle_char_store): Add missing integer_zerop
6265 2013-05-12 Joern Rennecke <joern.rennecke@embecosm.com>
6267 * config/epiphany/epiphany.c (epiphany_init): Check size of
6268 NUM_MODES_FOR_MODE_SWITCHING.
6269 (epiphany_expand_prologue):
6270 Remove CONFIG_REGNUM initial value handling code.
6271 (epiphany_optimize_mode_switching): Handle EPIPHANY_MSW_ENTITY_CONFIG.
6272 (epiphany_mode_needed, epiphany_mode_entry_exit): Likewise.
6273 (emit_set_fp_mode, epiphany_mode_after): Likewise.
6274 (epiphany_mode_needed) <Handle EPIPHANY_MSW_ENTITY_AND>:
6275 Don't return 1 for FP_MODE_NONE.
6276 * config/epiphany/epiphany.h (NUM_MODES_FOR_MODE_SWITCHING):
6277 Add value for EPIPHANY_MSW_ENTITY_CONFIG.
6278 (EPIPHANY_MSW_ENTITY_CONFIG, EPIPHANY_MSW_ENTITY_NUM): Define.
6279 * config/epiphany/epiphany.md (save_config): New pattern.
6281 2013-05-10 Sebastian Huber <sebastian.huber@embedded-brains.de>
6283 * config/arm/t-rtems-eabi: Remove mthumb/march=armv7 multilib.
6284 Add mthumb/march=armv7-a multilib.
6285 Add mthumb/march=armv7-r multilib.
6286 Add mthumb/march=armv7-a/mfpu=neon/mfloat-abi=hard multilib.
6288 2013-05-10 Ralf Corsépius <ralf.corsepius@rtems.org>
6291 * config/v850/t-rtems: Add more multilibs.
6293 2013-05-10 Richard Biener <rguenther@suse.de>
6295 PR tree-optimization/57214
6296 * tree-ssa-loop-ivcanon.c (propagate_constants_for_unrolling): Do
6297 not propagate from SSA names that occur in abnormal PHI nodes.
6299 2013-05-10 Alan Modra <amodra@gmail.com>
6302 * varasm.c (default_elf_select_section): Move !DECL_P check..
6303 (get_named_section): ..to here before calling get_section_name.
6305 (default_section_type_flags): Add DECL_P check.
6306 * config/i386/winnt.c (i386_pe_section_type_flags): Likewise.
6307 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags): Likewise.
6309 2013-05-09 Joern Rennecke <joern.rennecke@embecosm.com>
6311 * config/epiphany/epiphany.c (epiphany_expand_prologue):
6312 When using gen_stack_adjust_str with a register offset, add a
6313 REG_FRAME_RELATED_EXPR note.
6315 2013-05-09 Martin Jambor <mjambor@suse.cz>
6318 * ipa-prop.h (ipa_agg_replacement_value): New flag by_ref.
6319 * ipa-cp.c (find_aggregate_values_for_callers_subset): Fill in the
6320 by_ref flag of ipa_agg_replacement_value structures.
6321 (known_aggs_to_agg_replacement_list): Likewise.
6322 * ipa-prop.c (write_agg_replacement_chain): Stream by_ref flag.
6323 (read_agg_replacement_chain): Likewise.
6324 (ipcp_transform_function): Also check that by_ref flags match.
6326 2013-05-08 Diego Novillo <dnovillo@google.com>
6332 2012-08-17 Diego Novillo <dnovillo@google.com>
6335 * configure.ac: Add libintl.h to AC_CHECK_HEADERS list.
6336 * config.in: Regenerate.
6337 * configure: Regenerate.
6338 * intl.h: Always include libintl.h if HAVE_LIBINTL_H is
6341 2013-05-08 Paolo Carlini <paolo.carlini@oracle.com>
6343 PR tree-optimization/57200
6344 * tree-ssa-loop-niter.c (do_warn_aggressive_loop_optimizations):
6345 Only call inform if the preceding warning_at returns true.
6347 2013-05-07 Jakub Jelinek <jakub@redhat.com>
6349 PR tree-optimization/57149
6350 * tree-ssa-uninit.c (uninit_undefined_value_p): New inline.
6351 (can_skip_redundant_opnd, compute_uninit_opnds_pos,
6352 collect_phi_def_edges, execute_late_warn_uninitialized): Use
6353 uninit_undefined_value_p instead of ssa_undefined_value_p.
6356 * expr.c (expand_expr_addr_expr_1): Handle COMPOUND_LITERAL_EXPR
6357 for modifier == EXPAND_INITIALIZER.
6359 2013-05-07 Richard Biener <rguenther@suse.de>
6361 Backport from mainline
6362 2013-05-06 Richard Biener <rguenther@suse.de>
6364 PR tree-optimization/57185
6365 * tree-parloops.c (add_field_for_reduction): Handle anonymous
6368 2013-04-19 Richard Biener <rguenther@suse.de>
6370 PR tree-optimization/57000
6371 * tree-ssa-reassoc.c (pass_reassoc): Add TODO_update_ssa_only_virtuals.
6373 2013-05-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6376 2013-05-03 Michael Meissner <meissner@linux.vnet.ibm.com>
6379 * config/rs6000/rs6000.h (HARD_REGNO_CALLER_SAVE_MODE): Use DFmode
6380 to save TFmode registers and DImode to save TImode registers for
6381 caller save operations.
6382 (HARD_REGNO_CALL_PART_CLOBBERED): TFmode and TDmode do not need to
6383 mark being partially clobbered since they only use the first
6386 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): TFmode
6387 and TDmode only use the upper 64-bits of each VSX register.
6389 2013-05-06 Oleg Endo <olegendo@gcc.gnu.org>
6392 * config/sh/sh.md (tstsi_t_zero_extract_eq): Use QIHISIDI mode iterator.
6394 2013-05-06 Uros Bizjak <ubizjak@gmail.com>
6396 Backport from mainline
6397 2013-05-06 Uros Bizjak <ubizjak@gmail.com>
6400 * config/i386/i386.c (add_parameter_dependencies): Add dependence
6401 between "first_arg" and "insn", not "last" and "insn".
6403 2013-05-03 Jakub Jelinek <jakub@redhat.com>
6405 PR rtl-optimization/57130
6406 * combine.c (make_compound_operation) <case SUBREG>: Pass
6407 SET instead of COMPARE as in_code to the recursive call
6410 Backported from mainline
6411 2013-04-26 Jakub Jelinek <jakub@redhat.com>
6413 PR tree-optimization/57051
6414 * fold-const.c (const_binop): Handle VEC_LSHIFT_EXPR
6415 and VEC_RSHIFT_EXPR if shift count is a multiple of element
6418 2013-04-12 Marc Glisse <marc.glisse@inria.fr>
6420 * fold-const.c (fold_binary_loc): Call const_binop also for mixed
6421 vector-scalar operations.
6423 2013-05-03 Marek Polacek <polacek@redhat.com>
6425 Backport from mainline
6426 2013-04-25 Marek Polacek <polacek@redhat.com>
6428 PR tree-optimization/57066
6429 * builtins.c (fold_builtin_logb): Return +Inf for -Inf.
6431 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6433 Backport from mainline
6434 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6436 * lra-constraints.c (process_alt_operands): Add checking alt
6437 number to choose the best alternative.
6439 2013-05-01 Vladimir Makarov <vmakarov@redhat.com>
6442 * lra-constraints.c (best_small_class_operands_num): Remove.
6443 (process_alt_operands): Remove small_class_operands_num. Take
6444 small classes operands into losers and only if the operand is not
6445 matched. Modify debugging output.
6446 (curr_insn_transform): Remove best_small_class_operands_num.
6449 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6451 Backport from mainline
6452 2013-04-29 Vladimir Makarov <vmakarov@redhat.com>
6455 * lra-constraints.c (process_alt_operands): Discourage a bit more
6456 using memory for pseudos. Print cost dump for alternatives.
6457 Modify cost values for conflicts with early clobbers.
6458 (curr_insn_transform): Spill pseudos reassigned to NO_REGS.
6460 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6462 Backport from mainline
6463 2013-04-24 Vladimir Makarov <vmakarov@redhat.com>
6465 PR rtl-optimizations/57046
6466 * lra-constraints (split_reg): Set up lra_risky_transformations_p
6467 for multi-reg splits.
6469 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6471 Backport from mainline
6472 2013-04-22 Vladimir Makarov <vmakarov@redhat.com>
6475 * lra-eliminations.c (mark_not_eliminable): Prevent elimination of
6476 a set sp if no stack realignment.
6478 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6480 Backport from mainline
6481 2013-04-18 Vladimir Makarov <vmakarov@redhat.com>
6483 PR rtl-optimization/56999
6484 * lra-coalesce.c (coalescable_pseudo_p): Remove 2nd parameter and
6486 (lra_coalesce): Remove split_origin_bitmap and related code.
6487 * lra.c (lra): Coalesce after undoing inheritance. Recreate live
6488 ranges if necessary.
6490 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6492 Backport from mainline
6493 2013-04-19 Vladimir Makarov <vmakarov@redhat.com>
6495 PR rtl-optimization/56847
6496 * lra-constraints.c (process_alt_operands): Discourage alternative
6497 with non-matche doffsettable memory constraint fro memory with
6500 2013-05-02 Ian Bolton <ian.bolton@arm.com>
6502 Backport from mainline
6503 2013-03-28 Ian Bolton <ian.bolton@arm.com>
6505 * config/aarch64/aarch64.md (aarch64_can_eliminate): Keep frame
6506 record only when desired or required.
6508 2013-04-30 Jakub Jelinek <jakub@redhat.com>
6510 PR tree-optimization/57104
6511 * tsan.c (instrument_expr): Don't instrument accesses to
6512 DECL_HARD_REGISTER VAR_DECLs.
6514 2013-04-30 Uros Bizjak <ubizjak@gmail.com>
6516 Backport from mainline
6517 2013-04-29 Uros Bizjak <ubizjak@gmail.com>
6520 * config/i386/i386.md (*zero_extendsidi2_rex64): Add "!" to m->?*y
6522 (*zero_extendsidi2): Ditto.
6524 Backport from mainline
6525 2013-04-29 Uros Bizjak <ubizjak@gmail.com>
6528 * config/i386/i386.c (ix86_expand_vec_perm): Validize constant memory.
6530 2013-04-29 Richard Biener <rguenther@suse.de>
6533 * tree-cfg.c (move_stmt_op): Fix condition under which to update
6535 (move_stmt_r): Remove redundant checking.
6537 2013-04-29 Christian Bruel <christian.bruel@st.com>
6540 * sh.md (tstsi_t_zero_extract_eq): Set mode for operand 0.
6542 2013-04-29 Jakub Jelinek <jakub@redhat.com>
6544 PR tree-optimization/57083
6545 * tree-vrp.c (extract_range_from_binary_expr_1): For LSHIFT_EXPR with
6546 non-singleton shift count range, zero extend low_bound for uns case.
6548 2013-04-28 Eric Botcazou <ebotcazou@gcc.gnu.org>
6550 * stor-layout.c (finalize_size_functions): Allocate a structure and
6551 reset cfun before dumping the functions.
6553 2013-04-27 Jakub Jelinek <jakub@redhat.com>
6556 * config/i386/i386.c (ix86_expand_mul_widen_evenodd): Don't
6557 use xop_pmacsdqh if uns_p.
6558 * config/i386/sse.md (xop_rotr<mode>3): Fix up computation of
6559 the immediate rotate count.
6561 2013-04-25 Jakub Jelinek <jakub@redhat.com>
6563 PR rtl-optimization/57003
6564 * regcprop.c (copyprop_hardreg_forward_1): If ksvd.ignore_set_reg,
6565 call note_stores with kill_clobbered_value callback again after
6566 killing regs_invalidated_by_call.
6568 2013-04-25 Ian Bolton <ian.bolton@arm.com>
6570 Backported from mainline.
6571 2013-03-22 Ian Bolton <ian.bolton@arm.com>
6573 * config/aarch64/aarch64.c (aarch64_print_operand): New
6574 format specifier for printing a constant in hex.
6575 * config/aarch64/aarch64.md (insv_imm<mode>): Use the X
6576 format specifier for printing second operand.
6578 2013-04-24 James Greenhalgh <james.greenhalgh@arm.com>
6580 Backported from mainline.
6581 2013-04-24 James Greenhalgh <james.greenhalgh@arm.com>
6583 * config/aarch64/arm_neon.h (vld1<q>_lane*): Fix constraints.
6584 (vld1<q>_dup_<sufp><8, 16, 32, 64>): Likewise.
6585 (vld1<q>_<sufp><8, 16, 32, 64>): Likewise.
6587 2013-04-24 Greta Yorsh <Greta.Yorsh@arm.com>
6589 Backported from mainline.
6591 * config/arm/arm.c (load_multiple_sequence): Require SP
6592 as base register for loads if SP is in the register list.
6594 2013-04-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6595 Steven Bosscher <steven@gcc.gnu.org>
6597 Backported from mainline.
6598 PR rtl-optimization/56605
6599 * loop-iv.c (implies_p): Handle equal RTXs and subregs.
6601 2013-04-22 Marek Polacek <polacek@redhat.com>
6603 Backported from mainline.
6604 2013-04-22 Marek Polacek <polacek@redhat.com>
6607 * tsan.c (instrument_expr): Don't instrument expression
6608 in case its size is zero.
6610 2013-04-22 Yufeng Zhang <yufeng.zhang@arm.com>
6612 Backported from mainline.
6613 2013-04-10 Yufeng Zhang <yufeng.zhang@arm.com>
6614 * config/aarch64/aarch64.c (aarch64_print_extension): New function.
6615 (aarch64_start_file): Use the new function.
6617 2013-04-18 Jakub Jelinek <jakub@redhat.com>
6619 PR tree-optimization/56984
6620 * tree-vrp.c (register_edge_assert_for_2): For (x >> M) < N
6621 and (x >> M) >= N don't register any assertion if N << M is the
6624 2013-04-17 David Edelsohn <dje.gcc@gmail.com>
6627 * config/rs6000/vsx.md (vsx_mov<mode>): Add j->r alternative.
6629 2013-04-15 Jakub Jelinek <jakub@redhat.com>
6631 PR tree-optimization/56962
6632 * gimple-ssa-strength-reduction.c (record_increment): Only set
6633 initializer if gimple_assign_rhs_code is {,POINTER_}PLUS_EXPR and
6634 either rhs1 or rhs2 is equal to c->base_expr.
6636 2013-04-15 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
6638 * emit-rtl.c (reset_all_used_flags): New function.
6639 (verify_rtl_sharing): Call reset_all_used_flags before and after
6640 performing the checks.
6642 2013-04-15 Eric Botcazou <ebotcazou@adacore.com>
6645 * config/sparc/sparc.c (enum sparc_mode_class): Add H_MODE value.
6646 (S_MODES): Set H_MODE bit.
6647 (SF_MODES): Set only S_MODE and SF_MODE bits.
6648 (DF_MODES): Set SF_MODES and only D_MODE and DF_MODE bits.
6649 (sparc_init_modes) <MODE_INT>: Set H_MODE bit for sub-word modes.
6650 <MODE_VECTOR_INT>: Do not set SF_MODE for sub-word modes.
6651 <MODE_FLOAT>: Likewise.
6653 2013-04-12 Vladimir Makarov <vmakarov@redhat.com>
6656 * config/i386/i386.c (ix86_hard_regno_mode_ok): Add
6657 lra_in_progress for return.
6659 2013-04-12 Jakub Jelinek <jakub@redhat.com>
6661 PR tree-optimization/56918
6662 PR tree-optimization/56920
6663 * fold-const.c (int_const_binop_1): Use op1.mul_with_sign (op2, ...)
6664 instead of op1 - op2. Pass 2 * TYPE_PRECISION (type) as second
6665 argument to rshift method.
6667 2013-04-12 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
6669 * ifcvt.c (end_ifcvt_sequence): Mark a and b for unsharing as
6672 2013-04-11 Marek Polacek <polacek@redhat.com>
6674 PR tree-optimization/48184
6675 * params.def (PARAM_ALIGN_THRESHOLD): Increase the minimum
6678 2013-04-11 James Greenhalgh <james.greenhalgh@arm.com>
6680 Backported from mainline.
6681 2013-04-11 James Greenhalgh <james.greenhalgh@arm.com>
6683 * config/aarch64/aarch64-simd.md (aarch64_vcond_internal): Fix
6684 floating-point vector comparisons against 0.
6686 2013-04-11 Jakub Jelinek <jakub@redhat.com>
6688 PR tree-optimization/56899
6689 * fold-const.c (extract_muldiv_1): Apply distributive law
6690 only if TYPE_OVERFLOW_WRAPS (ctype).
6692 2013-04-10 David S. Miller <davem@davemloft.net>
6694 * config/sparc/sparc.h (ASM_CPU_SPEC): Pass -Av8 if -mcpu=supersparc
6695 or -mcpu=hypersparc.
6697 2013-04-10 Jakub Jelinek <jakub@redhat.com>
6699 Backported from mainline
6700 2013-04-09 Jakub Jelinek <jakub@redhat.com>
6703 * omp-low.c (expand_omp_for_generic, expand_omp_for_static_nochunk,
6704 expand_omp_for_static_chunk): Use simple_p = true in
6705 force_gimple_operand_gsi calls when assigning to addressable decls.
6707 2013-04-09 Marek Polacek <polacek@redhat.com>
6709 PR tree-optimization/48762
6710 * params.def (PARAM_MAX_CSE_INSNS): Increase the minimum
6713 2013-04-08 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
6715 * config/s390/s390.c (s390_expand_insv): Only accept insertions
6718 2013-04-08 Marek Polacek <polacek@redhat.com>
6720 PR rtl-optimization/48182
6721 * params.def (PARAM_MIN_CROSSJUMP_INSNS): Increase the minimum
6724 2013-04-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
6727 * config/pa/pa.c (legitimize_pic_address): Before incrementing label
6728 nuses, make sure we have a label.
6730 2013-04-04 Ian Lance Taylor <iant@google.com>
6732 Backport from mainline:
6733 * doc/standards.texi (Standards): The Go frontend supports the Go
6734 1 language standard.
6736 2013-04-04 Marek Polacek <polacek@redhat.com>
6738 Backport from mainline:
6739 2013-04-04 Marek Polacek <polacek@redhat.com>
6741 PR tree-optimization/48186
6742 * predict.c (maybe_hot_frequency_p): Return false if
6743 HOT_BB_FREQUENCY_FRACTION is 0.
6744 (cgraph_maybe_hot_edge_p): Likewise.
6746 2013-04-04 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6748 Backport from mainline:
6749 2013-03-25 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6752 * config/arm/iterators.md (v_cmp_result): New mode attribute.
6753 * config/arm/neon.md (vcond<mode><mode>): Handle unordered cases.
6755 2013-04-04 Richard Biener <rguenther@suse.de>
6757 PR tree-optimization/56837
6758 * tree-loop-distribution.c (classify_partition): For non-zero
6759 values require that the value has the same precision as its
6760 mode to be useful as memset value.
6762 2013-04-03 Roland McGrath <mcgrathr@google.com>
6764 Backport from mainline:
6765 2013-03-26 Roland McGrath <mcgrathr@google.com>
6767 * config/arm/arm.c (arm_print_operand: case 'w'): Use fputs rather
6768 than fprintf with a non-constant, non-format string.
6770 2013-04-03 Marek Polacek <polacek@redhat.com>
6772 Backport from mainline:
6773 2013-04-03 Marek Polacek <polacek@redhat.com>
6776 * tsan.c (instrument_func_exit): Allow BUILT_IN_RETURN
6779 2013-04-03 Richard Biener <rguenther@suse.de>
6781 PR tree-optimization/56817
6782 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely):
6784 (tree_unroll_loops_completely_1): ... new function to manually
6785 walk the loop tree, properly defering outer loops of unrolled
6786 loops to later iterations.
6788 2013-04-02 Jakub Jelinek <jakub@redhat.com>
6790 PR rtl-optimization/56745
6791 * ifcvt.c (cond_exec_find_if_block): Don't try to optimize
6792 if then_bb has no successors and else_bb is EXIT_BLOCK_PTR.
6794 2013-04-02 Wei Mi <wmi@google.com>
6796 * config/i386/i386.c (ix86_rtx_costs): Set proper rtx cost for
6797 ashl<mode>3_mask, *<shift_insn><mode>3_mask and
6798 *<rotate_insn><mode>3_mask in i386.md.
6800 2013-04-01 Wei Mi <wmi@google.com>
6802 * config/i386/i386.md (*ashl<mode>3_mask): Rewrite as define_insn.
6803 Truncate operand 2 using %b asm operand modifier.
6804 (*<shift_insn><mode>3_mask): Ditto.
6805 (*<rotate_insn><mode>3_mask): Ditto.
6807 2013-04-01 Uros Bizjak <ubizjak@gmail.com>
6809 * config/i386/i386.md (*movsf_internal): Change type of
6810 alternatives 3,4 to imov.
6812 2013-03-29 Paolo Carlini <paolo.carlini@oracle.com>
6815 * doc/invoke.texi ([-fwhole-program]): Fix typo.
6817 2013-03-29 Kirill Yukhin <kirill.yukhin@intel.com>
6819 * config/i386/avx2intrin.h (_mm256_broadcastsi128_si256):
6820 Fix declaration name.
6822 2013-03-28 Gerald Pfeifer <gerald@pfeifer.com>
6824 * doc/invoke.texi (AVR Options): Tweak link for AVR-LibC user manual.
6825 * doc/extend.texi (Named Address Spaces): Ditto.
6826 (Variable Attributes): Ditto.
6828 2013-03-28 Eric Botcazou <ebotcazou@adacore.com>
6830 * toplev.c (process_options): Do not disable -fomit-frame-pointer on a
6831 general basis if unwind info is requested and ACCUMULATE_OUTGOING_ARGS
6834 2013-03-27 Walter Lee <walt@tilera.com>
6836 Backport from mainline:
6837 2013-03-27 Walter Lee <walt@tilera.com>
6839 * config/tilegx/tilegx.c (tilegx_expand_prologue): Avoid
6840 double-decrement of next_scratch_regno.
6842 2013-03-27 Walter Lee <walt@tilera.com>
6844 Backport from mainline:
6845 2013-03-27 Walter Lee <walt@tilera.com>
6847 * config/tilegx/tilegx.md (insn_v1mulu): Fix predicates on
6849 (insn_v1mulus): Ditto.
6850 (insn_v2muls): Ditto.
6852 2013-03-27 Walter Lee <walt@tilera.com>
6854 Backport from mainline:
6855 2013-03-27 Walter Lee <walt@tilera.com>
6857 * config/tilegx/tilegx.h (ASM_OUTPUT_ADDR_VEC_ELT): Delete
6859 (ASM_OUTPUT_ADDR_DIFF_ELT): Ditto.
6861 2013-03-27 Walter Lee <walt@tilera.com>
6863 Backport from mainline:
6864 2013-03-27 Walter Lee <walt@tilera.com>
6866 * config/tilegx/tilegx.md (*sibcall_insn): Fix type atribute for jr.
6867 (*sibcall_value): Ditto.
6869 2013-03-27 Walter Lee <walt@tilera.com>
6871 Backport from mainline:
6872 2013-03-27 Walter Lee <walt@tilera.com>
6874 * config/tilegx/tilegx.md (insn_mnz_<mode>): Replaced by ...
6875 (insn_mnz_v8qi): ... this ...
6876 (insn_mnz_v4hi): ... and this. Replace (const_int 0) with the
6878 (insn_v<n>mnz): Replaced by ...
6879 (insn_v1mnz): ... this ...
6880 (insn_v2mnz): ... and this. Replace (const_int 0) with the vector
6882 (insn_mz_<mode>): Replaced by ...
6883 (insn_mz_v8qi): ... this ...
6884 (insn_mz_v4hi): ... and this. Replace (const_int 0) with the
6886 (insn_v<n>mz): Replaced by ...
6887 (insn_v1mz): ... this ...
6888 (insn_v2mz): ... and this. Replace (const_int 0) with the vector
6891 2013-03-26 Eric Botcazou <ebotcazou@adacore.com>
6893 * doc/invoke.texi (SPARC options): Remove -mlittle-endian.
6895 2013-03-26 Sebastian Huber <sebastian.huber@embedded-brains.de>
6897 * config/rtems.opt: Add -pthread option.
6899 2013-03-26 Sofiane Naci <sofiane.naci@arm.com>
6901 * config/aarch64/aarch64.c (aarch64_classify_address): Support
6902 PC-relative load in SI modes and above only.
6904 2013-03-26 Walter Lee <walt@tilera.com>
6906 Backport from mainline:
6907 2013-03-26 Walter Lee <walt@tilera.com>
6909 * config/tilegx/tilegx.h (PROFILE_BEFORE_PROLOGUE): Define.
6910 * config/tilegx/tilepro.h (PROFILE_BEFORE_PROLOGUE): Define.
6912 2013-03-26 Walter Lee <walt@tilera.com>
6914 Backport from mainline:
6915 2013-03-25 Walter Lee <walt@tilera.com>
6917 * config/tilegx/tilegx-builtins.h (enum tilegx_builtin): Add
6918 TILEGX_INSN_SHUFFLEBYTES1.
6919 * config/tilegx/tilegx.c (tilegx_builtin_info): Add entry for
6921 (tilegx_builtins): Ditto.
6922 * config/tilegx/tilegx.md (insn_shufflebytes1): New pattern.
6924 2013-03-26 Walter Lee <walt@tilera.com>
6926 Backport from mainline:
6927 2013-03-25 Walter Lee <walt@tilera.com>
6929 * config/tilegx/tilegx.c (expand_set_cint64_one_inst): Inline
6930 tests for constraint J, K, N, P.
6932 2013-03-26 Walter Lee <walt@tilera.com>
6934 Backport from mainline:
6935 2013-03-25 Walter Lee <walt@tilera.com>
6937 * config/tilegx/tilegx.c (tilegx_asm_preferred_eh_data_format):
6938 Use indirect/pcrel encoding.
6939 * config/tilepro/tilepro.c (tilepro_asm_preferred_eh_data_format):
6942 2013-03-25 Richard Biener <rguenther@suse.de>
6945 * tree-eh.c (lower_eh_must_not_throw): Strip BLOCKs from the
6946 must-not-throw stmt location.
6948 2012-03-22 Jakub Jelinek <jakub@redhat.com>
6950 * BASE-VER: Set to 4.8.1.
6951 * DEV-PHASE: Set to prerelease.
6953 2013-03-22 Release Manager
6955 * GCC 4.8.0 released.
6957 2013-03-21 Walter Lee <walt@tilera.com>
6959 * config/tilegx/sync.md (atomic_test_and_set): New pattern.
6961 2013-03-21 Mark Wielaard <mjw@redhat.com>
6963 * dwarf2out.c (size_of_aranges): Skip DECL_IGNORED_P functions.
6965 2013-03-20 Richard Biener <rguenther@suse.de>
6967 PR tree-optimization/56661
6968 * tree-ssa-sccvn.c (visit_use): Only value-number calls if
6969 the result does not have to be distinct.
6971 2013-03-20 Jakub Jelinek <jakub@redhat.com>
6973 PR tree-optimization/56635
6974 * tree-ssa-phiopt.c (cond_if_else_store_replacement_1): Give up
6975 if lhs of then_assign and else_assign don't have compatible types.
6977 2013-03-17 Jakub Jelinek <jakub@redhat.com>
6980 * config/arm/arm.h (REG_CLASS_NAMES): Add "SFP_REG" and "AFP_REG"
6981 class names. Remove trailing comma after "ALL_REGS".
6983 2013-03-16 Jakub Jelinek <jakub@redhat.com>
6985 * DEV-PHASE: Set to prerelease.
6987 2013-03-14 Andi Kleen <ak@linux.intel.com>
6990 * doc/extend.texi: Document __ATOMIC_HLE_ACQUIRE,
6991 __ATOMIC_HLE_RELEASE. Document __builtin_ia32 TSX intrincs.
6992 Document _x* TSX intrinsics.
6994 2013-03-14 Edgar E. Iglesias <edgar.iglesias@xilinx.com>
6995 David Holsgrove <david.holsgrove@xilinx.com>
6997 * configure.ac: Add MicroBlaze TLS support detection.
6998 * configure: Regenerate.
6999 * config/microblaze/microblaze-protos.h
7000 (microblaze_cannot_force_const_mem, microblaze_tls_referenced_p,
7001 symbol_mentioned_p, label_mentioned_p): Add prototypes.
7002 * config/microblaze/microblaze.c (microblaze_address_type): Add
7003 ADDRESS_TLS and tls_reloc address types.
7004 (microblaze_address_info): Add tls_reloc.
7005 (TARGET_HAVE_TLS): Define.
7006 (get_tls_get_addr, microblaze_tls_symbol_p, microblaze_tls_operand_p_1,
7007 microblaze_tls_referenced_p, microblaze_cannot_force_const_mem,
7008 symbol_mentioned_p, label_mentioned_p, tls_mentioned_p,
7009 load_tls_operand, microblaze_call_tls_get_addr,
7010 microblaze_legitimize_tls_address): New functions.
7011 (microblaze_classify_unspec): Handle UNSPEC_TLS.
7012 (get_base_reg): Use microblaze_tls_symbol_p.
7013 (microblaze_classify_address): Handle TLS.
7014 (microblaze_legitimate_pic_operand): Use symbol_mentioned_p,
7015 label_mentioned_p and microblaze_tls_referenced_p.
7016 (microblaze_legitimize_address): Handle TLS.
7017 (microblaze_address_insns): Handle ADDRESS_TLS.
7018 (pic_address_needs_scratch): Handle TLS.
7019 (print_operand_address): Handle TLS.
7020 (microblaze_expand_prologue): Check TLS_NEEDS_GOT.
7021 (microblaze_expand_move): Handle TLS.
7022 (microblaze_legitimate_constant_p): Check
7023 microblaze_cannot_force_const_mem and microblaze_tls_symbol_p.
7024 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
7025 * config/microblaze/microblaze.h (TLS_NEEDS_GOT): Define
7026 (PIC_OFFSET_TABLE_REGNUM): Set.
7027 * config/microblaze/linux.h (TLS_NEEDS_GOT): Define.
7028 * config/microblaze/microblaze.md (UNSPEC_TLS): Define.
7029 (addsi3, movsi_internal2, movdf_internal): Update constraints
7030 * config/microblaze/predicates.md (arith_plus_operand): Define
7031 (move_operand): Redefine as move_src_operand, check
7032 microblaze_tls_referenced_p.
7034 2013-03-14 Ian Bolton <ian.bolton@arm.com>
7036 * config/aarch64/aarch64.md: (*and<mode>3nr_compare0): Use CC_NZ.
7037 (*and_<SHIFT:optab><mode>3nr_compare0): Likewise.
7039 2013-03-14 Ian Bolton <ian.bolton@arm.com>
7041 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return correct
7044 2013-03-14 Jakub Jelinek <jakub@redhat.com>
7046 PR tree-optimization/53265
7047 * common.opt (Waggressive-loop-optimizations): New option.
7048 * tree-ssa-loop-niter.c: Include tree-pass.h.
7049 (do_warn_aggressive_loop_optimizations): New function.
7050 (record_estimate): Call it. Don't add !is_exit bounds to loop->bounds
7051 if number_of_latch_executions returned constant.
7052 (estimate_numbers_of_iterations_loop): Call number_of_latch_executions
7053 early. If number_of_latch_executions returned constant, set
7054 nb_iterations_upper_bound back to it.
7055 * cfgloop.h (struct loop): Add warned_aggressive_loop_optimizations
7057 * Makefile.in (tree-ssa-loop-niter.o): Depend on $(TREE_PASS_H).
7058 * doc/invoke.texi (-Wno-aggressive-loop-optimizations): Document.
7060 * config/aarch64/t-aarch64-linux (MULTARCH_DIRNAME): Remove.
7061 (MULTILIB_OSDIRNAMES): Set.
7062 * genmultilib: If defaultosdirname doesn't start with :: , set
7063 defaultosdirname2 instead, clear it and emit two . multilib_raw
7064 entries instead of just one.
7066 2013-03-14 Kaz Kojima <kkojima@gcc.gnu.org>
7068 * config/sh/linux.h (TARGET_DEFAULT): Remove MASK_USERMODE.
7069 (SUBTARGET_OVERRIDE_OPTIONS): Set TARGET_USERMODE as default.
7070 * config/sh/netbsd-elf.h (TARGET_DEFAULT): Remove MASK_USERMODE.
7071 (SUBTARGET_OVERRIDE_OPTIONS): New.
7073 2013-03-13 Oleg Endo <olegendo@gcc.gnu.org>
7076 * config/sh/sh.opt (FPU_SINGLE_ONLY): New mask.
7077 (musermode): Convert to Var(TARGET_USERMODE).
7078 * config/sh/sh.h (SELECT_SH2A_SINGLE_ONLY, SELECT_SH4_SINGLE_ONLY,
7079 MASK_ARCH): Add MASK_FPU_SINGLE_ONLY.
7080 * config/sh/sh.c (sh_option_override): Use
7081 TARGET_FPU_DOUBLE || TARGET_FPU_SINGLE_ONLY for call-fp case.
7082 * config/sh/sh.md (udivsi3_i1, divsi3_i1): Remove ! TARGET_SH4
7084 (udivsi3_i4, divsi3_i4): Use TARGET_FPU_DOUBLE condition instead of
7086 (udivsi3_i4_single, divsi3_i4_single): Use
7087 TARGET_FPU_SINGLE_ONLY || TARGET_FPU_DOUBLE instead of TARGET_HARD_SH4.
7089 2013-03-13 Dave Korn <dave.korn.cygwin@....>
7091 * config/i386/cygwin.h (SHARED_LIBGCC_SPEC): Make shared libgcc the
7094 2013-03-13 Richard Biener <rguenther@suse.de>
7096 PR tree-optimization/56608
7097 * tree-vect-slp.c (vect_schedule_slp): Do not remove scalar
7098 calls when vectorizing basic-blocks.
7100 2013-03-13 Jakub Jelinek <jakub@redhat.com>
7103 * config.gcc: On arm, mips, sh and sparc add vxworks-dummy.h to
7106 2013-03-12 Jakub Jelinek <jakub@redhat.com>
7108 * doc/invoke.texi (-Waddr-space-convert): Move into the table earlier.
7110 2013-03-11 Jan Hubicka <jh@suse.cz>
7113 * lto-streamer-out.c (output_symbol_p): Skip references from
7114 constructors of external variables.
7116 2013-03-11 Jan Hubicka <jh@suse.cz>
7119 * valtrack.c (cleanup_auto_inc_dec): Unshare clobbers originating
7121 * emit-rtl.c (verify_rtx_sharing): Likewise.
7122 (copy_insn_1): Likewise.
7123 * rtl.c (copy_rtx): Likewise.
7125 2013-03-11 Georg-Johann Lay <avr@gjlay.de>
7128 * config/avr/avr.c (avr_print_operand): Add space after '%c' in
7129 output_operand_lossage message.
7131 2013-03-11 Richard Earnshaw <rearnsha@arm.com>
7134 * arm.c (shift_op): Validate RTL pattern on the fly.
7135 (arm_print_operand, case 'S'): Don't use shift_operator to validate
7138 2013-03-10 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
7141 * config/pa/pa.md (call_value): Check for calls to powf and direct to
7142 new call patterns that clobber %fr12.
7143 (call_val_powf, call_val_powf_pic, call_val_powf_64bit): New insn,
7144 split and postreload patterns.
7145 * config/pa/pa.c (pa_conditional_register_usage): Revert marking
7146 registers %fr12 and %fr12R as call used.
7148 2013-03-09 Steven Bosscher <steven@gcc.gnu.org>
7150 * dse.c (delete_dead_store_insn): Respect TDF_DETAILS.
7151 (canon_address, record_store, replace_read, check_mem_read_rtx,
7152 scan_insn, dse_step1, dse_step2_init, dse_step2_spill,
7153 dse_step4, dse_step5_nospill, dse_step5_spill, dse_step6,
7154 rest_of_handle_dse): Likewise.
7156 2013-03-09 Richard Sandiford <rdsandiford@googlemail.com>
7159 * tree.h (tree_optimization_option): Rename target_optabs to optabs.
7161 (TREE_OPTIMIZATION_OPTABS): Update after previous field change.
7162 (TREE_OPTIMIZATION_BASE_OPTABS): New macro.
7163 (save_optabs_if_changed): Replace with...
7164 (init_tree_optimization_optabs): ...this.
7165 * optabs.c (save_optabs_if_changed): Rename to...
7166 (init_tree_optimization_optabs): ...this. Take the optimization node
7167 as argument. Do nothing if the base optabs are already correct.
7168 Reuse the existing TREE_OPTIMIZATION_OPTABS memory if we need
7169 to recompute optabs.
7170 * function.h (function): Remove optabs field.
7171 * function.c (invoke_set_current_function_hook): Call
7172 init_tree_optimization_optabs. Use the result to initialize
7175 2013-02-27 Aldy Hernandez <aldyh@redhat.com>
7177 * trans-mem.c (expand_transaction): Do not set PR_INSTRUMENTEDCODE
7178 if GTMA_HAS_NO_INSTRUMENTATION.
7179 (generate_tm_state): Keep GTMA_HAS_NO_INSTRUMENTATION bit.
7180 (ipa_tm_transform_transaction): Set GTMA_HAS_NO_INSTRUMENTATION.
7181 * gimple.h (GTMA_HAS_NO_INSTRUMENTATION): Define.
7182 * gimple-pretty-print.c (dump_gimple_transaction): Handle
7183 GTMA_HAS_NO_INSTRUMENTATION.
7185 2013-03-08 Jakub Jelinek <jakub@redhat.com>
7187 * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Don't link against
7190 2013-03-08 Marek Polacek <polacek@redhat.com>
7191 Jakub Jelinek <jakub@redhat.com>
7193 PR tree-optimization/56478
7194 * predict.c (is_comparison_with_loop_invariant_p): Change the
7195 type of loop_step to tree.
7196 (predict_loops): Adjust.
7197 (predict_iv_comparison): Perform the computations on double_ints.
7199 2013-03-08 Richard Biener <rguenther@suse.de>
7201 PR tree-optimization/56570
7202 * tree-cfg.c (verify_expr_location_1): Verify locations for
7204 * tree-sra.c (create_access_replacement): Strip locations
7205 from DECL_DEBUG_EXPRs.
7207 2013-03-08 Richard Biener <rguenther@suse.de>
7209 * tree-inline.c (expand_call_inline): Do not associate
7210 a BLOCK with the location in BLOCK_SOURCE_LOCATION.
7211 * tree-cfg.c (verify_location): Verify BLOCK_SOURCE_LOCATION.
7213 2013-03-08 Richard Biener <rguenther@suse.de>
7215 * tree-ssa-ter.c (is_replaceable_p): Do not TER across location
7216 or block changes with -Og. Fix for location / block encoding
7217 changes and PHI arguments with locations.
7219 2013-03-07 Steven Bosscher <steven@gcc.gnu.org>
7221 * bitmap.c (struct bitmap_descriptor_d): Use unsigned HOST_WIDEST_INT
7223 (struct output_info): Likewise.
7224 (register_overhead): Remove bad gcc_assert.
7225 (bitmap_find_bit): If there is only a single bitmap element, do not
7226 count a miss as a search.
7227 (print_statistics): Update for counter type changes.
7228 (dump_bitmap_statistics): Likewise. Print headers such that they
7229 are properly lined up with the printed counters.
7231 2013-03-07 Jakub Jelinek <jakub@redhat.com>
7233 PR tree-optimization/56559
7234 * tree-ssa-reassoc.c (zero_one_operation): When looking at rhs2,
7235 check that it has only a single use.
7237 2013-03-07 Richard Biener <rguenther@suse.de>
7239 * doc/invoke.texi (fwhole-program): Discourage use in combination
7242 2013-03-06 Jakub Jelinek <jakub@redhat.com>
7244 * config/arm/t-arm (TM_H, OPTIONS_H_EXTRA): Add arm-cores.def.
7246 PR tree-optimization/56539
7247 * tree-tailcall.c (adjust_return_value_with_ops): Use GSI_SAME_STMT
7248 instead of GSI_CONTINUE_LINKING as last argument to
7249 force_gimple_operand_gsi. Adjust function comment.
7251 * config/aarch64/t-aarch64 (TM_H, OPTIONS_H_EXTRA): Add
7255 * expr.c (expand_cond_expr_using_cmove): When expanding cmove in
7256 promoted mode, convert the result back to the original mode.
7258 2013-03-06 Richard Biener <rguenther@suse.de>
7261 * tree-into-ssa.c (insert_phi_nodes_for): Add dumping.
7262 (insert_updated_phi_nodes_compare_uids): New function.
7263 (update_ssa): Sort symbols_to_rename after UID before
7264 traversing it to insert PHI nodes.
7266 2013-03-06 Richard Biener <rguenther@suse.de>
7269 * tree-vect-data-refs.c (vect_can_force_dr_alignment_p):
7270 Do not adjust alignment of DECL_IN_CONSTANT_POOL decls.
7273 2013-02-13 Richard Biener <rguenther@suse.de>
7276 * varasm.c (output_constant_def_1): Get the decl representing
7277 the constant as argument.
7278 (output_constant_def): Wrap output_constant_def_1.
7279 (make_decl_rtl): Use output_constant_def_1 with the decl
7280 representing the constant.
7281 (build_constant_desc): Optionally re-use a decl already
7282 representing the constant.
7283 (tree_output_constant_def): Adjust.
7285 2013-03-06 Joey Ye <joey.ye@arm.com>
7288 * gcc.c (convert_white_space): New function.
7289 (main): Handles white space in function name.
7291 2013-03-06 Oleg Endo <olegendo@gcc.gnu.org>
7294 * config/sh/sh.c (sh_option_override): Check for TARGET_DYNSHIFT
7295 instead of TARGET_SH2 for call-table case. Do not set sh_div_strategy
7296 to SH_DIV_CALL_TABLE for TARGET_SH2.
7297 * config.gcc (sh_multilibs): Add m2 and m2a to sh*-*-linux* multilib
7299 * doc/invoke.texi (SH options): Document mdiv= call-div1, call-fp,
7302 2013-03-05 Sterling Augustine <saugustine@google.com>
7303 Cary Coutant <ccoutant@google.com>
7306 * dwarf2out.c (resolve_addr): Don't call
7307 remove_loc_list_addr_table_entries a second time for the same
7310 2013-03-05 Jakub Jelinek <jakub@redhat.com>
7313 * cfgexpand.c (expand_debug_parm_decl): Call copy_rtx on incoming.
7314 (avoid_complex_debug_insns): New function.
7315 (expand_debug_locations): Call it.
7317 PR rtl-optimization/56484
7318 * ifcvt.c (noce_process_if_block): If else_bb is NULL, avoid extending
7319 lifetimes of hard registers on small register class machines.
7321 2013-03-05 David Holsgrove <david.holsgrove@xilinx.com>
7323 * config/microblaze/microblaze-protos.h: Rename
7324 microblaze_is_interrupt_handler to microblaze_is_interrupt_variant.
7325 * config/microblaze/microblaze.c (microblaze_attribute_table): Add
7327 (microblaze_fast_interrupt_function_p): New function.
7328 (microblaze_is_interrupt_handler): Rename to
7329 microblaze_is_interrupt_variant and add fast_interrupt check.
7330 (microblaze_must_save_register): Use microblaze_is_interrupt_variant.
7331 (save_restore_insns): Likewise.
7332 (compute_frame_size): Likewise.
7333 (microblaze_function_prologue): Add FAST_INTERRUPT_NAME.
7334 (microblaze_globalize_label): Likewise.
7335 * config/microblaze/microblaze.h: Define FAST_INTERRUPT_NAME.
7336 * config/microblaze/microblaze.md: Use wrapper
7337 microblaze_is_interrupt_variant.
7339 2013-03-05 Kai Tietz <ktietz@redhat.com>
7341 * sdbout.c (sdbout_one_type): Switch to current function's section
7342 supporting cold/hot.
7344 2013-03-05 David Holsgrove <david.holsgrove@xilinx.com>
7346 * doc/invoke.texi (MicroBlaze): Add -mbig-endian, -mlittle-endian,
7349 2013-03-05 Jakub Jelinek <jakub@redhat.com>
7352 * ggc-common.c (gt_pch_save): For ENABLE_VALGRIND_CHECKING,
7353 if VALGRIND_GET_VBITS is defined, temporarily make object
7354 memory all defined, and restore previous valgrind addressability
7355 and definability afterwards. Free this_object at the end.
7358 * lra.c (lra): Call lra_clear_live_ranges if live_p,
7359 right before calling lra_create_live_ranges, also call it
7360 when clearing live_p. Only call lra_clear_live_ranges
7361 at the end if live_p.
7364 * sched-deps.c (delete_dep_node): Free DEP_REPLACE.
7366 2013-03-05 Richard Biener <rguenther@suse.de>
7368 PR tree-optimization/56521
7369 * tree-ssa-sccvn.c (set_value_id_for_result): Always initialize
7372 2013-03-05 Steven Bosscher <steven@gcc.gnu.org>
7375 * except.h (remove_unreachable_eh_regions): New prototype.
7376 * except.c (remove_eh_handler_splicer): New function, split out
7377 of remove_eh_handler.
7378 (remove_eh_handler): Use remove_eh_handler_splicer. Add comment
7379 warning about running it on many EH regions one at a time.
7380 (remove_unreachable_eh_regions_worker): New function, walk the
7381 EH tree in depth-first order and remove non-marked regions.
7382 (remove_unreachable_eh_regions): New function.
7383 * tree-eh.c (mark_reachable_handlers): New function, split out
7384 from remove_unreachable_handlers.
7385 (remove_unreachable_handlers): Use mark_reachable_handlers and
7386 remove_unreachable_eh_regions.
7387 (remove_unreachable_handlers_no_lp): Use mark_reachable_handlers
7388 and remove_unreachable_eh_regions.
7390 2013-03-05 Richard Biener <rguenther@suse.de>
7393 * loop-init.c (fix_loop_structure): Remove loops in two stages,
7394 not freeing them until the end.
7396 2013-03-05 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
7398 * config/s390/s390.h: Define DWARF2_ASM_LINE_DEBUG_INFO.
7400 2013-03-05 Richard Biener <rguenther@suse.de>
7402 PR tree-optimization/56270
7403 * tree-vect-slp.c (vect_schedule_slp): Clear vectorized stmts
7404 of loads after scheduling an SLP instance.
7406 2013-03-05 Jakub Jelinek <jakub@redhat.com>
7408 * Makefile.in (dg_target_exps): Add aarch64.exp, epiphany.exp and
7410 (check_gcc_parallelize): Run guality.exp as a separate job from
7411 vect.exp with unsorted.exp and $(dg_target_exps) separately from
7412 struct-layout-1.exp with stackalign.exp.
7414 * alias.c (init_alias_analysis): Clear reg_known_equiv_p bitmap.
7417 * tree-vect-slp.c (vect_supported_load_permutation_p): Free
7418 load_index sbitmap even if some bit in it isn't set.
7421 * tree-ssa-loop-niter.c (bb_queue): Remove typedef.
7422 (discover_iteration_bound_by_body_walk): Change queues to
7423 vec<vec<basic_block> > and queue to vec<basic_block>. Fix up
7424 spelling in comment. Call safe_push on queues[bound_index] directly.
7425 Release queues[queue_index] in every iteration unconditionally.
7426 Release bounds vector.
7429 * tree-vect-stmts.c (free_stmt_vec_info_vec): Call
7430 free_stmt_vec_info on any left-over stmt_vec_info in the vector.
7431 * tree-vect-loop.c (vect_create_epilog_for_reduction): Release
7434 2013-03-05 Richard Biener <rguenther@suse.de>
7437 * tree-inline.c (remap_blocks_to_null): New function.
7438 (expand_call_inline): When expanding a call stmt without
7439 an associated block inline remap all callee blocks to NULL.
7441 2013-03-05 Jakub Jelinek <jakub@redhat.com>
7443 PR rtl-optimization/56494
7444 * simplify-rtx.c (simplify_truncation): If C is narrower than A,
7445 optimize (truncate:A (subreg:B (truncate:C X) 0)) into
7446 (subreg:A (truncate:C X) 0) instead of (truncate:A X).
7449 * sel-sched-ir.c (free_sched_pools): Release
7450 succs_info_pool.stack[succs_info_pool.max_top] vectors too
7451 if succs_info_pool.max_top isn't -1.
7454 * opts.c (opts_obstack, opts_concat): Moved to...
7455 * opts-common.c (opts_obstack, opts_concat): ... here.
7457 2013-03-04 Jakub Jelinek <jakub@redhat.com>
7460 * diagnostic.c (diagnostic_append_note): Save and restore old prefix.
7462 2013-03-04 Martin Jambor <mjambor@suse.cz>
7464 * tree-dfa.c (get_or_create_ssa_default_def): Use parameter fn in
7465 all appropriate places.
7467 2013-01-04 Eric Botcazou <ebotcazou@adacore.com>
7469 PR tree-optimization/56424
7470 * ipa-split.c (split_function): Do not set the RSO flag if result is
7471 not by reference and its type is a register type.
7473 2013-03-04 David Holsgrove <david.holsgrove@xilinx.com>
7475 * config/microblaze/microblaze.c (microblaze_valid_pic_const): New.
7476 (microblaze_legitimate_pic_operand): Likewise.
7477 * config/microblaze/microblaze.h (LEGITIMATE_PIC_OPERAND_P): Call
7478 new function microblaze_legitimate_pic_operand.
7479 * config/microblaze/microblaze-protos.h
7480 (microblaze_legitimate_pic_operand): Declare.
7482 2013-03-04 Edgar E. Iglesias <edgar.iglesias@gmail.com>
7484 * config/microblaze/predicates.md (call_insn_simple_operand):
7485 New predicate for supported rtx code types.
7486 * config/microblaze/microblaze.md (call_internal1): Use
7487 call_insn_simple_operand predicate.
7489 2013-03-04 Jakub Jelinek <jakub@redhat.com>
7492 * tree-loop-distribution.c (ldist_gen): Call partition_free after each
7493 partitions.ordered_remove.
7496 * tree-vect-stmts.c (vectorizable_conversion): Don't call
7497 vec_oprnds0.create (1) for modifier == NONE.
7500 * tree-vect-stmts.c (vectorizable_shift): Don't call create methods
7501 on vec_oprnds0 or vec_oprnds1 before loop, only call it on
7502 vec_oprnds1 right before pushing anything to it for
7506 * tree-vect-loop.c (destroy_loop_vec_info): For !clean_stmts, just
7507 set nbbs to 0 instead of having separate code path.
7508 (vect_analyze_loop_form): Call destroy_loop_vec_info with true
7509 instead of false as last argument if returning NULL.
7511 2013-03-03 Sandra Loosemore <sandra@codesourcery.com>
7513 * target.def (TARGET_OPTION_VALID_ATTRIBUTE_P): Update comments;
7514 the attribute is now called "target" instead of "option".
7515 (TARGET_OPTION_PRAGMA_PARSE): Likewise, for the pragma.
7516 * doc/tm.texi.in (Target Attributes): Likewise document the correct
7517 attribute/pragma name for TARGET_OPTION_VALID_P and
7518 TARGET_OPTION_PRAGMA_PARSE. Also copy-edit and correct markup.
7519 * doc/tm.texi: Regenerated.
7521 2013-03-02 David Holsgrove <david.holsgrove@xilinx.com>
7523 * config/microblaze/microblaze.c:
7524 Check mcpu, pcmp requirement and set TARGET_REORDER to 0 if not met.
7525 * config/microblaze/microblaze.h: Add -mxl-reorder to
7527 * config/microblaze/microblaze.md: New bswapsi2 and bswaphi2.
7528 instructions emitted if TARGET_REORDER.
7529 * config/microblaze/microblaze.opt: New option -mxl-reorder set to 1
7530 or 0 for -m/-mno case, but initialises as 2 to detect default use case
7533 2013-03-01 Xinliang David Li <davidxl@google.com>
7535 * tree-ssa-uninit.c (compute_control_dep_chain): Limit post-dom
7538 2013-03-01 Jakub Jelinek <jakub@redhat.com>
7541 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Release path
7542 vector even when returning true. Fix up function comment formatting.
7545 * ira-build.c (ira_loop_nodes_count): New variable.
7546 (create_loop_tree_nodes): Initialize it.
7547 (finish_loop_tree_nodes): Use it instead of looking at current_loops.
7550 * tree-vect-data-refs.c (vect_permute_store_chain): Avoid using copy
7551 method on dr_chain and result_chain.
7552 * tree-vect-stmts.c (vectorizable_store): Only call
7553 result_chain.create if j == 0.
7556 * tree-vect-stmts.c (vect_create_vectorized_promotion_stmts): Call
7557 vec_oprnds0->release (); rather than vec_oprnds0->truncate (0)
7558 before overwriting it.
7560 2013-03-01 Tobias Burnus <burnus@net-b.de>
7562 * doc/extended.texi (C Extensions): Change order in @menu
7564 (Other MIPS Built-in Functions): Move last MIPS entry before
7565 "picoChip Built-in Functions".
7566 (SH Built-in Functions): Move after RX Built-in Functions.
7567 * doc/gcc.texi (Introduction): Change order in @menu to match @node.
7568 * doc/md.texi (Constraints): Ditto.
7569 * gty.texi (Type Information): Ditto.
7570 (User-provided marking routines for template types): Make subsection.
7571 * doc/invoke.texi (AArch64 Options): Move before
7572 "Adapteva Epiphany Options".
7574 2013-02-28 Konstantin Serebryany <konstantin.s.serebryany@gmail.com>
7575 Jakub Jelinek <jakub@redhat.com>
7578 * asan.c (gate_asan): Lookup no_sanitize_address instead of
7579 no_address_safety_analysis attribute.
7580 * doc/extend.texi (no_address_safety_attribute): Rename to
7581 no_sanitize_address attribute, mention no_address_safety_analysis
7582 attribute as deprecated alias.
7584 2013-02-28 Jakub Jelinek <jakub@redhat.com>
7587 * tree-vectorizer.h (vect_get_slp_defs): Change 3rd argument
7588 type to vec<vec<tree> > *.
7589 * tree-vect-slp.c (vect_get_slp_defs): Likewise. Change vec_defs
7590 to be vec<tree> instead of vec<tree> *, set vec_defs
7591 to vNULL and call vec_defs.create (number_of_vects), adjust other
7593 * tree-vect-stmts.c (vect_get_vec_defs, vectorizable_call,
7594 vectorizable_condition): Adjust vect_get_slp_defs callers.
7596 2013-02-28 James Greenhalgh <james.greenhalgh@arm.com>
7598 * config/aarch64/aarch64.c
7599 (aarch64_float_const_representable): Remove unused variable.
7601 2013-02-28 James Greenhalgh <james.greenhalgh@arm.com>
7603 * config/aarch64/aarch64.c (aarch64_mangle_type): Make static.
7605 2013-02-28 James Greenhalgh <james.greenhalgh@arm.com>
7607 * config/aarch64/aarch64-builtins.c
7608 (aarch64_init_simd_builtins): Make static.
7610 2013-02-28 James Greenhalgh <james.greenhalgh@arm.com>
7612 * config/aarch64/aarch64.c
7613 (aarch64_simd_make_constant): Make static.
7615 2013-02-28 Martin Jambor <mjambor@suse.cz>
7617 * tree-sra.c (load_assign_lhs_subreplacements): Do not put replacements
7618 with no initialization to the RHS of debug statements.
7620 2013-02-28 Martin Jambor <mjambor@suse.cz>
7622 PR tree-optimization/56294
7623 * tree-sra.c (analyze_access_subtree): Create replacement declarations.
7625 (get_access_replacement): Do not call create_access_replacement.
7626 Assert a replacement exists.
7627 (get_repl_default_def_ssa_name): Create the replacement declaration
7630 2013-02-28 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7632 * config/arm/arm.c (arm_output_mi_thunk): Call final_start_function and
7635 2013-02-28 Marek Polacek <polacek@redhat.com>
7637 PR rtl-optimization/56466
7638 * loop-unroll.c (unroll_and_peel_loops): Call fix_loop_structure
7639 if we're changing a loop.
7640 (peel_loops_completely): Likewise.
7642 2013-02-28 Paolo Carlini <paolo.carlini@oracle.com>
7645 * doc/invoke.texi ([-Wctor-dtor-privacy]): Complete.
7647 2013-02-28 Georg-Johann Lay <avr@gjlay.de>
7650 * config/avr/avr.c (avr_init_builtins): Use 'n' instead of empty
7651 macro parameters with: FX_FTYPE_FX, FX_FTYPE_FX_INT, INT_FTYPE_FX,
7652 INTX_FTYPE_FX, FX_FTYPE_INTX.
7653 * config/avr/builtins.def: Adjust respective DEF_BUILTIN.
7655 2013-02-28 Georg-Johann Lay <avr@gjlay.de>
7657 * avr/avr-mcus.def (ata5272, ata5505, attiny1634, ata6285)
7658 (ata6286, atmega8a, atmega48pa, ata5790, ata5790n, ata5795)
7659 (atmega164pa, atmega165pa, atmega168pa, atmega16hva, atmega16hvb)
7660 (atmega16hvbrevb, atmega16m1, atmega16u4, atmega26hvg, atmega32a)
7661 (atmega32a, atmega3250pa, atmega3290pa, atmega32c1, atmega32m1)
7662 (atmega32u4, atmega32u6, atmega64a, atmega6490a, atmega6490p)
7663 (atmega64c1, atmega64m1, atmega64rfa2, atmega64rfr2, atmega32hvb)
7664 (atmega32hvbrevb, atmega16hva2, atmega48hvf, at90pwm161)
7665 (atmega128a, atmega1284, atmxt112sl, atmxt224, atmxt224e)
7666 (atmxt336s, atxmega16a4u, atxmega16c4, atxmega32a4u, atxmega32c4)
7667 (atxmega32e5, atxmega64a3u, atxmega64a4u, atxmega64b1, atxmega64b3)
7668 (atxmega64c3, atxmega64d4, atxmega128a3u, atxmega128b1)
7669 (atxmega128b3, atxmega128c3, atxmega128d4, atmxt540s, atmxt540sreva)
7670 (atxmega192a3u, atxmega192c3, atxmega256a3u, atxmega256c3)
7671 (atxmega384c3, atxmega384d3, atxmega128a4u): New AVR_MCU.
7672 (avrxmega6): Increase max flash segments from 5 to 6.
7673 * config/avr/t-multilib: Regenerate.
7674 * config/avr/avr-tables.opt: Regenerate.
7675 * doc/avr-mmcu.texi: Regenerate.
7677 2013-02-28 Georg-Johann Lay <avr@gjlay.de>
7679 * config/avr/avr.h (device_to_arch): Rename to device_to_ld.
7680 (avr_device_to_arch): Rename to avr_device_to_ld.
7681 (avr_device_to_as): New prototype.
7682 (EXTRA_SPEC_FUNCTIONS): Add device_to_as.
7683 (ASM_SPEC): Use device_to_as to get -mmcu= and -mno-skip-bug=.
7684 * config/avr/driver-avr.c (avr_device_to_as): New.
7685 (avr_device_to_arch): Rename to avr_device_to_ld.
7687 2013-02-27 Jakub Jelinek <jakub@redhat.com>
7690 * tree-vect-data-refs.c (vect_permute_load_chain): Avoid using copy
7691 method on dr_chain and result_chain.
7694 * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Call
7695 pointer_set_destroy on not_executed_last_iteration.
7698 * tree-vect-loop.c (vectorizable_reduction): Release vect_defs vector.
7701 * ipa-pure-const.c (propagate): Use FOR_EACH_FUNCTION instead of
7702 FOR_EACH_DEFINED_FUNCTION when freeing state.
7705 * df-scan.c (df_insn_delete): Use df_scan_free_mws_vec before
7707 (df_insn_rescan_debug_internal): Use df_scan_free_mws_vec before
7711 * ipa-cp.c (decide_whether_version_node): Call vec_free on
7712 known_aggs[i].items and release known_aggs vector.
7715 * ipa-reference.c (propagate): Free node_info even for alias nodes.
7717 2013-02-27 Edgar E. Iglesias <edgar.iglesias@gmail.com>
7719 * config/microblaze/microblaze.c (microblaze_emit_compare):
7720 Use xor for EQ/NE comparisions.
7721 * config/microblaze/microblaze.md (cstoresf4): Add constraints
7722 (cbranchsf4): Adjust operator to comparison_operator.
7724 2013-02-27 Jakub Jelinek <jakub@redhat.com>
7727 * tree-flow.h (edge_var_map_vector): Change into va_heap, vl_embed
7729 * tree-ssa.c (redirect_edge_var_map_add): Use vec_safe_reserve and
7730 vec_safe_push, always update *slot.
7731 (redirect_edge_var_map_clear): Use vec_free.
7732 (redirect_edge_var_map_dup): Use vec_safe_copy and vec_safe_reserve.
7733 (free_var_map_entry): Use vec_free.
7734 * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Use
7735 FOR_EACH_VEC_SAFE_ELT instead of FOR_EACH_VEC_ELT.
7737 2013-02-27 Andrey Belevantsev <abel@ispras.ru>
7740 * sel-sched-ir.c (merge_expr): Also change vinsn of merged expr
7741 when the may_trap_p bit of the exprs being merged differs.
7742 Reorder tests for speculativeness in the logical and operator.
7744 2013-02-27 Jakub Jelinek <jakub@redhat.com>
7746 * incpath.c (add_standard_paths): Use reconcat instead of concat
7747 where appropriate and avoid leaking memory.
7749 * opts.h: Include obstack.h.
7750 (opts_concat): New prototype.
7751 (opts_obstack): New declaration.
7752 * opts.c (opts_concat): New function.
7753 (opts_obstack): New variable.
7754 (init_options_struct): Call gcc_init_obstack on opts_obstack.
7755 (finish_options): Use opts_concat instead of concat
7756 and XOBNEWVEC instead of XNEWVEC.
7757 * opts-common.c (generate_canonical_option, decode_cmdline_option,
7758 generate_option): Likewise.
7759 * Makefile.in (OPTS_H): Depend on $(OBSTACK_H).
7760 * lto-wrapper.c (main): Call gcc_init_obstack on opts_obstack.
7763 * stmt.c (expand_switch_as_decision_tree_p): If flag_pic
7764 and ASM_OUTPUT_ADDR_DIFF_ELT isn't defined, return true.
7766 2013-02-26 Jakub Jelinek <jakub@redhat.com>
7769 * lra-spills.c (lra_spill): Free spill_hard_reg at the end.
7771 2013-02-26 Joern Rennecke <joern.rennecke@embecosm.com>
7773 * config/arm/arm.c (const_ok_for_dimode_op): Back out last change.
7774 (arm_block_move_unaligned_straight): Likewise.
7775 (arm_adjust_block_mem): Likewise.
7777 2013-02-26 Joern Rennecke <joern.rennecke@embecosm.com>
7780 * config/lm32/lm32.c (gen_int_relational): Remove unused variables
7781 temp, cond and label.
7782 * config/lm32/lm32.md (ashlsi3): Remove unused variable one.
7785 * config/c6x/c6x.c (dbx_register_map): Change to unsigned.
7786 * config/c6x/c6x.h (dbx_register_map): Update declaration.
7789 * config/cr16/cr16-protos.h: Move end of RTX_CODE guard below end
7790 of prologue/epilogue functions.
7793 * config/tilegx/tilegx.c (tilegx_expand_prologue):
7794 Remove unused variable cfa_offset.
7795 * config/tilepro/tilepro.c (tilepro_expand_prologue): Likewise.
7798 * config/mn10300/mn10300.c (mn10300_expand_epilogue): Avoid offset
7799 type promotion to unsigned.
7802 * config/arm/arm.c (const_ok_for_dimode_op): Make code consistent
7803 for HOST_WIDE_INT of 32 bit / same size as int.
7804 (arm_block_move_unaligned_straight): Likewise.
7805 (arm_adjust_block_mem): Likewise.
7808 * config/mep/t-mep (mep-pragma.o): Use ALL_COMPILERFLAGS instead of
7811 2013-02-26 Marek Polacek <polacek@redhat.com>
7813 PR tree-optimization/56426
7814 * tree-ssa-loop.c (tree_ssa_loop_init): Always call
7817 2013-02-26 Richard Biener <rguenther@suse.de>
7820 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Remove
7821 unused variable loops.
7823 2013-02-26 Jakub Jelinek <jakub@redhat.com>
7825 PR tree-optimization/56448
7826 * fold-const.c (operand_equal_p) <case tcc_reference>: Don't look at
7827 TREE_SIDE_EFFECTS if flags contain OEP_CONSTANT_ADDRESS_OF.
7828 Clear OEP_CONSTANT_ADDRESS_OF from flags before recursing on second or
7829 later operands of the references, or even first operand for
7830 INDIRECT_REF, TARGET_MEM_REF or MEM_REF.
7832 PR tree-optimization/56443
7833 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): For
7834 overaligned types, pass TYPE_UNSIGNED (scalar_type) as second argument
7835 to type_for_mode langhook.
7837 2013-02-25 Matt Turner <mattst88@gmail.com>
7839 * doc/invoke.texi: Document r4700.
7841 2013-02-25 Richard Biener <rguenther@suse.de>
7843 PR tree-optimization/56175
7844 * tree-ssa-forwprop.c (hoist_conversion_for_bitop_p): New predicate,
7846 (simplify_bitwise_binary): ... here. Also guard the conversion
7847 of (type) X op CST to (type) (X op ((type-x) CST)) with it.
7849 2013-02-25 Catherine Moore <clm@codesourcery.com>
7852 2013-02-24 Catherine Moore <clm@codesourcery.com>
7853 Maciej W. Rozycki <macro@codesourcery.com>
7854 Tom de Vries <tom@codesourcery.com>
7855 Nathan Sidwell <nathan@codesourcery.com>
7856 Iain Sandoe <iain@codesourcery.com>
7857 Nathan Froyd <froydnj@codesourcery.com>
7858 Chao-ying Fu <fu@mips.com>
7860 * doc/extend.texi (micromips, nomicromips, nocompression):
7861 Document new function attributes.
7862 * doc/invoke.texi (minterlink-compressed, mmicromips,
7863 m14k, m14ke, m14kec): Document new options.
7864 (minterlink-mips16): Update documentation.
7865 * doc/md.texi (ZC, ZD): Document new constraints.
7866 * configure.ac (gcc_cv_as_micromips): Check if linker
7867 supports the .set micromips directive.
7868 * configure: Regenerate.
7869 * config.in: Regenerate.
7870 * config/mips/mips-tables.opt: Regenerate.
7871 * config/mips/micromips.md: New file.
7872 * constraints.md (ZC, AD): New constraints.
7873 * config/mips/predicates.md (movep_src_register): New predicate.
7874 (movep_src_operand): New predicate.
7875 (non_volatile_mem_operand): New predicate.
7876 * config/mips/mips.md (multimem): New type.
7877 (length): Differentiate between 17-bit and 18-bit branch offsets.
7878 (MOVEP1, MOVEP2): New mode iterator.
7879 (mov_<load>l): Use ZC constraint.
7880 (mov_<load>r): Likewise.
7881 (mov_<store>l): Likewise.
7882 (mov_<store>r): Likewise.
7883 (*branch_equality<mode>_inverted): Add microMIPS support.
7884 (*branch_equality<mode>): Likewise.
7885 (*jump_absolute): Likewise.
7886 (indirect_jump_<mode>): Likewise.
7887 (tablejump_<mode>): Likewise.
7888 (<optab>_internal): Likewise.
7889 (sibcall_internal): Likewise.
7890 (sibcall_value_internal): Likewise.
7891 (prefetch): Use constraint ZD.
7892 * config/mips/mips.opt (minterlink-compressed): New option.
7893 (minterlink-mips16): Now an alias for minterlink-compressed.
7894 (mmicromips): New option.
7895 * config/mips/sync.md (sync_compare_and_swap<mode>): Use ZR constraint.
7896 (compare_and_swap_12): Likewise.
7897 (sync_add<mode>): Likewise.
7898 (sync_<optab>_12): Likewise.
7899 (sync_old_<optab>_12): Likewise.
7900 (sync_new_<optab>_12): Likewise.
7901 (sync_nand_12): Likewise.
7902 (sync_old_nand_12): Likewise.
7903 (sync_new_nand_12): Likewise.
7904 (sync_sub<mode>): Likewise.
7905 (sync_old_add<mode>): Likewise.
7906 (sync_old_sub<mode>): Likewise.
7907 (sync_new_add<mode>): Likewise.
7908 (sync_new_sub<mode>): Likewise.
7909 (sync_<optab><mode>): Likewise.
7910 (sync_old_<optab><mode>): Likewise.
7911 (sync_new_<optab><mode>): Likewise.
7912 (sync_nand<mode>): Likewise.
7913 (sync_old_nand<mode>): Likewise.
7914 (sync_new_nand<mode>): Likewise.
7915 (sync_lock_test_and_set<mode>): Likewise.
7916 (test_and_set_12): Likewise.
7917 (atomic_compare_and_swap<mode>): Likewise.
7918 (atomic_exchange<mode>_llsc): Likewise.
7919 (atomic_fetch_add<mode>_llsc): Likewise.
7920 * config/mips/mips-cpus.def (m14kc, m14k): New processors.
7921 * config/mips/mips-protos.h (umips_output_save_restore): New prototype.
7922 (umips_save_restore_pattern_p): Likewise.
7923 (umips_load_store_pair_p): Likewise.
7924 (umips_output_load_store_pair): Likewise.
7925 (umips_movep_target_p): Likewise.
7926 (umips_12bit_offset_address_p): Likewise.
7927 * config/mips/mips.c (MIPS_MAX_FIRST_STEP): Update for microMIPS.
7928 (mips_base_mips16): Rename this...
7929 (mips_base_compression_flags): ...to this. Update all uses.
7930 (mips_attribute_table): Add micromips, nomicromips and nocompression.
7931 (mips_mips16_decl_p): Delete.
7932 (mips_nomips16_decl_p): Delete.
7933 (mips_get_compress_on_flags): New function.
7934 (mips_get_compress_off_flags): New function.
7935 (mips_get_compress_mode): New function.
7936 (mips_get_compress_on_name): New function.
7937 (mips_get_compress_off_name): New function.
7938 (mips_insert_attributes): Support multiple compression types.
7939 (mips_merge_decl_attributes): Likewise.
7940 (umips_12bit_offset_address_p): New function.
7941 (mips_start_function_definition): Emit .set micromips directive.
7942 (mips_call_may_need_jalx_p): New function.
7943 (mips_function_ok_for_sibcall): Add microMIPS support.
7944 (mips_print_operand_punctuation): Support short delay slots and
7946 (umips_swm_mask, umips_swm_encoding): New.
7947 (umips_build_save_restore): New function.
7948 (mips_for_each_saved_gpr_and_fpr): Add microMIPS support.
7949 (was_mips16_p): Remove.
7950 (old_compression_mode): New.
7951 (mips_set_compression_mode): New function.
7952 (mips_set_current_function): Add microMIPS support.
7953 (mips_option_override): Likewise.
7954 (umips_save_restore_pattern_p): New function.
7955 (umips_output_save_restore): New function.
7956 (umips_load_store_pair_p_1): New function.
7957 (umips_load_store_pair_p): New function.
7958 (umips_output_load_store_pair_1): New function.
7959 (umips_output_load_store_pair): New function.
7960 (umips_movep_target_p) New function.
7961 (mips_prepare_pch_save): Add microMIPS support.
7962 * config/mips/mips.h (TARGET_COMPRESSION): New.
7963 (TARGET_CPU_CPP_BUILTINS): Update macro
7964 to use new compression flags and to support microMIPS.
7965 (MIPS_ISA_LEVEL_SPEC): Add m14k processors.
7966 (MIPS_ARCH_FLOAT_SPEC): Likewise.
7967 (ISA_HAS_LWXS): Include TARGET_MICROMIPS.
7968 (ISA_HAS_LOAD_DELAY): Exclude TARGET_MICROMIPS.
7969 (ASM_SPEC): Support mmicromips and mno-micromips.
7970 (M16STORE_REG_P): New macro.
7971 (MIPS_CALL): Support TARGET_MICROMIPS.
7972 (MICROMIPS_J): New macro.
7973 (mips_base_mips16): Rename this...
7974 (mips_base_compression_flags): ...to this.
7975 (UMIPS_12BIT_OFFSET_P): New macro.
7976 * config/mips/t-sde: (MULTILIB_OPTIONS): Add microMIPS.
7977 (MULTILIB_DIRNAMES): Likewise.
7979 2013-02-25 Tom de Vries <tom@codesourcery.com>
7981 PR rtl-optimization/56131
7982 * insn-notes.def (INSN_NOTE_BASIC_BLOCK): Update comment.
7983 * cfgrtl.c (delete_insn): Don't reorder NOTE_INSN_DELETED_LABEL and
7984 NOTE_INSN_BASIC_BLOCK if BLOCK_FOR_INSN == NULL.
7986 2013-02-25 Tobias Burnus <burnus@net-b.de>
7988 * doc/invoke.texi (-fsanitize=): Move from optimization
7989 to debugging options.
7991 2013-02-25 Andrey Belevantsev <abel@ispras.ru>
7993 * sched-deps.c (sched_analyze_insn): Fix typo in comment.
7995 2013-02-25 Andrey Belevantsev <abel@ispras.ru>
7996 Alexander Monakov <amonakov@ispras.ru>
7999 * sched-deps.c (sched_analyze_insn): When reg_pending_barrier,
8000 flush pending lists also on non-jumps. Adjust comment.
8002 2013-02-24 Catherine Moore <clm@codesourcery.com>
8003 Maciej W. Rozycki <macro@codesourcery.com>
8004 Tom de Vries <tom@codesourcery.com>
8005 Nathan Sidwell <nathan@codesourcery.com>
8006 Iain Sandoe <iain@codesourcery.com>
8007 Nathan Froyd <froydnj@codesourcery.com>
8008 Chao-ying Fu <fu@mips.com>
8010 * doc/extend.texi (micromips, nomicromips, nocompression):
8011 Document new function attributes.
8012 * doc/invoke.texi (minterlink-compressed, mmicromips,
8013 m14k, m14ke, m14kec): Document new options.
8014 (minterlink-mips16): Update documentation.
8015 * doc/md.texi (ZC, ZD): Document new constraints.
8016 * configure.ac (gcc_cv_as_micromips): Check if linker
8017 supports the .set micromips directive.
8018 * configure: Regenerate.
8019 * config.in: Regenerate.
8020 * config/mips/mips-tables.opt: Regenerate.
8021 * config/mips/micromips.md: New file.
8022 * constraints.md (ZC, AD): New constraints.
8023 * config/mips/predicates.md (movep_src_register): New predicate.
8024 (movep_src_operand): New predicate.
8025 (non_volatile_mem_operand): New predicate.
8026 * config/mips/mips.md (multimem): New type.
8027 (length): Differentiate between 17-bit and 18-bit branch offsets.
8028 (MOVEP1, MOVEP2): New mode iterator.
8029 (mov_<load>l): Use ZC constraint.
8030 (mov_<load>r): Likewise.
8031 (mov_<store>l): Likewise.
8032 (mov_<store>r): Likewise.
8033 (*branch_equality<mode>_inverted): Add microMIPS support.
8034 (*branch_equality<mode>): Likewise.
8035 (*jump_absolute): Likewise.
8036 (indirect_jump_<mode>): Likewise.
8037 (tablejump_<mode>): Likewise.
8038 (<optab>_internal): Likewise.
8039 (sibcall_internal): Likewise.
8040 (sibcall_value_internal): Likewise.
8041 (prefetch): Use constraint ZD.
8042 * config/mips/mips.opt (minterlink-compressed): New option.
8043 (minterlink-mips16): Now an alias for minterlink-compressed.
8044 (mmicromips): New option.
8045 * config/mips/sync.md (sync_compare_and_swap<mode>): Use ZR constraint.
8046 (compare_and_swap_12): Likewise.
8047 (sync_add<mode>): Likewise.
8048 (sync_<optab>_12): Likewise.
8049 (sync_old_<optab>_12): Likewise.
8050 (sync_new_<optab>_12): Likewise.
8051 (sync_nand_12): Likewise.
8052 (sync_old_nand_12): Likewise.
8053 (sync_new_nand_12): Likewise.
8054 (sync_sub<mode>): Likewise.
8055 (sync_old_add<mode>): Likewise.
8056 (sync_old_sub<mode>): Likewise.
8057 (sync_new_add<mode>): Likewise.
8058 (sync_new_sub<mode>): Likewise.
8059 (sync_<optab><mode>): Likewise.
8060 (sync_old_<optab><mode>): Likewise.
8061 (sync_new_<optab><mode>): Likewise.
8062 (sync_nand<mode>): Likewise.
8063 (sync_old_nand<mode>): Likewise.
8064 (sync_new_nand<mode>): Likewise.
8065 (sync_lock_test_and_set<mode>): Likewise.
8066 (test_and_set_12): Likewise.
8067 (atomic_compare_and_swap<mode>): Likewise.
8068 (atomic_exchange<mode>_llsc): Likewise.
8069 (atomic_fetch_add<mode>_llsc): Likewise.
8070 * config/mips/mips-cpus.def (m14kc, m14k): New processors.
8071 * config/mips/mips-protos.h (umips_output_save_restore): New prototype.
8072 (umips_save_restore_pattern_p): Likewise.
8073 (umips_load_store_pair_p): Likewise.
8074 (umips_output_load_store_pair): Likewise.
8075 (umips_movep_target_p): Likewise.
8076 (umips_12bit_offset_address_p): Likewise.
8077 * config/mips/mips.c (MIPS_MAX_FIRST_STEP): Update for microMIPS.
8078 (mips_base_mips16): Rename this...
8079 (mips_base_compression_flags): ...to this. Update all uses.
8080 (mips_attribute_table): Add micromips, nomicromips and nocompression.
8081 (mips_mips16_decl_p): Delete.
8082 (mips_nomips16_decl_p): Delete.
8083 (mips_get_compress_on_flags): New function.
8084 (mips_get_compress_off_flags): New function.
8085 (mips_get_compress_mode): New function.
8086 (mips_get_compress_on_name): New function.
8087 (mips_get_compress_off_name): New function.
8088 (mips_insert_attributes): Support multiple compression types.
8089 (mips_merge_decl_attributes): Likewise.
8090 (umips_12bit_offset_address_p): New function.
8091 (mips_start_function_definition): Emit .set micromips directive.
8092 (mips_call_may_need_jalx_p): New function.
8093 (mips_function_ok_for_sibcall): Add microMIPS support.
8094 (mips_print_operand_punctuation): Support short delay slots and
8096 (umips_swm_mask, umips_swm_encoding): New.
8097 (umips_build_save_restore): New function.
8098 (mips_for_each_saved_gpr_and_fpr): Add microMIPS support.
8099 (was_mips16_p): Remove.
8100 (old_compression_mode): New.
8101 (mips_set_compression_mode): New function.
8102 (mips_set_current_function): Add microMIPS support.
8103 (mips_option_override): Likewise.
8104 (umips_save_restore_pattern_p): New function.
8105 (umips_output_save_restore): New function.
8106 (umips_load_store_pair_p_1): New function.
8107 (umips_load_store_pair_p): New function.
8108 (umips_output_load_store_pair_1): New function.
8109 (umips_output_load_store_pair): New function.
8110 (umips_movep_target_p) New function.
8111 (mips_prepare_pch_save): Add microMIPS support.
8112 * config/mips/mips.h (TARGET_COMPRESSION): New.
8113 (TARGET_CPU_CPP_BUILTINS): Update macro
8114 to use new compression flags and to support microMIPS.
8115 (MIPS_ISA_LEVEL_SPEC): Add m14k processors.
8116 (MIPS_ARCH_FLOAT_SPEC): Likewise.
8117 (ISA_HAS_LWXS): Include TARGET_MICROMIPS.
8118 (ISA_HAS_LOAD_DELAY): Exclude TARGET_MICROMIPS.
8119 (ASM_SPEC): Support mmicromips and mno-micromips.
8120 (M16STORE_REG_P): New macro.
8121 (MIPS_CALL): Support TARGET_MICROMIPS.
8122 (MICROMIPS_J): New macro.
8123 (mips_base_mips16): Rename this...
8124 (mips_base_compression_flags): ...to this.
8125 (UMIPS_12BIT_OFFSET_P): New macro.
8126 * config/mips/t-sde: (MULTILIB_OPTIONS): Add microMIPS.
8127 (MULTILIB_DIRNAMES): Likewise.
8129 2013-02-24 Jakub Jelinek <jakub@redhat.com>
8132 * target-globals.c (save_target_globals): For init_reg_sets and
8133 target_reinit remporarily set this_fn_optabs to this_target_optabs.
8135 2013-02-22 James Grennahlgh <james.greenhalgh@arm.com>
8137 * config/aarch64/aarch64-simd-builtins.def: Add copyright header.
8138 * config/aarch64/t-aarch64
8139 (aarch64-builtins.o): Depend on aarch64-simd-builtins.def.
8141 2013-02-22 Vladimir Makarov <vmakarov@redhat.com>
8144 * lra-constraints.c (process_alt_operands): Reload operand
8145 conflicting with earlier clobber only if no more other conflicting
8148 2013-02-22 Jakub Jelinek <jakub@redhat.com>
8151 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Link in libasan_preinit.o
8152 if not linking a shared library.
8154 2013-02-22 Seth LaForge <sethml@google.com>
8156 * config.gcc (arm*-*-eabi*): Treat arm*eb as big-endian.
8158 2013-02-22 Greta Yorsh <Greta.Yorsh@arm.com>
8160 * config/arm/arm.md (split for extendsidi): Update condition.
8161 (zero_extend<mode>di2,extend<mode>di2): Add an alternative.
8162 * config/arm/iterators.md (qhs_extenddi_cstr): Likewise.
8163 (qhs_zextenddi_cstr): Likewise.
8165 2013-02-21 Jakub Jelinek <jakub@redhat.com>
8168 * expmed.c (EXACT_POWER_OF_2_OR_ZERO_P): Do subtraction in uhwi, to
8169 avoid signed wrapping.
8170 (expand_mult): Handle properly multiplication by
8171 ((dword_type) -1) << (BITS_PER_WORD - 1). Improve multiplication by
8172 ((dword_type) 1) << (BITS_PER_WORD - 1). Avoid undefined behavior
8173 in the compiler if coeff is HOST_WIDE_INT_MIN.
8174 (expand_divmod): Don't make ext_op1 static, change it's type to uhwi.
8175 Avoid undefined behavior in -INTVAL (op1).
8177 PR rtl-optimization/50339
8178 * lower-subreg.h (struct lower_subreg_choices): Add splitting_ashiftrt
8180 * lower-subreg.c (compute_splitting_shift): Handle ASHIFTRT.
8181 (compute_costs): Call compute_splitting_shift also for ASHIFTRT
8182 into splitting_ashiftrt field.
8183 (find_decomposable_shift_zext, resolve_shift_zext): Handle also
8185 (dump_choices): Fix up printing LSHIFTRT choices, print ASHIFTRT
8188 2013-02-20 Aldy Hernandez <aldyh@redhat.com>
8191 * trans-mem.c (execute_tm_mark): Do not expand transactions that
8192 are sure to go irrevocable.
8194 2013-02-21 Hans-Peter Nilsson <hp@axis.com>
8196 * doc/rtl.texi (vec_concat, vec_duplicate): Mention that
8197 scalars are valid operands.
8199 2013-02-21 Martin Jambor <mjambor@suse.cz>
8201 PR tree-optimization/56310
8202 * ipa-cp.c (agg_replacements_to_vector): New parameter index, copy
8203 only matching indices and non-negative final offsets.
8204 (intersect_aggregates_with_edge): Pass src_idx to
8205 agg_replacements_to_vector. Pass src_idx insstead of index to
8206 intersect_with_agg_replacements.
8208 2013-02-21 Martin Jambor <mjambor@suse.cz>
8210 * ipa-cp.c (good_cloning_opportunity_p): Dump the real threshold
8211 instead of hard-wired defaults.
8213 2013-02-21 Maciej W. Rozycki <macro@codesourcery.com>
8215 * doc/invoke.texi (MIPS Options): Update documentation of the
8216 floating-point multiply-accumulate instruction restrictions.
8218 2013-02-21 Kostya Serebryany <kcc@google.com>
8220 * config/i386/i386.c (ix86_asan_shadow_offset): Use 0x7fff8000 as
8221 asan_shadow_offset on x86_64 linux.
8223 2013-02-21 Richard Biener <rguenther@suse.de>
8225 PR tree-optimization/56415
8227 2013-02-11 Richard Biener <rguenther@suse.de>
8229 PR tree-optimization/56273
8230 * tree-vrp.c (simplify_cond_using_ranges): Disable for the
8233 2013-02-21 Jakub Jelinek <jakub@redhat.com>
8236 * doc/invoke.texi (-fdump-rtl-pro_and_epilogue): Use @item
8240 * expr.c (expand_expr_real_1) <case TARGET_MEM_REF, MEM_REF>: Don't
8241 use movmisalign or extract_bit_field for EXPAND_MEMORY modifier.
8243 2013-02-20 Jan Hubicka <jh@suse.cz>
8245 PR tree-optimization/56265
8246 * ipa-prop.c (ipa_make_edge_direct_to_target): Fixup callgraph when
8247 target is referenced for first time.
8249 2013-02-20 Richard Biener <rguenther@suse.de>
8251 * tree-call-cdce.c (tree_call_cdce): Do not remove unused locals.
8252 * tree-ssa-forwprop.c (ssa_forward_propagate_and_combine): Likewise.
8253 * tree-ssa-dce.c (perform_tree_ssa_dce): Likewise.
8254 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Do
8255 not return anything.
8256 (rename_ssa_copies): Do not remove unused locals.
8257 * tree-ssa-ccp.c (do_ssa_ccp): Likewise.
8258 * tree-ssanames.c (pass_release_ssa_names): Remove unused locals first.
8259 * passes.c (execute_function_todo): Do not schedule unused locals
8260 removal if cleanup_tree_cfg did something.
8261 * tree-ssa-live.c (remove_unused_locals): Dump statistics
8262 about the number of removed locals.
8264 2013-02-20 Richard Biener <rguenther@suse.de>
8266 PR tree-optimization/56398
8267 * tree-vect-loop-manip.c (adjust_debug_stmts): Skip
8270 2013-02-20 Martin Jambor <mjambor@suse.cz>
8272 PR tree-optimization/55334
8273 * ipa-cp.c (initialize_node_lattices): Disable IPA-CP through and to
8274 restricted pointers to arrays.
8276 2013-02-20 Richard Biener <rguenther@suse.de>
8277 Jakub Jelinek <jakub@redhat.com>
8279 PR tree-optimization/56396
8280 * tree-ssa-ccp.c (n_const_val): New static variable.
8281 (get_value): Return NULL for SSA names we don't have a lattice
8283 (ccp_initialize): Initialize n_const_val.
8284 * tree-ssa-copy.c (n_copy_of): New static variable.
8285 (init_copy_prop): Initialize n_copy_of.
8286 (get_value): Return NULL_TREE for SSA names we don't have a
8289 2013-02-20 Martin Jambor <mjambor@suse.cz>
8291 * ipa-cp.c (initialize_node_lattices): Fix dumping condition.
8293 2013-02-20 Richard Biener <rguenther@suse.de>
8295 * genpreds.c (write_lookup_constraint): Do not compare first
8296 letter of the constraint again.
8298 2013-02-20 Richard Biener <rguenther@suse.de>
8300 * tree-ssa-loop-ivopts.c (alloc_use_cost_map): Use bitmap_count_bits
8302 (get_use_iv_cost): Terminate hashtable walk when coming across
8305 2013-02-20 Igor Zamyatin <igor.zamyatin@intel.com>
8307 * config/i386/i386.c (initial_ix86_tune_features): Turn on fp
8308 reassociation for avx2 targets.
8310 2012-02-19 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8312 * config/microblaze/microblaze.c: microblaze_has_clz = 0
8313 Add version check for v8.10.a to enable microblaze_has_clz
8314 * config/microblaze/microblaze.h: Add TARGET_HAS_CLZ as combined
8315 version and TARGET_PATTERN_COMPARE check
8316 * config/microblaze/microblaze.md: New clzsi2 instruction
8318 2012-02-19 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8320 * config/microblaze/microblaze.md (call_value_intern): Check symbol is
8321 function before branching.
8323 2012-02-19 Andrey Belevantsev <abel@ispras.ru>
8325 * sel-sched-dump.c (dump_insn_rtx_flags): Explicitly set
8327 (dump_insn_rtx_1): Pass PATTERN (insn) to str_pattern_slim.
8329 2012-02-19 Andrey Belevantsev <abel@ispras.ru>
8332 * sel-sched.c: Include ira.h.
8333 (implicit_clobber_conflict_p): New function.
8334 (moveup_expr): Use it.
8335 * Makefile.in (sel-sched.o): Depend on ira.h.
8337 2013-02-19 Richard Biener <rguenther@suse.de>
8339 PR tree-optimization/56384
8340 * tree-ssa-sccvn.h (struct vn_phi_s): Add type member.
8341 (vn_hash_type): Split out from ...
8342 (vn_hash_constant_with_type): ... here.
8343 * tree-ssa-sccvn.c (vn_phi_compute_hash): Use vn_hash_type.
8344 (vn_phi_eq): Compare types from vn_phi_s structure.
8345 (vn_phi_lookup): Populate vn_phi_s type.
8346 (vn_phi_insert): Likewise.
8348 2013-02-19 Jakub Jelinek <jakub@redhat.com>
8350 PR tree-optimization/56350
8351 * tree-vect-loop.c (vectorizable_reduction): If orig_stmt, return false
8352 if haven't found reduction or nested cycle operand, rather than
8353 asserting we must find it.
8355 PR tree-optimization/56381
8356 * tree-ssa-pre.c (create_expression_by_pieces): Fix up last argument
8359 2013-02-18 Aldy Hernandez <aldyh@redhat.com>
8360 Jakub Jelinek <jakub@redhat.com>
8363 * genopinit.c (raw_optab_handler): Use this_fn_optabs.
8364 (swap_optab_enable): Same.
8365 (init_all_optabs): Use argument instead of global.
8366 * tree.h (struct tree_optimization_option): New field target_optabs.
8367 * expr.h (init_all_optabs): Add argument to prototype.
8368 (TREE_OPTIMIZATION_OPTABS): New.
8369 (save_optabs_if_changed): Protoize.
8370 * optabs.h: Declare this_fn_optabs.
8371 * optabs.c (save_optabs_if_changed): New.
8372 Declare this_fn_optabs.
8373 (init_optabs): Add argument to init_all_optabs() call.
8374 * function.c (invoke_set_current_function_hook): Handle per
8376 * function.h (struct function): New field optabs.
8377 * config/mips/mips.c (mips_set_mips16_mode): Handle when
8378 optimization_current_node has changed.
8379 * target-globals.h (save_target_globals_default_opts): Protoize.
8380 * target-globals.c (save_target_globals_default_opts): New.
8382 2013-02-18 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
8385 * config/pa/pa.c (pa_conditional_register_usage): On HP-UX, mark
8386 registers %fr12 and %fr12R as call used.
8389 * config/pa/predicates.md (base14_operand): Except for BLKmode, QImode
8390 and HImode, require all displacements to be an integer multiple of
8392 * config/pa/pa.c (pa_legitimate_address_p): For REG+BASE addresses,
8393 only allow QImode and HImode when reload is in progress and strict is
8394 true. Likewise for symbolic addresses. Use base14_operand to check
8395 displacements in REG+BASE addresses.
8397 2013-02-18 Richard Biener <rguenther@suse.de>
8399 PR tree-optimization/56366
8400 * tree-vect-loop.c (get_initial_def_for_induction): Properly
8401 handle sign-conversion of outer-loop initial induction value.
8403 2013-02-18 Richard Biener <rguenther@suse.de>
8406 * cfghooks.c (merge_blocks): If we merge a latch into another
8407 block adjust references to it.
8408 * cfgloop.c (flow_loops_find): Reset latch before recomputing it.
8409 (verify_loop_structure): Verify that a recorded latch is in fact
8412 2013-02-18 Richard Biener <rguenther@suse.de>
8414 PR tree-optimization/56321
8415 * tree-ssa-reassoc.c (propagate_op_to_single_use): Properly
8416 order SSA name release and virtual operand unlinking.
8418 2013-02-17 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8420 * config/microblaze/microblaze.md (save_stack_block): Define.
8421 (restore_stack_block): Likewise.
8423 2013-02-16 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8425 * config/microblaze/linux.h (TARGET_SUPPORTS_PIC): Define as 1.
8426 * config/microblaze/microblaze.h (TARGET_SUPPORTS_PIC): Define as 1.
8427 * config/microblaze/microblaze.c (microblaze_option_override):
8428 Bail out early for PIC modes when target does not support PIC.
8430 2013-02-16 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8432 * config/microblaze/microblaze.c (microblaze_asm_trampoline_template):
8433 Replace with a microblaze version.
8434 (microblaze_trampoline_init): Adapt for microblaze.
8435 * config/microblaze/microblaze.h (TRAMPOLINE_SIZE): Adapt for
8438 2013-02-16 Jakub Jelinek <jakub@redhat.com>
8439 Dodji Seketeli <dodji@redhat.com>
8442 * asan.c (get_mem_refs_of_builtin_call): White space and style
8444 (instrument_mem_region_access): Do not forget to always put
8445 instrumentation of the of 'base' and 'base + len' in a "if (len !=
8446 0) statement, even for cases where either 'base' or 'base + len'
8447 are not instrumented -- because they have been previously
8448 instrumented. Simplify the logic by putting all the statements
8449 instrument 'base + len' inside a sequence, and then insert that
8450 sequence right before the current insertion point. Then, to
8451 instrument 'base + len', just get an iterator on that statement.
8452 And do not forget to update the pointer to iterator the function
8453 received as argument.
8455 2013-02-15 Vladimir Makarov <vmakarov@redhat.com>
8457 PR rtl-optimization/56348
8458 * lra-assigns.c (reload_pseudo_compare_func): Prefer bigger pseudos.
8460 2013-02-15 Steven Bosscher <steven@gcc.gnu.org>
8462 * graph.c (start_graph_dump): Print dumpfile base as digraph label.
8463 (clean_graph_dump_file): Pass base to start_graph_dump.
8465 2013-02-14 Richard Henderson <rth@redhat.com>
8468 * lower-subreg.c (simple_move): Check dest mode instead of src mode.
8470 2013-02-14 Steven Bosscher <steven@gcc.gnu.org>
8472 * collect2-aix.h: Define F_LOADONLY.
8474 2013-02-14 Richard Biener <rguenther@suse.de>
8477 * varasm.c (output_constant_def_1): Get the decl representing
8478 the constant as argument.
8479 (output_constant_def): Wrap output_constant_def_1.
8480 (make_decl_rtl): Use output_constant_def_1 with the decl
8481 representing the constant.
8482 (build_constant_desc): Optionally re-use a decl already
8483 representing the constant.
8484 (tree_output_constant_def): Adjust.
8486 2013-02-14 Dodji Seketeli <dodji@redhat.com>
8489 * asan.c (instrument_builtin_call): Really put the length of the
8490 second source argument into src1_len.
8492 2013-02-13 Jakub Jelinek <jakub@redhat.com>
8494 * asan.c (create_cond_insert_point): Add create_then_fallthru_edge
8495 argument. If it is false, don't create edge from then_bb to
8497 (insert_if_then_before_iter): Pass true to it.
8498 (build_check_stmt): Pass false to it.
8499 (transform_statements): Flush hash table only on extended basic
8500 block boundaries, rather than at the beginning of every bb.
8501 Don't flush hash table on nonfreeing_call_p calls.
8502 * tree-flow.h (nonfreeing_call_p): New prototype.
8503 * tree-ssa-phiopt.c (nonfreeing_call_p): No longer static.
8505 2013-02-13 David S. Miller <davem@davemloft.net>
8507 * expmed.c (expand_shift_1): Only strip scalar integer subregs.
8509 2013-02-13 Vladimir Makarov <vmakarov@redhat.com>
8512 * ira.c (max_regno_before_ira): Move from ...
8514 (fix_reg_equiv_init): Use max_regno_before_ira instead of
8517 2013-02-13 Jakub Jelinek <jakub@redhat.com>
8519 * config/i386/i386.c (ix86_asan_shadow_offset): Revert last change.
8521 2013-02-13 Richard Biener <rguenther@suse.de>
8524 * gimple-streamer-out.c (output_gimple_stmt): Undo wrapping
8525 globals in MEM_REFs.
8527 2013-02-13 Richard Biener <rguenther@suse.de>
8529 * loop-init.c (loop_optimizer_init): Clear loop state when
8530 re-initializing preserved loops.
8531 * loop-unswitch.c (unswitch_single_loop): Return whether
8532 we unswitched the loop. Do not verify loop state here.
8533 (unswitch_loops): When we unswitched a loop discover new
8536 2013-02-13 Kostya Serebryany <kcc@google.com>
8538 * config/i386/i386.c: Use 0x7fff8000 as asan_shadow_offset on x86_64
8540 * sanitizer.def: Rename __asan_init to __asan_init_v1.
8542 2013-02-12 Dodji Seketeli <dodji@redhat.com>
8544 Avoid instrumenting duplicated memory access in the same basic block
8545 * Makefile.in (asan.o): Add new dependency on hash-table.h
8546 * asan.c (struct asan_mem_ref, struct mem_ref_hasher): New types.
8547 (asan_mem_ref_init, asan_mem_ref_get_end, get_mem_ref_hash_table)
8548 (has_stmt_been_instrumented_p, empty_mem_ref_hash_table)
8549 (free_mem_ref_resources, has_mem_ref_been_instrumented)
8550 (has_stmt_been_instrumented_p, update_mem_ref_hash_table)
8551 (get_mem_ref_of_assignment): New functions.
8552 (get_mem_refs_of_builtin_call): Extract from
8553 instrument_builtin_call and tweak a little bit to make it fit with
8555 (instrument_builtin_call): Use the new
8556 get_mem_refs_of_builtin_call. Use gimple_call_builtin_p instead
8557 of is_gimple_builtin_call.
8558 (instrument_derefs, instrument_mem_region_access): Insert the
8559 instrumented memory reference into the hash table.
8560 (maybe_instrument_assignment): Renamed instrument_assignment into
8561 this, and change it to advance the iterator when instrumentation
8562 actually happened and return true in that case. This makes it
8563 homogeneous with maybe_instrument_assignment, and thus give a
8564 chance to callers to be more 'regular'.
8565 (transform_statements): Clear the memory reference hash table
8566 whenever we enter a new BB, when we cross a function call, or when
8567 we are done transforming statements. Use
8568 maybe_instrument_assignment instead of instrumentation. No more
8569 need to special case maybe_instrument_assignment and advance the
8570 iterator after calling it; it's now handled just like
8571 maybe_instrument_call. Update comment.
8573 2013-02-13 Richard Biener <rguenther@suse.de>
8575 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc):
8576 Fix loop discovery code.
8578 2013-02-12 Vladimir Makarov <vmakarov@redhat.com>
8581 * lra-constraints.c (process_alt_operands): Match early clobber
8582 operand with itself. Check conflicts with earlyclobber only if
8583 the operand is not reloaded. Prefer to reload conflicting operand
8584 if earlyclobber and matching operands are the same.
8586 2013-02-12 Richard Biener <rguenther@suse.de>
8589 * lto-streamer-out.c (write_symbol): Do not output symbols
8590 for hard register variables.
8592 2013-02-12 Georg-Johann Lay <avr@gjlay.de>
8595 * config/avr/avr-dimode.md (umulsidi3, mulsidi3): New expanders.
8596 (umulsidi3_insn, mulsidi3_insn): New insns.
8598 2013-02-12 Christophe Lyon <christophe.lyon@linaro.org>
8600 * config/arm/arm-protos.h (struct cpu_vec_costs): New struct type.
8601 (struct tune_params): Add vec_costs field.
8602 * config/arm/arm.c (arm_builtin_vectorization_cost)
8603 (arm_add_stmt_cost): New functions.
8604 (TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST)
8605 (TARGET_VECTORIZE_ADD_STMT_COST): Define.
8606 (arm_default_vec_cost): New struct of type cpu_vec_costs.
8607 (arm_slowmul_tune, arm_fastmul_tune, arm_strongarm_tune)
8608 (arm_xscale_tune, arm_9e_tune, arm_v6t2_tune, arm_cortex_tune)
8609 (arm_cortex_a15_tune, arm_cortex_a5_tune, arm_cortex_a9_tune)
8610 (arm_v6m_tune, arm_fa726te_tune): Define new vec_costs field.
8612 2013-02-12 Richard Biener <rguenther@suse.de>
8615 * gimple-streamer-in.c (input_gimple_stmt): Strip MEM_REFs off
8616 decls again if possible.
8618 2013-02-12 Richard Biener <rguenther@suse.de>
8621 * tree-ssa.c (verify_ssa_name): Fix check, move
8622 SSA_NAME_IN_FREE_LIST check up.
8624 2013-02-12 Jakub Jelinek <jakub@redhat.com>
8625 Steven Bosscher <steven@gcc.gnu.org>
8627 PR rtl-optimization/56151
8628 * optabs.c (add_equal_note): Don't return 0 if target is a MEM,
8629 equal to op0 or op1, and last_insn pattern is CODE operation
8630 with MEM dest and one of the operands matches that MEM.
8632 2013-02-11 Sriraman Tallam <tmsriramgoogle.com>
8634 * doc/extend.texi: Document Function Multiversioning and "default"
8635 parameter string to target attribute.
8636 * config/i386/i386.c (get_builtin_code_for_version): Return 0 if
8637 target attribute parameter is "default".
8638 (ix86_compare_version_priority): Remove checks for target attribute.
8639 (ix86_mangle_function_version_assembler_name): Change error to sorry.
8640 Remove check for target attribute equal to NULL. Add assert.
8641 (ix86_generate_version_dispatcher_body): Change error to sorry.
8643 2013-02-11 Iain Sandoe <iain@codesourcery.com>
8644 Jack Howarth <howarth@bromo.med.uc.edu>
8645 Patrick Marlier <patrick.marlier@gmail.com>
8648 * config/darwin.h: Replace ENDFILE_SPEC with TM_DESTRUCTOR and
8649 define ENDFILE_SPEC as TM_DESTRUCTOR.
8650 * config/i386/darwin.h (ENDFILE_SPEC): Use TM_DESTRUCTOR.
8652 2013-02-11 Alexander Potapenko <glider@google.com>
8653 Jack Howarth <howarth@bromo.med.uc.edu>
8654 Jakub Jelinek <jakub@redhat.com>
8657 * config/darwin.c (cdtor_record): Rename ctor_record.
8658 (sort_cdtor_records): Rename sort_ctor_records.
8659 (finalize_dtors): New routine to sort destructors by
8660 priority before use in assemble_integer.
8661 (machopic_asm_out_destructor): Use finalize_dtors if needed.
8663 2013-02-11 Uros Bizjak <ubizjak@gmail.com>
8665 PR rtl-optimization/56275
8666 * simplify-rtx.c (avoid_constant_pool_reference): Check that
8667 offset is non-negative and less than cmode size before
8668 calling simplify_subreg.
8670 2013-02-11 Richard Biener <rguenther@suse.de>
8672 PR tree-optimization/56264
8673 * cfgloop.h (fix_loop_structure): Adjust prototype.
8674 * loop-init.c (fix_loop_structure): Return the number of
8675 newly discovered loops.
8676 * tree-cfgcleanup.c (repair_loop_structures): When new loops
8677 are discovered, do a full loop-closed SSA rewrite.
8679 2013-02-11 Richard Biener <rguenther@suse.de>
8681 PR tree-optimization/56273
8682 * tree-vrp.c (simplify_cond_using_ranges): Disable for the
8684 (check_array_ref): Fix missing newline in dumps.
8685 (search_for_addr_array): Likewise.
8687 2013-02-09 David Edelsohn <dje.gcc@gmail.com>
8689 * config/rs6000/aix61.h (OS_MISSING_ALTIVEC): Undefine.
8691 2013-02-09 Jakub Jelinek <jakub@redhat.com>
8694 * config/rs6000/rs6000.h (ASSEMBLER_DIALECT): Define.
8696 2013-02-08 Vladimir Makarov <vmakarov@redhat.com>
8698 PR rtl-optimization/56246
8699 * lra-constraints.c (simplify_operand_subreg): Try to reuse
8701 * lra.c (lra): Clear lra_optional_reload_pseudos only when all
8702 constraints are satisfied.
8704 2013-02-08 Jeff Law <law@redhat.com>
8707 * emit-rtl.c (reg_is_parm_p): New function.
8708 * regs.h (reg_is_parm_p): New prototype.
8709 * ira-conflicts.c (ira_build_conflicts): Allow parameters in
8710 callee-clobbered registers.
8712 2013-02-08 Michael Meissner <meissner@linux.vnet.ibm.com>
8715 * config/rs6000/rs6000.c (rs6000_builtin_vectorized_libmass):
8716 If there is no implicit builtin declaration, just return NULL.
8718 2013-02-08 Uros Bizjak <ubizjak@gmail.com>
8720 * config/i386/sse.md (FMAMODEM): New mode iterator.
8721 (fma<mode>4, fms<mode>4, fnma<mode>4, fnms<mode>4): Use FMAMODEM
8722 mode iterator. Do not use TARGET_SSE_MATH in insn constraint.
8724 2013-02-08 Uros Bizjak <ubizjak@gmail.com>
8726 * config/i386/gnu-user.h (TARGET_CAN_SPLIT_STACK): Define only
8727 when HAVE_GAS_CFI_PERSONALITY_DIRECTIVE is set.
8728 * config/i386/gnu-user64.h (TARGET_CAN_SPLIT_STACK): Ditto.
8730 2013-02-08 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8732 * config.gcc (microblaze*-linux*): Add TARGET_BIG_ENDIAN_DEFAULT.
8733 (microblaze*-*-elf): Likewise.
8734 * config/microblaze/linux.h: Add -mbig-endian / -mlittle-endian to
8736 * config/microblaze/microblaze-c.c: Add builtin defines for
8737 _LITTLE_ENDIAN and _BIG_ENDIAN.
8738 * config/microblaze/microblaze.h: Add TARGET_ENDIAN_DEFAULT and
8739 add to TARGET_DEFAULT flags.
8740 Expand ASM_SPEC and LINK_SPEC.
8741 Update BYTES_BIG_ENDIAN and WORDS_BIG_ENDIAN.
8742 * config/microblaze/microblaze.md: Update extendsidi2 and
8743 movdi_internal instructions to use low-order / high-order reg
8745 * config/microblaze/microblaze.opt: Add mbig-endian and mlittle-endian
8746 options and inversemask / mask of LITTLE_ENDIAN.
8747 * config/microblaze/t-microblaze: Expand multilib options to
8748 include mlittle-endian (le) and update exceptions patterns.
8750 2013-02-08 Jakub Jelinek <jakub@redhat.com>
8752 PR rtl-optimization/56195
8753 * lra-constraints.c (get_reload_reg): Don't reuse regs
8754 if they have smaller mode than requested, if they have
8755 wider mode than requested, try to return a SUBREG.
8757 PR tree-optimization/56250
8758 * fold-const.c (extract_muldiv_1) <case NEGATE_EXPR>: Don't optimize
8759 if type is unsigned and code isn't MULT_EXPR.
8761 2013-02-08 Georg-Johann Lay <avr@gjlay.de>
8763 PR tree-optimization/56064
8764 * fixed-value.c (fixed_from_double_int): Sign/zero extend payload
8765 bits according to mode.
8766 * fixed-value.h (fixed_from_double_int)
8767 (const_fixed_from_double_int): Adjust comments.
8769 2013-02-08 Richard Biener <rguenther@suse.de>
8772 * lto-streamer.h (struct data_in): Remove current_file, current_line
8773 and current_col members.
8774 * lto-streamer-out.c (lto_output_location): Stream changed bits
8775 en-block for efficiency.
8776 * lto-streamer-in.c (clear_line_info): Remove.
8777 (lto_input_location): Cache current file, line and column
8778 globally via local statics. Read changed bits en-block.
8779 (input_function): Do not call clear_line_info.
8780 (lto_read_body): Likewise.
8781 (lto_input_toplevel_asms): Likewise.
8783 2013-02-08 Michael Matz <matz@suse.de>
8785 PR tree-optimization/52448
8786 * tree-ssa-phiopt.c (struct name_to_bb): Add phase member.
8787 (nt_call_phase): New static.
8788 (add_or_mark_expr): Only mark accesses with newer phase than any
8790 (nonfreeing_call_p): New.
8791 (nt_init_block): Update nt_call_phase, mark blocks as visited.
8792 (nt_fini_block): Keep blocks marked as visited.
8793 (get_non_trapping): Initialize nt_call_phase, and reset aux pointer.
8795 2013-02-08 Richard Biener <rguenther@suse.de>
8797 * ira.c (ira): Free broken dominator information.
8799 2013-02-08 Uros Bizjak <ubizjak@gmail.com>
8801 * config/i386/i386.c (ix86_spill_class): Use INTEGER_CLASS_P macro.
8803 2013-02-08 Marek Polacek <polacek@redhat.com>
8805 * cfgloop.c (verify_loop_structure): Add more checking of headers.
8807 2013-02-08 Richard Biener <rguenther@suse.de>
8810 * cfgloop.h (flow_loops_find): Adjust.
8811 (bb_loop_header_p): Declare.
8812 * cfgloop.c (bb_loop_header_p): New function split out from ...
8813 (flow_loops_find): ... here. Adjust function signature,
8814 support incremental loop structure update.
8815 (verify_loop_structure): Cleanup. Verify a loop is a loop.
8816 * cfgloopmanip.c (fix_loop_structure): Move ...
8817 * loop-init.c (fix_loop_structure): ... here.
8818 (apply_loop_flags): Split out from ...
8819 (loop_optimizer_init): ... here.
8820 (fix_loop_structure): Use apply_loop_flags. Use flow_loops_find
8821 in incremental mode, only remove dead loops here.
8823 2013-02-08 Georg-Johann Lay <avr@gjlay.de>
8826 * config/avr/avr.md (unspec) <UNSPEC_ROUND>: Add.
8827 * config/avr/avr-fixed.md (ALL4QA, ALL124QA): New mode iterators.
8828 (round<mode>3, round<mode>3_const): New expanders for fixed-mode.
8829 (*round<mode>3.libgcc): New insns for fixed-modes.
8830 * config/avr/builtins.def (ABSxx): Use a non-NULL LIBNAME.
8831 (ROUNDxx, COUNTLSxx, BITSxx, xxBITS): New DEF_BUILTINs.
8832 (ROUNDFX, COUNTLSFX, ABSFX): New DEF_BUILTINs.
8833 * config/avr/stdfix.h (absFX, bitsFX, FXbits): Remove inline
8834 implementations. Define to __builtin_avr_absFX,
8835 __builtin_avr_bitsFX, __builtin_avr_FXbits, respectively.
8836 (roundFX, countlsFX): Define to __builtin_avr_roundFX,
8837 __builtin_avr_countlsFX, respectively.
8838 * config/avr/avr-c.c (target.h): Include it.
8839 (enum avr_builtin_id): New enum.
8840 (avr_resolve_overloaded_builtin): New static function.
8841 (avr_register_target_pragmas): Use it to set
8842 targetm.resolve_overloaded_builtin.
8843 * config/avr/avr.c (avr_init_builtins): Supply myriads of local
8844 tree nodes used by DEF_BUILTIN.
8845 (avr_expand_builtin) <AVR_BUILTIN_ROUNDxx>: Sanity-check them.
8846 (avr_fold_builtin) <AVR_BUILTIN_BITSxx>: Fold to VIEW_COVERT_EXPR.
8847 <AVR_BUILTIN_xxBITS>: Same.
8849 2013-02-08 Richard Biener <rguenther@suse.de>
8851 * cfgloop.c (verify_loop_structure): Properly handle
8852 a loop exiting to another loop header.
8853 * ira-int.h (ira_loops): Remove.
8854 * ira.c (ira_loops): Remove.
8855 (ira): Use loop_optimizer_init and loop_optimizer_finalize.
8856 (do_reload): Use loop_optimizer_finalize.
8857 * ira-build.c (create_loop_tree_nodes): Use get_loops and
8858 number_of_loops to access the loop tree.
8859 (more_one_region_p): Likewise.
8860 (finish_loop_tree_nodes): Likewise.
8861 (rebuild_regno_allocno_maps): Likewise.
8862 (mark_loops_for_removal): Likewise.
8863 (mark_all_loops_for_removal): Likewise.
8864 (remove_unnecessary_regions): Likewise.
8865 (ira_build): Likewise.
8866 * ira-emit.c (setup_entered_from_non_parent_p): Likewise.
8868 2013-02-08 Richard Biener <rguenther@suse.de>
8870 * Makefile.in (tree-tailcall.o): Add $(CFGLOOP_H) dependency.
8871 * ipa-pure-const.c (analyze_function): Avoid calling
8872 mark_irreducible_loops twice.
8873 * tree-tailcall.c (tree_optimize_tail_calls_1): Mark loops for fixup.
8875 2013-02-07 David S. Miller <davem@davemloft.net>
8877 * dwarf2out.c (based_loc_descr): Perform leaf register remapping
8879 * var-tracking.c (vt_add_function_parameter): Test the presence of
8880 HAVE_window_save properly and do not remap argument registers when
8881 we have a leaf function.
8883 2013-02-07 Uros Bizjak <ubizjak@gmail.com>
8886 * ggc-page.c (ggc_print_statistics): Use HOST_LONG_LONG_FORMAT
8888 * config/i386/i386.c (ix86_print_operand): Ditto.
8890 2013-02-07 Vladimir Makarov <vmakarov@redhat.com>
8892 * lra-constraints.c (process_alt_operands): Fix recently added comment.
8894 2013-02-07 Vladimir Makarov <vmakarov@redhat.com>
8896 PR rtl-optimization/56225
8897 * lra-constraints.c (process_alt_operands): Check that reload hard
8898 reg can hold value for strict_low_part.
8900 2013-02-07 Jakub Jelinek <jakub@redhat.com>
8903 * dwarf2out.c (dwarf2_debug_hooks): Set end_function hook to
8904 dwarf2out_end_function.
8905 (in_first_function_p, maybe_at_text_label_p,
8906 first_loclabel_num_not_at_text_label): New variables.
8907 (dwarf2out_var_location): In the first function find out
8908 lowest loclabel_num N where .LVLN is known not to be equal to .Ltext0.
8909 (find_empty_loc_ranges_at_text_label, dwarf2out_end_function): New
8912 2013-02-07 Eric Botcazou <ebotcazou@adacore.com>
8914 PR rtl-optimization/56178
8915 * cse.c (cse_insn): Do not create a REG_EQUAL note if the source is a
8916 SUBREG of a register. Tidy up related block of code.
8917 * fwprop.c (forward_propagate_and_simplify): Do not create a REG_EQUAL
8918 note if the source is a register or a SUBREG of a register.
8920 2013-02-07 Jakub Jelinek <jakub@redhat.com>
8923 * config/rs6000/rs6000.md (ptrm): New mode attr.
8924 (call_indirect_aix<ptrsize>, call_indirect_aix<ptrsize>_nor11,
8925 call_value_indirect_aix<pttrsize>,
8926 call_value_indirect_aix<pttrsize>_nor11): Use <ptrm> instead of
8929 2013-02-07 Michael Haubenwallner <michael.haubenwallner@salomon.at>
8931 * collect2.c (main): Set aix64_flag for -G and -bsvr4 too, disable
8932 if -bnortl. Convert to strcmp and strncmp.
8934 2013-02-07 Alan Modra <amodra@gmail.com>
8937 * config/rs6000/rs6000.c (mem_operand_gpr): Check that LO_SUM
8938 addresses won't wrap when offsetting.
8939 (rs6000_secondary_reload): Provide secondary reloads needed for
8940 wrapping LO_SUM addresses.
8942 2013-02-06 Thomas Schwinge <thomas@codesourcery.com>
8944 * config/gnu.h (GNU_USER_TARGET_OS_CPP_BUILTINS): Never define
8945 MACH, just __MACH__.
8947 2013-02-06 Richard Biener <rguenther@suse.de>
8949 * tracer.c (tracer): Mark loops with LOOPS_NEED_FIXUP
8950 instead of calling fix_loop_structure.
8952 2013-02-06 Jakub Jelinek <jakub@redhat.com>
8955 * omp-low.c (use_pointer_for_field): Return false if
8956 lower_send_shared_vars doesn't generate any copy-out code.
8958 2013-02-06 Tom de Vries <tom@codesourcery.com>
8960 PR rtl-optimization/56131
8961 * cfgrtl.c (delete_insn): Use NOTE_BASIC_BLOCK instead of BLOCK_FOR_INSN
8962 to get the bb of a NOTE_INSN_BASIC_BLOCK. Handle the case that the bb
8963 of the label is NULL. Add comment.
8965 2013-02-05 Jakub Jelinek <jakub@redhat.com>
8967 * tree.h (struct tree_decl_with_vis): Remove thread_local field.
8970 * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Define.
8971 (STATIC_LIBTSAN_LIBS): Likewise.
8972 * gcc.c (ADD_STATIC_LIBTSAN_LIBS, LIBTSAN_EARLY_SPEC): Define.
8973 (LIBTSAN_SPEC): Add ADD_STATIC_LIBTSAN_LIBS, if LIBTSAN_EARLY_SPEC
8974 is defined, don't add anything else beyond that.
8975 (SANITIZER_EARLY_SPEC, SANITIZER_SPEC): Define.
8976 (LINK_COMMAND_SPEC): Use them.
8978 PR tree-optimization/56205
8979 * tree-stdarg.c (check_all_va_list_escapes): Return true if
8980 there are any PHI nodes that set non-va_list_escape_vars SSA_NAME
8981 and some va_list_escape_vars SSA_NAME appears in some PHI argument.
8983 2013-02-05 Richard Biener <rguenther@suse.de>
8985 PR tree-optimization/53342
8986 PR tree-optimization/53185
8987 * tree-vectorizer.h (vect_check_strided_load): Remove.
8988 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Do
8989 not disallow peeling for vectorized strided loads.
8990 (vect_check_strided_load): Make static and simplify.
8991 (vect_analyze_data_refs): Adjust.
8992 * tree-vect-stmts.c (vectorizable_load): Handle peeled loops
8993 correctly when vectorizing strided loads.
8995 2013-02-05 Richard Biener <rguenther@suse.de>
8997 * doc/install.texi: Refer to ISL, not PPL.
8999 2013-02-05 Jan Hubicka <jh@suse.cz>
9001 PR tree-optimization/55789
9002 * params.def (PARAM_EARLY_INLINER_MAX_ITERATIONS): Drop to 1.
9004 2013-02-05 Jan Hubicka <jh@suse.cz>
9006 PR tree-optimization/55789
9007 * cgraphclones.c (cgraph_remove_node_and_inline_clones): Remove
9008 the dead call anyway.
9010 2013-02-05 Eric Botcazou <ebotcazou@adacore.com>
9013 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Add missing guard.
9015 2013-02-04 Alexander Potapenko <glider@google.com>
9016 Jack Howarth <howarth@bromo.med.uc.edu>
9017 Jakub Jelinek <jakub@redhat.com>
9020 * config/darwin.c (sort_ctor_records): Stabilized qsort
9021 on constructor priority by using original position.
9022 (finalize_ctors): New routine to sort constructors by
9023 priority before use in assemble_integer.
9024 (machopic_asm_out_constructor): Use finalize_ctors if needed.
9026 2013-02-04 Jakub Jelinek <jakub@redhat.com>
9029 * config/i386/winnt.c (i386_pe_assemble_visibility): Don't warn
9030 about visibility on artificial decls.
9031 * config/sol2.c (solaris_assemble_visibility): Likewise.
9033 2013-02-04 Kai Tietz <ktietz@redhat.com>
9036 * config/i386/i386.c (function_value_ms_64): Add additional valtype
9037 argument and improve checking of return-argument types for 16-byte
9039 (ix86_function_value_1): Add additional valtype argument on call
9040 of function_value_64.
9041 (return_in_memory_ms_64): Sync 16-byte sized mode handling with
9042 handling infunction_value_64 function.
9044 2013-02-04 Matthew Gretton-Dann <matthew.gretton-dann@linaro.org>
9046 * reload.c (subst_reloads): Fix DEBUG_RELOAD build issue.
9048 2013-02-04 Richard Biener <rguenther@suse.de>
9050 PR tree-optimization/56188
9051 * tree-ssa-structalias.c (label_visit): Consider case with
9052 initially non-empty points-to set.
9053 (perform_var_substitution): Dump node mapping and clean up.
9055 2013-02-04 Richard Guenther <rguenther@suse.de>
9058 * lto-symtab.c (lto_symtab_merge_decls_1): Make non-builtin
9059 node prevail as last resort.
9060 (lto_symtab_merge_decls): Remove guard on LTRANS here.
9061 (lto_symtab_prevailing_decl): Builtins are their own prevailing decl.
9063 2013-02-04 Richard Biener <rguenther@suse.de>
9065 PR tree-optimization/56113
9066 * tree-ssa-structalias.c (equiv_class_lookup, equiv_class_add):
9068 (equiv_class_lookup_or_add): ... this.
9069 (label_visit): Adjust and fix error in previous patch.
9070 (perform_var_substitution): Adjust.
9072 2013-02-03 Oleg Endo <olegendo@gcc.gnu.org>
9074 * config/sh/divtab.c: Fix formatting and comments throughout the file.
9075 * config/sh/sh4-300.md: Likewise.
9076 * config/sh/sh4a.md: Likewise.
9077 * config/sh/constraints.md: Likewise.
9078 * config/sh/sh.md: Likewise.
9079 * config/sh/netbsd-elf.h: Likewise.
9080 * config/sh/predicates.md: Likewise.
9081 * config/sh/sh-protos.h: Likewise.
9082 * config/sh/ushmedia.h: Likewise.
9083 * config/sh/linux.h: Likewise.
9084 * config/sh/sh.c: Likewise.
9085 * config/sh/superh.h: Likewise.
9086 * config/sh/elf.h: Likewise.
9087 * config/sh/sh4.md: Likewise.
9088 * config/sh/sh.h: Likewise.
9090 2013-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
9092 * config/pa/constraints.md: Adjust unused letters. Change "T"
9093 constraint to match_test floating_point_store_memory_operand().
9094 * config/pa/predicates.md (reg_plus_base_memory_operand): New.
9095 (base14_operand): New.
9096 (floating_point_store_memory_operand): New.
9097 (integer_store_memory_operand): Revise to use base14_operand and
9098 reg_plus_base_memory_operand.
9099 (move_dest_operand): Allow symbolic_memory_operands.
9100 (symbolic_memory_operand): Check for LO_SOM.
9101 (symbolic_operand): Change default case to break.
9102 * config/pa/pa.md: Remove unamed DFmode and SFmode patterns to force
9103 CONST_DOUBLE values to be reloaded by putting them into memory when
9104 the destination is a floating point register.
9105 (movdf): Remove code to handle CONST_DOUBLE.
9107 (reload_indf_r1): New.
9108 (reload_insf_r1): New.
9109 Consistently use "Q" and "T" constraints with integer and floating
9110 point move instructions, respectively.
9111 (movdi): Remove FAIL.
9112 Change predicate for source operand unamed DImode move from
9113 general_operand to move_src_operand.
9114 (umulsidi3): Change predicate for destination operand to
9116 Likewise for similar unamed patterns.
9117 * config/pa/pa-protos.h (pa_legitimize_reload_address): Declare.
9118 * config/pa/pa.c (pa_symbolic_expression_p): Remove extra parenthesis.
9119 (hppa_legitimize_address): Simplify mask calculation.
9120 (pa_emit_move_sequence): Revised handling of secondary reloads from
9121 REG+D addresses for floating point loads and stores. Directly handle
9122 loading CONST0_RTX (mode) to a floating point register.
9123 (pa_secondary_reload): Handle reloading DF and SFmode constant values
9124 to floating point registers. Don't restrict secondary reloads to
9125 floating point registers to integer modes. Revise some comments and
9127 (TARGET_LEGITIMATE_ADDRESS_P): Define.
9128 (pa_legitimate_address_p): New.
9129 (pa_legitimize_reload_address): New.
9130 * config/pa/pa.h (STRICT_REG_OK_FOR_INDEX_P): New.
9131 (STRICT_REG_OK_FOR_BASE_P): New.
9132 (GO_IF_LEGITIMATE_ADDRESS): Delete. Update some related comments.
9133 (LEGITIMIZE_RELOAD_ADDRESS): Revise to use pa_legitimize_reload_address.
9135 2013-02-03 David Edelsohn <dje.gcc@gmail.com>
9136 Andrew Dixie <andrewd@gentrack.com>
9138 * collect2.c (GCC_CHECK_HDR): Do not scan objects with F_LOADONLY
9141 2013-02-03 Richard Sandiford <rdsandiford@googlemail.com>
9143 * expmed.c (extract_bit_field_1): Pass the full width of the
9144 structure to get_best_reg_extraction_insn.
9146 2013-02-01 David Edelsohn <dje.gcc@gmail.com>
9149 * configure.ac (use_cxa_atexit): Add AIX.
9150 * configure: Regenerate.
9152 * config/rs6000/aix61.h (STARTFILE_SPEC): Add crtcxa.o.
9154 2013-02-01 Jakub Jelinek <jakub@redhat.com>
9157 * final.c (need_profile_function): New variable.
9158 (final_start_function): Drop ATTRIBUTE_UNUSED from first argument.
9159 If first of NOTE_INSN_BASIC_BLOCK or NOTE_INSN_FUNCTION_BEG
9160 is only preceeded by NOTE_INSN_VAR_LOCATION or NOTE_INSN_DELETED
9161 notes, targetm.asm_out.function_prologue doesn't emit anything,
9162 HAVE_prologue and profiler should be emitted before prologue,
9163 set need_profile_function instead of emitting it.
9164 (final_scan_insn): If need_profile_function, emit
9165 profile_function on the first NOTE_INSN_BASIC_BLOCK or
9166 NOTE_INSN_FUNCTION_BEG note.
9168 2013-02-01 Richard Henderson <rth@redhat.com>
9170 * config/rs6000/rs6000.md (smulditi3): New.
9173 * config/alpha/alpha.md (umulditi3): New.
9175 2013-02-01 David Edelsohn <dje.gcc@gmail.com>
9177 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_COMMON): Use floor_log2.
9178 (ASM_OUTPUT_ALIGNED_LOCAL): New.
9180 2013-02-01 Richard Biener <rguenther@suse.de>
9182 PR tree-optimization/56113
9183 * tree-ssa-structalias.c (label_visit): Reduce work for
9184 single-predecessor nodes.
9186 2013-02-01 Eric Botcazou <ebotcazou@adacore.com>
9188 * fold-const.c (make_range_step) <TRUTH_NOT_EXPR>: Bail out if the
9189 range isn't testing for zero.
9191 2013-01-31 Steven Bosscher <steven@gcc.gnu.org>
9194 * fwprop.c (fwprop_init): Set up loops without CFG modifications.
9196 2013-01-31 Hiroyuki Ono <hiroyuki.ono.jc@renesas.com>
9197 Nick Clifton <nickc@redhat.com>
9199 * config/v850/constraints.md (Q): Define as a memory constraint.
9200 * config/v850/predicates.md (label_ref_operand): New predicate.
9201 (e3v5_shift_operand): New predicate.
9202 (ior_operator): New predicate.
9203 * config/v850/t-v850: Add e3v5 multilib.
9204 * config/v850/v850-protos.h (v850_adjust_insn_length): Prototype.
9205 (v850_gen_movdi): Prototype.
9206 * config/v850/v850.c: Add support for e3v5 architecture.
9207 Rename all uses of TARGET_V850E || TARGET_V850E2_ALL to
9209 (construct_save_jarl): Add e3v5 long JARL support.
9210 (v850_adjust_insn_length): New function. Adjust length of call
9211 insns when using e3v5 instructions.
9212 (v850_gen_movdi): New function: Generate instructions to move a
9214 * config/v850/v850.h (TARGET_CPU_v850e3v5): Define.
9215 (CPP_SPEC): Define __v850e3v5__ as appropriate.
9216 (TARGET_USE_FPU): Enable for e3v5.
9217 (CONST_OK_FOR_W): New macro.
9218 (ADJUST_INSN_LENGTH): Define.
9219 * config/v850/v850.md (UNSPEC_LOOP): Define.
9220 (attr cpu): Add v850e3v5.
9221 Rename all uses of TARGET_V850E2 to TARGET_V850E2V3_UP.
9222 (movdi): New pattern.
9223 (movdi_internal): New pattern.
9224 (cbranchsf4): Conditionalize on TARGET_USE_FPU.
9225 (cbranchdf4): Conditionalize on TARGET_USE_FPU.
9226 (cstoresf4): Likewise.
9227 (cstoredf4): Likewise.
9228 (insv): New pattern.
9229 (rotlso3_a): New pattern.
9230 (rotlsi3_b): New pattern
9231 (rotlsi3_v850e3v5): New pattern.
9232 (doloop_begin): New pattern.
9233 (fix_loop_counter): New pattern.
9234 (doloop_end): New pattern.
9235 (branch_normal): Add e3v5 long branch support.
9236 (branch_invert): Likewise.
9237 (branch_z_normal): Likewise.
9238 (branch_z_invert): Likewise.
9239 (branch_nz_normal): Likewise.
9240 (branch_nz_invert): Likewise.
9241 (call_internal_short): Add e3v5 register-indirect JARL support.
9242 (call_internal_long): Likewise.
9243 (call_value_internal_short): Likewise.
9244 (call_value_internal_long): Likewise.
9245 * config/v850/v850.opt (mv850e3v5, mv850e2v4): New options.
9246 (mloop): New option.
9247 * config.gcc: Add support for configuring v840e3v5 target.
9248 * doc/invoke.texi: Document new v850 specific command line options.
9250 2013-01-31 Paul Koning <ni1d@arrl.net>
9254 * dwarf2out.c (prune_unused_types_mark): Mark all of parent's
9255 children if parent is a class.
9256 (prune_unused_types_prune): Don't add DW_AT_declaration.
9258 2013-01-31 Richard Biener <rguenther@suse.de>
9260 PR tree-optimization/56157
9261 * tree-vect-slp.c (vect_get_slp_defs): More thoroughly try to
9262 match up operand with SLP child.
9264 2013-01-31 Jason Merrill <jason@redhat.com>
9267 * dwarf2out.c (gen_struct_or_union_type_die): Always schedule template
9268 parameters the first time.
9269 (gen_scheduled_generic_parms_dies): Check completeness here.
9271 2013-01-31 Richard Biener <rguenther@suse.de>
9274 * common.opt (faggressive-loop-optimizations): New flag,
9276 * doc/invoke.texi (faggressive-loop-optimizations): Document.
9277 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Guard
9278 infer_loop_bounds_from_undefined by it.
9280 2013-01-31 Richard Biener <rguenther@suse.de>
9282 PR tree-optimization/56150
9283 * tree-ssa-loop-manip.c (find_uses_to_rename_stmt): Do not
9284 visit virtual operands.
9285 (find_uses_to_rename_bb): Likewise.
9287 2013-01-31 Richard Biener <rguenther@suse.de>
9289 PR tree-optimization/56150
9290 * tree-ssa-tail-merge.c (gimple_equal_p): Properly handle
9291 mixed store non-store stmts.
9293 2013-01-30 Jakub Jelinek <jakub@redhat.com>
9296 * gcc.c (LIBASAN_SPEC): Define just to ADD_STATIC_LIBASAN_LIBS if
9297 LIBASAN_EARLY_SPEC is defined.
9298 (LIBASAN_EARLY_SPEC): Define to empty string if not already defined.
9299 (LINK_COMMAND_SPEC): Add LIBASAN_EARLY_SPEC for -fsanitize=address,
9301 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Define.
9304 * config/i386/i386.c (ix86_valid_target_attribute_inner_p): Diagnose
9305 invalid args instead of ICEing on it.
9306 (ix86_valid_target_attribute_tree): Return error_mark_node if
9307 ix86_valid_target_attribute_inner_p failed.
9308 (ix86_valid_target_attribute_p): Return false only if
9309 ix86_valid_target_attribute_tree returned error_mark_node. Allow
9310 target("default") attribute.
9311 (sorted_attr_string): Change argument from const char * to tree,
9312 merge in all target attribute arguments rather than just one.
9313 Formatting fix. Use XNEWVEC instead of xmalloc and XDELETEVEC
9314 instead of free. Avoid using strcat.
9315 (ix86_mangle_function_version_assembler_name): Mangle
9316 target("default") as if no target attribute is present. Adjust
9317 sorted_attr_string caller. Avoid leaking memory. Use XNEWVEC
9318 instead of xmalloc and XDELETEVEC instead of free.
9319 (ix86_function_versions): Don't return true if one of the decls
9320 doesn't have target attribute. If they don't and one of the decls
9321 is DECL_FUNCTION_VERSIONED, report an error. Adjust
9322 sorted_attr_string caller. Use XDELETEVEC instead of free.
9323 (ix86_supports_function_versions): Remove.
9324 (make_name): Fix up formatting.
9325 (make_dispatcher_decl): Remove resolver_name and its initialization.
9326 Avoid leaking memory.
9327 (is_function_default_version): Return true if there is
9328 target("default") attribute rather than no target attribute at all.
9329 (make_resolver_func): Avoid leaking memory.
9330 (ix86_generate_version_dispatcher_body): Likewise.
9331 (TARGET_OPTION_SUPPORTS_FUNCTION_VERSIONS): Remove.
9332 * target.def (supports_function_versions): Remove.
9333 * doc/tm.texi.in (SUPPORTS_FUNCTION_VERSIONS): Remove.
9334 * doc/tm.texi: Regenerated.
9336 2013-01-30 Vladimir Makarov <vmakarov@redhat.com>
9338 PR rtl-optimization/56144
9339 * lra-constraints.c (get_reload_reg): Don't reuse reload pseudo
9340 for values with side effects.
9342 2013-01-30 Richard Biener <rguenther@suse.de>
9344 * sparseset.h (sparseset_bit_p): Use gcc_checking_assert.
9345 (sparseset_pop): Likewise.
9346 * cfganal.c (compute_idf): Likewise. Increase work-stack size
9347 to be able to use quick_push in the worker loop.
9349 2013-01-30 Marek Polacek <polacek@redhat.com>
9351 * cfgcleanup.c (cleanup_cfg): Don't mark affected BBs.
9353 2013-01-30 Richard Biener <rguenther@suse.de>
9356 * lto-symtab.c (lto_symtab_merge_decls_1): Guard DECL_BUILT_IN check.
9358 2013-01-30 Georg-Johann Lay <avr@gjlay.de>
9360 PR tree-optimization/56064
9361 * fixed-value.c (fixed_from_double_int): New function.
9362 * fixed-value.h (fixed_from_double_int): New prototype.
9363 (const_fixed_from_double_int): New static inline function.
9364 * fold-const.c (native_interpret_fixed): New static function.
9365 (native_interpret_expr) <FIXED_POINT_TYPE>: Use it.
9366 (can_native_interpret_type_p) <FIXED_POINT_TYPE>: Return true.
9367 (native_encode_fixed): New static function.
9368 (native_encode_expr) <FIXED_CST>: Use it.
9369 (native_interpret_int): Move double_int worker code to...
9370 * double-int.c (double_int::from_buffer): ...this new static method.
9371 * double-int.h (double_int::from_buffer): Prototype it.
9373 2013-01-30 Richard Biener <rguenther@suse.de>
9375 * tree-ssa-structalias.c (final_solutions, final_solutions_obstack):
9376 New pointer-map and obstack.
9377 (init_alias_vars): Allocate pointer-map and obstack.
9378 (delete_points_to_sets): Free them.
9379 (find_what_var_points_to): Cache result.
9380 (find_what_p_points_to): Adjust for changed interface of
9381 find_what_var_points_to.
9382 (compute_points_to_sets): Likewise.
9383 (ipa_pta_execute): Likewise.
9385 2013-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9387 * configure.ac (HAVE_AS_SPARC_NOBITS): New test.
9388 * configure: Regenerate.
9389 * config.in: Regenerate.
9390 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section): Emit
9391 #nobits/#progbits if supported.
9393 2013-01-29 Oleg Endo <olegendo@gcc.gnu.org>
9396 * config/sh/sh.md (bclr_m2a, bset_m2a, bst_m2a, bld_m2a, bldsign_m2a,
9397 bld_reg, *bld_regqi, band_m2a, bandreg_m2a, bor_m2a, borreg_m2a,
9398 bxor_m2a, bxorreg_m2a): Add satisfies_constraint_K03 condition.
9400 2013-01-29 Greta Yorsh <Greta.Yorsh@arm.com>
9402 * config/arm/cortex-a7.md (cortex_a7_neon, cortex_a7_all): Remove.
9403 (cortex_a7_idiv): Use cortex_a7_both instead of cortex_a7_all.
9405 2013-01-29 Greta Yorsh <Greta.Yorsh@arm.com>
9407 * config/arm/arm.c (cortexa7_younger): Return true for TYPE_CALL.
9408 * config/arm/cortex-a7.md (cortex_a7_call): Update required units.
9410 2013-01-29 Greta Yorsh <Greta.Yorsh@arm.com>
9412 * config/arm/arm-protos.h (arm_mac_accumulator_is_result): New
9414 * config/arm/arm.c (arm_mac_accumulator_is_result): New function.
9415 * config/arm/cortex-a7.md: New bypasses using
9416 arm_mac_accumulator_is_result.
9418 2013-01-29 Greta Yorsh <Greta.Yorsh@arm.com>
9420 * config/arm/cortex-a7.md (cortex_a7_neon_mul): New reservation.
9421 (cortex_a7_neon_mla): Likewise.
9422 (cortex_a7_fpfmad): New reservation.
9423 (cortex_a7_fpmacs): Use ffmas and update required units.
9424 (cortex_a7_fpmuld): Update required units and latency.
9425 (cortex_a7_fpmacd): Likewise.
9426 (cortex_a7_fdivs, cortex_a7_fdivd): Likewise.
9427 (cortex_a7_neon). Likewise.
9428 (bypass) Update participating units.
9430 2013-01-29 Greta Yorsh <Greta.Yorsh@arm.com>
9432 * config/arm/arm.md (type): Add ffmas and ffmad to "type" attribute.
9433 * config/arm/vfp.md (fma,fmsub,fnmsub,fnmadd): Change type
9435 * config/arm/vfp11.md (vfp_farith): Use ffmas.
9436 (vfp_fmul): Use ffmad.
9437 * config/arm/cortex-r4f.md (cortex_r4_fmacs): Use ffmas.
9438 (cortex_r4_fmacd): Use ffmad.
9439 * config/arm/cortex-m4-fpu.md (cortex_m4_fmacs): Use ffmas.
9440 * config/arm/cortex-a9.md (cortex_a9_fmacs): Use ffmas.
9441 (cortex_a9_fmacd): Use ffmad.
9442 * config/arm/cortex-a8-neon.md (cortex_a8_vfp_macs): Use ffmas.
9443 (cortex_a8_vfp_macd): Use ffmad.
9444 * config/arm/cortex-a5.md (cortex_a5_fpmacs): Use ffmas.
9445 (cortex_a5_fpmacd): Use ffmad.
9446 * config/arm/cortex-a15-neon.md (cortex_a15_vfp_macs) Use ffmas.
9447 (cortex_a15_vfp_macd): Use ffmad.
9448 * config/arm/arm1020e.md (v10_fmul): Use ffmas and ffmad.
9450 2013-01-29 Jason Merrill <jason@redhat.com>
9453 * varasm.c (default_assemble_visibility): Don't warn about
9454 visibility on artificial decls.
9456 2013-01-29 Richard Biener <rguenther@suse.de>
9458 PR tree-optimization/56113
9459 * tree-ssa-structalias.c (equiv_class_lookup): Also return
9461 (label_visit): Free duplicate bitmaps and record the leader instead.
9462 (perform_var_substitution): Adjust.
9464 2013-01-29 Richard Biener <rguenther@suse.de>
9466 PR tree-optimization/55270
9467 * tree-ssa-dom.c (eliminate_degenerate_phis): If we changed
9468 the CFG, schedule loops for fixup.
9470 2013-01-29 Nick Clifton <nickc@redhat.com>
9472 * config/rl78/rl78.c (rl78_regno_mode_code_ok_for_base_p): Allow
9475 2013-01-28 Leif Ekblad <leif@rdos.net>
9477 * config.gcc (i[34567]86-*-rdos*, x86_64-*-rdos*): New targets.
9478 * config/i386/i386.h (TARGET_RDOS): New macro.
9479 (DEFAULT_LARGE_SECTION_THRESHOLD): New macro.
9480 * config/i386/i386.c (ix86_option_override_internal): For 64bit
9481 TARGET_RDOS, set ix86_cmodel to CM_MEDIUM_PIC and flag_pic to 1.
9482 * config/i386/i386.opt (mlarge-data-threshold): Initialize to
9483 DEFAULT_LARGE_SECTION_THRESHOLD.
9484 * config/i386/i386.md (R14_REG, R15_REG): New constants.
9485 * config/i386/rdos.h: New file.
9486 * config/i386/rdos64.h: New file.
9488 2013-01-28 Bernd Schmidt <bernds@codesourcery.com>
9491 * reload.c (find_valid_class_1): Use in_hard_reg_set_p instead of
9494 2013-01-28 Jakub Jelinek <jakub@redhat.com>
9496 PR rtl-optimization/56117
9497 * sched-deps.c (sched_analyze_2) <case PREFETCH>: For use_cselib
9498 call cselib_lookup_from_insn on the MEM before calling
9499 add_insn_mem_dependence.
9501 2013-01-28 Richard Biener <rguenther@suse.de>
9503 * tree-inline.c (remap_gimple_stmt): Do not assing a BLOCK
9504 to a stmt that didn't have one.
9505 (copy_phis_for_bb): Likewise for PHI arguments.
9506 (copy_debug_stmt): Likewise for debug stmts.
9508 2013-01-28 Richard Biener <rguenther@suse.de>
9510 PR tree-optimization/56034
9511 * tree-loop-distribution.c (enum partition_kind): Add PKIND_REDUCTION.
9512 (partition_builtin_p): Adjust.
9513 (generate_code_for_partition): Handle PKIND_REDUCTION. Assert
9514 it is the last partition.
9515 (rdg_flag_uses): Check SSA_NAME_IS_DEFAULT_DEF before looking
9516 up the vertex for the definition.
9517 (classify_partition): Classify whether a partition is a
9518 PKIND_REDUCTION, thus has uses outside of the loop.
9519 (ldist_gen): Inherit PKIND_REDUCTION when merging partitions.
9520 Merge all PKIND_REDUCTION partitions into the last partition.
9521 (tree_loop_distribution): Seed partitions from reductions as well.
9523 2013-01-28 Jakub Jelinek <jakub@redhat.com>
9525 PR tree-optimization/56125
9526 * tree-ssa-math-opts.c (gimple_expand_builtin_pow): Don't optimize
9527 pow(x,c) into sqrt(x) * powi(x, n/2) or
9528 1.0 / (sqrt(x) * powi(x, abs(n/2))) if c is an integer or when
9529 optimizing for size.
9530 Don't optimize pow(x,c) into powi(x, n/3) * powi(cbrt(x), n%3) or
9531 1.0 / (powi(x, abs(n)/3) * powi(cbrt(x), abs(n)%3)) if 2c is an
9534 PR tree-optimization/56094
9535 * gimplify.c (force_gimple_operand_1): Temporarily set input_location
9536 to UNKNOWN_LOCATION while gimplifying expr.
9538 2013-01-27 Uros Bizjak <ubizjak@gmail.com>
9541 * config/i386/i386.md (*movabs<mode>_1): Add square brackets around
9542 operand 0 in movabs insn template for -masm=intel asm alternative.
9543 (*movabs<mode>_2): Ditto for operand 1.
9545 2013-01-26 David Holsgrove <david.holsgrove@xilinx.com>
9548 * config.gcc (microblaze*-linux*): Add tmake_file to allow building
9551 2013-01-26 Edgar E. Iglesias <edgar.iglesias@gmail.com>
9553 * config.gcc (microblaze*-*-*): Rename microblaze*-*-elf, update
9556 2013-01-25 Naveen H.S <Naveen.Hurugalawadi@caviumnetworks.com>
9558 * config/aarch64/aarch64.c (TARGET_FIXED_CONDITION_CODE_REGS):
9559 Undef to avoid warning.
9561 2013-01-25 Michael Haubenwallner <michael.haubenwallner@salomon.at>
9563 * configure.ac (gcc_cv_ld_static_dynamic): Define for AIX native ld.
9564 * configure: Regenerate.
9566 2013-01-25 Jakub Jelinek <jakub@redhat.com>
9568 PR tree-optimization/56098
9569 * tree-ssa-phiopt.c (nt_init_block): Don't call add_or_mark_expr
9570 for stmts with volatile ops.
9571 (cond_store_replacement): Don't optimize if assign has volatile ops.
9572 (cond_if_else_store_replacement_1): Don't optimize if either
9573 then_assign or else_assign have volatile ops.
9574 (hoist_adjacent_loads): Don't optimize if either def1 or def2 have
9577 2013-01-25 Georg-Johann Lay <avr@gjlay.de>
9579 * doc/invoke.texi (AVR Built-in Macros): Document __XMEGA__.
9581 2013-01-25 Georg-Johann Lay <avr@gjlay.de>
9583 * doc/extend.texi (Example of asm with clobbered asm reg): Fix
9584 missing ':' in asm example.
9586 2013-01-25 Tejas Belagod <tejas.belagod@arm.com>
9588 * config/aarch64/aarch64-simd-builtins.def: Separate sq<r>dmulh_lane
9589 entries into lane and laneq entries.
9590 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_lane<mode>):
9591 Remove AdvSIMD scalar modes.
9592 (aarch64_sq<r>dmulh_laneq<mode>): New.
9593 (aarch64_sq<r>dmulh_lane<mode>): New RTL pattern for Scalar AdvSIMD
9595 * config/aarch64/arm_neon.h: Fix all the vq<r>dmulh_lane* intrinsics'
9596 builtin implementations to relfect changes in RTL in aarch64-simd.md.
9597 * config/aarch64/iterators.md (VCOND): New.
9600 2013-01-25 Georg-Johann Lay <avr@gjlay.de>
9603 * config/avr/builtins.def (DEF_BUILTIN): Add LIBNAME argument.
9604 Add NULL LIBNAME argument to existing definitions.
9605 (ABSHR, ABSR, ABSLR, ABSLLR, ABSHK, ABSK, ABSLK, ABSLLK): New.
9606 * config/avr/avr-c.c (DEF_BUILTIN): Add LIBNAME argument.
9607 * config/avr/avr.c (DEF_BUILTIN): Same.
9608 (avr_init_builtins): Pass down LIBNAME to add_builtin_function.
9609 (avr_expand_builtin): Expand to a vanilla call if a libgcc
9610 implementation is available (DECL_ASSEMBLER_NAME is set).
9611 (avr_fold_absfx): New static function.
9612 (avr_fold_builtin): Use it to handle: AVR_BUILTIN_ABSHR,
9613 AVR_BUILTIN_ABSR, AVR_BUILTIN_ABSLR, AVR_BUILTIN_ABSLLR,
9614 AVR_BUILTIN_ABSHK, AVR_BUILTIN_ABSK, AVR_BUILTIN_ABSLK,
9616 * config/avr/stdfix.h (abshr, absr, abslr, absllr)
9617 (abshk, absk, abslk, absllk): Provide as static inline functions.
9619 2013-01-25 Marek Polacek <polacek@redhat.com>
9621 PR tree-optimization/56035
9622 * cfgloopmanip.c (fix_loop_structure): Remove redundant condition.
9624 2012-01-24 Uros Bizjak <ubizjak@gmail.com>
9626 * config/i386/i386.md (*movti_internal_rex64): Add (o,e) alternative.
9627 (*movtf_internal_rex64): Add (!o,C) alternative
9628 (*movxf_internal_rex64): Ditto.
9629 (*movdf_internal_rex64): Add (?r,C) and (?m,C) alternatives.
9631 2013-01-24 Shenghou Ma <minux.ma@gmail.com>
9633 * doc/invoke.texi: fix typo.
9634 * doc/objc.texi: fix typo.
9636 2013-01-24 Richard Sandiford <rdsandiford@googlemail.com>
9638 * config/mips/mips.md (*and<mode>3_mips16): Use the "W" constraint
9639 for the first two alternatives.
9641 2013-01-24 Diego Novillo <dnovillo@google.com>
9643 * Makefile.in (GGC): Remove. Replace all instances with ggc-page.o.
9644 (ggc-zone.o): Remove.
9645 * configure.ac: Remove option --with-gc.
9646 * configure: Re-generate.
9647 * doc/install.texi: Remove documentation for --with-gc.
9648 * gengtype.c (write_enum_defn): Remove. Update all users.
9649 (write_Types_process_field): Remove generation of gt_e_* argument.
9650 (output_type_enum): Remove. Update all users.
9651 (write_enum_defn): Remove. Update all users.
9652 (enum alloc_zone): Remove. Update all users.
9653 (write_splay_tree_allocator_def): Remove generation of gt_e_* argument.
9654 * ggc-common.c (ggc_splay_alloc): Remove first argument.
9656 (struct ptr_data): Remove field TYPE. Update all users.
9657 (gt_pch_note_object): Remove argument TYPE. Update all users.
9658 * ggc-internal.h (ggc_pch_alloc_object): Remove last argument.
9660 * ggc-none.c (ggc_alloc_typed_stat): Remove.
9661 (struct alloc_zone): Remove.
9662 (ggc_internal_alloc_zone_stat): Remove.
9663 (ggc_internal_cleared_alloc_zone_stat): Remove.
9664 * ggc-page.c (ggc_alloc_typed_stat): Remove.
9665 (ggc_pch_count_object): Remove last argument. Update all users.
9666 (ggc_pch_alloc_object): Remove last argument. Update all users.
9667 (struct alloc_zone): Remove.
9668 * ggc-zone.c: Remove.
9669 * ggc.h (gt_pch_note_object): Remove last argument. Update all users.
9670 (struct alloc_zone): Remove.
9671 (ggc_alloc_typed_stat): Remove.
9672 (ggc_alloc_typed): Remove.
9673 (ggc_splay_alloc): Remove first argument.
9674 (rtl_zone): Remove. Update all users.
9675 (tree_zone): Remove. Update all users.
9676 (tree_id_zone): Remove. Update all users.
9677 (ggc_internal_zone_alloc_stat): Remove. Update all users.
9678 (ggc_internal_zone_cleared_alloc_stat): Remove. Update all users.
9679 (ggc_internal_zone_vec_alloc_stat): Remove. Update all users.
9680 * tree-ssanames.c: Remove references to zone allocator in comments.
9682 2013-01-24 Georg-Johann Lay <avr@gjlay.de>
9684 * config/avr/avr.c (avr_out_fract): Make register numbers that
9685 might be outside of source operand signed.
9687 2013-01-24 Uros Bizjak <ubizjak@gmail.com>
9689 * config/i386/constraints.md (Yf): New constraint.
9690 * config/i386/i386.md (*movdf_internal_rex64): Use Yf*f instead
9691 of f constraint to conditionaly disable x87 register preferences.
9692 (*movdf_internal): Ditto.
9693 (*movsf_internal): Ditto.
9695 2013-01-24 Steven Bosscher <steven@gcc.gnu.org>
9698 * lra-assigns.c (assign_by_spills): Throw away the pattern of asms
9699 that have operands with impossible constraints.
9700 Add a FIXME for a speed-up opportunity.
9701 * lra-constraints.c (process_alt_operands): Verify that a class
9702 selected from constraints on asms is valid for the operand mode.
9703 (curr_insn_transform): Remove incorrect comment.
9705 2013-01-23 David Edelsohn <dje.gcc@gmail.com>
9707 * config/rs6000/rs6000.c (rs6000_delegitimize_address): Check that
9708 TOC operand is a valid symbol ref in the constant pool.
9710 2013-01-23 Edgar E. Iglesias <edgar.iglesias@gmail.com>
9712 * config/microblaze/linux.h: Add TARGET_OS_CPP_BUILTINS
9714 2013-01-23 Georg-Johann Lay <avr@gjlay.de>
9717 * config/avr/stdfix.h: New file.
9718 * t-avr (stdfix-gcc.h): New rule to build it.
9719 (EXTRA_HEADERS): Set it to install stdfix.h, stdfix-gcc.h.
9721 2013-01-23 Kostya Serebryany <kcc@google.com>
9723 * config/darwin.h: remove dependency on
9724 CoreFoundation (asan on Mac OS).
9726 2013-01-23 Jakub Jelinek <jakub@redhat.com>
9729 * config/arm/arm.md (cbranchdi4, cstoredi4): Use s_register_operand
9730 instead of cmpdi_operand for first comparison operand.
9731 Don't assert that comparison operands aren't both constants.
9733 2013-01-22 Jonathan Wakely <jwakely.gcc@gmail.com>
9735 * doc/install.texi (Downloading the Source): Update references to
9736 downloading separate components.
9738 2013-01-22 Jonathan Wakely <jwakely.gcc@gmail.com>
9740 * doc/extend.texi (__int128): Improve grammar.
9742 2013-01-22 Uros Bizjak <ubizjak@gmail.com>
9745 * config/i386/i386.md (*movti_internal_rex64): Change (o,riF)
9746 alternative to (o,r).
9747 (*movdi_internal_rex64): Remove (!o,n) alternative.
9748 (DImode immediate->memory splitter): Remove.
9749 (DImode immediate->memory peephole2): Remove.
9750 (movtf): Enable for TARGET_64BIT || TARGET_SSE.
9751 (*movtf_internal_rex64): Rename from *movtf_internal. Change (!o,F*r)
9752 alternative to (!o,*r).
9753 (*movtf_internal_sse): New pattern.
9754 (*movxf_internal_rex64): New pattern.
9755 (*movxf_internal): Disable for TARGET_64BIT.
9756 (*movdf_internal_rex64): Remove (!o,F) alternative.
9758 2013-01-22 Jakub Jelinek <jakub@redhat.com>
9761 * dumpfile.c (dump_loc): Only print loc if LOCATION_LOCUS (loc)
9762 isn't UNKNOWN_LOCATION nor BUILTINS_LOCATION.
9763 * tree-vect-loop-manip.c (find_loop_location): Also ignore
9764 stmt locations where LOCATION_LOCUS of the stmt location is
9765 UNKNOWN_LOCATION or BUILTINS_LOCATION.
9768 * config/i386/i386.md (UNSPEC_STOS): New.
9769 (strset_singleop, *strsetdi_rex_1, *strsetsi_1, *strsethi_1,
9770 *strsetqi_1): Add UNSPEC_STOS.
9772 2013-01-22 Paolo Carlini <paolo.carlini@oracle.com>
9775 * doc/invoke.texi: Remove left over -Wsynth example.
9777 2013-01-21 Jakub Jelinek <jakub@redhat.com>
9779 PR tree-optimization/56051
9780 * fold-const.c (fold_binary_loc): Don't fold
9781 X < (cast) (1 << Y) into (X >> Y) != 0 if cast is either
9782 a narrowing conversion, or widening conversion from signed
9785 2013-01-21 Uros Bizjak <ubizjak@gmail.com>
9787 PR rtl-optimization/56023
9788 * haifa-sched.c (fix_inter_tick): Do not update ticks of instructions,
9789 dependent on debug instruction.
9791 2013-01-21 Martin Jambor <mjambor@suse.cz>
9794 * function.c (allocate_struct_function): Call
9795 invoke_set_current_function_hook earlier.
9797 2013-01-21 Jakub Jelinek <jakub@redhat.com>
9799 * reload1.c (init_reload): Only initialize reload_obstack
9800 during the first call.
9802 2013-01-21 Marek Polacek <polacek@redhat.com>
9804 * cfgloop.c (verify_loop_structure): Fix up grammar.
9806 2013-01-21 Yi-Hsiu Hsu <ahsu@marvell.com>
9808 * config/arm/marvell-pj4.md (pj4_shift_conds, pj4_alu_shift,
9809 pj4_alu_shift_conds, pj4_shift): Handle simple_alu_shift.
9811 2013-01-21 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9814 * config/arm/marvell-pj4.md: Update copyright year.
9815 Fix up use of alu to alu_reg and simple_alu_imm.
9817 2013-01-21 Uros Bizjak <ubizjak@gmail.com>
9819 * config/i386/i386.md (enabled): Do not disable fma4 for TARGET_FMA.
9821 2013-01-20 Vladimir Makarov <vmakarov@redhat.com>
9824 * lra-constraints.c (curr_insn_transform): Don't reuse original
9825 insn for secondary memory move when memory mode should be different.
9827 2013-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
9829 * config/pa/pa.md (atomic_loaddi, atomic_loaddi_1, atomic_storedi,
9830 atomic_storedi_1): New patterns.
9832 2013-01-20 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
9834 btver2 pipeline descriptions.
9835 * config/i386/i386.c: Enable CPU_BTVER2 to use btver2 pipeline
9837 * config/i386/i386.md (btver2_decode): New type attributes.
9838 * config/i386/sse.md (btver2_decode, btver2_sse_attr): New
9840 * config/i386/btver2.md: New file describing btver2 pipelines.
9842 2013-01-19 Andrew Pinski <apinski@cavium.com>
9844 PR tree-optimization/52631
9845 * tree-ssa-sccvn (visit_use): Before looking up the original
9846 statement, try looking up the simplified expression.
9848 2013-01-19 Anthony Green <green@moxielogic.com>
9850 * config/moxie/moxie.c (moxie_expand_prologue): Set
9851 current_function_static_stack_size.
9853 2013-01-18 Jakub Jelinek <jakub@redhat.com>
9855 PR tree-optimization/56029
9856 * tree-phinodes.c (reserve_phi_args_for_new_edge): Set
9857 gimple_phi_arg_location for the new arg to UNKNOWN_LOCATION.
9859 2013-01-18 Sharad Singhai <singhai@google.com>
9861 PR tree-optimization/55995
9862 * dumpfile.c (dump_loc): Print location only if available.
9863 * tree-vectorizer.c (increase_alignment): Intialize vect_location.
9865 2013-01-18 Vladimir Makarov <vmakarov@redhat.com>
9868 * lra-constraints.c (curr_insn_transform): Reuse original insn for
9869 secondary memory move.
9870 (inherit_reload_reg): Use rclass instead of cl for
9871 check_secondary_memory_needed_p.
9873 2013-01-18 Jakub Jelinek <jakub@redhat.com>
9876 * expr.c (expand_expr_real_2) <case COMPLEX_EXPR>: Handle
9877 the case where writing real complex part of target modifies op1.
9879 2013-01-18 James Greenhalgh <james.greenhalgh@arm.com>
9881 * config/aarch64/aarch64-simd.md
9882 (aarch64_vcond_internal<mode>): Handle unordered cases.
9883 * config/aarch64/iterators.md (v_cmp_result): New.
9885 2013-01-18 Yi-Hsiu Hsu <ahsu@marvell.com>
9886 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9888 * config/arm/marvell-pj4.md: New file.
9889 * config/arm/arm.c (arm_issue_rate): Add marvell_pj4.
9890 * config/arm/arm.md (generic_sched): Add marvell_pj4.
9891 (generic_vfp): Likewise.
9892 * config/arm/arm-cores.def: Add marvell-pj4.
9893 * config/arm/arm-tune.md: Regenerate.
9894 * config/arm/arm-tables.opt: Regenerate.
9895 * config/arm/bpabi.h (BE8_LINK_SPEC): Add marvell_pj4.
9896 * doc/invoke.texi: Document marvell-pj4.
9898 2013-01-18 Tejas Belagod <tejas.belagod@arm.com>
9900 * config/aarch64/arm_neon.h: Map scalar types to standard types.
9902 2013-01-18 Alexandre Oliva <aoliva@redhat.com>
9907 * var-tracking.c (negative_power_of_two_p): New.
9908 (global_get_addr_cache, local_get_addr_cache): New.
9909 (get_addr_from_global_cache, get_addr_from_local_cache): New.
9910 (vt_canonicalize_addr): Rewrite using the above. Adjust the
9912 (vt_stack_offset_p): Remove.
9913 (vt_canon_true_dep): Always canonicalize loc's address.
9914 (clobber_overlapping_mems): Make sure we have a MEM.
9915 (local_get_addr_clear_given_value): New.
9916 (val_reset): Clear local cached entries.
9917 (compute_bb_dataflow): Create and release the local cache.
9918 Disable duplicate MEMs clobbering.
9919 (emit_notes_in_bb): Clobber MEMs likewise.
9920 (vt_emit_notes): Create and release the local cache.
9921 (vt_initialize, vt_finalize): Create and release the global
9922 cache, respectively.
9923 * alias.c (rtx_equal_for_memref_p): Compare operands of ENTRY_VALUEs.
9925 2013-01-18 Alexandre Oliva <aoliva@redhat.com>
9928 * tree-mudflap.c (mudflap_finish_file): Skip deferred decls
9929 not found in the symtab.
9931 2013-01-18 Alexandre Oliva <aoliva@redhat.com>
9934 PR rtl-optimization/55547
9935 PR rtl-optimization/53827
9938 * alias.c (offset_overlap_p): New, factored out of...
9939 (memrefs_conflict_p): ... this. Use absolute sizes. Retain
9940 the conservative special case for symbolic constants. Don't
9941 adjust zero sizes on alignment.
9943 2013-01-18 Bernd Schmidt <bernds@codesourcery.com>
9945 PR rtl-optimization/52573
9946 * regrename.c (build_def_use): Ignore REG_DEAD notes if there is a
9947 REG_UNUSED for the same register.
9949 2013-01-17 Richard Biener <rguenther@suse.de>
9950 Marek Polacek <polacek@redhat.com>
9952 PR rtl-optimization/55833
9953 * loop-unswitch.c (unswitch_loops): Move loop verification...
9954 (unswitch_single_loop): ...here. Call mark_irreducible_loops.
9955 * cfgloopmanip.c (fix_loop_placement): Add IRRED_INVALIDATED parameter.
9956 Set it to true when we're removing a loop from hierarchy tree in
9957 an irreducible region.
9958 (fix_bb_placements): Adjust caller.
9959 (fix_loop_placements): Likewise.
9961 2013-01-17 Georg-Johann Lay <avr@gjlay.de>
9963 * config/avr/builtins.def (DEF_BUILTIN): Factor out
9964 "__builtin_avr_" from NAME, turn NAME to an uppercase identifier.
9965 Factor out 'CODE_FOR_' from ICODE, use 'nothing' instead of '-1'.
9966 Remove ID. Adjust comments.
9967 * config/avr/avr-c.c (avr_builtin_name): Remove.
9968 (avr_cpu_cpp_builtins): Use DEF_BUILTIN instead of for-loop.
9969 * config/avr/avr.c (avr_tolower): New static function.
9970 (DEF_BUILTIN): Remove parameter ID. Prefix ICODE by 'CODE_FOR_'.
9971 Stringify NAME, prefix it with "__builtin_avr_" and lowercase it.
9972 (avr_expand_builtin): Assert insn_code != CODE_FOR_nothing for
9975 2013-01-17 Jan Hubicka <jh@suse.cz>
9977 PR tree-optimization/55273
9978 * loop-iv.c (iv_number_of_iterations): Consider zero iteration case.
9980 2013-01-17 Uros Bizjak <ubizjak@gmail.com>
9983 * config/i386/sync.md (atomic_store<mode>): Always generate SWImode
9984 store through atomic_store<mode>_1.
9985 (atomic_store<mode>_1): Macroize insn using SWI mode iterator.
9987 2013-01-17 Martin Jambor <mjambor@suse.cz>
9989 PR tree-optimizations/55264
9990 * ipa-inline-transform.c (can_remove_node_now_p_1): Never return true
9991 for virtual methods.
9992 * ipa.c (symtab_remove_unreachable_nodes): Never return true for
9993 virtual methods before inlining is over.
9994 * cgraph.h (cgraph_only_called_directly_or_aliased_p): Return false for
9996 * cgraphclones.c (cgraph_create_virtual_clone): Mark clones as
9999 2013-01-16 Vladimir Makarov <vmakarov@redhat.com>
10001 PR rtl-optimization/56005
10002 * sched-deps.c (sched_analyze_2): Check deps->readonly for adding
10003 pending reads for prefetch.
10005 2013-01-16 Ian Bolton <ian.bolton@arm.com>
10007 * config/aarch64/aarch64.md
10008 (*cstoresi_neg_uxtw): New pattern.
10009 (*cmovsi_insn_uxtw): New pattern.
10010 (*<optab>si3_uxtw): New pattern.
10011 (*<LOGICAL:optab>_<SHIFT:optab>si3_uxtw): New pattern.
10012 (*<optab>si3_insn_uxtw): New pattern.
10013 (*bswapsi2_uxtw): New pattern.
10015 2013-01-16 Richard Biener <rguenther@suse.de>
10017 * tree-inline.c (tree_function_versioning): Remove set but
10018 never used variable.
10020 2013-01-16 Richard Biener <rguenther@suse.de>
10022 PR tree-optimization/55964
10023 * tree-flow.h (rename_variables_in_loop): Remove.
10024 (rename_variables_in_bb): Likewise.
10025 * tree-loop-distribution.c (update_phis_for_loop_copy): Remove.
10026 (copy_loop_before): Adjust and delete update-ssa status.
10027 * tree-vect-loop-manip.c (rename_variables_in_bb): Make static.
10028 (rename_variables_in_bb): Likewise. Properly walk over predecessors.
10029 (rename_variables_in_loop): Remove.
10030 (slpeel_update_phis_for_duplicate_loop): Likewise.
10031 (slpeel_tree_duplicate_loop_to_edge_cfg): Handle nested loops,
10032 use available cfg machinery instead of duplicating it.
10033 Update PHI nodes and perform poor-mans SSA update here.
10034 (slpeel_tree_peel_loop_to_edge): Adjust.
10036 2013-01-16 Richard Biener <rguenther@suse.de>
10038 PR tree-optimization/54767
10039 PR tree-optimization/53465
10040 * tree-vrp.c (vrp_meet_1): Revert original fix for PR53465.
10041 (vrp_visit_phi_node): For PHI arguments coming via backedges
10042 drop all symbolical range information.
10043 (execute_vrp): Compute backedges.
10045 2013-01-16 Richard Biener <rguenther@suse.de>
10047 * doc/install.texi: Update CLooG and ISL requirements to
10050 2013-01-16 Christian Bruel <christian.bruel@st.com>
10053 * config/sh/sh.c (sh_expand_prologue): Postpone new_stack mem symbol.
10054 (broken_move): Handle UNSPECV_SP_SWITCH_B.
10055 * config/sh/sh.md (sp_switch_1): Use set (reg:SI SP_REG).
10057 2013-01-16 DJ Delorie <dj@redhat.com>
10059 * config/sh/sh.md (UNSPECV_SP_SWITCH_B): New.
10060 (UNSPECV_SP_SWITCH_E): New.
10061 (sp_switch_1): Change to an unspec.
10062 (sp_switch_2): Change to an unspec. Don't use post-inc when we
10065 2013-01-16 Uros Bizjak <ubizjak@gmail.com>
10067 * emit-rtl.c (need_atomic_barrier_p): Mask memory model argument
10068 with MEMMODEL_MASK before comparing with MEMMODEL_* memory types.
10069 * optabs.c (maybe_emit_sync_lock_test_and_set): Ditto.
10070 (expand_mem_thread_fence): Ditto.
10071 (expand_mem_signal_fence): Ditto.
10072 (expand_atomic_load): Ditto.
10073 (expand_atomic_store): Ditto.
10075 2013-01-16 Alexandre Oliva <aoliva@redhat.com>
10077 PR rtl-optimization/55547
10078 PR rtl-optimization/53827
10081 * alias.c (memrefs_conflict_p): Set sizes to negative after
10084 2013-01-15 Jakub Jelinek <jakub@redhat.com>
10087 * function.c (thread_prologue_and_epilogue_insns): Always
10088 add crtl->drap_reg to set_up_by_prologue.set, even if
10089 stack_realign_drap is false.
10091 2013-01-15 Jan-Benedict Glaw <jbglaw@lug-owl.de>
10093 * config/vax/vax.md (add<mode>3, sub<mode>3, mul<mode>3, div<mode>3,
10094 and<mode>3, *and<mode>_const_int, ior<mode>3, xor<mode>3, ashrsi3,
10095 *call): Fix indention.
10097 2013-01-15 Tom de Vries <tom@codesourcery.com>
10100 * optabs.c (widen_operand): Use gen_lowpart instead of gen_rtx_SUBREG.
10103 2013-01-15 Vladimir Makarov <vmakarov@redhat.com>
10105 PR rtl-optimization/55153
10106 * sched-deps.c (sched_analyze_2): Add pending reads for prefetch.
10108 2013-01-15 Martin Jambor <mjambor@suse.cz>
10110 PR tree-optimization/55920
10111 * tree-sra.c (analyze_access_subtree): Do not mark non-removable
10112 accesses as grp_to_be_debug_replaced.
10114 2013-01-15 Jakub Jelinek <jakub@redhat.com>
10116 PR tree-optimization/55920
10117 * tree-sra.c (sra_modify_assign): If for lacc->grp_to_be_debug_replaced
10118 there is non-useless type conversion needed from debug rhs to lhs,
10119 use build_debug_ref_for_model and/or VIEW_CONVERT_EXPR.
10121 2013-01-15 Joseph Myers <joseph@codesourcery.com>
10122 Mikael Pettersson <mikpe@it.uu.se>
10125 * config/arm/arm.h (ADDR_VEC_ALIGN): Align SImode jump tables for
10127 (ASM_OUTPUT_CASE_LABEL): Remove.
10128 (ASM_OUTPUT_BEFORE_CASE_LABEL): Define to empty.
10129 * final.c (shorten_branches): Update alignment of labels before
10130 jump tables if CASE_VECTOR_SHORTEN_MODE.
10132 2013-01-15 Richard Biener <rguenther@suse.de>
10135 * system.h: Do not include gmp.h for building host tools.
10137 2013-01-15 Richard Biener <rguenther@suse.de>
10139 PR middle-end/55882
10140 * emit-rtl.c (set_mem_attributes_minus_bitpos): Correctly
10141 account for bitpos when computing alignment.
10143 2013-01-15 Vladimir Yakovlev <vladimir.b.yakovlev@intel.com>
10145 * config/i386/i386-c.c (ix86_target_macros_internal): New case.
10146 (ix86_target_macros_internal): Likewise.
10148 * config/i386/i386.c (m_CORE2I7): Removed.
10149 (m_CORE_HASWELL): New macro.
10150 (m_CORE_ALL): Likewise.
10151 (initial_ix86_tune_features): m_CORE2I7 is replaced by m_CORE_ALL.
10152 (initial_ix86_arch_features): Likewise.
10153 (processor_target_table): Initializations for Core avx2.
10154 (cpu_names): New names "core-avx2".
10155 (ix86_option_override_internal): Changed PROCESSOR_COREI7 by
10156 PROCESSOR_CORE_HASWELL.
10157 (ix86_issue_rate): New case.
10158 (ia32_multipass_dfa_lookahead): Likewise.
10159 (ix86_sched_init_global): Likewise.
10161 * config/i386/i386.h (TARGET_HASWELL): New macro.
10162 (target_cpu_default): New TARGET_CPU_DEFAULT_haswell.
10163 (processor_type): New PROCESSOR_HASWELL.
10165 2013-01-15 Jakub Jelinek <jakub@redhat.com>
10167 PR tree-optimization/55955
10168 * tree-vect-loop.c (vectorizable_reduction): Give up early on
10169 *SHIFT_EXPR and *ROTATE_EXPR codes.
10171 PR tree-optimization/48766
10172 * opts.c (common_handle_option): For -fwrapv disable -ftrapv, for
10173 -ftrapv disable -fwrapv.
10175 2013-01-14 Georg-Johann Lay <avr@gjlay.de>
10178 * config/avr/avr-c.c (avr_cpu_cpp_builtins): Define __FLASH
10179 etc. to 1 and not to __flash.
10180 Use LL suffix for __INT24_MAX__ with -mint8.
10181 Use ULL suffix for __UINT24_MAX__ with -mint8.
10183 2013-01-14 Georg-Johann Lay <avr@gjlay.de>
10185 * config/avr/avr-arch.h
10186 (struct base_arch_s): Use typedef avr_arch_t instead.
10187 (struct arch_info_s): Use typedef avr_arch_info_t instead.
10188 (struct mcu_type_s): Use typedef avr_mcu_t instead.
10189 * config/avr/avr.c: Same.
10190 * config/avr/avr-devices.c: Same.
10191 * config/avr/driver-avr.c: Same.
10192 * config/avr/gen-avr-mmcu-texi.c: Same.
10193 * config/avr/avr-mcus.def: Adjust comment.
10195 2013-01-14 Tejas Belagod <tejas.belagod@arm.com>
10197 * config/aarch64/aarch64-simd.md (*aarch64_simd_ld1r<mode>): New.
10198 * config/aarch64/iterators.md (VALLDI): New.
10200 2013-01-14 Uros Bizjak <ubizjak@gmail.com>
10201 Andi Kleen <ak@linux.intel.com>
10204 * config/i386/sync.md (atomic_store<mode>_1): New pattern.
10205 (atomic_store<mode>): Call atomic_store<mode>_1 for IX86_HLE_RELEASE
10208 2013-01-14 Georg-Johann Lay <avr@gjlay.de>
10210 * config/avr/avr-stdint.h: Remove trailing blanks.
10211 * config/avr/avr-log.h: Same.
10212 * config/avr/avr-arch.h: Same.
10213 * config/avr/avr-devices.c: Same.
10214 * config/avr/avr-dimode.md: Same.
10215 * config/avr/predicates.md: Same.
10216 * config/avr/avr-c.c: Same. And fix typo.
10218 * config/avr/avr-protos.h: Same. And:
10219 (function_arg_regno_p): Rename to avr_function_arg_regno_p.
10220 (init_cumulative_args): Rename to avr_init_cumulative_args.
10221 (expand_prologue): Rename to avr_expand_prologue.
10222 (expand_epilogue): Rename to avr_expand_epilogue.
10223 (adjust_insn_length): Rename to avr_adjust_insn_length.
10224 (notice_update_cc): Rename to avr_notice_update_cc.
10225 (final_prescan_insn): Rename to avr_final_prescan_insn.
10226 * config/avr/avr.c: Same.
10227 * config/avr/avr.h: Same.
10228 * config/avr/avr.md: Remove trailing blanks.
10229 (prologue): Use avr_expand_prologue.
10230 (epilogue, sibcall_epilogue): Use avr_expand_epilogue.
10232 2013-01-14 Richard Biener <rguenther@suse.de>
10234 * tree-cfg.c (verify_expr_location, verify_expr_location_1,
10235 verify_location, collect_subblocks): New functions.
10236 (verify_gimple_in_cfg): Verify that locations only reference
10237 BLOCKs in the functions BLOCK tree.
10239 2013-01-14 Richard Biener <rguenther@suse.de>
10241 * tree-cfgcleanup.c (remove_forwarder_block): Unshare propagated
10243 * graphite-sese-to-poly.c (insert_out_of_ssa_copy): Properly
10245 (insert_out_of_ssa_copy_on_edge): Likewise.
10246 (rewrite_close_phi_out_of_ssa): Likewise.
10247 * tree-ssa.c (insert_debug_temp_for_var_def): Properly unshare
10249 * tree-ssa-pre.c (insert_into_preds_of_block): Properly unshare
10250 propagated constants.
10251 * tree-cfg.c (tree_node_can_be_shared): Handled component-refs
10254 2013-01-14 Georg-Johann Lay <avr@gjlay.de>
10256 * config/avr/avr-modes.def: Add GPL copyright notice.
10258 2013-01-13 Uros Bizjak <ubizjak@gmail.com>
10260 * config/i386/sync.md (mem_thread_fence): Mask operands[0] with
10261 MEMMODEL_MASK to determine memory model.
10262 (atomic_store<mode>): Ditto from operands[2].
10263 * config/i386/i386.c (ix86_memmodel_check): Declare "strong" as bool.
10265 2013-01-13 Jakub Jelinek <jakub@redhat.com>
10268 * gimple-fold.c (get_symbol_constant_value): Call unshare_expr.
10269 (fold_gimple_assign): Don't call unshare_expr here.
10270 (fold_ctor_reference): Call unshare_expr.
10272 2013-01-13 Terry Guo <terry.guo@arm.com>
10274 * Makefile.in (s-mlib): New argument MULTILIB_REUSE.
10275 * doc/fragments.texi: Document MULTILIB_REUSE.
10276 * gcc.c (multilib_reuse): New internal spec.
10277 (set_multilib_dir): Also search multilib from multilib_reuse.
10278 * genmultilib (tmpmultilib3): Refactor code.
10279 (tmpmultilib4): Ditto.
10280 (multilib_reuse): New multilib argument.
10282 2013-01-13 Richard Sandiford <rdsandiford@googlemail.com>
10284 * Makefile.in: Update copyright.
10286 2013-01-12 Tom de Vries <tom@codesourcery.com>
10288 PR middle-end/55890
10289 * calls.c (expand_call): Check if arg_nr is valid.
10291 2013-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10293 * doc/extend.texi (X86 Built-in Functions): Add whitespace in
10294 __builtin_ia32_paddb256 and __builtin_ia32_pavgb256
10295 documentation. Add missing '__' in front of
10296 __builtin_ia32_packssdw256.
10298 2013-01-11 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
10301 * config/s390/s390.c (s390_preferred_reload_class): Do not return
10302 NO_REGS for larl operands.
10303 (s390_reload_larl_operand): Use s390_load_address instead of
10306 2013-01-11 Richard Biener <rguenther@suse.de>
10308 * tree-cfg.c (verify_node_sharing_1): Split out from ...
10309 (verify_node_sharing): ... here.
10310 (verify_gimple_in_cfg): Use verify_node_sharing_1 for walk_tree.
10312 2013-01-11 Eric Botcazou <ebotcazou@adacore.com>
10314 * configure.ac (Tree checking): Set TREECHECKING to yes if enabled.
10315 Substitute TREECHECKING.
10316 * configure: Regenerate.
10317 * Makefile.in (TREECHECKING): New.
10319 2013-01-11 Richard Guenther <rguenther@suse.de>
10321 PR tree-optimization/44061
10322 * tree-vrp.c (extract_range_basic): Compute zero as
10323 value-range for __builtin_constant_p of function parameters.
10325 2013-01-10 Richard Sandiford <rdsandiford@googlemail.com>
10327 Update copyright years.
10329 2013-01-10 Vladimir Makarov <vmakarov@redhat.com>
10331 PR rtl-optimization/55672
10332 * lra-eliminations.c (mark_not_eliminable): Permit addition with
10333 const to be eliminable.
10335 2013-01-10 David Edelsohn <dje.gcc@gmail.com>
10337 * configure.ac (HAVE_AS_TLS): Add check for powerpc-ibm-aix.
10338 * configure: Regenerate.
10340 2013-01-10 Richard Biener <rguenther@suse.de>
10342 * builtins.c (expand_builtin_init_trampoline): Use set_mem_attributes.
10344 2013-01-10 Richard Biener <rguenther@suse.de>
10347 * tree-into-ssa.c (rewrite_add_phi_arguments): Do not set
10348 locations for virtual PHI arguments.
10349 (rewrite_update_phi_arguments): Likewise.
10351 2013-01-10 Joel Sherrill <joel.sherrill@OARcorp.com>
10353 * config/v850/rtems.h (ASM_SPEC): Pass -m8byte-align and -mgcc-abi
10356 2013-01-10 Jakub Jelinek <jakub@redhat.com>
10358 PR tree-optimization/55921
10359 * tree-complex.c (expand_complex_asm): New function.
10360 (expand_complex_operations_1): Call it for GIMPLE_ASM.
10362 2013-01-10 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
10365 * config/s390/s390.c (s390_symref_operand_p)
10366 (s390_loadrelative_operand_p): Merge the two functions.
10367 (s390_check_qrst_address, print_operand_address): Add parameters
10368 to s390_loadrelative_operand_p invokation.
10369 (s390_check_symref_alignment): Use s390_loadrelative_operand_p.
10370 (s390_reload_larl_operand, s390_secondary_reload): Use
10371 s390_loadrelative_operand_p instead of s390_symref_operand_p.
10372 (legitimize_pic_address): Handle @GOTENT and @PLT + addend.
10374 2013-01-09 Mike Stump <mikestump@comcast.net>
10376 * dse.c (record_store): Remove unnecessary assert.
10378 2013-01-09 Jan Hubicka <jh@suse.cz>
10380 PR tree-optimization/55569
10381 * cfgloopmanip.c (scale_loop_profile): Make ITERATION_BOUND gcov_type.
10382 * cfgloop.h (scale_loop_profile): Likewise.
10384 2013-01-09 Jan Hubicka <jh@suse.cz>
10387 * ipa-inline.c (ipa_inline): Remove extern inlines and virtual
10389 * cgraphclones.c (cgraph_clone_node): Cpoy also LTO file data.
10391 2013-01-09 Richard Sandiford <rdsandiford@googlemail.com>
10393 PR middle-end/55114
10394 * expr.h (maybe_emit_group_store): Declare.
10395 * expr.c (maybe_emit_group_store): New function.
10396 * builtins.c (expand_builtin_int_roundingfn): Call it.
10397 (expand_builtin_int_roundingfn_2): Likewise.
10399 2013-01-09 Vladimir Makarov <vmakarov@redhat.com>
10401 PR rtl-optimization/55829
10402 * lra-constraints.c (match_reload): Add code for absent output.
10403 (curr_insn_transform): Add code for reloads of matched inputs
10406 2013-01-09 Uros Bizjak <ubizjak@gmail.com>
10408 * config/i386/sse.md (*vec_interleave_highv2df): Change mode
10409 attribute of movddup insn to DF.
10410 (*vec_interleave_lowv2df): Ditto.
10411 (vec_dupv2df): Ditto.
10413 2013-01-09 Jan Hubicka <jh@suse.cz>
10415 PR tree-optimiation/55875
10416 * tree-ssa-loop-niter.c (number_of_iterations_cond): Add
10417 EVERY_ITERATION parameter.
10418 (number_of_iterations_exit): Check if exit is executed every iteration.
10419 (idx_infer_loop_bounds): Similarly here.
10420 (n_of_executions_at_most): Simplify
10421 to only test for cases where statement is dominated by the
10422 particular bound; handle correctly the "postdominance" test.
10423 (scev_probably_wraps_p): Use max loop iterations info
10424 as a global bound first.
10426 2013-01-09 Nguyen Duy Dat <dat.nguyen.yn@rvc.renesas.com>
10427 Nick Clifton <nickc@redhat.com>
10429 * config/v850/v850.md (cbranchsf4): New pattern.
10430 (cstoresf4): New pattern.
10431 (cbranchdf4): New pattern.
10432 (cstoredf4): New pattern.
10433 (movsicc): Disallow floating point comparisons.
10434 (cmpsf_le_insn): Fix order of operators.
10435 (cmpsf_lt_insn): Likewise.
10436 (cmpsf_eq_insn): Likewise.
10437 (cmpdf_le_insn): Likewise.
10438 (cmpdf_lt_insn): Likewise.
10439 (cmpdf_eq_insn): Likewise.
10440 (cmpsf_ge_insn): Use LE comparison.
10441 (cmpdf_ge_insn): Likewise.
10442 (cmpsf_gt_insn): Use LT comparison.
10443 (cmpdf_gt_insn): Likewise.
10444 (cmpsf_ne_insn): Delete pattern.
10445 (cmpdf_ne_insn): Delete pattern.
10446 * config/v850/v850.c (v850_gen_float_compare): Use
10447 gen_cmpdf_eq_insn for NE comparison.
10448 (v850_float_z_comparison_operator)
10449 (v850_float_nz_comparison_operator): Move from here ...
10450 * config/v850/predicates.md: ... to here. Move GT and GE
10451 comparisons into v850_float_z_comparison_operator.
10452 * config/v850/v850-protos.h (v850_float_z_comparison_operator):
10454 (v850_float_nz_comparison_operator): Likewise.
10456 2013-01-09 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
10458 * config/pa/pa.c (pa_emit_move_sequence): Replace calls to gen_insv
10459 with calls to gen_insvsi/gen_insvdi.
10461 2013-01-09 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
10463 * config/i386/i386.c (initial_ix86_tune_features): Set up
10464 X86_TUNE_AVX128_OPTIMAL for m_BTVER2.
10466 2013-01-09 Steven Bosscher <steven@gcc.gnu.org>
10467 Jakub Jelinek <jakub@redhat.com>
10469 PR tree-optimization/48189
10470 * predict.c (predict_loops): If max is 0, don't call compare_tree_int.
10471 If nitercst is 0, don't predict the exit edge.
10473 2013-01-08 Naveen H.S <Naveen.Hurugalawadi@caviumnetworks.com>
10475 * config/aarch64/aarch64.c (aarch64_print_operand): Replace %r
10476 in asm_fprintf with reg_names.
10477 (aarch64_print_operand_address): Likewise.
10478 (aarch64_return_addr): Likewise.
10479 * config/aarch64/aarch64.h (ASM_FPRINTF_EXTENSIONS): Remove.
10481 2013-01-08 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
10483 * config/pa/pa.h (VAL_U6_BITS_P): Define.
10484 (INT_U6_BITS): Likewise.
10485 * config/pa/predicates.md (uint6_operand): New predicate.
10486 (shift5_operand, shift6_operand): Likewise.
10487 * config/pa/pa.md (lshrsi3, rotrsi3): Use shift5_operand instead of
10489 (lshrdi3): Use shift6_operand.
10490 (shrpsi4, shrpdi4): New insn patterns.
10491 (extzv): Delete expander.
10492 (extzvsi, extzvdi): New expanders. Use uint5_operand and uint6_operand
10493 predicates in unamed zero extract patterns. Tighten common constraint.
10494 (extv): Delete expander.
10495 (extvsi, extvdi): New expanders. Use uint5_operand and uint6_operand
10496 predicates in unamed sign extract patterns. Tighten common constraint.
10497 (insv): Delete expander.
10498 (insvsi, insvdi): New expanders. Use uint5_operand and uint6_operand
10499 predicates in unamed insert patterns. Tighten common constraint.
10500 Change uint32_operand predicate to uint6_operand predicate in unamed
10501 DImode pattern to insert constant values of type 1...1xxxx.
10503 2013-01-04 Jan Hubicka <jh@suse.cz>
10505 PR tree-optimization/55823
10506 * ipa-prop.c (update_indirect_edges_after_inlining): Fix ordering
10509 2013-01-08 Jakub Jelinek <jakub@redhat.com>
10510 Uros Bizjak <ubizjak@gmail.com>
10512 PR rtl-optimization/55845
10513 * df-problems.c (can_move_insns_across): Stop scanning at
10514 volatile_insn_p source instruction or give up if
10515 across_from .. across_to range contains any volatile_insn_p
10518 2013-01-08 Tejas Belagod <tejas.belagod@arm.com>
10520 * config/aarch64/aarch64-simd.md (vec_init<mode>): New.
10521 * config/aarch64/aarch64-protos.h (aarch64_expand_vector_init):
10523 * config/aarch64/aarch64.c (aarch64_simd_dup_constant,
10524 aarch64_simd_make_constant, aarch64_expand_vector_init): New.
10526 2013-01-08 Jakub Jelinek <jakub@redhat.com>
10529 * asan.c (asan_clear_shadow): New function.
10530 (asan_emit_stack_protection): Use it.
10532 2013-01-08 Tejas Belagod <tejas.belagod@arm.com>
10534 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_<su>mult_lo_<mode>,
10535 aarch64_simd_vec_<su>mult_hi_<mode>): Separate instruction and operand
10536 with tab instead of space.
10538 2013-01-08 Nick Clifton <nickc@redhat.com>
10540 * config/rl78/rl78.c (rl78_expand_prologue): Always select
10541 register bank 0 at the start of an interrupt handler.
10542 * config/rl78/rl78.md (mulsi3_g13): Correct values for MDBL and
10545 2013-01-08 James Greenhalgh <james.greenhalgh@arm.com>
10547 * config/aarch64/aarch64-simd.md
10548 (aarch64_simd_bsl<mode>_internal): Add floating-point modes.
10549 (aarch64_simd_bsl): Likewise.
10550 (aarch64_vcond_internal<mode>): Likewise.
10551 (vcond<mode><mode>): Likewise.
10552 (aarch64_cm<cmp><mode>): Fix constraints, add new modes.
10553 * config/aarch64/iterators.md (V_cmp_result): Add V2DF.
10555 2013-01-08 James Greenhalgh <james.greenhalgh@arm.com>
10557 * config/aarch64/aarch64-builtins.c
10558 (aarch64_builtin_vectorized_function): Handle sqrt, sqrtf.
10560 2013-01-08 Martin Jambor <mjambor@suse.cz>
10563 * tree-sra.c (analyze_access_subtree): Return true also after
10564 potentially creating a debug-only replacement.
10566 2013-01-08 Jakub Jelinek <jakub@redhat.com>
10568 PR middle-end/55890
10569 * tree-ssa-ccp.c (evaluate_stmt): Use gimple_call_builtin_p.
10571 PR tree-optimization/54120
10572 * tree-vrp.c (range_fits_type_p): Don't allow
10573 src_precision < precision from signed vr to unsigned_p
10574 if vr->min or vr->max is negative.
10575 (simplify_float_conversion_using_ranges): Test can_float_p
10576 against CODE_FOR_nothing.
10578 2013-01-08 Jakub Jelinek <jakub@redhat.com>
10579 Richard Biener <rguenther@suse.de>
10581 PR middle-end/55851
10582 * fold-const.c (int_binop_types_match_p): Allow all INTEGRAL_TYPE_P
10583 types instead of just INTEGER_TYPE types.
10585 2013-01-07 Mark Kettenis <kettenis@openbsd.org>
10587 * config/i386/openbsdelf.h (LIBGCC2_HAS_TF_MODE, LIBGCC2_TF_CEXT,
10590 2013-01-07 Steve Ellcey <sellcey@mips.com>
10593 * config/mips/mips.opt: Change mad to mmad to match documentation.
10595 2013-01-07 Georg-Johann Lay <avr@gjlay.de>
10598 * doc/extend.texi (AVR Named Address Spaces): __memx goes into
10599 .progmemx.data now.
10601 2013-01-07 Georg-Johann Lay <avr@gjlay.de>
10604 * config/avr/avr.h (ADDR_SPACE_COUNT): New enum.
10605 (avr_addrspace_t): Add .section_name field.
10606 * config/avr/avr.c (progmem_section): Use ADDR_SPACE_COUNT as
10608 (avr_addrspace): Same. Initialize .section_name. Remove last
10609 NULL entry. Put __memx into .progmemx.data.
10610 (progmem_section_prefix): Remove.
10611 (avr_asm_init_sections): No need to initialize progmem_section.
10612 (avr_asm_named_section): Use avr_addrspace[].section_name to get
10613 section name prefix.
10614 (avr_asm_select_section): Ditto. And use get_unnamed_section to
10615 retrieve the progmem section.
10616 * avr-c.c (avr_cpu_cpp_builtins): Use ADDR_SPACE_COUNT as loop
10617 boundary to run over avr_addrspace[].
10618 (avr_register_target_pragmas): Ditto.
10620 2013-01-06 Jakub Jelinek <jakub@redhat.com>
10622 * varasm.c (output_constant_def_contents): For asan_protect_global
10623 protected strings, adjust DECL_ALIGN if needed, before testing for
10625 (place_block_symbol): Adjust size for asan protected STRING_CSTs if
10626 TREE_CONSTANT_POOL_ADDRESS_P. Increase alignment for asan protected
10628 (output_object_block): For asan protected decls, emit asan padding
10629 after their contents.
10630 * asan.c (asan_protect_global): Don't check TREE_ASM_WRITTEN here.
10631 (asan_finish_file): Test it here instead.
10633 2013-01-07 Nick Clifton <nickc@redhat.com>
10634 Matthias Klose <doko@debian.org>
10635 Doug Kwan <dougkwan@google.com>
10636 H.J. Lu <hongjiu.lu@intel.com>
10639 * collect2.c (main): Support -fuse-ld=bfd and -fuse-ld=gold.
10641 * common.opt: Add fuse-ld=bfd and fuse-ld=gold.
10643 * gcc.c (LINK_COMMAND_SPEC): Pass -fuse-ld=* to collect2.
10645 * opts.c (comman_handle_option): Ignore -fuse-ld=bfd and -fuse-ld=gold.
10647 * doc/invoke.texi: Document -fuse-ld=bfd and -fuse-ld=gold.
10649 2013-01-07 Georg-Johann Lay <avr@gjlay.de>
10652 * doc/install.texi (Cross-Compiler-Specific Options): Document
10655 2013-01-07 Tejas Belagod <tejas.belagod@arm.com>
10657 * config/aarch64/arm_neon.h (vmovn_high_is16, vmovn_high_s32,
10658 vmovn_high_s64, vmovn_high_u16, vmovn_high_u32, vmovn_high_u64,
10659 vqmovn_high_s16, vqmovn_high_s32, vqmovn_high_s64, vqmovn_high_u16,
10660 vqmovn_high_u32, vqmovn_high_u64, vqmovun_high_s16, vqmovun_high_s32,
10661 vqmovun_high_s64): Fix source operand number and update copyright.
10663 2013-01-07 Richard Biener <rguenther@suse.de>
10665 PR middle-end/55890
10666 * gimple.h (gimple_call_builtin_p): New overload.
10667 * gimple.c (validate_call): New function.
10668 (gimple_call_builtin_p): Likewise.
10669 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
10670 Use gimple_call_builtin_p.
10671 (find_func_clobbers): Likewise.
10672 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
10673 (strlen_optimize_stmt): Likewise.
10675 2013-01-07 James Greenhalgh <james.greenhalgh@arm.com>
10677 * config/aarch64/arm_neon.h (vld1_dup_*): Make argument const.
10678 (vld1q_dup_*): Likewise.
10679 (vld1_*): Likewise.
10680 (vld1q_*): Likewise.
10681 (vld1_lane_*): Likewise.
10682 (vld1q_lane_*): Likewise.
10684 2013-01-07 Richard Biener <rguenther@suse.de>
10686 * lto-streamer.h (LTO_minor_version): Bump to 2.
10688 2013-01-07 James Greenhalgh <james.greenhalgh@arm.com>
10690 * config/aarch64/aarch64-protos.h
10691 (aarch64_const_double_zero_rtx_p): Rename to...
10692 (aarch64_float_const_zero_rtx_p): ...this.
10693 (aarch64_float_const_representable_p): New.
10694 (aarch64_output_simd_mov_immediate): Likewise.
10695 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>): Refactor
10696 move immediate case.
10697 * config/aarch64/aarch64.c
10698 (aarch64_const_double_zero_rtx_p): Rename to...
10699 (aarch64_float_const_zero_rtx_p): ...this.
10700 (aarch64_print_operand): Allow printing of new constants.
10701 (aarch64_valid_floating_const): New.
10702 (aarch64_legitimate_constant_p): Check for valid floating-point
10704 (aarch64_simd_valid_immediate): Likewise.
10705 (aarch64_vect_float_const_representable_p): New.
10706 (aarch64_float_const_representable_p): Likewise.
10707 (aarch64_simd_imm_zero_p): Also allow for floating-point 0.0.
10708 (aarch64_output_simd_mov_immediate): New.
10709 * config/aarch64/aarch64.md (*movsf_aarch64): Add new alternative.
10710 (*movdf_aarch64): Likewise.
10711 * config/aarch64/constraints.md (Ufc): New.
10712 (Y): call aarch64_float_const_zero_rtx.
10713 * config/aarch64/predicates.md (aarch64_fp_compare_operand): New.
10715 2013-01-07 Richard Biener <rguenther@suse.de>
10717 PR tree-optimization/55888
10718 PR tree-optimization/55862
10719 * tree-ssa-pre.c (phi_translate_1): Revert previous change.
10720 (valid_in_sets): Check if a NAME has a leader in AVAIL_OUT,
10721 not if it is contained therein.
10723 2013-01-07 Georg-Johann Lay <avr@gjlay.de>
10725 * config/avr/t-avr: Typo.
10727 2013-01-07 Georg-Johann Lay <avr@gjlay.de>
10730 * config/avr/t-avr: Don't automatically rebuild
10731 $(srcdir)/config/avr/t-multilib
10732 $(srcdir)/config/avr/avr-tables.opt
10733 $(srcdir)/doc/avr-mmcu.texi
10734 (avr-mcus): New phony target to build them on request.
10735 (s-avr-mlib, s-avr-mmcu-texi): Remove.
10736 * avr/avr-mcus.def: Adjust comments.
10738 2013-01-07 Uros Bizjak <ubizjak@gmail.com>
10740 * config/i386/i386.c (DEFAULT_PCC_STRUCT_RETURN): Remove.
10742 2013-01-06 Richard Sandiford <rdsandiford@googlemail.com>
10744 * file-find.c, file-find.h, realmpfr.c: Add FSF as copyright holder.
10746 2013-01-06 Richard Sandiford <rdsandiford@googlemail.com>
10748 * config/tilepro/gen-mul-tables.cc: Put copyright on one line.
10750 2013-01-05 David Edelsohn <dje.gcc@gmail.com>
10752 * config/rs6000/aix53.h (LIB_SPEC): Add -lpthreads when compiling
10753 to generate profiling.
10754 * config/rs6000/aix64.h (LIB_SPEC): Same.
10756 2013-01-04 Andrew Pinski <apinski@cavium.com>
10758 * config/aarch64/aarch64.c (aarch64_fixed_condition_code_regs):
10760 (TARGET_FIXED_CONDITION_CODE_REGS): Define.
10762 2013-01-04 Uros Bizjak <ubizjak@gmail.com>
10764 * config/i386/i386.c (ix86_legitimize_address): Call convert_to_mode
10766 (ix86_expand_move): Ditto.
10767 (ix86_zero_extend_to_Pmode): Ditto.
10768 (ix86_expand_call): Ditto.
10769 (ix86_expand_special_args_builtin): Ditto.
10770 (ix86_expand_builtin): Ditto.
10772 2013-01-04 Richard Biener <rguenther@suse.de>
10774 PR tree-optimization/55862
10775 * tree-ssa-pre.c (phi_translate_1): Valueize SSA names after
10776 translating them through PHI nodes.
10778 2013-01-04 Martin Jambor <mjambor@suse.cz>
10780 PR tree-optimization/55755
10781 * tree-sra.c (sra_modify_assign): Do not check that an access has no
10782 children when trying to avoid producing a VIEW_CONVERT_EXPR.
10784 2013-01-04 Marek Polacek <polacek@redhat.com>
10786 PR middle-end/55859
10787 * opts.c (default_options_optimization): Clarify error message.
10789 2013-01-04 Richard Biener <rguenther@suse.de>
10791 PR middle-end/55863
10792 * fold-const.c (split_tree): Undo -X - 1 to ~X folding for
10795 2013-01-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
10798 * config/pa/pa.md (movsi): Revert previous change.
10799 * config/pa/pa.c (pa_legitimate_constant_p): Reject all TLS symbol
10802 2013-01-03 Richard Henderson <rth@redhat.com>
10804 * config/i386/i386.c (ix86_expand_move): Always assign to op1
10805 after eliminating TLS symbols.
10807 2013-01-03 Marc Glisse <marc.glisse@inria.fr>
10810 * graphite-interchange.c (pdr_stride_in_loop): Use gmp_fprintf.
10811 * graphite-poly.c (debug_gmp_value): Likewise.
10813 2013-01-03 Uros Bizjak <ubizjak@gmail.com>
10816 * config/i386/i386-c.c (ix86_target_macros_internal): Depending on
10817 selected code model, define __code_mode_small__, __code_model_medium__,
10818 __code_model_large__, __code_model_32__ or __code_model_kernel__.
10819 * config/i386/cpuid.h (__cpuid, __cpuid_count) [__i386__]: Prefix
10820 xchg temporary register with %k. Declare temporary register as
10822 [__x86_64__]: For medium and large code models, preserve %rbx register.
10824 2013-01-03 Richard Biener <rguenther@suse.de>
10826 * tree-data-ref.c (dump_conflict_function): Use less vertical spacing.
10827 (dump_subscript): Adjust.
10828 (finalize_ddr_dependent): Do not dump redundant info.
10829 (analyze_siv_subscript): Adjust.
10830 (subscript_dependence_tester): Likewise.
10831 (compute_affine_dependence): Likewise.
10833 2013-01-03 Richard Biener <rguenther@suse.de>
10836 2013-01-03 Richard Biener <rguenther@suse.de>
10838 PR tree-optimization/55857
10839 * tree-vect-stmts.c (vectorizable_load): Do not setup
10840 re-alignment for invariant loads.
10842 2013-01-02 Richard Biener <rguenther@suse.de>
10844 * tree-vect-stmts.c (vectorizable_load): When vectorizing an
10845 invariant load do not generate a vector load from the scalar location.
10847 2013-01-03 Richard Biener <rguenther@suse.de>
10849 * tree-vect-loop.c (vect_analyze_loop_form): Clarify reason
10850 for not vectorizing.
10851 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Do
10852 not build INDIRECT_REFs, call get_name once only.
10853 (vect_create_data_ref_ptr): Likewise. Dump base object kind
10854 based on DR_BASE_OBJECT, not DR_BASE_ADDRESS.
10856 2013-01-03 Richard Biener <rguenther@suse.de>
10858 PR tree-optimization/55857
10859 * tree-vect-stmts.c (vectorizable_load): Do not setup
10860 re-alignment for invariant loads.
10862 2013-01-03 Richard Biener <rguenther@suse.de>
10865 * lto-symtab.c (lto_symtab_merge_decls_1): As last resort, always
10866 prefer a built-in decl.
10868 2013-01-03 Jakub Jelinek <jakub@redhat.com>
10870 * gcc.c (process_command): Update copyright notice dates.
10871 * gcov.c (print_version): Likewise.
10872 * gcov-dump.c (print_version): Likewise.
10874 PR rtl-optimization/55838
10875 * loop-iv.c (iv_number_of_iterations): Call lowpart_subreg on
10876 iv0.step, iv1.step and step.
10878 2013-01-03 Jakub Jelinek <jakub@redhat.com>
10879 Marc Glisse <marc.glisse@inria.fr>
10881 PR tree-optimization/55832
10882 * fold-const.c (fold_binary_loc): For ABS_EXPR<x> >= 0 and
10883 ABS_EXPR<x> < 0 folding use constant_boolean_node instead of
10884 integer_{one,zero}_node.
10886 2013-01-03 Jakub Jelinek <jakub@redhat.com>
10889 * params.def (PARAM_MAX_VARTRACK_REVERSE_OP_SIZE): New param.
10890 * var-tracking.c (reverse_op): Don't add reverse ops to
10891 VALUEs that have already
10892 PARAM_VALUE (PARAM_MAX_VARTRACK_REVERSE_OP_SIZE) or longer locs list.
10894 2013-01-02 Gerald Pfeifer <gerald@pfeifer.com>
10896 * doc/contrib.texi: Note years as release manager for Mark Mitchell.
10898 2013-01-02 Teresa Johnson <tejohnson@google.com>
10900 * dumpfile.c (dump_loc): Print filename with location.
10901 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Use
10902 new location_t parameter to emit complete unroll message with
10903 new dump framework.
10904 (canonicalize_loop_induction_variables): Compute loops location
10905 and pass to try_unroll_loop_completely.
10906 * loop-unroll.c (report_unroll_peel): New function.
10907 (peel_loops_completely): Use new dump format with location
10908 for main dumpfile message, and invoke report_unroll_peel on success.
10909 (decide_unrolling_and_peeling): Ditto.
10910 (decide_peel_once_rolling): Remove old dumpfile message subsumed
10911 by report_unroll_peel.
10912 (decide_peel_completely): Ditto.
10913 (decide_unroll_constant_iterations): Ditto.
10914 (decide_unroll_runtime_iterations): Ditto.
10915 (decide_peel_simple): Ditto.
10916 (decide_unroll_stupid): Ditto.
10917 * cfgloop.c (get_loop_location): New function.
10918 * cfgloop.h (get_loop_location): Declare.
10920 2013-01-02 Sriraman Tallam <tmsriram@google.com>
10922 * config/i386/i386.c (fold_builtin_cpu): Remove unnecessary checks for
10925 2013-01-02 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
10927 PR middle-end/55198
10928 * expr.c (expand_expr_real_1): Don't use bitfield extraction for non
10929 BLKmode objects when EXPAND_MEMORY is specified.
10931 2013-01-02 Sriraman Tallam <tmsriram@google.com>
10933 * config/i386/i386.c (ix86_get_function_versions_dispatcher): Fix bug
10935 (fold_builtin_cpu): Do not share cpu model decls across statements.
10937 2013-01-02 Jason Merrill <jason@redhat.com>
10940 * tree.c (build_array_type_1): Revert earlier change.
10942 2013-01-02 Yufeng Zhang <yufeng.zhang@arm.com>
10944 * config/aarch64/aarch64-cores.def: Add entries for "cortex-a53" and
10946 * config/aarch64/aarch64-tune.md: Re-generate.
10948 2013-01-02 Richard Biener <rguenther@suse.de>
10950 * tree-vect-stmts.c (vectorizable_load): When vectorizing an
10951 invariant load do not generate a vector load from the scalar location.
10953 2013-01-02 Richard Biener <rguenther@suse.de>
10956 * configure.ac: Add $GMPINC to CFLAGS/CXXFLAGS.
10957 * configure: Regenerate.
10959 2013-01-02 Richard Sandiford <rdsandiford@googlemail.com>
10961 * builtins.c (expand_builtin_mathfn, expand_builtin_mathfn_2)
10962 (expand_builtin_mathfn_ternary, expand_builtin_mathfn_3)
10963 (expand_builtin_int_roundingfn_2): Keep the original target around
10964 for the fallback case.
10966 2013-01-02 Richard Sandiford <rdsandiford@googlemail.com>
10968 * tree-vrp.c (range_fits_type_p): Require the MSB of the double_int
10969 to be clear for sign changes.
10971 2013-01-01 Jan Hubicka <jh@suse.cz>
10973 * ipa-inline-analysis.c: Fix formatting.
10975 2013-01-01 Jakub Jelinek <jakub@redhat.com>
10977 PR tree-optimization/55831
10978 * tree-vect-loop.c (get_initial_def_for_induction): Use
10979 gsi_after_labels instead of gsi_start_bb.
10981 Copyright (C) 2013 Free Software Foundation, Inc.
10983 Copying and distribution of this file, with or without modification,
10984 are permitted in any medium without royalty provided the copyright
10985 notice and this notice are preserved.