1 2014-05-22 Release Manager
5 2014-05-15 Peter Bergner <bergner@vnet.ibm.com>
8 2014-05-15 Peter Bergner <bergner@vnet.ibm.com>
11 * config/rs6000/htmxlintrin.h (_HTM_TBEGIN_STARTED): New define.
12 (__TM_simple_begin): Use it.
13 (__TM_begin): Likewise.
15 2014-05-14 Eric Botcazou <ebotcazou@adacore.com>
17 * config/sparc/sparc-protos.h (sparc_absnegfloat_split_legitimate):
19 * config/sparc/sparc.c (sparc_absnegfloat_split_legitimate): Likewise.
20 * config/sparc/sparc.md (fptype_ut699): New attribute.
21 (in_branch_delay): Return false if -mfix-ut699 is specified and
22 fptype_ut699 is set to single.
23 (truncdfsf2): Add fptype_ut699 attribute.
24 (fix_truncdfsi2): Likewise.
25 (floatsisf2): Change fptype attribute.
26 (fix_truncsfsi2): Likewise.
27 (negtf2_notv9): Delete.
28 (negtf2_v9): Likewise.
29 (negtf2_hq): New instruction.
30 (negtf2): New instruction and splitter.
31 (negdf2_notv9): Rewrite.
32 (abstf2_notv9): Delete.
33 (abstf2_hq_v9): Likewise.
34 (abstf2_v9): Likewise.
35 (abstf2_hq): New instruction.
36 (abstf2): New instruction and splitter.
37 (absdf2_notv9): Rewrite.
39 2014-05-14 Matthias Klose <doko@ubuntu.com>
42 2014-05-08 Manuel López-Ibáñez <manu@gcc.gnu.org>
43 Matthias Klose <doko@ubuntu.com>
46 * optc-gen.awk: Fix option handling for -Wunused-parameter.
48 2014-05-13 Peter Bergner <bergner@vnet.ibm.com>
50 * doc/sourcebuild.texi: (dfp_hw): Document.
51 (p8vector_hw): Likewise.
52 (powerpc_eabi_ok): Likewise.
53 (powerpc_elfv2): Likewise.
54 (powerpc_htm_ok): Likewise.
55 (ppc_recip_hw): Likewise.
58 2014-05-12 Senthil Kumar Selvaraj <senthil_kumar.selvaraj@atmel.com>
60 Backport from mainline
61 2014-05-12 Senthil Kumar Selvaraj <senthil_kumar.selvaraj@atmel.com>
64 * config/avr/avr.c (avr_out_store_psi): Use correct constant
67 2014-05-09 Georg-Johann Lay <avr@gjlay.de>
69 Backport from 2014-05-09 trunk r210267
72 * config/avr/avr.md (cc): Add new attribute set_vzn.
73 (addqi3, addqq3, adduqq3, subqi3, subqq3, subuqq3, negqi2) [cc]:
74 Set cc insn attribute to set_vzn instead of set_zn for alternatives
76 * config/avr/avr.c (avr_notice_update_cc): Handle SET_VZN.
77 (avr_out_plus_1): ADIW sets cc0 to CC_SET_CZN.
78 INC, DEC and ADD+ADC set cc0 to CC_CLOBBER.
80 2014-05-09 Richard Sandiford <rdsandiford@googlemail.com>
82 * builtins.c (expand_builtin_setjmp_receiver): Emit a use of
83 the hard frame pointer. Synchronize commentary with mainline.
84 * cse.c (cse_insn): Only check for volatile asms.
85 * cselib.c (cselib_process_insn): Likewise.
86 * dse.c (scan_insn): Likewise.
87 * stmt.c (expand_nl_goto_receiver): Emit a use and a clobber of
88 the hard frame pointer.
90 2014-05-08 Manuel López-Ibáñez <manu@gcc.gnu.org>
91 Matthias Klose <doko@ubuntu.com>
94 * optc-gen.awk: Fix option handling for -Wunused-parameter.
96 2014-05-08 Uros Bizjak <ubizjak@gmail.com>
99 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_RTM
102 2014-05-08 Charles Baylis <charles.baylis@linaro.org>
104 Backport from mainline
105 2014-04-07 Charles Baylis <charles.baylis@linaro.org>
108 * config/arm/arm.h (ASM_OUTPUT_CASE_END): Remove.
109 (LABEL_ALIGN_AFTER_BARRIER): Align barriers which occur after
112 2014-05-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
114 * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
115 Do not define target_cpu_default2 to generic.
117 2014-05-06 Richard Biener <rguenther@suse.de>
119 Backport from mainline
120 2014-04-14 Richard Biener <rguenther@suse.de>
123 * fold-const.c (negate_expr_p): Don't negate directional rounding
125 (fold_negate_expr): Likewise.
127 2014-05-06 Richard Biener <rguenther@suse.de>
129 Backport from mainline
130 2014-04-17 Richard Biener <rguenther@suse.de>
133 * tree-ssa-propagate.c (valid_gimple_rhs_p): Only allow effective
134 boolean results for comparisons.
136 2014-04-07 Richard Biener <rguenther@suse.de>
138 PR tree-optimization/60766
139 * tree-ssa-loop-ivopts.c (cand_value_at): Compute in an
141 (may_eliminate_iv): Convert cand_value_at result to desired
144 2014-04-23 Richard Biener <rguenther@suse.de>
146 PR tree-optimization/60903
147 * tree-ssa-loop-im.c (execute_sm_if_changed): Properly apply
148 IRREDUCIBLE_LOOP loop flags to newly created BBs and edges.
150 2014-05-05 Richard Biener <rguenther@suse.de>
152 Backport from mainline
153 2014-04-23 Richard Biener <rguenther@suse.de>
156 * tree-inline.c (declare_return_variable): Use mark_addressable.
158 2014-04-07 Richard Biener <rguenther@suse.de>
161 * tree-ssa-operands.c (maybe_add_call_vops): Also add VDEFs
163 * tree-cfgcleanup.c (fixup_noreturn_call): Do not remove VDEFs.
165 2014-04-14 Richard Biener <rguenther@suse.de>
167 PR tree-optimization/59817
168 PR tree-optimization/60453
169 * graphite-scop-detection.c (graphite_can_represent_scev): Complete
170 recursion to catch all CHRECs in the scalar evolution and restrict
171 the predicate for the remains appropriately.
173 2014-04-17 Richard Biener <rguenther@suse.de>
175 PR tree-optimization/60836
176 * tree-vect-loop.c (vect_create_epilog_for_reduction): Force
177 initial PHI args to be gimple values.
179 2014-05-05 Jakub Jelinek <jakub@redhat.com>
181 Backported from mainline
182 2014-04-25 Jakub Jelinek <jakub@redhat.com>
184 PR tree-optimization/60960
185 * tree-vect-generic.c (expand_vector_operation): Only call
186 expand_vector_divmod if type's mode satisfies VECTOR_MODE_P.
188 2014-05-04 Peter Bergner <bergner@vnet.ibm.com>
190 * config/rs6000/rs6000.h (RS6000_BTM_HARD_FLOAT): New define.
191 (RS6000_BTM_COMMON): Add RS6000_BTM_HARD_FLOAT.
192 (TARGET_EXTRA_BUILTINS): Add TARGET_HARD_FLOAT.
193 * config/rs6000/rs6000-builtin.def (BU_MISC_1):
194 Use RS6000_BTM_HARD_FLOAT.
195 (BU_MISC_2): Likewise.
196 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Handle
197 RS6000_BTM_HARD_FLOAT.
198 (rs6000_option_override_internal): Enforce -mhard-float if -mhard-dfp
200 (rs6000_invalid_builtin): Add hard floating builtin support.
201 (rs6000_expand_builtin): Relax the gcc_assert to allow the new
203 (rs6000_builtin_mask_names): Add RS6000_BTM_HARD_FLOAT.
205 2014-05-03 Joey Ye <joey.ye@arm.com>
207 Backport from mainline r209463
208 2014-04-17 Joey Ye <joey.ye@arm.com>
210 * opts.c (OPT_fif_conversion, OPT_fif_conversion2): Disable for Og.
212 2014-05-03 Oleg Endo <olegendo@gcc.gnu.org>
214 Back port from mainline
216 * config/sh/sh.c: Include stdlib headers before everything else.
218 2014-05-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
220 PR tree-optimization/60930
221 * gimple-ssa-strength-reduction.c (create_mul_imm_cand): Reject
222 creating a multiply candidate by folding two constant
223 multiplicands when the result overflows.
225 2014-05-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
227 * config/aarch64/aarch64.h (TARGET_SIMD): Take AARCH64_ISA_SIMD
229 (TARGET_FLOAT): Take AARCH64_ISA_FP into account.
231 2014-04-30 Michael Meissner <meissner@linux.vnet.ibm.com>
233 Back port from mainline
234 2014-04-24 Michael Meissner <meissner@linux.vnet.ibm.com>
236 * doc/extend.texi (PowerPC Built-in Functions): Document new
237 powerpc extended divide, bcd, pack/unpack 128-bit, builtin
239 (PowerPC AltiVec/VSX Built-in Functions): Likewise.
241 * config/rs6000/predicates.md (const_0_to_3_operand): New
242 predicate to match 0..3 integer constants.
244 * config/rs6000/rs6000-builtin.def (BU_DFP_MISC_1): Add new macros
245 to support adding miscellaneous builtin functions.
246 (BU_DFP_MISC_2): Likewise.
247 (BU_P7_MISC_1): Likewise.
248 (BU_P7_MISC_2): Likewise.
249 (BU_P8V_MISC_3): Likewise.
250 (BU_MISC_1): Likewise.
251 (BU_MISC_2): Likewise.
252 (DIVWE): Add extended divide builtin functions.
260 (DXEX): Add decimal floating-point builtin functions.
272 (CDTBCD): Add new BCD builtin functions.
276 (BCDADD_LT): Likewise.
277 (BCDADD_EQ): Likewise.
278 (BCDADD_GT): Likewise.
279 (BCDADD_OV): Likewise.
281 (BCDSUB_LT): Likewise.
282 (BCDSUB_EQ): Likewise.
283 (BCDSUB_GT): Likewise.
284 (BCDSUB_OV): Likewise.
285 (PACK_TD): Add new pack/unpack 128-bit type builtin functions.
286 (UNPACK_TD): Likewise.
288 (UNPACK_TF): Likewise.
289 (UNPACK_TF_0): Likewise.
290 (UNPACK_TF_1): Likewise.
291 (PACK_V1TI): Likewise.
292 (UNPACK_V1TI): Likewise.
294 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
295 support for decimal floating point builtin functions.
296 (rs6000_expand_ternop_builtin): Add checks for the new builtin
297 functions that take constant arguments.
298 (rs6000_invalid_builtin): Add decimal floating point builtin
300 (rs6000_init_builtins): Setup long double, _Decimal64, and
301 _Decimal128 types for new builtin functions.
302 (builtin_function_type): Set the unsigned flags appropriately for
303 the new builtin functions.
304 (rs6000_opt_masks): Add support for decimal floating point builtin
307 * config/rs6000/rs6000.h (RS6000_BTM_DFP): Add support for decimal
308 floating point builtin functions.
309 (RS6000_BTM_COMMON): Likewise.
310 (RS6000_BTI_long_double): Likewise.
311 (RS6000_BTI_dfloat64): Likewise.
312 (RS6000_BTI_dfloat128): Likewise.
313 (long_double_type_internal_node): Likewise.
314 (dfloat64_type_internal_node): Likewise.
315 (dfloat128_type_internal_node): Likewise.
317 * config/rs6000/altivec.h (UNSPEC_BCDADD): Add support for ISA
318 2.07 bcd arithmetic instructions.
319 (UNSPEC_BCDSUB): Likewise.
320 (UNSPEC_BCD_OVERFLOW): Likewise.
321 (UNSPEC_BCD_ADD_SUB): Likewise.
322 (bcd_add_sub): Likewise.
323 (BCD_TEST): Likewise.
324 (bcd<bcd_add_sub>): Likewise.
325 (bcd<bcd_add_sub>_test): Likewise.
326 (bcd<bcd_add_sub>_test2): Likewise.
327 (bcd<bcd_add_sub>_<code>): Likewise.
328 (peephole2 for combined bcd ops): Likewise.
330 * config/rs6000/dfp.md (UNSPEC_DDEDPD): Add support for new
331 decimal floating point builtin functions.
332 (UNSPEC_DENBCD): Likewise.
333 (UNSPEC_DXEX): Likewise.
334 (UNSPEC_DIEX): Likewise.
335 (UNSPEC_DSCLI): Likewise.
336 (UNSPEC_DSCRI): Likewise.
337 (D64_D128): Likewise.
338 (dfp_suffix): Likewise.
339 (dfp_ddedpd_<mode>): Likewise.
340 (dfp_denbcd_<mode>): Likewise.
341 (dfp_dxex_<mode>): Likewise.
342 (dfp_diex_<mode>): Likewise.
343 (dfp_dscli_<mode>): Likewise.
344 (dfp_dscri_<mode>): Likewise.
346 * config/rs6000/rs6000.md (UNSPEC_ADDG6S): Add support for new BCD
348 (UNSPEC_CDTBCD): Likewise.
349 (UNSPEC_CBCDTD): Likewise.
350 (UNSPEC_DIVE): Add support for new extended divide builtin
352 (UNSPEC_DIVEO): Likewise.
353 (UNSPEC_DIVEU): Likewise.
354 (UNSPEC_DIVEUO): Likewise.
355 (UNSPEC_UNPACK_128BIT): Add support for new builtin functions to
356 pack/unpack 128-bit types.
357 (UNSPEC_PACK_128BIT): Likewise.
358 (idiv_ldiv): New mode attribute to set the 32/64-bit divide type.
359 (udiv<mode>3): Use idiv_ldiv mode attribute.
360 (div<mode>3): Likewise.
361 (addg6s): Add new BCD builtin functions.
364 (UNSPEC_DIV_EXTEND): Add support for new extended divide
366 (div_extend): Likewise.
367 (div<div_extend>_<mode>"): Likewise.
368 (FP128_64): Add support for new builtin functions to pack/unpack
370 (unpack<mode>): Likewise.
371 (unpacktf_0): Likewise.
372 (unpacktf_1): Likewise.
373 (unpack<mode>_dm): Likewise.
374 (unpack<mode>_nodm): Likewise.
375 (pack<mode>): Likewise.
376 (unpackv1ti): Likewise.
377 (packv1ti): Likewise.
379 2014-04-29 Pat Haugen <pthaugen@us.ibm.com>
381 Backport from mainline
382 2014-04-17 Pat Haugen <pthaugen@us.ibm.com>
384 * config/rs6000/rs6000.md (addti3, subti3): New.
386 2014-04-28 Pat Haugen <pthaugen@us.ibm.com>
388 Backport from mainline
389 2014-04-28 Pat Haugen <pthaugen@us.ibm.com>
391 * config/rs6000/sync.md (AINT mode_iterator): Move definition.
392 (loadsync_<mode>): Change mode.
393 (load_quadpti, store_quadpti): New.
394 (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
395 * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
397 2014-04-28 Eric Botcazou <ebotcazou@adacore.com>
399 * configure.ac: Tweak GAS check for LEON instructions on SPARC.
400 * configure: Regenerate.
401 * config/sparc/sparc.opt (muser-mode): New option.
402 * config/sparc/sync.md (atomic_compare_and_swap<mode>_1): Do not enable
404 (atomic_compare_and_swap_leon3_1): New instruction for LEON3.
405 * doc/invoke.texi (SPARC options): Document -muser-mode.
407 2014-04-25 Eric Botcazou <ebotcazou@adacore.com>
410 * config/sparc/sparc.md (ashlsi3_extend): Delete.
412 2014-04-22 Michael Meissner <meissner@linux.vnet.ibm.com>
414 Back port from main line:
415 2014-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
417 * config/rs6000/rs6000-builtins.def (VBPERMQ): Add vbpermq builtin
420 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
423 * config/rs6000/altivec.md (UNSPEC_VBPERMQ): Add support for the
425 (altivec_vbpermq): Likewise.
428 * config/rs6000/altivec.h (vec_xxsldwi): Add missing define to
429 enable use of xxsldwi and xxpermdi builtin functions.
430 (vec_xxpermdi): Likewise.
432 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
433 Document use of vec_xxsldwi and vec_xxpermdi builtins.
435 2014-04-23 Uros Bizjak <ubizjak@gmail.com>
437 Backport from mainline
438 2014-04-21 Uros Bizjak <ubizjak@gmail.com>
441 * config/i386/i386.c (ix86_expand_builtin)
442 <case IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Use temporary
443 register for target RTX.
444 <case IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Ditto.
446 2014-04-23 Richard Biener <rguenther@suse.de>
448 Backport from mainline
449 2014-04-02 Richard Biener <rguenther@suse.de>
452 * optabs.c (expand_abs_nojump): Honor flag_trapv only for
453 MODE_INTs. Properly use negv_optab.
454 (expand_abs): Likewise.
456 2014-04-03 Richard Biener <rguenther@suse.de>
458 PR tree-optimization/60740
459 * graphite-scop-detection.c (stmt_simple_for_scop_p): Iterate
460 over all GIMPLE_COND operands.
462 2014-04-23 Richard Biener <rguenther@suse.de>
465 * gimplify.c (gimple_regimplify_operands): Update the
468 2014-04-21 Michael Meissner <meissner@linux.vnet.ibm.com>
470 Back port from the trunk, subversion id 209546.
472 2014-04-21 Michael Meissner <meissner@linux.vnet.ibm.com>
475 * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64 case):
476 If mode is DDmode and TARGET_E500_DOUBLE allow move.
478 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print some
479 more debug information for E500 if -mdebug=reg.
481 2014-04-18 Richard Henderson <rth@redhat.com>
483 * config/aarch64/aarch64.c (aarch64_register_move_cost): Pass a mode
484 to GET_MODE_SIZE, not a reg_class_t.
486 2014-04-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
488 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust for
490 (vsx_xxmrglw_<mode>): Likewise.
492 2014-04-15 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
495 Revert the following patch
497 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
499 Back port mainline subversion id 209025.
500 2014-04-02 Michael Meissner <meissner@linux.vnet.ibm.com>
503 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
504 software floating point or no floating point registers, do not
505 allow any type in the FPRs. Eliminate a test for SPE SIMD types
506 in GPRs that occurs after we tested for GPRs that would never be
509 * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
510 Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
511 since the FMOVE64 type is DFmode/DDmode. If TARGET_E500_DOUBLE,
512 specifically allow DDmode, since that does not use the SPE SIMD
515 2014-04-10 Vladimir Makarov <vmakarov@redhat.com>
517 PR rtl-optimization/60769
518 * lra-constraints.c (simplify_operand_subreg): Force reload of
519 paradoxical subreg if it is not in the class contents.
521 2014-04-10 Jakub Jelinek <jakub@redhat.com>
523 Backport from mainline
524 2014-03-12 Jakub Jelinek <jakub@redhat.com>
525 Marc Glisse <marc.glisse@inria.fr>
527 PR tree-optimization/60502
528 * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
529 instead of build_low_bits_mask.
531 2013-06-13 Marc Glisse <marc.glisse@inria.fr>
533 * tree.c (build_all_ones_cst): New function.
534 * tree.h (build_all_ones_cst): Declare it.
536 2013-05-10 Marc Glisse <marc.glisse@inria.fr>
538 * tree.c (build_minus_one_cst): New function.
539 * tree.h (build_minus_one_cst): Declare new function.
541 2014-04-10 Jakub Jelinek <jakub@redhat.com>
543 Backport from mainline
544 2014-03-28 Jakub Jelinek <jakub@redhat.com>
547 * config/i386/i386.c (ix86_copy_addr_to_reg): Call copy_addr_to_reg
548 also if addr has VOIDmode.
550 2014-03-17 Jakub Jelinek <jakub@redhat.com>
553 * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
554 note creation for the 2010-08-31 changes.
556 2014-03-06 Jakub Jelinek <jakub@redhat.com>
557 Meador Inge <meadori@codesourcery.com>
560 * config/arm/arm.c (arm_tls_symbol_p): Remove.
561 (arm_legitimize_address): Call legitimize_tls_address for any
562 arm_tls_referenced_p expression, handle constant addend. Call it
563 before testing for !TARGET_ARM.
564 (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
566 2014-04-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
568 Backport from mainline r208750
569 2014-03-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
571 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
572 pattern for vector nor instead of subtract from splat(-1).
573 (altivec_expand_vec_perm_const_le): Likewise.
575 Backport from mainline r209235
576 2014-04-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
578 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Use vnand
579 instead of vnor to exploit possible fusion opportunity in the
581 (altivec_expand_vec_perm_const_le): Likewise.
583 2014-04-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
585 Revert following patch
586 2014-04-08 Pat Haugen <pthaugen@us.ibm.com>
588 Backport from mainline
589 2014-04-08 Pat Haugen <pthaugen@us.ibm.com>
591 * config/rs6000/sync.md (AINT mode_iterator): Move definition.
592 (loadsync_<mode>): Change mode.
593 (load_quadpti, store_quadpti): New.
594 (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
595 * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
597 2014-04-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
599 Backport from mainline r202642
600 2013-09-17 Alan Modra <amodra@gmail.com>
603 * config/rs6000/driver-rs6000.c (elf_platform): Revert 2013-06-11
606 2014-04-08 Pat Haugen <pthaugen@us.ibm.com>
608 Backport from mainline
609 2014-04-08 Pat Haugen <pthaugen@us.ibm.com>
611 * config/rs6000/sync.md (AINT mode_iterator): Move definition.
612 (loadsync_<mode>): Change mode.
613 (load_quadpti, store_quadpti): New.
614 (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
615 * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
617 2014-04-07 Martin Jambor <mjambor@suse.cz>
620 * ipa-cp.c (propagate_constants_accross_call): Do not propagate
623 2014-04-07 Dominique d'Humieres <dominiq@lps.ens.fr>
625 Backport from mainline
626 2013-09-14 Iain Sandoe <iains@gcc.gnu.org>
629 * config/darwin.c (darwin_objc2_section): Note if ObjC Metadata
631 (darwin_objc1_section): Likewise.
632 (darwin_file_end): Emit Image Info section when required.
634 2014-04-05 Alan Modra <amodra@gmail.com>
637 2014-01-28 Alan Modra <amodra@gmail.com>
638 * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
639 * configure.ac <recursive call for build != host>: Define
640 GENERATOR_FILE. Comment. Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
641 and LD_FOR_BUILD too.
642 * configure: Regenerate.
644 2014-04-04 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
646 Backport from mainline r208895:
647 2014-03-28 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
649 * config/rs6000/rs6000.c (fusion_gpr_load_p): Refuse optimization
650 if it would clobber the stack pointer, even temporarily.
652 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
654 Back port from main line:
655 2014-04-01 Michael Meissner <meissner@linux.vnet.ibm.com>
657 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
660 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
662 Back port mainline subversion id 209025.
663 2014-04-02 Michael Meissner <meissner@linux.vnet.ibm.com>
666 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
667 software floating point or no floating point registers, do not
668 allow any type in the FPRs. Eliminate a test for SPE SIMD types
669 in GPRs that occurs after we tested for GPRs that would never be
672 * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
673 Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
674 since the FMOVE64 type is DFmode/DDmode. If TARGET_E500_DOUBLE,
675 specifically allow DDmode, since that does not use the SPE SIMD
678 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
680 Backport from mainline r205308
681 2013-11-23 David Edelsohn <dje.gcc@gmail.com>
683 * config/rs6000/rs6000.c (IN_NAMED_SECTION): New macro.
684 (rs6000_xcoff_select_section): Place decls with stricter alignment
686 (rs6000_xcoff_unique_section): Allow unique sections for
687 uninitialized data with strict alignment.
689 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
691 Backport from mainline
692 2013-11-15 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
694 * gcc/configure: Regenerate.
696 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
699 2013-04-25 Alan Modra <amodra@gmail.com>
702 * config/rs6000/rs6000.md (rotlsi3_internal7): Rename to
703 rotlsi3_internal7le and condition on !BYTES_BIG_ENDIAN.
704 (rotlsi3_internal8be): New BYTES_BIG_ENDIAN insn.
705 Repeat for many other rotate/shift and mask patterns using subregs.
707 (ashrdisi3_noppc64): Rename to ashrdisi3_noppc64be and condition
710 2013-06-07 Alan Modra <amodra@gmail.com>
712 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
713 override user -mfp-in-toc.
714 (offsettable_ok_by_alignment): Consider just the current access
715 rather than the whole object, unless BLKmode. Handle
716 CONSTANT_POOL_ADDRESS_P constants that lack a decl too.
717 (use_toc_relative_ref): Allow CONSTANT_POOL_ADDRESS_P constants
719 * config/rs6000/linux64.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Don't
720 override user -mfp-in-toc or -msum-in-toc. Default to
721 -mno-fp-in-toc for -mcmodel=medium.
723 2013-06-18 Alan Modra <amodra@gmail.com>
725 * config/rs6000/rs6000.h (enum data_align): New.
726 (LOCAL_ALIGNMENT, DATA_ALIGNMENT): Use rs6000_data_alignment.
727 (DATA_ABI_ALIGNMENT): Define.
728 (CONSTANT_ALIGNMENT): Correct comment.
729 * config/rs6000/rs6000-protos.h (rs6000_data_alignment): Declare.
730 * config/rs6000/rs6000.c (rs6000_data_alignment): New function.
732 2013-07-11 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
734 * config/rs6000/rs6000.md (""*tls_gd_low<TLSmode:tls_abi_suffix>"):
735 Require GOT register as additional operand in UNSPEC.
736 ("*tls_ld_low<TLSmode:tls_abi_suffix>"): Likewise.
737 ("*tls_got_dtprel_low<TLSmode:tls_abi_suffix>"): Likewise.
738 ("*tls_got_tprel_low<TLSmode:tls_abi_suffix>"): Likewise.
739 ("*tls_gd<TLSmode:tls_abi_suffix>"): Update splitter.
740 ("*tls_ld<TLSmode:tls_abi_suffix>"): Likewise.
741 ("tls_got_dtprel_<TLSmode:tls_abi_suffix>"): Likewise.
742 ("tls_got_tprel_<TLSmode:tls_abi_suffix>"): Likewise.
744 2014-01-23 Pat Haugen <pthaugen@us.ibm.com>
746 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
747 force flag_ira_loop_pressure if set via command line.
749 2014-02-06 Alan Modra <amodra@gmail.com>
752 * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
753 change SDmode to DDmode when lra_in_progress.
755 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
759 2014-03-12 Michael Meissner <meissner@linux.vnet.ibm.com>
761 * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
765 (VEC_base): Likewise.
766 (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
767 registers, we need to swap double words in little endian mode.
769 * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
770 to be a container mode for 128-bit integer operations added in ISA
771 2.07. Unlike TImode and PTImode, the preferred register set is
772 the Altivec/VMX registers for the 128-bit operations.
774 * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
776 (rs6000_split_128bit_ok_p): Likewise.
778 * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
779 macros for creating ISA 2.07 normal and overloaded builtin
780 functions with 3 arguments.
781 (BU_P8V_OVERLOAD_3): Likewise.
782 (VPERM_1T): Add support for V1TImode in 128-bit vector operations
783 for use as overloaded functions.
784 (VPERM_1TI_UNS): Likewise.
785 (VSEL_1TI): Likewise.
786 (VSEL_1TI_UNS): Likewise.
787 (ST_INTERNAL_1ti): Likewise.
788 (LD_INTERNAL_1ti): Likewise.
789 (XXSEL_1TI): Likewise.
790 (XXSEL_1TI_UNS): Likewise.
791 (VPERM_1TI): Likewise.
792 (VPERM_1TI_UNS): Likewise.
793 (XXPERMDI_1TI): Likewise.
795 (LXVD2X_V1TI): Likewise.
796 (STXVD2X_V1TI): Likewise.
797 (VEC_INIT_V1TI): Likewise.
798 (VEC_SET_V1TI): Likewise.
799 (VEC_EXT_V1TI): Likewise.
800 (EQV_V1TI): Likewise.
801 (NAND_V1TI): Likewise.
802 (ORC_V1TI): Likewise.
803 (VADDCUQ): Add support for 128-bit integer arithmetic instructions
804 added in ISA 2.07. Add both normal 'altivec' builtins, and the
808 (VADDEUQM): Likewise.
809 (VADDECUQ): Likewise.
810 (VSUBEUQM): Likewise.
811 (VSUBECUQ): Likewise.
813 * config/rs6000/rs6000-c.c (__int128_type): New static to hold
814 __int128_t and __uint128_t types.
815 (__uint128_type): Likewise.
816 (altivec_categorize_keyword): Add support for vector __int128_t,
817 vector __uint128_t, vector __int128, and vector unsigned __int128
818 as a container type for TImode operations that need to be done in
819 VSX/Altivec registers.
820 (rs6000_macro_to_expand): Likewise.
821 (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
822 to support 128-bit integer instructions vaddcuq, vadduqm,
823 vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
824 (altivec_resolve_overloaded_builtin): Add support for V1TImode.
826 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
827 for V1TImode, and set up preferences to use VSX/Altivec
828 registers. Setup VSX reload handlers.
829 (rs6000_debug_reg_global): Likewise.
830 (rs6000_init_hard_regno_mode_ok): Likewise.
831 (rs6000_preferred_simd_mode): Likewise.
832 (vspltis_constant): Do not allow V1TImode as easy altivec
834 (easy_altivec_constant): Likewise.
835 (output_vec_const_move): Likewise.
836 (rs6000_expand_vector_set): Convert V1TImode set and extract to
838 (rs6000_expand_vector_extract): Likewise.
839 (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
841 (rs6000_const_vec): Add support for V1TImode.
842 (rs6000_emit_le_vsx_load): Swap double words when loading or
843 storing TImode/V1TImode.
844 (rs6000_emit_le_vsx_store): Likewise.
845 (rs6000_emit_le_vsx_move): Likewise.
846 (rs6000_emit_move): Add support for V1TImode.
847 (altivec_expand_ld_builtin): Likewise.
848 (altivec_expand_st_builtin): Likewise.
849 (altivec_expand_vec_init_builtin): Likewise.
850 (altivec_expand_builtin): Likewise.
851 (rs6000_init_builtins): Add support for V1TImode type. Add
852 support for ISA 2.07 128-bit integer builtins. Define type names
853 for the VSX/Altivec vector types.
854 (altivec_init_builtins): Add support for overloaded vector
855 functions with V1TImode type.
856 (rs6000_preferred_reload_class): Prefer Altivec registers for
858 (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
860 (rs6000_split_128bit_ok_p): Likewise.
861 (rs6000_handle_altivec_attribute): Create V1TImode from vector
862 __int128_t and vector __uint128_t.
864 * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
872 (VS_scalar): Likewise.
873 (VS_double): Likewise.
874 (vsx_set_v1ti): New builtin function to create V1TImode from
877 * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say
878 whether we support the ISA 2.07 128-bit integer arithmetic
880 (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
881 (enum rs6000_builtin_type_index): Add fields to hold V1TImode
882 and TImode types for use with the builtin functions.
883 (V1TI_type_node): Likewise.
884 (unsigned_V1TI_type_node): Likewise.
885 (intTI_type_internal_node): Likewise.
886 (uintTI_type_internal_node): Likewise.
888 * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA
889 2.07 128-bit builtin functions.
890 (UNSPEC_VADDEUQM): Likewise.
891 (UNSPEC_VADDECUQ): Likewise.
892 (UNSPEC_VSUBCUQ): Likewise.
893 (UNSPEC_VSUBEUQM): Likewise.
894 (UNSPEC_VSUBECUQ): Likewise.
895 (VM): Add V1TImode to vector mode iterators.
898 (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
899 (altivec_vaddcuq): Likewise.
900 (altivec_vsubuqm): Likewise.
901 (altivec_vsubcuq): Likewise.
902 (altivec_vaddeuqm): Likewise.
903 (altivec_vaddecuq): Likewise.
904 (altivec_vsubeuqm): Likewise.
905 (altivec_vsubecuq): Likewise.
907 * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
909 (BOOL_128): Likewise.
910 (BOOL_REGS_OUTPUT): Likewise.
911 (BOOL_REGS_OP1): Likewise.
912 (BOOL_REGS_OP2): Likewise.
913 (BOOL_REGS_UNARY): Likewise.
914 (BOOL_REGS_AND_CR0): Likewise.
916 * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
917 128-bit integer builtin support.
918 (vec_vadduqm): Likewise.
919 (vec_vaddecuq): Likewise.
920 (vec_vaddeuqm): Likewise.
921 (vec_vsubecuq): Likewise.
922 (vec_vsubeuqm): Likewise.
923 (vec_vsubcuq): Likewise.
924 (vec_vsubuqm): Likewise.
926 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
927 Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
928 vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
929 128-bit integer add/subtract to ISA 2.07.
931 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
933 Apply mainline r207798
934 2014-02-26 Alan Modra <amodra@gmail.com>
937 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
938 find_replacement on parts of insn rtl that might be reloaded.
940 Backport from mainline r208287
941 2014-03-03 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
943 * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
944 reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
945 constraint on constants to permit them being loaded into
946 GENERAL_REGS or BASE_REGS.
948 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
950 Backport from mainline r207699.
951 2014-02-11 Michael Meissner <meissner@linux.vnet.ibm.com>
954 * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
955 for VSX/Altivec vectors that land in GPR registers.
957 Backport from mainline r207808.
958 2014-02-15 Michael Meissner <meissner@linux.vnet.ibm.com>
961 * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
962 (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
963 into 64-bit and 32-bit moves. On 64-bit moves, add support for
964 using direct move instructions on ISA 2.07. Also adjust
965 instruction length for 64-bit.
966 (mov<mode>_64bit, TFmode/TDmode): Likewise.
967 (mov<mode>_32bit, TFmode/TDmode): Likewise.
969 Backport from mainline r207868.
970 2014-02-18 Michael Meissner <meissner@linux.vnet.ibm.com>
973 * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
974 Split 64-bit moves into 2 patterns. Do not allow the use of
975 direct move for TDmode in little endian, since the decimal value
976 has little endian bytes within a word, but the 64-bit pieces are
977 ordered in a big endian fashion, and normal subreg's of TDmode are
979 (mov<mode>_64bit_dm): Likewise.
980 (movtd_64bit_nodm): Likewise.
982 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
984 Backport from mainline r207658
985 2014-02-06 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
987 * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
988 big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
989 -mcall-openbsd, or -mcall-linux.
990 (CC1_ENDIAN_BIG_SPEC): Remove.
991 (CC1_ENDIAN_LITTLE_SPEC): Remove.
992 (CC1_ENDIAN_DEFAULT_SPEC): Remove.
993 (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
994 (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
995 and %cc1_endian_default.
996 * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
998 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1000 Little Endian Vector API Support
1001 Backport from mainline r206443
1002 2014-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1004 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1005 two duplicate entries.
1007 Backport from mainline r206494
1008 2014-01-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1010 * doc/invoke.texi: Add -maltivec={be,le} options, and document
1011 default element-order behavior for -maltivec.
1012 * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
1013 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
1014 that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
1015 when targeting big endian, at least for now.
1016 * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
1018 Backport from mainline r206541
1019 2014-01-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1021 * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
1023 Backport from mainline r206590
1024 2014-01-13 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1026 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1027 Implement -maltivec=be for vec_insert and vec_extract.
1029 Backport from mainline r206641
1030 2014-01-15 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
1032 * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
1033 and vmulosh rather than call gen_vec_widen_smult_*.
1034 (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
1035 than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
1036 (vec_widen_smult_even_v16qi): Likewise.
1037 (vec_widen_umult_even_v8hi): Likewise.
1038 (vec_widen_smult_even_v8hi): Likewise.
1039 (vec_widen_umult_odd_v16qi): Likewise.
1040 (vec_widen_smult_odd_v16qi): Likewise.
1041 (vec_widen_umult_odd_v8hi): Likewise.
1042 (vec_widen_smult_odd_v8hi): Likewise.
1043 (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
1044 vmuloub rather than call gen_vec_widen_umult_*.
1045 (vec_widen_umult_lo_v16qi): Likewise.
1046 (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
1047 vmulosb rather than call gen_vec_widen_smult_*.
1048 (vec_widen_smult_lo_v16qi): Likewise.
1049 (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
1050 rather than call gen_vec_widen_umult_*.
1051 (vec_widen_umult_lo_v8hi): Likewise.
1052 (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
1053 rather than call gen_vec_widen_smult_*.
1054 (vec_widen_smult_lo_v8hi): Likewise.
1056 Backport from mainline r207062
1057 2014-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1059 * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
1060 correction for little endian...
1061 * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
1064 Backport from mainline r207262
1065 2014-01-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1067 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Use
1068 CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
1069 * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
1070 -maltivec=be with LE targets.
1071 (vsx_mergeh_<mode>): Likewise.
1072 * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New
1074 (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
1075 (altivec_vmrghb): Replace with define_expand and new
1076 *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE
1078 (altivec_vmrghb_direct): New define_insn.
1079 (altivec_vmrghh): Replace with define_expand and new
1080 *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE
1082 (altivec_vmrghh_direct): New define_insn.
1083 (altivec_vmrghw): Replace with define_expand and new
1084 *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE
1086 (altivec_vmrghw_direct): New define_insn.
1087 (*altivec_vmrghsf): Adjust for endianness.
1088 (altivec_vmrglb): Replace with define_expand and new
1089 *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE
1091 (altivec_vmrglb_direct): New define_insn.
1092 (altivec_vmrglh): Replace with define_expand and new
1093 *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE
1095 (altivec_vmrglh_direct): New define_insn.
1096 (altivec_vmrglw): Replace with define_expand and new
1097 *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE
1099 (altivec_vmrglw_direct): New define_insn.
1100 (*altivec_vmrglsf): Adjust for endianness.
1101 (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1102 (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1103 (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1104 (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1105 (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1106 (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1107 (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1108 (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1110 Backport from mainline r207318
1111 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1113 * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
1114 gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
1115 remove element index adjustment for endian (now handled in vsx.md
1117 (altivec_expand_vec_perm_const): Use
1118 gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
1119 * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
1120 (vsx_xxspltw_<mode>): Adjust element index for little endian.
1121 * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
1122 define_expand and a new define_insn *altivec_vspltb_internal;
1123 adjust for -maltivec=be on a little endian target.
1124 (altivec_vspltb_direct): New.
1125 (altivec_vsplth): Divide into a define_expand and a new
1126 define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
1127 little endian target.
1128 (altivec_vsplth_direct): New.
1129 (altivec_vspltw): Divide into a define_expand and a new
1130 define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
1131 little endian target.
1132 (altivec_vspltw_direct): New.
1133 (altivec_vspltsf): Divide into a define_expand and a new
1134 define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
1135 a little endian target.
1137 Backport from mainline r207326
1138 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1140 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
1141 unused variable "field".
1142 * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
1143 (vsx_mergeh_<mode>): Likewise.
1144 * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
1145 (altivec_vmrghh): Likewise.
1146 (altivec_vmrghw): Likewise.
1147 (altivec_vmrglb): Likewise.
1148 (altivec_vmrglh): Likewise.
1149 (altivec_vmrglw): Likewise.
1150 (altivec_vspltb): Add missing uses.
1151 (altivec_vsplth): Likewise.
1152 (altivec_vspltw): Likewise.
1153 (altivec_vspltsf): Likewise.
1155 Backport from mainline r207414
1156 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1158 * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
1159 (altivec_vsumsws): Add handling for -maltivec=be with a little
1161 (altivec_vsumsws_direct): New.
1162 (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
1163 gen_altivec_vsumsws.
1165 Backport from mainline r207415
1166 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1168 * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
1169 for vector types other than V16QImode.
1170 * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
1171 define_expand, and call altivec_expand_vec_perm_le when producing
1172 code with little endian element order.
1173 (*altivec_vperm_<mode>_internal): New insn having previous
1174 behavior of altivec_vperm_<mode>.
1175 (altivec_vperm_<mode>_uns): Change to a define_expand, and call
1176 altivec_expand_vec_perm_le when producing code with little endian
1178 (*altivec_vperm_<mode>_uns_internal): New insn having previous
1179 behavior of altivec_vperm_<mode>_uns.
1181 Backport from mainline r207520
1182 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1184 * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
1185 (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
1186 (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
1187 (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
1188 gen_altivec_vpkuwum.
1189 (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
1191 (altivec_vpks<VI_char>ss): Likewise.
1192 (altivec_vpks<VI_char>us): Likewise.
1193 (altivec_vpku<VI_char>us): Likewise.
1194 (altivec_vpku<VI_char>um): Likewise.
1195 (altivec_vpku<VI_char>um_direct): New (copy of
1196 altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
1198 (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
1199 target is little endian and -maltivec=be is not specified.
1200 (*altivec_vupkhs<VU_char>_direct): New (copy of
1201 altivec_vupkhs<VU_char> that always emits vupkhs*, for internal
1203 (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
1204 target is little endian and -maltivec=be is not specified.
1205 (*altivec_vupkls<VU_char>_direct): New (copy of
1206 altivec_vupkls<VU_char> that always emits vupkls*, for internal
1208 (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
1209 little endian and -maltivec=be is not specified.
1210 (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
1211 little endian and -maltivec=be is not specified.
1213 Backport from mainline r207521
1214 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1216 * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
1217 generation for -maltivec=be.
1218 (altivec_vsumsws): Simplify redundant test.
1220 Backport from mainline r207525
1221 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1223 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
1224 CODE_FOR_altivec_vpku[hw]um to
1225 CODE_FOR_altivec_vpku[hw]um_direct.
1226 * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
1227 UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
1228 (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
1229 UNSPEC_VUNPACK_LO_SIGN_DIRECT.
1231 Backport from mainline r207814.
1232 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1234 * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
1237 Backport from mainline r207815.
1238 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1240 * config/rs6000/altivec.md (p8_vmrgew): Handle little endian
1242 (p8_vmrgow): Likewise.
1244 Backport from mainline r207919.
1245 2014-02-19 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1247 * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
1250 Backport from mainline 208019
1251 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1253 * config/rs6000/altivec.md (altivec_lvxl): Rename as
1254 *altivec_lvxl_<mode>_internal and use VM2 iterator instead of
1256 (altivec_lvxl_<mode>): New define_expand incorporating
1257 -maltivec=be semantics where needed.
1258 (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
1259 (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
1260 semantics where needed.
1261 (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
1262 (altivec_stvx_<mode>): New define_expand incorporating
1263 -maltivec=be semantics where needed.
1264 (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
1265 VM2 iterator instead of V4SI.
1266 (altivec_stvxl_<mode>): New define_expand incorporating
1267 -maltivec=be semantics where needed.
1268 * config/rs6000/rs6000-builtin.def: Add new built-in definitions
1269 LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
1270 LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI,
1271 STVX_V2DF, STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI,
1272 STVXL_V2DF, STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI,
1274 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
1275 ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
1276 similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
1277 ALTIVEC_BUILTIN_STVXL.
1278 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New
1280 (altivec_expand_stvx_be): Likewise.
1281 * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
1282 (altivec_expand_lvx_be): Likewise.
1283 (altivec_expand_stvx_be): Likewise.
1284 (altivec_expand_builtin): Add cases for
1285 ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
1286 ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
1287 (altivec_init_builtins): Add definitions for
1288 __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
1289 __builtin_altivec_stvx_<mode>, and
1290 __builtin_altivec_stvxl_<mode>.
1292 Backport from mainline 208021
1293 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1295 * config/rs6000/altivec.md (altivec_vsumsws): Replace second
1297 (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
1298 gen_altivec_vsumsws.
1300 Backport from mainline 208049
1301 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1303 * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
1304 define_insn with define_expand and new define_insn
1305 *altivec_lve<VI_char>x_internal.
1306 (altivec_stve<VI_char>x): Replace define_insn with define_expand
1307 and new define_insn *altivec_stve<VI_char>x_internal.
1308 * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
1310 * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
1312 (altivec_expand_stvex_be): New function.
1314 Backport from mainline
1315 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1316 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
1319 Backport from mainline
1320 2014-02-25 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1321 * config/rs6000/vector.md (*vector_unordered<mode>): Change split
1322 to use canonical form for nor<mode>3.
1324 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1326 Backport from mainline
1327 2014-02-04 Michael Meissner <meissner@linux.vnet.ibm.com>
1329 * config/rs6000/rs6000.opt (-mlra): Add switch to enable the LRA
1332 * config/rs6000/rs6000.c (TARGET_LRA_P): Add support for -mlra to
1333 enable the LRA register allocator. Back port the changes from the
1334 trunk to enable LRA.
1335 (rs6000_legitimate_offset_address_p): Likewise.
1336 (legitimate_lo_sum_address_p): Likewise.
1337 (use_toc_relative_ref): Likewise.
1338 (rs6000_legitimate_address_p): Likewise.
1339 (rs6000_emit_move): Likewise.
1340 (rs6000_secondary_memory_needed_mode): Likewise.
1341 (rs6000_alloc_sdmode_stack_slot): Likewise.
1342 (rs6000_lra_p): Likewise.
1344 * config/rs6000/sync.md (load_lockedti): Copy TI/PTI variables by
1345 64-bit parts to force the register allocator to allocate even/odd
1346 register pairs for the quad word atomic instructions.
1347 (store_conditionalti): Likewise.
1349 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1351 Back port from mainline
1352 2014-01-23 Michael Meissner <meissner@linux.vnet.ibm.com>
1355 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1356 -mquad-memory-atomic. Update -mquad-memory documentation to say
1357 it is only used for non-atomic loads/stores.
1359 * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
1360 -mquad-memory or -mquad-memory-atomic switches.
1362 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
1363 -mquad-memory-atomic to ISA 2.07 support.
1365 * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
1366 to separate support of normal quad word memory operations (ldq,
1367 stq) from the atomic quad word memory operations.
1369 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
1370 support to separate non-atomic quad word operations from atomic
1371 quad word operations. Disable non-atomic quad word operations in
1372 little endian mode so that we don't have to swap words after the
1373 load and before the store.
1374 (quad_load_store_p): Add comment about atomic quad word support.
1375 (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
1376 options printed with -mdebug=reg.
1378 * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
1379 -mquad-memory-atomic as the test for whether we have quad word
1380 atomic instructions.
1381 (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic,
1382 -mquad-memory, or -mp8-vector are used, allow byte/half-word
1385 * config/rs6000/sync.md (load_lockedti): Insure that the address
1386 is a proper indexed or indirect address for the lqarx instruction.
1387 On little endian systems, swap the hi/lo registers after the lqarx
1389 (load_lockedpti): Use indexed_or_indirect_operand predicate to
1390 insure the address is valid for the lqarx instruction.
1391 (store_conditionalti): Insure that the address is a proper indexed
1392 or indirect address for the stqcrx. instruction. On little endian
1393 systems, swap the hi/lo registers before doing the stqcrx.
1395 (store_conditionalpti): Use indexed_or_indirect_operand predicate to
1396 insure the address is valid for the stqcrx. instruction.
1398 * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
1399 Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
1400 type of quad memory support is available.
1402 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1404 Apply mainline r202190, powerpc64le multilibs and multiarch dir
1405 2013-09-03 Alan Modra <amodra@gmail.com>
1407 * config.gcc (powerpc*-*-linux*): Add support for little-endian
1408 multilibs to big-endian target and vice versa.
1409 * config/rs6000/t-linux64: Use := assignment on all vars.
1410 (MULTILIB_EXTRA_OPTS): Remove fPIC.
1411 (MULTILIB_OSDIRNAMES): Specify using mapping from multilib_options.
1412 * config/rs6000/t-linux64le: New file.
1413 * config/rs6000/t-linux64bele: New file.
1414 * config/rs6000/t-linux64lebe: New file.
1416 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1418 Back port from mainline
1419 2014-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
1422 * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
1423 endian support, remove tests for WORDS_BIG_ENDIAN.
1424 (p8_mfvsrd_3_<mode>): Likewise.
1425 (reload_gpr_from_vsx<mode>): Likewise.
1426 (reload_gpr_from_vsxsf): Likewise.
1427 (p8_mfvsrd_4_disf): Likewise.
1429 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1431 Backport from mainline
1432 2013-04-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1435 * config/rs6000/rs6000.c (rs6000_emit_swdiv_high_precision): Remove.
1436 (rs6000_emit_swdiv_low_precision): Remove.
1437 (rs6000_emit_swdiv): Rewrite to handle between one and four
1438 iterations of Newton-Raphson generally; modify required number of
1439 iterations for some cases.
1440 * config/rs6000/rs6000.h (RS6000_RECIP_HIGH_PRECISION_P): Remove.
1442 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1444 Backport from mainline
1445 2013-08-19 Peter Bergner <bergner@vnet.ibm.com>
1446 Jakub Jelinek <jakub@redhat.com>
1448 * builtins.def (BUILT_IN_FABSD32): New DFP ABS builtin.
1449 (BUILT_IN_FABSD64): Likewise.
1450 (BUILT_IN_FABSD128): Likewise.
1451 * builtins.c (expand_builtin): Add support for
1452 new DFP ABS builtins.
1453 (fold_builtin_1): Likewise.
1454 * config/rs6000/dfp.md
1455 (*abstd2_fpr): Handle non-overlapping destination
1456 and source operands.
1457 (*nabstd2_fpr): Likewise.
1459 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1461 Apply mainline r205060.
1462 2013-11-20 Alan Modra <amodra@gmail.com>
1463 * config/rs6000/sysv4.h (CC1_ENDIAN_LITTLE_SPEC): Define as empty.
1464 * config/rs6000/rs6000.c (rs6000_option_override_internal): Default
1465 to strict alignment on older processors when little-endian.
1466 * config/rs6000/linux64.h (PROCESSOR_DEFAULT64): Default to power8
1469 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1471 POWER ELFv2 ABI Support
1472 Backport from mainline r204842:
1474 2013-11-15 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1476 * doc/invoke.texi (-mabi=elfv1, -mabi=elfv2): Document.
1478 Backport from mainline r204809:
1480 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1482 * config/rs6000/sysv4le.h (LINUX64_DEFAULT_ABI_ELFv2): Define.
1484 Backport from mainline r204808:
1486 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1487 Alan Modra <amodra@gmail.com>
1489 * config/rs6000/rs6000.h (RS6000_SAVE_AREA): Handle ABI_ELFv2.
1490 (RS6000_SAVE_TOC): Remove.
1491 (RS6000_TOC_SAVE_SLOT): New macro.
1492 * config/rs6000/rs6000.c (rs6000_parm_offset): New function.
1493 (rs6000_parm_start): Use it.
1494 (rs6000_function_arg_advance_1): Likewise.
1495 (rs6000_emit_prologue): Use RS6000_TOC_SAVE_SLOT.
1496 (rs6000_emit_epilogue): Likewise.
1497 (rs6000_call_aix): Likewise.
1498 (rs6000_output_function_prologue): Do not save/restore r11
1499 around calling _mcount for ABI_ELFv2.
1501 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1502 Alan Modra <amodra@gmail.com>
1504 * config/rs6000/rs6000-protos.h (rs6000_reg_parm_stack_space):
1506 * config/rs6000/rs6000.h (RS6000_REG_SAVE): Remove.
1507 (REG_PARM_STACK_SPACE): Call rs6000_reg_parm_stack_space.
1508 * config/rs6000/rs6000.c (rs6000_parm_needs_stack): New function.
1509 (rs6000_function_parms_need_stack): Likewise.
1510 (rs6000_reg_parm_stack_space): Likewise.
1511 (rs6000_function_arg): Do not replace BLKmode by Pmode when
1512 returning a register argument.
1514 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1515 Michael Gschwind <mkg@us.ibm.com>
1517 * config/rs6000/rs6000.h (FP_ARG_MAX_RETURN): New macro.
1518 (ALTIVEC_ARG_MAX_RETURN): Likewise.
1519 (FUNCTION_VALUE_REGNO_P): Use them.
1520 * config/rs6000/rs6000.c (TARGET_RETURN_IN_MSB): Define.
1521 (rs6000_return_in_msb): New function.
1522 (rs6000_return_in_memory): Handle ELFv2 homogeneous aggregates.
1523 Handle aggregates of up to 16 bytes for ELFv2.
1524 (rs6000_function_value): Handle ELFv2 homogeneous aggregates.
1526 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1527 Michael Gschwind <mkg@us.ibm.com>
1529 * config/rs6000/rs6000.h (AGGR_ARG_NUM_REG): Define.
1530 * config/rs6000/rs6000.c (rs6000_aggregate_candidate): New function.
1531 (rs6000_discover_homogeneous_aggregate): Likewise.
1532 (rs6000_function_arg_boundary): Handle homogeneous aggregates.
1533 (rs6000_function_arg_advance_1): Likewise.
1534 (rs6000_function_arg): Likewise.
1535 (rs6000_arg_partial_bytes): Likewise.
1536 (rs6000_psave_function_arg): Handle BLKmode arguments.
1538 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1539 Michael Gschwind <mkg@us.ibm.com>
1541 * config/rs6000/rs6000.h (AGGR_ARG_NUM_REG): Define.
1542 * config/rs6000/rs6000.c (rs6000_aggregate_candidate): New function.
1543 (rs6000_discover_homogeneous_aggregate): Likewise.
1544 (rs6000_function_arg_boundary): Handle homogeneous aggregates.
1545 (rs6000_function_arg_advance_1): Likewise.
1546 (rs6000_function_arg): Likewise.
1547 (rs6000_arg_partial_bytes): Likewise.
1548 (rs6000_psave_function_arg): Handle BLKmode arguments.
1550 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1552 * config/rs6000/rs6000.c (machine_function): New member
1554 (rs6000_emit_prologue): Set r2_setup_needed if necessary.
1555 (rs6000_output_mi_thunk): Set r2_setup_needed.
1556 (rs6000_output_function_prologue): Output global entry point
1557 prologue and local entry point marker if needed for ABI_ELFv2.
1558 Output -mprofile-kernel code here.
1559 (output_function_profiler): Do not output -mprofile-kernel
1560 code here; moved to rs6000_output_function_prologue.
1561 (rs6000_file_start): Output ".abiversion 2" for ABI_ELFv2.
1563 (rs6000_emit_move): Do not handle dot symbols for ABI_ELFv2.
1564 (rs6000_output_function_entry): Likewise.
1565 (rs6000_assemble_integer): Likewise.
1566 (rs6000_elf_encode_section_info): Likewise.
1567 (rs6000_elf_declare_function_name): Do not create dot symbols
1568 or .opd section for ABI_ELFv2.
1570 (rs6000_trampoline_size): Update for ABI_ELFv2 trampolines.
1571 (rs6000_trampoline_init): Likewise.
1572 (rs6000_elf_file_end): Call file_end_indicate_exec_stack
1575 (rs6000_call_aix): Handle ELFv2 indirect calls. Do not check
1576 for function descriptors in ABI_ELFv2.
1578 * config/rs6000/rs6000.md ("*call_indirect_aix<mode>"): Support
1579 on ABI_AIX only, not ABI_ELFv2.
1580 ("*call_value_indirect_aix<mode>"): Likewise.
1581 ("*call_indirect_elfv2<mode>"): New pattern.
1582 ("*call_value_indirect_elfv2<mode>"): Likewise.
1584 * config/rs6000/predicates.md ("symbol_ref_operand"): Do not
1585 check for function descriptors in ABI_ELFv2.
1586 ("current_file_function_operand"): Likewise.
1588 * config/rs6000/ppc-asm.h [__powerpc64__ && _CALL_ELF == 2]:
1590 (FUNC_NAME): Define ELFv2 variant.
1591 (JUMP_TARGET): Likewise.
1592 (FUNC_START): Likewise.
1593 (HIDDEN_FUNC): Likewise.
1594 (FUNC_END): Likeiwse.
1596 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1598 * config.gcc [powerpc*-*-* | rs6000-*-*]: Support --with-abi=elfv1
1599 and --with-abi=elfv2.
1600 * config/rs6000/option-defaults.h (OPTION_DEFAULT_SPECS): Add "abi".
1601 * config/rs6000/rs6000.opt (mabi=elfv1): New option.
1602 (mabi=elfv2): Likewise.
1603 * config/rs6000/rs6000-opts.h (enum rs6000_abi): Add ABI_ELFv2.
1604 * config/rs6000/linux64.h (DEFAULT_ABI): Do not hard-code to AIX_ABI
1606 (ELFv2_ABI_CHECK): New macro.
1607 (SUBSUBTARGET_OVERRIDE_OPTIONS): Use it to decide whether to set
1608 rs6000_current_abi to ABI_AIX or ABI_ELFv2.
1609 (GLIBC_DYNAMIC_LINKER64): Support ELFv2 ld.so version.
1610 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Predefine
1611 _CALL_ELF and __STRUCT_PARM_ALIGN__ if appropriate.
1613 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Handle ABI_ELFv2.
1614 (debug_stack_info): Likewise.
1615 (rs6000_file_start): Treat ABI_ELFv2 the same as ABI_AIX.
1616 (rs6000_legitimize_tls_address): Likewise.
1617 (rs6000_conditional_register_usage): Likewise.
1618 (rs6000_emit_move): Likewise.
1619 (init_cumulative_args): Likewise.
1620 (rs6000_function_arg_advance_1): Likewise.
1621 (rs6000_function_arg): Likewise.
1622 (rs6000_arg_partial_bytes): Likewise.
1623 (rs6000_output_function_entry): Likewise.
1624 (rs6000_assemble_integer): Likewise.
1625 (rs6000_savres_strategy): Likewise.
1626 (rs6000_stack_info): Likewise.
1627 (rs6000_function_ok_for_sibcall): Likewise.
1628 (rs6000_emit_load_toc_table): Likewise.
1629 (rs6000_savres_routine_name): Likewise.
1630 (ptr_regno_for_savres): Likewise.
1631 (rs6000_emit_prologue): Likewise.
1632 (rs6000_emit_epilogue): Likewise.
1633 (rs6000_output_function_epilogue): Likewise.
1634 (output_profile_hook): Likewise.
1635 (output_function_profiler): Likewise.
1636 (rs6000_trampoline_size): Likewise.
1637 (rs6000_trampoline_init): Likewise.
1638 (rs6000_elf_output_toc_section_asm_op): Likewise.
1639 (rs6000_elf_encode_section_info): Likewise.
1640 (rs6000_elf_reloc_rw_mask): Likewise.
1641 (rs6000_elf_declare_function_name): Likewise.
1642 (rs6000_function_arg_boundary): Treat ABI_ELFv2 the same as ABI_AIX,
1643 except that rs6000_compat_align_parm is always assumed false.
1644 (rs6000_gimplify_va_arg): Likewise.
1645 (rs6000_call_aix): Update comment.
1646 (rs6000_sibcall_aix): Likewise.
1647 * config/rs6000/rs6000.md ("tls_gd_aix<TLSmode:tls_abi_suffix>"):
1648 Treat ABI_ELFv2 the same as ABI_AIX.
1649 ("*tls_gd_call_aix<TLSmode:tls_abi_suffix>"): Likewise.
1650 ("tls_ld_aix<TLSmode:tls_abi_suffix>"): Likewise.
1651 ("*tls_ld_call_aix<TLSmode:tls_abi_suffix>"): Likewise.
1652 ("load_toc_aix_si"): Likewise.
1653 ("load_toc_aix_di"): Likewise.
1655 ("call_value"): Likewise.
1656 ("*call_local_aix<mode>"): Likewise.
1657 ("*call_value_local_aix<mode>"): Likewise.
1658 ("*call_nonlocal_aix<mode>"): Likewise.
1659 ("*call_value_nonlocal_aix<mode>"): Likewise.
1660 ("*call_indirect_aix<mode>"): Likewise.
1661 ("*call_value_indirect_aix<mode>"): Likewise.
1662 ("sibcall"): Likewise.
1663 ("sibcall_value"): Likewise.
1664 ("*sibcall_aix<mode>"): Likewise.
1665 ("*sibcall_value_aix<mode>"): Likewise.
1666 * config/rs6000/predicates.md ("symbol_ref_operand"): Likewise.
1667 ("current_file_function_operand"): Likewise.
1669 Backport from mainline r204807:
1671 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1673 * config/rs6000/rs6000.c (rs6000_arg_partial_bytes): Simplify logic
1674 by making use of the fact that for vector / floating point arguments
1675 passed both in VRs/FPRs and in the fixed parameter area, the partial
1676 bytes mechanism is in fact not used.
1678 Backport from mainline r204806:
1680 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1682 * config/rs6000/rs6000.c (rs6000_psave_function_arg): New function.
1683 (rs6000_finish_function_arg): Likewise.
1684 (rs6000_function_arg): Use rs6000_psave_function_arg and
1685 rs6000_finish_function_arg to handle both vector and floating
1686 point arguments that are also passed in GPRs / the stack.
1688 Backport from mainline r204805:
1690 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1692 * config/rs6000/rs6000.c (USE_FP_FOR_ARG_P): Remove TYPE argument.
1693 (USE_ALTIVEC_FOR_ARG_P): Likewise.
1694 (rs6000_darwin64_record_arg_advance_recurse): Update uses.
1695 (rs6000_function_arg_advance_1):Likewise.
1696 (rs6000_darwin64_record_arg_recurse): Likewise.
1697 (rs6000_function_arg): Likewise.
1698 (rs6000_arg_partial_bytes): Likewise.
1700 Backport from mainline r204804:
1702 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1704 * config/rs6000/rs6000.c (rs6000_option_override_internal): Replace
1705 "DEFAULT_ABI != ABI_AIX" test by testing for ABI_V4 or ABI_DARWIN.
1706 (rs6000_savres_strategy): Likewise.
1707 (rs6000_return_addr): Likewise.
1708 (rs6000_emit_load_toc_table): Replace "DEFAULT_ABI != ABI_AIX" by
1709 testing for ABI_V4 (since ABI_DARWIN is impossible here).
1710 (rs6000_emit_prologue): Likewise.
1711 (legitimate_lo_sum_address_p): Simplify DEFAULT_ABI test.
1712 (rs6000_elf_declare_function_name): Remove duplicated test.
1713 * config/rs6000/rs6000.md ("load_toc_v4_PIC_1"): Explicitly test
1714 for ABI_V4 (instead of "DEFAULT_ABI != ABI_AIX" test).
1715 ("load_toc_v4_PIC_1_normal"): Likewise.
1716 ("load_toc_v4_PIC_1_476"): Likewise.
1717 ("load_toc_v4_PIC_1b"): Likewise.
1718 ("load_toc_v4_PIC_1b_normal"): Likewise.
1719 ("load_toc_v4_PIC_1b_476"): Likewise.
1720 ("load_toc_v4_PIC_2"): Likewise.
1721 ("load_toc_v4_PIC_3b"): Likewise.
1722 ("load_toc_v4_PIC_3c"): Likewise.
1723 * config/rs6000/rs6000.h (RS6000_REG_SAVE): Simplify DEFAULT_ABI test.
1724 (RS6000_SAVE_AREA): Likewise.
1725 (FP_ARG_MAX_REG): Likewise.
1726 (RETURN_ADDRESS_OFFSET): Likewise.
1727 * config/rs6000/sysv.h (TARGET_TOC): Test for ABI_V4 instead
1729 (SUBTARGET_OVERRIDE_OPTIONS): Likewise.
1730 (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
1732 Backport from mainline r204803:
1734 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1736 * config/rs6000/rs6000.c (rs6000_call_indirect_aix): Rename to ...
1737 (rs6000_call_aix): ... this. Handle both direct and indirect calls.
1738 Create call insn directly instead of via various gen_... routines.
1739 Mention special registers used by the call in CALL_INSN_FUNCTION_USAGE.
1740 (rs6000_sibcall_aix): New function.
1741 * config/rs6000/rs6000.md (TOC_SAVE_OFFSET_32BIT): Remove.
1742 (TOC_SAVE_OFFSET_64BIT): Likewise.
1743 (AIX_FUNC_DESC_TOC_32BIT): Likewise.
1744 (AIX_FUNC_DESC_TOC_64BIT): Likewise.
1745 (AIX_FUNC_DESC_SC_32BIT): Likewise.
1746 (AIX_FUNC_DESC_SC_64BIT): Likewise.
1747 ("call" expander): Call rs6000_call_aix.
1748 ("call_value" expander): Likewise.
1749 ("call_indirect_aix<ptrsize>"): Replace this pattern ...
1750 ("call_indirect_aix<ptrsize>_nor11"): ... and this pattern ...
1751 ("*call_indirect_aix<mode>"): ... by this insn pattern.
1752 ("call_value_indirect_aix<ptrsize>"): Replace this pattern ...
1753 ("call_value_indirect_aix<ptrsize>_nor11"): ... and this pattern ...
1754 ("*call_value_indirect_aix<mode>"): ... by this insn pattern.
1755 ("*call_nonlocal_aix32", "*call_nonlocal_aix64"): Replace by ...
1756 ("*call_nonlocal_aix<mode>"): ... this pattern.
1757 ("*call_value_nonlocal_aix32", "*call_value_nonlocal_aix64"): Replace
1758 ("*call_value_nonlocal_aix<mode>"): ... by this pattern.
1759 ("*call_local_aix<mode>"): New insn pattern.
1760 ("*call_value_local_aix<mode>"): Likewise.
1761 ("sibcall" expander): Call rs6000_sibcall_aix.
1762 ("sibcall_value" expander): Likewise. Move earlier in file.
1763 ("*sibcall_nonlocal_aix<mode>"): Replace by ...
1764 ("*sibcall_aix<mode>"): ... this pattern.
1765 ("*sibcall_value_nonlocal_aix<mode>"): Replace by ...
1766 ("*sibcall_value_aix<mode>"): ... this pattern.
1767 * config/rs6000/rs6000-protos.h (rs6000_call_indirect_aix): Remove.
1768 (rs6000_call_aix): Add prototype.
1769 (rs6000_sibcall_aix): Likewise.
1771 Backport from mainline r204799:
1773 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1775 * config/rs6000/rs6000.c (rs6000_emit_prologue): Do not place a
1776 RTX_FRAME_RELATED_P marker on the UNSPEC_MOVESI_FROM_CR insn.
1777 Instead, add USEs of all modified call-saved CR fields to the
1778 insn storing the result to the stack slot, and provide an
1779 appropriate REG_FRAME_RELATED_EXPR for that insn.
1780 * config/rs6000/rs6000.md ("*crsave"): New insn pattern.
1781 * config/rs6000/predicates.md ("crsave_operation"): New predicate.
1783 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1785 ELFv2 ABI Call Support
1786 Backport from mainline r204798:
1788 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1789 Alan Modra <amodra@gmail.com>
1791 * function.c (assign_parms): Use all.reg_parm_stack_space instead
1792 of re-evaluating REG_PARM_STACK_SPACE target macro.
1793 (locate_and_pad_parm): New parameter REG_PARM_STACK_SPACE. Use it
1794 instead of evaluating target macro REG_PARM_STACK_SPACE every time.
1795 (assign_parm_find_entry_rtl): Update call.
1796 * calls.c (initialize_argument_information): Update call.
1797 (emit_library_call_value_1): Likewise.
1798 * expr.h (locate_and_pad_parm): Update prototype.
1800 Backport from mainline r204797:
1802 2013-11-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1804 * calls.c (store_unaligned_arguments_into_pseudos): Skip PARALLEL
1807 Backport from mainline r197003:
1809 2013-03-23 Eric Botcazou <ebotcazou@adacore.com>
1811 * calls.c (expand_call): Add missing guard to code handling return
1812 of non-BLKmode structures in MSB.
1813 * function.c (expand_function_end): Likewise.
1815 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1817 Backport from mainline r201750.
1818 Note: Default setting of -mcompat-align-parm inverted!
1820 2013-08-14 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1823 * doc/invoke.texi: Add documentation of mcompat-align-parm
1825 * config/rs6000/rs6000.opt: Add mcompat-align-parm option.
1826 * config/rs6000/rs6000.c (rs6000_function_arg_boundary): For AIX
1827 and Linux, correct BLKmode alignment when 128-bit alignment is
1828 required and compatibility flag is not set.
1829 (rs6000_gimplify_va_arg): For AIX and Linux, honor specified
1830 alignment for zero-size arguments when compatibility flag is not
1833 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1835 Little Endian Vector Support
1836 Backport from mainline r205333
1837 2013-11-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1839 * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Correct
1842 Backport from mainline r205241
1843 2013-11-21 Bill Schmidt <wschmidt@vnet.ibm.com>
1845 * config/rs6000/vector.md (vec_pack_trunc_v2df): Revert previous
1846 little endian change.
1847 (vec_pack_sfix_trunc_v2df): Likewise.
1848 (vec_pack_ufix_trunc_v2df): Likewise.
1849 * config/rs6000/rs6000.c (rs6000_expand_interleave): Correct
1850 double checking of endianness.
1852 Backport from mainline r205146
1853 2013-11-20 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1855 * config/rs6000/vsx.md (vsx_set_<mode>): Adjust for little endian.
1856 (vsx_extract_<mode>): Likewise.
1857 (*vsx_extract_<mode>_one_le): New LE variant on
1858 *vsx_extract_<mode>_zero.
1859 (vsx_extract_v4sf): Adjust for little endian.
1861 Backport from mainline r205080
1862 2013-11-19 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1864 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Adjust
1865 V16QI vector splat case for little endian.
1867 Backport from mainline r205045:
1869 2013-11-19 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1871 * config/rs6000/vector.md ("mov<mode>"): Do not call
1872 rs6000_emit_le_vsx_move to move into or out of GPRs.
1873 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Assert
1874 source and destination are not GPR hard regs.
1876 Backport from mainline r204920
1877 2011-11-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1879 * config/rs6000/rs6000.c (rs6000_frame_related): Add split_reg
1880 parameter and use it in REG_FRAME_RELATED_EXPR note.
1881 (emit_frame_save): Call rs6000_frame_related with extra NULL_RTX
1883 (rs6000_emit_prologue): Likewise, but for little endian VSX
1884 stores, pass the source register of the store instead.
1886 Backport from mainline r204862
1887 2013-11-15 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1889 * config/rs6000/altivec.md (UNSPEC_VPERM_X, UNSPEC_VPERM_UNS_X):
1891 (altivec_vperm_<mode>): Revert earlier little endian change.
1892 (*altivec_vperm_<mode>_internal): Remove.
1893 (altivec_vperm_<mode>_uns): Revert earlier little endian change.
1894 (*altivec_vperm_<mode>_uns_internal): Remove.
1895 * config/rs6000/vector.md (vec_realign_load_<mode>): Revise
1898 Backport from mainline r204441
1899 2013-11-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1901 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1902 Remove restriction against use of VSX instructions when generating
1903 code for little endian mode.
1905 Backport from mainline r204440
1906 2013-11-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1908 * config/rs6000/altivec.md (mulv4si3): Ensure we generate vmulouh
1909 for both big and little endian.
1910 (mulv8hi3): Swap input operands for merge high and merge low
1911 instructions for little endian.
1913 Backport from mainline r204439
1914 2013-11-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1916 * config/rs6000/altivec.md (vec_widen_umult_even_v16qi): Change
1917 define_insn to define_expand that uses even patterns for big
1918 endian and odd patterns for little endian.
1919 (vec_widen_smult_even_v16qi): Likewise.
1920 (vec_widen_umult_even_v8hi): Likewise.
1921 (vec_widen_smult_even_v8hi): Likewise.
1922 (vec_widen_umult_odd_v16qi): Likewise.
1923 (vec_widen_smult_odd_v16qi): Likewise.
1924 (vec_widen_umult_odd_v8hi): Likewise.
1925 (vec_widen_smult_odd_v8hi): Likewise.
1926 (altivec_vmuleub): New define_insn.
1927 (altivec_vmuloub): Likewise.
1928 (altivec_vmulesb): Likewise.
1929 (altivec_vmulosb): Likewise.
1930 (altivec_vmuleuh): Likewise.
1931 (altivec_vmulouh): Likewise.
1932 (altivec_vmulesh): Likewise.
1933 (altivec_vmulosh): Likewise.
1935 Backport from mainline r204395
1936 2013-11-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1938 * config/rs6000/vector.md (vec_pack_sfix_trunc_v2df): Adjust for
1940 (vec_pack_ufix_trunc_v2df): Likewise.
1942 Backport from mainline r204363
1943 2013-11-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1945 * config/rs6000/altivec.md (vec_widen_umult_hi_v16qi): Swap
1946 arguments to merge instruction for little endian.
1947 (vec_widen_umult_lo_v16qi): Likewise.
1948 (vec_widen_smult_hi_v16qi): Likewise.
1949 (vec_widen_smult_lo_v16qi): Likewise.
1950 (vec_widen_umult_hi_v8hi): Likewise.
1951 (vec_widen_umult_lo_v8hi): Likewise.
1952 (vec_widen_smult_hi_v8hi): Likewise.
1953 (vec_widen_smult_lo_v8hi): Likewise.
1955 Backport from mainline r204350
1956 2013-11-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1958 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode> for VSX_D):
1959 Replace the define_insn_and_split with a define_insn and two
1960 define_splits, with the split after reload re-permuting the source
1961 register to its original value.
1962 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
1963 (*vsx_le_perm_store_v8hi): Likewise.
1964 (*vsx_le_perm_store_v16qi): Likewise.
1966 Backport from mainline r204321
1967 2013-11-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1969 * config/rs6000/vector.md (vec_pack_trunc_v2df): Adjust for
1972 Backport from mainline r204321
1973 2013-11-02 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
1975 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Adjust for
1978 Backport from mainline r203980
1979 2013-10-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1981 * config/rs6000/altivec.md (mulv8hi3): Adjust for little endian.
1983 Backport from mainline r203930
1984 2013-10-22 Bill Schmidt <wschmidt@vnet.ibm.com>
1986 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Reverse
1987 meaning of merge-high and merge-low masks for little endian; avoid
1988 use of vector-pack masks for little endian for mismatched modes.
1990 Backport from mainline r203877
1991 2013-10-20 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1993 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Adjust for
1995 (vec_unpacku_hi_v8hi): Likewise.
1996 (vec_unpacku_lo_v16qi): Likewise.
1997 (vec_unpacku_lo_v8hi): Likewise.
1999 Backport from mainline r203863
2000 2013-10-19 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2002 * config/rs6000/rs6000.c (vspltis_constant): Make sure we check
2003 all elements for both endian flavors.
2005 Backport from mainline r203714
2006 2013-10-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2008 * gcc/config/rs6000/vector.md (vec_unpacks_hi_v4sf): Correct for
2010 (vec_unpacks_lo_v4sf): Likewise.
2011 (vec_unpacks_float_hi_v4si): Likewise.
2012 (vec_unpacks_float_lo_v4si): Likewise.
2013 (vec_unpacku_float_hi_v4si): Likewise.
2014 (vec_unpacku_float_lo_v4si): Likewise.
2016 Backport from mainline r203713
2017 2013-10-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2019 * config/rs6000/vsx.md (vsx_concat_<mode>): Adjust output for LE.
2020 (vsx_concat_v2sf): Likewise.
2022 Backport from mainline r203458
2023 2013-10-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2025 * config/rs6000/vsx.md (*vsx_le_perm_load_v2di): Generalize to
2026 handle vector float as well.
2027 (*vsx_le_perm_load_v4si): Likewise.
2028 (*vsx_le_perm_store_v2di): Likewise.
2029 (*vsx_le_perm_store_v4si): Likewise.
2031 Backport from mainline r203457
2032 2013-10-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2034 * config/rs6000/vector.md (vec_realign_load<mode>): Generate vperm
2035 directly to circumvent subtract from splat{31} workaround.
2036 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_le): New
2038 * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): New.
2039 * config/rs6000/altivec.md (define_c_enum "unspec"): Add
2040 UNSPEC_VPERM_X and UNSPEC_VPERM_UNS_X.
2041 (altivec_vperm_<mode>): Convert to define_insn_and_split to
2042 separate big and little endian logic.
2043 (*altivec_vperm_<mode>_internal): New define_insn.
2044 (altivec_vperm_<mode>_uns): Convert to define_insn_and_split to
2045 separate big and little endian logic.
2046 (*altivec_vperm_<mode>_uns_internal): New define_insn.
2047 (vec_permv16qi): Add little endian logic.
2049 Backport from mainline r203247
2050 2013-10-07 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2052 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const_le): New.
2053 (altivec_expand_vec_perm_const): Call it.
2055 Backport from mainline r203246
2056 2013-10-07 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2058 * config/rs6000/vector.md (mov<mode>): Emit permuted move
2059 sequences for LE VSX loads and stores at expand time.
2060 * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_move): New
2062 * config/rs6000/rs6000.c (rs6000_const_vec): New.
2063 (rs6000_gen_le_vsx_permute): New.
2064 (rs6000_gen_le_vsx_load): New.
2065 (rs6000_gen_le_vsx_store): New.
2066 (rs6000_gen_le_vsx_move): New.
2067 * config/rs6000/vsx.md (*vsx_le_perm_load_v2di): New.
2068 (*vsx_le_perm_load_v4si): New.
2069 (*vsx_le_perm_load_v8hi): New.
2070 (*vsx_le_perm_load_v16qi): New.
2071 (*vsx_le_perm_store_v2di): New.
2072 (*vsx_le_perm_store_v4si): New.
2073 (*vsx_le_perm_store_v8hi): New.
2074 (*vsx_le_perm_store_v16qi): New.
2075 (*vsx_xxpermdi2_le_<mode>): New.
2076 (*vsx_xxpermdi4_le_<mode>): New.
2077 (*vsx_xxpermdi8_le_V8HI): New.
2078 (*vsx_xxpermdi16_le_V16QI): New.
2079 (*vsx_lxvd2x2_le_<mode>): New.
2080 (*vsx_lxvd2x4_le_<mode>): New.
2081 (*vsx_lxvd2x8_le_V8HI): New.
2082 (*vsx_lxvd2x16_le_V16QI): New.
2083 (*vsx_stxvd2x2_le_<mode>): New.
2084 (*vsx_stxvd2x4_le_<mode>): New.
2085 (*vsx_stxvd2x8_le_V8HI): New.
2086 (*vsx_stxvd2x16_le_V16QI): New.
2088 Backport from mainline r201235
2089 2013-07-24 Bill Schmidt <wschmidt@linux.ibm.com>
2090 Anton Blanchard <anton@au1.ibm.com>
2092 * config/rs6000/altivec.md (altivec_vpkpx): Handle little endian.
2093 (altivec_vpks<VI_char>ss): Likewise.
2094 (altivec_vpks<VI_char>us): Likewise.
2095 (altivec_vpku<VI_char>us): Likewise.
2096 (altivec_vpku<VI_char>um): Likewise.
2098 Backport from mainline r201208
2099 2013-07-24 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
2100 Anton Blanchard <anton@au1.ibm.com>
2102 * config/rs6000/vector.md (vec_realign_load_<mode>): Reorder input
2103 operands to vperm for little endian.
2104 * config/rs6000/rs6000.c (rs6000_expand_builtin): Use lvsr instead
2105 of lvsl to create the control mask for a vperm for little endian.
2107 Backport from mainline r201195
2108 2013-07-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2109 Anton Blanchard <anton@au1.ibm.com>
2111 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Reverse
2112 two operands for little-endian.
2114 Backport from mainline r201193
2115 2013-07-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2116 Anton Blanchard <anton@au1.ibm.com>
2118 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Correct
2119 selection of field for vector splat in little endian mode.
2121 Backport from mainline r201149
2122 2013-07-22 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
2123 Anton Blanchard <anton@au1.ibm.com>
2125 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Fix
2126 endianness when selecting field to splat.
2128 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2130 Backport from mainline r205123:
2132 2013-11-20 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
2134 * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Do not
2135 allow subregs of TDmode in FPRs of smaller size in little-endian.
2136 (rs6000_split_multireg_move): When splitting an access to TDmode
2137 in FPRs, do not use simplify_gen_subreg.
2139 Backport from mainline r204927:
2141 2013-11-17 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
2143 * config/rs6000/rs6000.c (rs6000_emit_move): Use low word of
2144 sdmode_stack_slot also in little-endian mode.
2146 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2149 Backport from mainline
2150 2013-12-03 Peter Bergner <bergner@vnet.ibm.com>
2152 * config/rs6000/htmintrin.h (_TEXASR_INSTRUCTION_FETCH_CONFLICT): Fix
2154 (_TEXASRU_INSTRUCTION_FETCH_CONFLICT): Likewise.
2156 Backport from mainline r205233.
2157 2013-11-21 Peter Bergner <bergner@vnet.ibm.com>
2159 * doc/extend.texi: Document htm builtins.
2161 Backport from mainline
2162 2013-07-17 Iain Sandoe <iain@codesourcery.com>
2164 * config/rs6000/darwin.h (REGISTER_NAMES): Add HTM registers.
2166 Backport from mainline
2167 2013-07-16 Peter Bergner <bergner@vnet.ibm.com>
2169 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
2170 enable extra ISA flags with TARGET_HTM.
2172 2013-07-16 Jakub Jelinek <jakub@redhat.com>
2173 Peter Bergner <bergner@vnet.ibm.com>
2175 * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTERS): Mention HTM
2176 registers in the comment.
2177 (DWARF_FRAME_REGISTERS): Subtract also the 3 HTM registers.
2178 (DWARF_REG_TO_UNWIND_COLUMN): Use DWARF_FRAME_REGISTERS
2179 rather than FIRST_PSEUDO_REGISTERS.
2181 * config.gcc (powerpc*-*-*): Install htmintrin.h and htmxlintrin.h.
2182 * config/rs6000/t-rs6000 (MD_INCLUDES): Add htm.md.
2183 * config/rs6000/rs6000.opt: Add -mhtm option.
2184 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Add OPTION_MASK_HTM.
2185 (ISA_2_7_MASKS_SERVER): Add OPTION_MASK_HTM.
2186 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
2187 __HTM__ if the HTM instructions are available.
2188 * config/rs6000/predicates.md (u3bit_cint_operand, u10bit_cint_operand)
2189 (htm_spr_reg_operand): New define_predicates.
2190 * config/rs6000/rs6000.md (define_attr "type"): Add htm.
2191 (TFHAR_REGNO, TFIAR_REGNO, TEXASR_REGNO): New define_constants.
2193 * config/rs6000/rs6000-builtin.def (BU_HTM_0, BU_HTM_1, BU_HTM_2)
2194 (BU_HTM_3, BU_HTM_SPR0, BU_HTM_SPR1): Add support macros for defining
2195 HTM builtin functions.
2196 * config/rs6000/rs6000.c (RS6000_BUILTIN_H): New macro.
2197 (rs6000_reg_names, alt_reg_names): Add HTM SPR register names.
2198 (rs6000_init_hard_regno_mode_ok): Add support for HTM instructions.
2199 (rs6000_builtin_mask_calculate): Likewise.
2200 (rs6000_option_override_internal): Likewise.
2201 (bdesc_htm): Add new HTM builtin support.
2202 (htm_spr_num): New function.
2203 (htm_spr_regno): Likewise.
2204 (rs6000_htm_spr_icode): Likewise.
2205 (htm_expand_builtin): Likewise.
2206 (htm_init_builtins): Likewise.
2207 (rs6000_expand_builtin): Add support for HTM builtin functions.
2208 (rs6000_init_builtins): Likewise.
2209 (rs6000_invalid_builtin, rs6000_opt_mask): Add support for -mhtm option.
2210 * config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mhtm.
2211 (TARGET_HTM, MASK_HTM): Define macros.
2212 (FIRST_PSEUDO_REGISTER): Adjust for new HTM SPR registers.
2213 (FIXED_REGISTERS): Likewise.
2214 (CALL_USED_REGISTERS): Likewise.
2215 (CALL_REALLY_USED_REGISTERS): Likewise.
2216 (REG_ALLOC_ORDER): Likewise.
2217 (enum reg_class): Likewise.
2218 (REG_CLASS_NAMES): Likewise.
2219 (REG_CLASS_CONTENTS): Likewise.
2220 (REGISTER_NAMES): Likewise.
2221 (ADDITIONAL_REGISTER_NAMES): Likewise.
2222 (RS6000_BTC_SPR, RS6000_BTC_VOID, RS6000_BTC_32BIT, RS6000_BTC_64BIT)
2223 (RS6000_BTC_MISC_MASK, RS6000_BTM_HTM): New macros.
2224 (RS6000_BTM_COMMON): Add RS6000_BTM_HTM.
2225 * config/rs6000/htm.md: New file.
2226 * config/rs6000/htmintrin.h: New file.
2227 * config/rs6000/htmxlintrin.h: New file.
2229 2014-04-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2233 2013-11-23 Alan Modra <amodra@gmail.com>
2234 * config/rs6000/vsx.md (fusion peepholes): Disable when !TARGET_VSX.
2236 Backport from mainline
2237 2013-11-12 Michael Meissner <meissner@linux.vnet.ibm.com>
2240 * config/rs6000/rs6000.md (movdi_internal32): Eliminate
2241 constraints that would allow DImode into the traditional Altivec
2242 registers, but cause undesirable code generation when loading 0 as
2244 (movdi_internal64): Likewise.
2245 (cmp<mode>_fpr): Do not use %x for CR register output.
2246 (extendsfdf2_fpr): Fix constraints when -mallow-upper-df and
2247 -mallow-upper-sf debug switches are used.
2249 Backport from mainline
2250 2013-10-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2252 * config/rs6000/rs6000.c (enum rs6000_reload_reg_type): Add new
2253 fields to the reg_addr array that describes the valid addressing
2254 mode for any register, general purpose registers, floating point
2255 registers, and Altivec registers.
2256 (FIRST_RELOAD_REG_CLASS): Likewise.
2257 (LAST_RELOAD_REG_CLASS): Likewise.
2258 (struct reload_reg_map_type): Likewise.
2259 (reload_reg_map_type): Likewise.
2260 (RELOAD_REG_VALID): Likewise.
2261 (RELOAD_REG_MULTIPLE): Likewise.
2262 (RELOAD_REG_INDEXED): Likewise.
2263 (RELOAD_REG_OFFSET): Likewise.
2264 (RELOAD_REG_PRE_INCDEC): Likewise.
2265 (RELOAD_REG_PRE_MODIFY): Likewise.
2266 (reg_addr): Likewise.
2267 (mode_supports_pre_incdec_p): New helper functions to say whether
2268 a given mode supports PRE_INC, PRE_DEC, and PRE_MODIFY.
2269 (mode_supports_pre_modify_p): Likewise.
2270 (rs6000_debug_vector_unit): Rearrange the -mdebug=reg output to
2271 print the valid address mode bits for each mode.
2272 (rs6000_debug_print_mode): Likewise.
2273 (rs6000_debug_reg_global): Likewise.
2274 (rs6000_setup_reg_addr_masks): New function to set up the address
2275 mask bits for each type.
2276 (rs6000_init_hard_regno_mode_ok): Use memset to clear arrays.
2277 Call rs6000_setup_reg_addr_masks to set up the address mask bits.
2278 (rs6000_legitimate_address_p): Use mode_supports_pre_incdec_p and
2279 mode_supports_pre_modify_p to determine if PRE_INC, PRE_DEC, and
2280 PRE_MODIFY are supported.
2281 (rs6000_output_move_128bit): Change to use {src,dest}_vmx_p for altivec
2282 registers, instead of {src,dest}_av_p.
2283 (rs6000_print_options_internal): Tweak the debug output slightly.
2285 Backport from mainline
2286 2013-10-03 Michael Meissner <meissner@linux.vnet.ibm.com>
2288 * config/rs6000/rs6000-builtin.def (XSRDPIM): Use floatdf2,
2289 ceildf2, btruncdf2, instead of vsx_* name.
2291 * config/rs6000/vsx.md (vsx_add<mode>3): Change arithmetic
2292 iterators to only do V2DF and V4SF here. Move the DF code to
2293 rs6000.md where it is combined with SF mode. Replace <VSv> with
2294 just 'v' since only vector operations are handled with these insns
2295 after moving the DF support to rs6000.md.
2296 (vsx_sub<mode>3): Likewise.
2297 (vsx_mul<mode>3): Likewise.
2298 (vsx_div<mode>3): Likewise.
2299 (vsx_fre<mode>2): Likewise.
2300 (vsx_neg<mode>2): Likewise.
2301 (vsx_abs<mode>2): Likewise.
2302 (vsx_nabs<mode>2): Likewise.
2303 (vsx_smax<mode>3): Likewise.
2304 (vsx_smin<mode>3): Likewise.
2305 (vsx_sqrt<mode>2): Likewise.
2306 (vsx_rsqrte<mode>2): Likewise.
2307 (vsx_fms<mode>4): Likewise.
2308 (vsx_nfma<mode>4): Likewise.
2309 (vsx_copysign<mode>3): Likewise.
2310 (vsx_btrunc<mode>2): Likewise.
2311 (vsx_floor<mode>2): Likewise.
2312 (vsx_ceil<mode>2): Likewise.
2313 (vsx_smaxsf3): Delete scalar ops that were moved to rs6000.md.
2314 (vsx_sminsf3): Likewise.
2315 (vsx_fmadf4): Likewise.
2316 (vsx_fmsdf4): Likewise.
2317 (vsx_nfmadf4): Likewise.
2318 (vsx_nfmsdf4): Likewise.
2319 (vsx_cmpdf_internal1): Likewise.
2321 * config/rs6000/rs6000.h (TARGET_SF_SPE): Define macros to make it
2322 simpler to select whether a target has SPE or traditional floating
2323 point support in iterators.
2324 (TARGET_DF_SPE): Likewise.
2325 (TARGET_SF_FPR): Likewise.
2326 (TARGET_DF_FPR): Likewise.
2327 (TARGET_SF_INSN): Macros to say whether floating point support
2328 exists for a given operation for expanders.
2329 (TARGET_DF_INSN): Likewise.
2331 * config/rs6000/rs6000.c (Ftrad): New mode attributes to allow
2332 combining of SF/DF mode operations, using both traditional and VSX
2340 (abs<mode>2): Combine SF/DF modes using traditional floating point
2341 instructions. Add support for using the upper DF registers with
2342 VSX support, and SF registers with power8-vector support. Update
2343 expanders for operations supported by both the SPE and traditional
2344 floating point units.
2345 (abs<mode>2_fpr): Likewise.
2346 (nabs<mode>2): Likewise.
2347 (nabs<mode>2_fpr): Likewise.
2348 (neg<mode>2): Likewise.
2349 (neg<mode>2_fpr): Likewise.
2350 (add<mode>3): Likewise.
2351 (add<mode>3_fpr): Likewise.
2352 (sub<mode>3): Likewise.
2353 (sub<mode>3_fpr): Likewise.
2354 (mul<mode>3): Likewise.
2355 (mul<mode>3_fpr): Likewise.
2356 (div<mode>3): Likewise.
2357 (div<mode>3_fpr): Likewise.
2358 (sqrt<mode>3): Likewise.
2359 (sqrt<mode>3_fpr): Likewise.
2360 (fre<Fs>): Likewise.
2361 (rsqrt<mode>2): Likewise.
2362 (cmp<mode>_fpr): Likewise.
2363 (smax<mode>3): Likewise.
2364 (smin<mode>3): Likewise.
2365 (smax<mode>3_vsx): Likewise.
2366 (smin<mode>3_vsx): Likewise.
2367 (negsf2): Delete SF operations that are merged with DF.
2374 (fmasf4_fpr): Likewise.
2375 (fmssf4_fpr): Likewise.
2376 (nfmasf4_fpr): Likewise.
2377 (nfmssf4_fpr): Likewise.
2378 (sqrtsf2): Likewise.
2379 (rsqrtsf_internal1): Likewise.
2380 (smaxsf3): Likewise.
2381 (sminsf3): Likewise.
2382 (cmpsf_internal1): Likewise.
2383 (copysign<mode>3_fcpsgn): Add VSX/power8-vector support.
2384 (negdf2): Delete DF operations that are merged with SF.
2386 (nabsdf2): Likewise.
2392 (rsqrtdf_internal1): Likewise.
2393 (fmadf4_fpr): Likewise.
2394 (fmsdf4_fpr): Likewise.
2395 (nfmadf4_fpr): Likewise.
2396 (nfmsdf4_fpr): Likewise.
2397 (sqrtdf2): Likewise.
2398 (smaxdf3): Likewise.
2399 (smindf3): Likewise.
2400 (cmpdf_internal1): Likewise.
2401 (lrint<mode>di2): Use TARGET_<MODE>_FPR macro.
2402 (btrunc<mode>2): Delete separate expander, and combine with the
2403 insn and add VSX instruction support. Use TARGET_<MODE>_FPR.
2404 (btrunc<mode>2_fpr): Likewise.
2405 (ceil<mode>2): Likewise.
2406 (ceil<mode>2_fpr): Likewise.
2407 (floor<mode>2): Likewise.
2408 (floor<mode>2_fpr): Likewise.
2409 (fma<mode>4_fpr): Combine SF and DF fused multiply/add support.
2410 Add support for using the upper registers with VSX and
2411 power8-vector. Move insns to be closer to the define_expands. On
2412 VSX systems, prefer the traditional form of FMA over the VSX
2413 version, since the traditional form allows the target not to
2414 overlap with the inputs.
2415 (fms<mode>4_fpr): Likewise.
2416 (nfma<mode>4_fpr): Likewise.
2417 (nfms<mode>4_fpr): Likewise.
2419 Backport from mainline
2420 2013-09-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2422 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Allow
2423 DFmode, DImode, and SFmode in the upper VSX registers based on the
2424 -mupper-regs-{df,sf} flags. Fix wu constraint to be ALTIVEC_REGS
2425 if -mpower8-vector. Combine -mvsx-timode handling with the rest
2426 of the VSX register handling.
2428 * config/rs6000/rs6000.md (f32_lv): Use %x0 for VSX regsters.
2430 (zero_extendsidi2_lfiwzx): Add support for loading into the
2431 Altivec registers with -mpower8-vector. Use wu/wv constraints to
2432 only do VSX memory options on Altivec registers.
2433 (extendsidi2_lfiwax): Likewise.
2434 (extendsfdf2_fpr): Likewise.
2435 (mov<mode>_hardfloat, SF/SD modes): Likewise.
2436 (mov<mode>_hardfloat32, DF/DD modes): Likewise.
2437 (mov<mode>_hardfloat64, DF/DD modes): Likewise.
2438 (movdi_internal64): Likewise.
2440 Backport from mainline
2441 2013-09-23 Michael Meissner <meissner@linux.vnet.ibm.com>
2443 * config/rs6000/rs6000.c (rs6000_vector_reload): Delete, combine
2444 reload helper function arrays into a single array reg_addr.
2445 (reload_fpr_gpr): Likewise.
2446 (reload_gpr_vsx): Likewise.
2447 (reload_vsx_gpr): Likewise.
2448 (struct rs6000_reg_addr): Likewise.
2449 (reg_addr): Likewise.
2450 (rs6000_debug_reg_global): Change rs6000_vector_reload,
2451 reload_fpr_gpr, reload_gpr_vsx, reload_vsx_gpr uses to reg_addr.
2452 (rs6000_init_hard_regno_mode_ok): Likewise.
2453 (rs6000_secondary_reload_direct_move): Likewise.
2454 (rs6000_secondary_reload): Likewise.
2456 * config/rs6000/rs6000.h (enum r6000_reg_class_enum): Add new
2457 constraints: wu, ww, and wy. Repurpose wv constraint added during
2458 power8 changes. Put wg constraint in alphabetical order.
2460 * config/rs6000/rs6000.opt (-mvsx-scalar-float): New debug switch
2461 for future work to add ISA 2.07 VSX single precision support.
2462 (-mvsx-scalar-double): Change default from -1 to 1, update
2463 documentation comment.
2464 (-mvsx-scalar-memory): Rename debug switch to -mupper-regs-df.
2465 (-mupper-regs-df): New debug switch to control whether DF values
2466 can go in the traditional Altivec registers.
2467 (-mupper-regs-sf): New debug switch to control whether SF values
2468 can go in the traditional Altivec registers.
2470 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print wu, ww,
2472 (rs6000_init_hard_regno_mode_ok): Use ssize_t instead of int for
2473 loop variables. Rename -mvsx-scalar-memory to -mupper-regs-df.
2474 Add new constraints, wu/ww/wy. Repurpose wv constraint.
2475 (rs6000_debug_legitimate_address_p): Print if we are running
2476 before, during, or after reload.
2477 (rs6000_secondary_reload): Add a comment.
2478 (rs6000_opt_masks): Add -mupper-regs-df, -mupper-regs-sf.
2480 * config/rs6000/constraints.md (wa constraint): Sort w<x>
2481 constraints. Update documentation string.
2482 (wd constraint): Likewise.
2483 (wf constraint): Likewise.
2484 (wg constraint): Likewise.
2485 (wn constraint): Likewise.
2486 (ws constraint): Likewise.
2487 (wt constraint): Likewise.
2488 (wx constraint): Likewise.
2489 (wz constraint): Likewise.
2490 (wu constraint): New constraint for ISA 2.07 SFmode scalar
2492 (ww constraint): Likewise.
2493 (wy constraint): Likewise.
2494 (wv constraint): Repurpose ISA 2.07 constraint that did not use in
2495 the previous submissions.
2496 * doc/md.texi (PowerPC and IBM RS6000): Likewise.
2498 Backport from mainline
2499 2013-10-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2502 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Only
2503 restrict TImode addresses to single indirect registers if both
2504 -mquad-memory and -mvsx-timode are used.
2505 (rs6000_output_move_128bit): Use quad_load_store_p to determine if
2506 we should emit load/store quad. Remove using %y for quad memory
2509 * config/rs6000/rs6000.md (mov<mode>_ppc64, TI/PTImode): Add
2510 constraints to allow load/store quad on machines where TImode is
2511 not allowed in VSX registers. Use 'n' instead of 'F' constraint
2512 for TImode to load integer constants.
2514 Backport from mainline
2515 2013-10-02 Michael Meissner <meissner@linux.vnet.ibm.com>
2518 * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Turn off
2519 setting -mvsx-timode by default until the underlying problem is
2521 (RS6000_CPU, power7 defaults): Likewise.
2524 2013-08-16 Michael Meissner <meissner@linux.vnet.ibm.com>
2527 * config/rs6000/predicates.md (fusion_gpr_mem_load): Allow the
2528 memory rtx to contain ZERO_EXTEND and SIGN_EXTEND.
2530 * config/rs6000/rs6000-protos.h (fusion_gpr_load_p): Pass operands
2531 array instead of each individual operand as a separate argument.
2532 (emit_fusion_gpr_load): Likewise.
2533 (expand_fusion_gpr_load): Add new function declaration.
2535 * config/rs6000/rs6000.c (fusion_gpr_load_p): Change the calling
2536 signature to have the operands passed as an array, instead of as
2537 separate arguments. Allow ZERO_EXTEND to be in the memory
2538 address, and also SIGN_EXTEND if -mpower8-fusion-sign. Do not
2539 depend on the register live/dead flags when peepholes are run.
2540 (expand_fusion_gpr_load): New function to be called from the
2541 peephole2 pass, to change the register that addis sets to be the
2543 (emit_fusion_gpr_load): Change the calling signature to have the
2544 operands passed as an array, instead of as separate arguments.
2545 Allow ZERO_EXTEND to be in the memory address, and also
2546 SIGN_EXTEND if -mpower8-fusion-sign.
2548 * config/rs6000/rs6000.md (UNSPEC_FUSION_GPR): Delete unused
2550 (power8 fusion peephole/peephole2): Rework the fusion peepholes to
2551 adjust the register addis loads up in the peephole2 pass. Do not
2552 depend on the register live/dead state when the peephole pass is
2556 2013-07-23 Michael Meissner <meissner@linux.vnet.ibm.com>
2558 * config/rs6000/vector.md (xor<mode>3): Move 128-bit boolean
2559 expanders to rs6000.md.
2560 (ior<mode>3): Likewise.
2561 (and<mode>3): Likewise.
2562 (one_cmpl<mode>2): Likewise.
2563 (nor<mode>3): Likewise.
2564 (andc<mode>3): Likewise.
2565 (eqv<mode>3): Likewise.
2566 (nand<mode>3): Likewise.
2567 (orc<mode>3): Likewise.
2569 * config/rs6000/rs6000-protos.h (rs6000_split_logical): New
2572 * config/rs6000/rs6000.c (rs6000_split_logical_inner): Add support
2573 to split multi-word logical operations.
2574 (rs6000_split_logical_di): Likewise.
2575 (rs6000_split_logical): Likewise.
2577 * config/rs6000/vsx.md (VSX_L2): Delete, no longer used.
2578 (vsx_and<mode>3_32bit): Move 128-bit logical insns to rs6000.md,
2579 and allow TImode operations in 32-bit.
2580 (vsx_and<mode>3_64bit): Likewise.
2581 (vsx_ior<mode>3_32bit): Likewise.
2582 (vsx_ior<mode>3_64bit): Likewise.
2583 (vsx_xor<mode>3_32bit): Likewise.
2584 (vsx_xor<mode>3_64bit): Likewise.
2585 (vsx_one_cmpl<mode>2_32bit): Likewise.
2586 (vsx_one_cmpl<mode>2_64bit): Likewise.
2587 (vsx_nor<mode>3_32bit): Likewise.
2588 (vsx_nor<mode>3_64bit): Likewise.
2589 (vsx_andc<mode>3_32bit): Likewise.
2590 (vsx_andc<mode>3_64bit): Likewise.
2591 (vsx_eqv<mode>3_32bit): Likewise.
2592 (vsx_eqv<mode>3_64bit): Likewise.
2593 (vsx_nand<mode>3_32bit): Likewise.
2594 (vsx_nand<mode>3_64bit): Likewise.
2595 (vsx_orc<mode>3_32bit): Likewise.
2596 (vsx_orc<mode>3_64bit): Likewise.
2598 * config/rs6000/rs6000.h (VLOGICAL_REGNO_P): Always allow vector
2599 logical types in GPRs.
2601 * config/rs6000/altivec.md (altivec_and<mode>3): Move 128-bit
2602 logical insns to rs6000.md, and allow TImode operations in
2604 (altivec_ior<mode>3): Likewise.
2605 (altivec_xor<mode>3): Likewise.
2606 (altivec_one_cmpl<mode>2): Likewise.
2607 (altivec_nor<mode>3): Likewise.
2608 (altivec_andc<mode>3): Likewise.
2610 * config/rs6000/rs6000.md (BOOL_128): New mode iterators and mode
2611 attributes for moving the 128-bit logical operations into
2613 (BOOL_REGS_OUTPUT): Likewise.
2614 (BOOL_REGS_OP1): Likewise.
2615 (BOOL_REGS_OP2): Likewise.
2616 (BOOL_REGS_UNARY): Likewise.
2617 (BOOL_REGS_AND_CR0): Likewise.
2618 (one_cmpl<mode>2): Add support for DI logical operations on
2619 32-bit, splitting the operations to 32-bit.
2623 (and<mode>3, 128-bit types): Rewrite 2013-06-06 logical operator
2624 changes to combine the 32/64-bit code, allow logical operations on
2625 TI mode in 32-bit, and to use similar match_operator patterns like
2626 scalar mode uses. Combine the Altivec and VSX code for logical
2627 operations, and move it here.
2628 (ior<mode>3, 128-bit types): Likewise.
2629 (xor<mode>3, 128-bit types): Likewise.
2630 (one_cmpl<mode>3, 128-bit types): Likewise.
2631 (nor<mode>3, 128-bit types): Likewise.
2632 (andc<mode>3, 128-bit types): Likewise.
2633 (eqv<mode>3, 128-bit types): Likewise.
2634 (nand<mode>3, 128-bit types): Likewise.
2635 (orc<mode>3, 128-bit types): Likewise.
2636 (and<mode>3_internal): Likewise.
2637 (bool<mode>3_internal): Likewise.
2638 (boolc<mode>3_internal1): Likewise.
2639 (boolc<mode>3_internal2): Likewise.
2640 (boolcc<mode>3_internal1): Likewise.
2641 (boolcc<mode>3_internal2): Likewise.
2642 (eqv<mode>3_internal1): Likewise.
2643 (eqv<mode>3_internal2): Likewise.
2644 (one_cmpl1<mode>3_internal): Likewise.
2646 Back port from mainline:
2647 2013-06-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2648 Pat Haugen <pthaugen@us.ibm.com>
2649 Peter Bergner <bergner@vnet.ibm.com>
2651 * lib/target-supports.exp (check_p8vector_hw_available) Add power8
2653 (check_effective_target_powerpc_p8vector_ok): Likewise.
2654 (is-effective-target): Likewise.
2655 (check_vect_support_and_set_flags): Likewise.
2657 Backport from mainline
2658 2013-07-31 Michael Meissner <meissner@linux.vnet.ibm.com>
2660 * config/rs6000/predicates.md (fusion_gpr_addis): New predicates
2661 to support power8 load fusion.
2662 (fusion_gpr_mem_load): Likewise.
2664 * config/rs6000/rs6000-modes.def (PTImode): Update a comment.
2666 * config/rs6000/rs6000-protos.h (fusion_gpr_load_p): New
2667 declarations for power8 load fusion.
2668 (emit_fusion_gpr_load): Likewise.
2670 * config/rs6000/rs6000.c (rs6000_option_override_internal): If
2671 tuning for power8, turn on fusion mode by default. Turn on sign
2672 extending fusion mode if normal fusion mode is on, and we are at
2674 (fusion_gpr_load_p): New function, return true if we can fuse an
2675 addis instruction with a dependent load to a GPR.
2676 (emit_fusion_gpr_load): Emit the instructions for power8 load
2679 * config/rs6000/vsx.md (VSX_M2): New iterator for fusion
2681 (VSX load fusion peepholes): New peepholes to fuse together an
2682 addi instruction with a VSX load instruction.
2684 * config/rs6000/rs6000.md (GPR load fusion peepholes): New
2685 peepholes to fuse an addis instruction with a load to a GPR base
2686 register. If we are supporting sign extending fusions, convert
2687 sign extending loads to zero extending loads and add an explicit
2690 Backport from mainline
2691 2013-07-18 Pat Haugen <pthaugen@us.ibm.com>
2693 * config/rs6000/rs6000.c (rs6000_option_override_internal): Adjust flag
2694 interaction for new Power8 flags and VSX.
2696 Back port from the trunk
2697 2013-06-28 Michael Meissner <meissner@linux.vnet.ibm.com>
2700 * config/rs6000/rs6000.h (MODES_TIEABLE_P): Do not allow PTImode
2701 to tie with any other modes. Eliminate Altivec vector mode tests,
2702 since these are a subset of ALTIVEC or VSX vector modes. Simplify
2703 code, to return 0 if testing MODE2 for a condition, if we've
2704 already tested MODE1 for the same condition.
2706 Backport from mainline
2707 2013-06-28 Pat Haugen <pthaugen@us.ibm.com>
2709 * config/rs6000/rs6000.md (define_insn ""): Fix insn type.
2711 Back port from the trunk
2712 2013-06-26 Michael Meissner <meissner@linux.vnet.ibm.com>
2713 Pat Haugen <pthaugen@us.ibm.com>
2714 Peter Bergner <bergner@vnet.ibm.com>
2716 * config/rs6000/power8.md: New.
2717 * config/rs6000/rs6000-cpus.def (RS6000_CPU table): Adjust processor
2718 setting for power8 entry.
2719 * config/rs6000/t-rs6000 (MD_INCLUDES): Add power8.md.
2720 * config/rs6000/rs6000.c (is_microcoded_insn, is_cracked_insn): Adjust
2721 test for Power4/Power5 only.
2722 (insn_must_be_first_in_group, insn_must_be_last_in_group): Add Power8
2724 (force_new_group): Adjust comment.
2725 * config/rs6000/rs6000.md: Include power8.md.
2727 Back port from the trunk
2728 2013-06-14 Michael Meissner <meissner@linux.vnet.ibm.com>
2731 * config/rs6000/rs6000.md (mov<mode>_ppc64): Call
2732 rs6000_output_move_128bit to handle emitting quad memory
2733 operations. Set attribute length to 8 bytes.
2735 Back port from the trunk
2736 2013-06-13 Michael Meissner <meissner@linux.vnet.ibm.com>
2738 * config/rs6000/rs6000.c (rs6000_option_override_internal): Move
2739 test for clearing quad memory on 32-bit later.
2741 Back port from the trunk
2743 2013-06-12 Michael Meissner <meissner@linux.vnet.ibm.com>
2744 Pat Haugen <pthaugen@us.ibm.com>
2745 Peter Bergner <bergner@vnet.ibm.com>
2747 * config/rs6000/rs6000.c (emit_load_locked): Add support for
2748 power8 byte, half-word, and quad-word atomic instructions.
2749 (emit_store_conditional): Likewise.
2750 (rs6000_expand_atomic_compare_and_swap): Likewise.
2751 (rs6000_expand_atomic_op): Likewise.
2753 * config/rs6000/sync.md (larx): Add new modes for power8.
2755 (AINT): New mode iterator to include TImode as well as normal
2756 integer modes on power8.
2757 (fetchop_pred): Use int_reg_operand instead of gpc_reg_operand so
2758 that VSX registers are not considered. Use AINT mode iterator
2759 instead of INT1 to allow inclusion of quad word atomic operations
2761 (load_locked<mode>): Likewise.
2762 (store_conditional<mode>): Likewise.
2763 (atomic_compare_and_swap<mode>): Likewise.
2764 (atomic_exchange<mode>): Likewise.
2765 (atomic_nand<mode>): Likewise.
2766 (atomic_fetch_<fetchop_name><mode>): Likewise.
2767 (atomic_nand_fetch<mode>): Likewise.
2768 (mem_thread_fence): Use gen_loadsync_<mode> instead of enumerating
2770 (ATOMIC): On power8, add QImode, HImode modes.
2771 (load_locked<QHI:mode>_si): Varients of load_locked for QI/HI
2772 modes that promote to SImode.
2773 (load_lockedti): Convert TImode arguments to PTImode, so that we
2774 get a guaranteed even/odd register pair.
2775 (load_lockedpti): Likewise.
2776 (store_conditionalti): Likewise.
2777 (store_conditionalpti): Likewise.
2779 * config/rs6000/rs6000.md (QHI): New mode iterator for power8
2780 atomic load/store instructions.
2783 Back port from the trunk
2785 2013-06-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2786 Pat Haugen <pthaugen@us.ibm.com>
2787 Peter Bergner <bergner@vnet.ibm.com>
2789 * config/rs6000/vector.md (GPR move splitter): Do not split moves
2790 of vectors in GPRS if they are direct moves or quad word load or
2793 * config/rs6000/rs6000-protos.h (rs6000_output_move_128bit): Add
2795 (direct_move_p): Likewise.
2796 (quad_load_store_p): Likewise.
2798 * config/rs6000/rs6000.c (enum rs6000_reg_type): Simplify register
2799 classes into bins based on the physical register type.
2800 (reg_class_to_reg_type): Likewise.
2801 (IS_STD_REG_TYPE): Likewise.
2802 (IS_FP_VECT_REG_TYPE): Likewise.
2803 (reload_fpr_gpr): Arrays to determine what insn to use if we can
2804 use direct move instructions.
2805 (reload_gpr_vsx): Likewise.
2806 (reload_vsx_gpr): Likewise.
2807 (rs6000_init_hard_regno_mode_ok): Precalculate the register type
2808 information that is a simplification of register classes. Also
2809 precalculate direct move reload helpers.
2810 (direct_move_p): New function to return true if the operation can
2811 be done as a direct move instruciton.
2812 (quad_load_store_p): New function to return true if the operation
2813 is a quad memory operation.
2814 (rs6000_legitimize_address): If quad memory, only allow register
2815 indirect for TImode addresses.
2816 (rs6000_legitimate_address_p): Likewise.
2817 (enum reload_reg_type): Delete, replace with rs6000_reg_type.
2818 (rs6000_reload_register_type): Likewise.
2819 (register_to_reg_type): Return register type.
2820 (rs6000_secondary_reload_simple_move): New helper function for
2821 secondary reload and secondary memory needed to identify anything
2822 that is a simple move, and does not need reloading.
2823 (rs6000_secondary_reload_direct_move): New helper function for
2824 secondary reload to identify cases that can be done with several
2825 instructions via the direct move instructions.
2826 (rs6000_secondary_reload_move): New helper function for secondary
2827 reload to identify moves between register types that can be done.
2828 (rs6000_secondary_reload): Add support for quad memory operations
2829 and for direct move.
2830 (rs6000_secondary_memory_needed): Likewise.
2831 (rs6000_debug_secondary_memory_needed): Change argument names.
2832 (rs6000_output_move_128bit): New function to return the move to
2833 use for 128-bit moves, including knowing about the various
2834 limitations of quad memory operations.
2836 * config/rs6000/vsx.md (vsx_mov<mode>): Add support for quad
2837 memory operations. call rs6000_output_move_128bit for the actual
2838 instruciton(s) to generate.
2839 (vsx_movti_64bit): Likewise.
2841 * config/rs6000/rs6000.md (UNSPEC_P8V_FMRGOW): New unspec values.
2842 (UNSPEC_P8V_MTVSRWZ): Likewise.
2843 (UNSPEC_P8V_RELOAD_FROM_GPR): Likewise.
2844 (UNSPEC_P8V_MTVSRD): Likewise.
2845 (UNSPEC_P8V_XXPERMDI): Likewise.
2846 (UNSPEC_P8V_RELOAD_FROM_VSX): Likewise.
2847 (UNSPEC_FUSION_GPR): Likewise.
2848 (FMOVE128_GPR): New iterator for direct move.
2849 (f32_lv): New mode attribute for load/store of SFmode/SDmode
2853 (zero_extend<mode>di2_internal1): Add support for power8 32-bit
2854 loads and direct move instructions.
2855 (zero_extendsidi2_lfiwzx): Likewise.
2856 (extendsidi2_lfiwax): Likewise.
2857 (extendsidi2_nocell): Likewise.
2858 (floatsi<mode>2_lfiwax): Likewise.
2860 (floatunssi<mode>2_lfiwzx): Likewise.
2862 (fix_trunc<mode>_stfiwx): Likewise.
2863 (fixuns_trunc<mode>_stfiwx): Likewise.
2864 (mov<mode>_hardfloat, 32-bit floating point): Likewise.
2865 (mov<move>_hardfloat64, 64-bit floating point): Likewise.
2866 (parity<mode>2_cmpb): Set length/type attr.
2867 (unnamed shift right patterns, mov<mode>_internal2): Change type attr
2868 for 'mr.' to fast_compare.
2869 (bpermd_<mode>): Change type attr to popcnt.
2870 (p8_fmrgow_<mode>): New insns for power8 direct move support.
2871 (p8_mtvsrwz_1): Likewise.
2872 (p8_mtvsrwz_2): Likewise.
2873 (reload_fpr_from_gpr<mode>): Likewise.
2874 (p8_mtvsrd_1): Likewise.
2875 (p8_mtvsrd_2): Likewise.
2876 (p8_xxpermdi_<mode>): Likewise.
2877 (reload_vsx_from_gpr<mode>): Likewise.
2878 (reload_vsx_from_gprsf): Likewise.
2879 (p8_mfvsrd_3_<mode>): LIkewise.
2880 (reload_gpr_from_vsx<mode>): Likewise.
2881 (reload_gpr_from_vsxsf): Likewise.
2882 (p8_mfvsrd_4_disf): Likewise.
2883 (multi-word GPR splits): Do not split direct moves or quad memory
2886 Backport from the trunk
2888 2013-06-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2889 Pat Haugen <pthaugen@us.ibm.com>
2890 Peter Bergner <bergner@vnet.ibm.com>
2892 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
2893 Document new power8 builtins.
2895 * config/rs6000/vector.md (and<mode>3): Add a clobber/scratch of a
2896 condition code register, to allow 128-bit logical operations to be
2897 done in the VSX or GPR registers.
2898 (nor<mode>3): Use the canonical form for nor.
2899 (eqv<mode>3): Add expanders for power8 xxleqv, xxlnand, xxlorc,
2900 vclz*, and vpopcnt* vector instructions.
2901 (nand<mode>3): Likewise.
2902 (orc<mode>3): Likewise.
2903 (clz<mode>2): LIkewise.
2904 (popcount<mode>2): Likewise.
2906 * config/rs6000/predicates.md (int_reg_operand): Rework tests so
2907 that only the GPRs are recognized.
2909 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2910 support for new power8 builtins.
2912 * config/rs6000/rs6000-builtin.def (xscvspdpn): Add new power8
2914 (xscvdpspn): Likewise.
2920 (vpopcnt): Likewise.
2921 (vpopcntb): Likewise.
2922 (vpopcnth): Likewise.
2923 (vpopcntw): Likewise.
2924 (vpopcntd): Likewise.
2929 (eqv_v16qi3): Likewise.
2930 (eqv_v8hi3): Likewise.
2931 (eqv_v4si3): Likewise.
2932 (eqv_v2di3): Likewise.
2933 (eqv_v4sf3): Likewise.
2934 (eqv_v2df3): Likewise.
2936 (nand_v16qi3): Likewise.
2937 (nand_v8hi3): Likewise.
2938 (nand_v4si3): Likewise.
2939 (nand_v2di3): Likewise.
2940 (nand_v4sf3): Likewise.
2941 (nand_v2df3): Likewise.
2943 (orc_v16qi3): Likewise.
2944 (orc_v8hi3): Likewise.
2945 (orc_v4si3): Likewise.
2946 (orc_v2di3): Likewise.
2947 (orc_v4sf3): Likewise.
2948 (orc_v2df3): Likewise.
2950 * config/rs6000/rs6000.c (rs6000_option_override_internal): Only
2951 allow power8 quad mode in 64-bit.
2952 (rs6000_builtin_vectorized_function): Add support to vectorize
2953 ISA 2.07 count leading zeros, population count builtins.
2954 (rs6000_expand_vector_init): On ISA 2.07 use xscvdpspn to form
2955 V4SF vectors instead of xscvdpsp to avoid IEEE related traps.
2956 (builtin_function_type): Add vgbbd builtin function which takes an
2958 (altivec_expand_vec_perm_const): Add support for new power8 merge
2961 * config/rs6000/vsx.md (VSX_L2): New iterator for 128-bit types,
2962 that does not include TImdoe for use with 32-bit.
2963 (UNSPEC_VSX_CVSPDPN): Support for power8 xscvdpspn and xscvspdpn
2965 (UNSPEC_VSX_CVDPSPN): Likewise.
2966 (vsx_xscvdpspn): Likewise.
2967 (vsx_xscvspdpn): Likewise.
2968 (vsx_xscvdpspn_scalar): Likewise.
2969 (vsx_xscvspdpn_directmove): Likewise.
2970 (vsx_and<mode>3): Split logical operations into 32-bit and
2971 64-bit. Add support to do logical operations on TImode as well as
2972 VSX vector types. Allow logical operations to be done in either
2973 VSX registers or in general purpose registers in 64-bit mode. Add
2974 splitters if GPRs were used. For AND, add clobber of CCmode to
2975 allow use of ANDI on GPRs. Rewrite nor to use the canonical RTL
2977 (vsx_and<mode>3_32bit): Likewise.
2978 (vsx_and<mode>3_64bit): Likewise.
2979 (vsx_ior<mode>3): Likewise.
2980 (vsx_ior<mode>3_32bit): Likewise.
2981 (vsx_ior<mode>3_64bit): Likewise.
2982 (vsx_xor<mode>3): Likewise.
2983 (vsx_xor<mode>3_32bit): Likewise.
2984 (vsx_xor<mode>3_64bit): Likewise.
2985 (vsx_one_cmpl<mode>2): Likewise.
2986 (vsx_one_cmpl<mode>2_32bit): Likewise.
2987 (vsx_one_cmpl<mode>2_64bit): Likewise.
2988 (vsx_nor<mode>3): Likewise.
2989 (vsx_nor<mode>3_32bit): Likewise.
2990 (vsx_nor<mode>3_64bit): Likewise.
2991 (vsx_andc<mode>3): Likewise.
2992 (vsx_andc<mode>3_32bit): Likewise.
2993 (vsx_andc<mode>3_64bit): Likewise.
2994 (vsx_eqv<mode>3_32bit): Add support for power8 xxleqv, xxlnand,
2995 and xxlorc instructions.
2996 (vsx_eqv<mode>3_64bit): Likewise.
2997 (vsx_nand<mode>3_32bit): Likewise.
2998 (vsx_nand<mode>3_64bit): Likewise.
2999 (vsx_orc<mode>3_32bit): Likewise.
3000 (vsx_orc<mode>3_64bit): Likewise.
3002 * config/rs6000/rs6000.h (VLOGICAL_REGNO_P): Update comment.
3004 * config/rs6000/altivec.md (UNSPEC_VGBBD): Add power8 vgbbd
3006 (p8_vmrgew): Add power8 vmrgew and vmrgow instructions.
3007 (p8_vmrgow): Likewise.
3008 (altivec_and<mode>3): Add clobber of CCmode to allow AND using
3009 GPRs to be split under VSX.
3010 (p8v_clz<mode>2): Add power8 count leading zero support.
3011 (p8v_popcount<mode>2): Add power8 population count support.
3012 (p8v_vgbbd): Add power8 gather bits by bytes by doubleword
3015 * config/rs6000/rs6000.md (eqv<mode>3): Add support for powerp eqv
3018 * config/rs6000/altivec.h (vec_eqv): Add defines to export power8
3020 (vec_nand): Likewise.
3021 (vec_vclz): Likewise.
3022 (vec_vclzb): Likewise.
3023 (vec_vclzd): Likewise.
3024 (vec_vclzh): Likewise.
3025 (vec_vclzw): Likewise.
3026 (vec_vgbbd): Likewise.
3027 (vec_vmrgew): Likewise.
3028 (vec_vmrgow): Likewise.
3029 (vec_vpopcnt): Likewise.
3030 (vec_vpopcntb): Likewise.
3031 (vec_vpopcntd): Likewise.
3032 (vec_vpopcnth): Likewise.
3033 (vec_vpopcntw): Likewise.
3037 2013-05-29 Michael Meissner <meissner@linux.vnet.ibm.com>
3038 Pat Haugen <pthaugen@us.ibm.com>
3039 Peter Bergner <bergner@vnet.ibm.com>
3041 * config/rs6000/vector.md (VEC_I): Add support for new power8 V2DI
3045 (vrotl<mode>3): Likewise.
3046 (vashl<mode>3): Likewise.
3047 (vlshr<mode>3): Likewise.
3048 (vashr<mode>3): Likewise.
3050 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3051 support for power8 V2DI builtins.
3053 * config/rs6000/rs6000-builtin.def (abs_v2di): Add support for
3054 power8 V2DI builtins.
3055 (vupkhsw): Likewise.
3056 (vupklsw): Likewise.
3057 (vaddudm): Likewise.
3062 (vpkudum): Likewise.
3063 (vpksdss): Likewise.
3064 (vpkudus): Likewise.
3065 (vpksdus): Likewise.
3070 (vsubudm): Likewise.
3071 (vcmpequd): Likewise.
3072 (vcmpgtsd): Likewise.
3073 (vcmpgtud): Likewise.
3074 (vcmpequd_p): Likewise.
3075 (vcmpgtsd_p): Likewise.
3076 (vcmpgtud_p): Likewise.
3077 (vupkhsw): Likewise.
3078 (vupklsw): Likewise.
3079 (vaddudm): Likewise.
3084 (vpksdss): Likewise.
3085 (vpksdus): Likewise.
3086 (vpkudum): Likewise.
3087 (vpkudus): Likewise.
3092 (vsubudm): Likewise.
3094 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Add
3095 support for power8 V2DI instructions.
3097 * config/rs6000/altivec.md (UNSPEC_VPKUHUM): Add support for
3098 power8 V2DI instructions. Combine pack and unpack insns to use an
3099 iterator for each mode. Check whether a particular mode supports
3100 Altivec instructions instead of just checking TARGET_ALTIVEC.
3101 (UNSPEC_VPKUWUM): Likewise.
3102 (UNSPEC_VPKSHSS): Likewise.
3103 (UNSPEC_VPKSWSS): Likewise.
3104 (UNSPEC_VPKUHUS): Likewise.
3105 (UNSPEC_VPKSHUS): Likewise.
3106 (UNSPEC_VPKUWUS): Likewise.
3107 (UNSPEC_VPKSWUS): Likewise.
3108 (UNSPEC_VPACK_SIGN_SIGN_SAT): Likewise.
3109 (UNSPEC_VPACK_SIGN_UNS_SAT): Likewise.
3110 (UNSPEC_VPACK_UNS_UNS_SAT): Likewise.
3111 (UNSPEC_VPACK_UNS_UNS_MOD): Likewise.
3112 (UNSPEC_VUPKHSB): Likewise.
3113 (UNSPEC_VUNPACK_HI_SIGN): Likewise.
3114 (UNSPEC_VUNPACK_LO_SIGN): Likewise.
3115 (UNSPEC_VUPKHSH): Likewise.
3116 (UNSPEC_VUPKLSB): Likewise.
3117 (UNSPEC_VUPKLSH): Likewise.
3119 (VI_char): Likewise.
3120 (VI_scalar): Likewise.
3121 (VI_unit): Likewise.
3123 (VP_small): Likewise.
3124 (VP_small_lc): Likewise.
3125 (VU_char): Likewise.
3126 (add<mode>3): Likewise.
3127 (altivec_vaddcuw): Likewise.
3128 (altivec_vaddu<VI_char>s): Likewise.
3129 (altivec_vadds<VI_char>s): Likewise.
3130 (sub<mode>3): Likewise.
3131 (altivec_vsubcuw): Likewise.
3132 (altivec_vsubu<VI_char>s): Likewise.
3133 (altivec_vsubs<VI_char>s): Likewise.
3134 (altivec_vavgs<VI_char>): Likewise.
3135 (altivec_vcmpbfp): Likewise.
3136 (altivec_eq<mode>): Likewise.
3137 (altivec_gt<mode>): Likewise.
3138 (altivec_gtu<mode>): Likewise.
3139 (umax<mode>3): Likewise.
3140 (smax<mode>3): Likewise.
3141 (umin<mode>3): Likewise.
3142 (smin<mode>3): Likewise.
3143 (altivec_vpkuhum): Likewise.
3144 (altivec_vpkuwum): Likewise.
3145 (altivec_vpkshss): Likewise.
3146 (altivec_vpkswss): Likewise.
3147 (altivec_vpkuhus): Likewise.
3148 (altivec_vpkshus): Likewise.
3149 (altivec_vpkuwus): Likewise.
3150 (altivec_vpkswus): Likewise.
3151 (altivec_vpks<VI_char>ss): Likewise.
3152 (altivec_vpks<VI_char>us): Likewise.
3153 (altivec_vpku<VI_char>us): Likewise.
3154 (altivec_vpku<VI_char>um): Likewise.
3155 (altivec_vrl<VI_char>): Likewise.
3156 (altivec_vsl<VI_char>): Likewise.
3157 (altivec_vsr<VI_char>): Likewise.
3158 (altivec_vsra<VI_char>): Likewise.
3159 (altivec_vsldoi_<mode>): Likewise.
3160 (altivec_vupkhsb): Likewise.
3161 (altivec_vupkhs<VU_char>): Likewise.
3162 (altivec_vupkls<VU_char>): Likewise.
3163 (altivec_vupkhsh): Likewise.
3164 (altivec_vupklsb): Likewise.
3165 (altivec_vupklsh): Likewise.
3166 (altivec_vcmpequ<VI_char>_p): Likewise.
3167 (altivec_vcmpgts<VI_char>_p): Likewise.
3168 (altivec_vcmpgtu<VI_char>_p): Likewise.
3169 (abs<mode>2): Likewise.
3170 (vec_unpacks_hi_v16qi): Likewise.
3171 (vec_unpacks_hi_v8hi): Likewise.
3172 (vec_unpacks_lo_v16qi): Likewise.
3173 (vec_unpacks_hi_<VP_small_lc>): Likewise.
3174 (vec_unpacks_lo_v8hi): Likewise.
3175 (vec_unpacks_lo_<VP_small_lc>): Likewise.
3176 (vec_pack_trunc_v8h): Likewise.
3177 (vec_pack_trunc_v4si): Likewise.
3178 (vec_pack_trunc_<mode>): Likewise.
3180 * config/rs6000/altivec.h (vec_vaddudm): Add defines for power8
3182 (vec_vmaxsd): Likewise.
3183 (vec_vmaxud): Likewise.
3184 (vec_vminsd): Likewise.
3185 (vec_vminud): Likewise.
3186 (vec_vpksdss): Likewise.
3187 (vec_vpksdus): Likewise.
3188 (vec_vpkudum): Likewise.
3189 (vec_vpkudus): Likewise.
3190 (vec_vrld): Likewise.
3191 (vec_vsld): Likewise.
3192 (vec_vsrad): Likewise.
3193 (vec_vsrd): Likewise.
3194 (vec_vsubudm): Likewise.
3195 (vec_vupkhsw): Likewise.
3196 (vec_vupklsw): Likewise.
3198 2013-05-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3199 Pat Haugen <pthaugen@us.ibm.com>
3200 Peter Bergner <bergner@vnet.ibm.com>
3202 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Add
3203 documentation for the power8 crypto builtins.
3205 * config/rs6000/t-rs6000 (MD_INCLUDES): Add crypto.md.
3207 * config/rs6000/rs6000-builtin.def (BU_P8V_AV_1): Add support
3208 macros for defining power8 builtin functions.
3209 (BU_P8V_AV_2): Likewise.
3210 (BU_P8V_AV_P): Likewise.
3211 (BU_P8V_VSX_1): Likewise.
3212 (BU_P8V_OVERLOAD_1): Likewise.
3213 (BU_P8V_OVERLOAD_2): Likewise.
3214 (BU_CRYPTO_1): Likewise.
3215 (BU_CRYPTO_2): Likewise.
3216 (BU_CRYPTO_3): Likewise.
3217 (BU_CRYPTO_OVERLOAD_1): Likewise.
3218 (BU_CRYPTO_OVERLOAD_2): Likewise.
3219 (XSCVSPDP): Fix typo, point to the correct instruction.
3220 (VCIPHER): Add power8 crypto builtins.
3221 (VCIPHERLAST): Likewise.
3222 (VNCIPHER): Likewise.
3223 (VNCIPHERLAST): Likewise.
3224 (VPMSUMB): Likewise.
3225 (VPMSUMH): Likewise.
3226 (VPMSUMW): Likewise.
3227 (VPERMXOR_V2DI): Likewise.
3228 (VPERMXOR_V4SI: Likewise.
3229 (VPERMXOR_V8HI: Likewise.
3230 (VPERMXOR_V16QI: Likewise.
3231 (VSHASIGMAW): Likewise.
3232 (VSHASIGMAD): Likewise.
3234 (VPERMXOR): Likewise.
3235 (VSHASIGMA): Likewise.
3237 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
3238 __CRYPTO__ if the crypto instructions are available.
3239 (altivec_overloaded_builtins): Add support for overloaded power8
3242 * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
3243 support for power8 crypto builtins.
3244 (builtin_function_type): Likewise.
3245 (altivec_init_builtins): Add support for builtins that take vector
3246 long long (V2DI) arguments.
3248 * config/rs6000/crypto.md: New file, define power8 crypto
3251 2013-05-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3252 Pat Haugen <pthaugen@us.ibm.com>
3253 Peter Bergner <bergner@vnet.ibm.com>
3255 * doc/invoke.texi (Option Summary): Add power8 options.
3256 (RS/6000 and PowerPC Options): Likewise.
3258 * doc/md.texi (PowerPC and IBM RS6000 constraints): Update to use
3259 constraints.md instead of rs6000.h. Reorder w* constraints. Add
3260 wm, wn, wr documentation.
3262 * gcc/config/rs6000/constraints.md (wm): New constraint for VSX
3263 registers if direct move instructions are enabled.
3264 (wn): New constraint for no registers.
3265 (wq): New constraint for quad word even GPR registers.
3266 (wr): New constraint if 64-bit instructions are enabled.
3267 (wv): New constraint if power8 vector instructions are enabled.
3268 (wQ): New constraint for quad word memory locations.
3270 * gcc/config/rs6000/predicates.md (const_0_to_15_operand): New
3271 constraint for 0..15 for crypto instructions.
3272 (gpc_reg_operand): If VSX allow registers in VSX registers as well
3273 as GPR and floating point registers.
3274 (int_reg_operand): New predicate to match only GPR registers.
3275 (base_reg_operand): New predicate to match base registers.
3276 (quad_int_reg_operand): New predicate to match even GPR registers
3277 for quad memory operations.
3278 (vsx_reg_or_cint_operand): New predicate to allow vector logical
3279 operations in both GPR and VSX registers.
3280 (quad_memory_operand): New predicate for quad memory operations.
3281 (reg_or_indexed_operand): New predicate for direct move support.
3283 * gcc/config/rs6000/rs6000-cpus.def (ISA_2_5_MASKS_EMBEDDED):
3284 Inherit from ISA_2_4_MASKS, not ISA_2_2_MASKS.
3285 (ISA_2_7_MASKS_SERVER): New mask for ISA 2.07 (i.e. power8).
3286 (POWERPC_MASKS): Add power8 options.
3287 (power8 cpu): Use ISA_2_7_MASKS_SERVER instead of specifying the
3290 * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
3291 Define _ARCH_PWR8 and __POWER8_VECTOR__ for power8.
3293 * gcc/config/rs6000/rs6000.opt (-mvsx-timode): Add documentation.
3294 (-mpower8-fusion): New power8 options.
3295 (-mpower8-fusion-sign): Likewise.
3296 (-mpower8-vector): Likewise.
3297 (-mcrypto): Likewise.
3298 (-mdirect-move): Likewise.
3299 (-mquad-memory): Likewise.
3301 * gcc/config/rs6000/rs6000.c (power8_cost): Initial definition for
3303 (rs6000_hard_regno_mode_ok): Make PTImode only match even GPR
3305 (rs6000_debug_reg_print): Print the base register class if
3307 (rs6000_debug_vector_unit): Add p8_vector.
3308 (rs6000_debug_reg_global): If -mdebug=reg, print power8 constraint
3309 definitions. Also print fusion state.
3310 (rs6000_init_hard_regno_mode_ok): Set up power8 constraints.
3311 (rs6000_builtin_mask_calculate): Add power8 builtin support.
3312 (rs6000_option_override_internal): Add support for power8.
3313 (rs6000_common_init_builtins): Add debugging for skipped builtins
3315 (rs6000_adjust_cost): Add power8 support.
3316 (rs6000_issue_rate): Likewise.
3317 (insn_must_be_first_in_group): Likewise.
3318 (insn_must_be_last_in_group): Likewise.
3319 (force_new_group): Likewise.
3320 (rs6000_register_move_cost): Likewise.
3321 (rs6000_opt_masks): Likewise.
3323 * config/rs6000/rs6000.h (ASM_CPU_POWER8_SPEC): If we don't have a
3324 power8 capable assembler, default to power7 options.
3325 (TARGET_DIRECT_MOVE): Likewise.
3326 (TARGET_CRYPTO): Likewise.
3327 (TARGET_P8_VECTOR): Likewise.
3328 (VECTOR_UNIT_P8_VECTOR_P): Define power8 vector support.
3329 (VECTOR_UNIT_VSX_OR_P8_VECTOR_P): Likewise.
3330 (VECTOR_MEM_P8_VECTOR_P): Likewise.
3331 (VECTOR_MEM_VSX_OR_P8_VECTOR_P): Likewise.
3332 (VECTOR_MEM_ALTIVEC_OR_VSX_P): Likewise.
3333 (TARGET_XSCVDPSPN): Likewise.
3334 (TARGET_XSCVSPDPN): Likewsie.
3335 (TARGET_SYNC_HI_QI): Likewise.
3336 (TARGET_SYNC_TI): Likewise.
3337 (MASK_CRYPTO): Likewise.
3338 (MASK_DIRECT_MOVE): Likewise.
3339 (MASK_P8_FUSION): Likewise.
3340 (MASK_P8_VECTOR): Likewise.
3341 (REG_ALLOC_ORDER): Move fr13 to be lower in priority so that the
3342 TFmode temporary used by some of the direct move instructions to
3343 get two FP temporary registers does not force creation of a stack
3345 (VLOGICAL_REGNO_P): Allow vector logical operations in GPRs.
3346 (MODES_TIEABLE_P): Move the VSX tests above the Altivec tests so
3347 that any VSX registers are tieable, even if they are also an
3348 Altivec vector mode.
3349 (r6000_reg_class_enum): Add wm, wr, wv constraints.
3350 (RS6000_BTM_P8_VECTOR): Power8 builtin support.
3351 (RS6000_BTM_CRYPTO): Likewise.
3352 (RS6000_BTM_COMMON): Likewise.
3354 * config/rs6000/rs6000.md (cpu attribute): Add power8.
3355 * config/rs6000/rs6000-opts.h (PROCESSOR_POWER8): Likewise.
3356 (enum rs6000_vector): Add power8 vector support.
3359 Backport from mainline
3360 2013-03-20 Pat Haugen <pthaugen@us.ibm.com>
3362 * config/rs6000/predicates.md (indexed_address, update_address_mem
3363 update_indexed_address_mem): New predicates.
3364 * config/rs6000/vsx.md (vsx_extract_<mode>_zero): Set correct "type"
3365 attribute for load/store instructions.
3366 * config/rs6000/dfp.md (movsd_store): Likewise.
3367 (movsd_load): Likewise.
3368 * config/rs6000/rs6000.md (zero_extend<mode>di2_internal1): Likewise.
3369 (unnamed HI->DI extend define_insn): Likewise.
3370 (unnamed SI->DI extend define_insn): Likewise.
3371 (unnamed QI->SI extend define_insn): Likewise.
3372 (unnamed QI->HI extend define_insn): Likewise.
3373 (unnamed HI->SI extend define_insn): Likewise.
3374 (unnamed HI->SI extend define_insn): Likewise.
3375 (extendsfdf2_fpr): Likewise.
3376 (movsi_internal1): Likewise.
3377 (movsi_internal1_single): Likewise.
3378 (movhi_internal): Likewise.
3379 (movqi_internal): Likewise.
3380 (movcc_internal1): Correct mnemonic for stw insn. Set correct "type"
3381 attribute for load/store instructions.
3382 (mov<mode>_hardfloat): Set correct "type" attribute for load/store
3384 (mov<mode>_softfloat): Likewise.
3385 (mov<mode>_hardfloat32): Likewise.
3386 (mov<mode>_hardfloat64): Likewise.
3387 (mov<mode>_softfloat64): Likewise.
3388 (movdi_internal32): Likewise.
3389 (movdi_internal64): Likewise.
3390 (probe_stack_<mode>): Likewise.
3392 Backport from mainline
3393 2013-03-20 Michael Meissner <meissner@linux.vnet.ibm.com>
3395 * config/rs6000/vector.md (VEC_R): Add 32-bit integer, binary
3396 floating point, and decimal floating point to reload iterator.
3398 * config/rs6000/constraints.md (wl constraint): New constraints to
3399 return FLOAT_REGS if certain options are used to reduce the number
3400 of separate patterns that exist in the file.
3401 (wx constraint): Likewise.
3402 (wz constraint): Likewise.
3404 * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
3405 -mdebug=reg, print wg, wl, wx, and wz constraints.
3406 (rs6000_init_hard_regno_mode_ok): Initialize new constraints.
3407 Initialize the reload functions for 64-bit binary/decimal floating
3409 (reg_offset_addressing_ok_p): If we are on a power7 or later, use
3410 LFIWZX and STFIWX to load/store 32-bit decimal types, and don't
3411 create the buffer on the stack to overcome not having a 32-bit
3413 (rs6000_emit_move): Likewise.
3414 (rs6000_secondary_memory_needed_rtx): Likewise.
3415 (rs6000_alloc_sdmode_stack_slot): Likewise.
3416 (rs6000_preferred_reload_class): On VSX, we can create SFmode 0.0f
3417 via xxlxor, just like DFmode 0.0.
3419 * config/rs6000/rs6000.h (TARGET_NO_SDMODE_STACK): New macro)
3420 (define as 1 if we are running on a power7 or newer.
3421 (enum r6000_reg_class_enum): Add new constraints.
3423 * config/rs6000/dfp.md (movsd): Delete, combine with binary
3424 floating point moves in rs6000.md. Combine power6x (mfpgpr) moves
3425 with other moves by using conditional constraits (wg). Use LFIWZX
3426 and STFIWX for loading SDmode on power7. Use xxlxor to create
3428 (movsd splitter): Likewise.
3429 (movsd_hardfloat): Likewise.
3430 (movsd_softfloat): Likewise.
3432 * config/rs6000/rs6000.md (FMOVE32): New iterators to combine
3433 binary and decimal floating point moves.
3434 (fmove_ok): New attributes to combine binary and decimal floating
3435 point moves, and to combine power6x (mfpgpr) moves along normal
3437 (real_value_to_target): Likewise.
3444 (movsf): Combine binary and decimal floating point moves. Combine
3445 power6x (mfpgpr) moves with other moves by using conditional
3446 constraits (wg). Use LFIWZX and STFIWX for loading SDmode on
3448 (mov<mode> for SFmode/SDmode); Likewise.
3449 (SFmode/SDmode splitters): Likewise.
3450 (movsf_hardfloat): Likewise.
3451 (mov<mode>_hardfloat for SFmode/SDmode): Likewise.
3452 (movsf_softfloat): Likewise.
3453 (mov<mode>_softfloat for SFmode/SDmode): Likewise.
3455 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wl)
3456 (wx and wz constraints.
3458 * config/rs6000/constraints.md (wg constraint): New constraint to
3459 return FLOAT_REGS if -mmfpgpr (power6x) was used.
3461 * config/rs6000/rs6000.h (enum r6000_reg_class_enum): Add wg
3464 * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
3465 -mdebug=reg, print wg, wl, wx, and wz constraints.
3466 (rs6000_init_hard_regno_mode_ok): Initialize new constraints.
3467 Initialize the reload functions for 64-bit binary/decimal floating
3469 (reg_offset_addressing_ok_p): If we are on a power7 or later, use
3470 LFIWZX and STFIWX to load/store 32-bit decimal types, and don't
3471 create the buffer on the stack to overcome not having a 32-bit
3473 (rs6000_emit_move): Likewise.
3474 (rs6000_secondary_memory_needed_rtx): Likewise.
3475 (rs6000_alloc_sdmode_stack_slot): Likewise.
3476 (rs6000_preferred_reload_class): On VSX, we can create SFmode 0.0f
3477 via xxlxor, just like DFmode 0.0.
3480 * config/rs6000/dfp.md (movdd): Delete, combine with binary
3481 floating point moves in rs6000.md. Combine power6x (mfpgpr) moves
3482 with other moves by using conditional constraits (wg). Use LFIWZX
3483 and STFIWX for loading SDmode on power7.
3484 (movdd splitters): Likewise.
3485 (movdd_hardfloat32): Likewise.
3486 (movdd_softfloat32): Likewise.
3487 (movdd_hardfloat64_mfpgpr): Likewise.
3488 (movdd_hardfloat64): Likewise.
3489 (movdd_softfloat64): Likewise.
3491 * config/rs6000/rs6000.md (FMOVE64): New iterators to combine
3492 64-bit binary and decimal floating point moves.
3493 (FMOVE64X): Likewise.
3494 (movdf): Combine 64-bit binary and decimal floating point moves.
3495 Combine power6x (mfpgpr) moves with other moves by using
3496 conditional constraits (wg).
3497 (mov<mode> for DFmode/DDmode): Likewise.
3498 (DFmode/DDmode splitters): Likewise.
3499 (movdf_hardfloat32): Likewise.
3500 (mov<mode>_hardfloat32 for DFmode/DDmode): Likewise.
3501 (movdf_softfloat32): Likewise.
3502 (movdf_hardfloat64_mfpgpr): Likewise.
3503 (movdf_hardfloat64): Likewise.
3504 (mov<mode>_hardfloat64 for DFmode/DDmode): Likewise.
3505 (movdf_softfloat64): Likewise.
3506 (mov<mode>_softfloat64 for DFmode/DDmode): Likewise.
3507 (reload_<mode>_load): Move to later in the file so they aren't in
3508 the middle of the floating point move insns.
3509 (reload_<mode>_store): Likewise.
3511 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wg
3514 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print out wg
3515 constraint if -mdebug=reg.
3516 (rs6000_initi_hard_regno_mode_ok): Enable wg constraint if
3517 -mfpgpr. Enable using dd reload support if needed.
3519 * config/rs6000/dfp.md (movtd): Delete, combine with 128-bit
3520 binary and decimal floating point moves in rs6000.md.
3521 (movtd_internal): Likewise.
3523 * config/rs6000/rs6000.md (FMOVE128): Combine 128-bit binary and
3524 decimal floating point moves.
3526 (movtf_internal): Likewise.
3527 (mov<mode>_internal, TDmode/TFmode): Likewise.
3528 (movtf_softfloat): Likewise.
3529 (mov<mode>_softfloat, TDmode/TFmode): Likewise.
3531 * config/rs6000/rs6000.md (movdi_mfpgpr): Delete, combine with
3532 movdi_internal64, using wg constraint for move direct operations.
3533 (movdi_internal64): Likewise.
3535 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print
3536 MODES_TIEABLE_P for selected modes. Print the numerical value of
3537 the various virtual registers. Use GPR/FPR first/last values)
3538 (instead of hard coding the register numbers. Print which modes
3539 have reload functions registered.
3540 (rs6000_option_override_internal): If -mdebug=reg, trace the
3541 options settings before/after setting cpu, target and subtarget
3543 (rs6000_secondary_reload_trace): Improve the RTL dump for
3544 -mdebug=addr and for secondary reload failures in
3545 rs6000_secondary_reload_inner.
3546 (rs6000_secondary_reload_fail): Likewise.
3547 (rs6000_secondary_reload_inner): Likewise.
3549 * config/rs6000/rs6000.md (FIRST_GPR_REGNO): Add convenience
3550 macros for first/last GPR and FPR registers.
3551 (LAST_GPR_REGNO): Likewise.
3552 (FIRST_FPR_REGNO): Likewise.
3553 (LAST_FPR_REGNO): Likewise.
3555 * config/rs6000/vector.md (mul<mode>3): Use the combined macro
3556 VECTOR_UNIT_ALTIVEC_OR_VSX_P instead of separate calls to
3557 VECTOR_UNIT_ALTIVEC_P and VECTOR_UNIT_VSX_P.
3558 (vcond<mode><mode>): Likewise.
3559 (vcondu<mode><mode>): Likewise.
3560 (vector_gtu<mode>): Likewise.
3561 (vector_gte<mode>): Likewise.
3562 (xor<mode>3): Don't allow logical operations on TImode in 32-bit
3563 to prevent the compiler from converting DImode operations to
3565 (ior<mode>3): Likewise.
3566 (and<mode>3): Likewise.
3567 (one_cmpl<mode>2): Likewise.
3568 (nor<mode>3): Likewise.
3569 (andc<mode>3): Likewise.
3571 * config/rs6000/constraints.md (wt constraint): New constraint
3572 that returns VSX_REGS if TImode is allowed in VSX registers.
3574 * config/rs6000/predicates.md (easy_fp_constant): 0.0f is an easy
3577 * config/rs6000/rs6000-modes.def (PTImode): Define, PTImode is
3578 similar to TImode, but it is restricted to being in the GPRs.
3580 * config/rs6000/rs6000.opt (-mvsx-timode): New switch to allow
3581 TImode to occupy a single VSX register.
3583 * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Default to
3584 -mvsx-timode for power7/power8.
3585 (power7 cpu): Likewise.
3586 (power8 cpu): Likewise.
3588 * config/rs6000/rs6000.c (rs6000_hard_regno_nregs_internal): Make
3589 sure that TFmode/TDmode take up two registers if they are ever
3590 allowed in the upper VSX registers.
3591 (rs6000_hard_regno_mode_ok): If -mvsx-timode, allow TImode in VSX
3593 (rs6000_init_hard_regno_mode_ok): Likewise.
3594 (rs6000_debug_reg_global): Add debugging for PTImode and wt
3595 constraint. Print if LRA is turned on.
3596 (rs6000_option_override_internal): Give an error if -mvsx-timode
3597 and VSX is not enabled.
3598 (invalid_e500_subreg): Handle PTImode, restricting it to GPRs. If
3599 -mvsx-timode, restrict TImode to reg+reg addressing, and PTImode
3600 to reg+offset addressing. Use PTImode when checking offset
3601 addresses for validity.
3602 (reg_offset_addressing_ok_p): Likewise.
3603 (rs6000_legitimate_offset_address_p): Likewise.
3604 (rs6000_legitimize_address): Likewise.
3605 (rs6000_legitimize_reload_address): Likewise.
3606 (rs6000_legitimate_address_p): Likewise.
3607 (rs6000_eliminate_indexed_memrefs): Likewise.
3608 (rs6000_emit_move): Likewise.
3609 (rs6000_secondary_reload): Likewise.
3610 (rs6000_secondary_reload_inner): Handle PTImode. Allow 64-bit
3611 reloads to fpr registers to continue to use reg+offset addressing)
3612 (but 64-bit reloads to altivec registers need reg+reg addressing.
3613 Drop test for PRE_MODIFY, since VSX loads/stores no longer support
3614 it. Treat LO_SUM like a PLUS operation.
3615 (rs6000_secondary_reload_class): If type is 64-bit, prefer to use
3616 FLOAT_REGS instead of VSX_RGS to allow use of reg+offset
3618 (rs6000_cannot_change_mode_class): Do not allow TImode in VSX
3619 registers to share a register with a smaller sized type, since VSX
3620 puts scalars in the upper 64-bits.
3621 (print_operand): Add support for PTImode.
3622 (rs6000_register_move_cost): Use VECTOR_MEM_VSX_P instead of
3623 VECTOR_UNIT_VSX_P to catch types that can be loaded in VSX
3624 registers, but don't have arithmetic support.
3625 (rs6000_memory_move_cost): Add test for VSX.
3626 (rs6000_opt_masks): Add -mvsx-timode.
3628 * config/rs6000/vsx.md (VSm): Change to use 64-bit aligned moves
3631 (VSr): Use wt constraint for TImode.
3632 (VSv): Drop TImode support.
3633 (vsx_movti): Delete, replace with versions for 32-bit and 64-bit.
3634 (vsx_movti_64bit): Likewise.
3635 (vsx_movti_32bit): Likewise.
3636 (vec_store_<mode>): Use VSX iterator instead of vector iterator.
3637 (vsx_and<mode>3): Delete use of '?' constraint on inputs, just put
3638 one '?' on the appropriate output constraint. Do not allow TImode
3639 logical operations on 32-bit systems.
3640 (vsx_ior<mode>3): Likewise.
3641 (vsx_xor<mode>3): Likewise.
3642 (vsx_one_cmpl<mode>2): Likewise.
3643 (vsx_nor<mode>3): Likewise.
3644 (vsx_andc<mode>3): Likewise.
3645 (vsx_concat_<mode>): Likewise.
3646 (vsx_xxpermdi_<mode>): Fix thinko for non V2DF/V2DI modes.
3648 * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Map from
3649 OPTION_MASK_VSX_TIMODE.
3650 (enum rs6000_reg_class_enum): Add RS6000_CONSTRAINT_wt.
3651 (STACK_SAVEAREA_MODE): Use PTImode instead of TImode.
3653 * config/rs6000/rs6000.md (INT mode attribute): Add PTImode.
3654 (TI2 iterator): New iterator for TImode, PTImode.
3655 (wd mode attribute): Add values for vector types.
3656 (movti_string): Replace TI move operations with operations for
3657 TImode and PTImode. Add support for TImode being allowed in VSX
3659 (mov<mode>_string, TImode/PTImode): Likewise.
3660 (movti_ppc64): Likewise.
3661 (mov<mode>_ppc64, TImode/PTImode): Likewise.
3662 (TI mode splitters): Likewise.
3664 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wt
3667 2014-04-04 Richard Biener <rguenther@suse.de>
3669 * tree-ssanames.c (make_ssa_name_fn): Fix assert.
3671 2014-04-02 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3673 * config/s390/s390.c (s390_expand_insv): Use GET_MODE_BITSIZE.
3675 2014-04-01 Richard Biener <rguenther@suse.de>
3677 * gimple.h (struct gimple_statement_base): Align subcode to
3680 2014-04-01 Sebastian Huber <sebastian.huber@embedded-brains.de>
3682 * doc/invoke.texi (mapp-regs): Clarify.
3684 2014-03-31 H.J. Lu <hongjiu.lu@intel.com>
3686 PR rtl-optimization/60700
3687 Backport from mainline
3688 2013-07-30 Zhenqiang Chen <zhenqiang.chen@linaro.org>
3690 PR rtl-optimization/57637
3691 * function.c (move_insn_for_shrink_wrap): Also check the
3692 GEN set of the LIVE problem for the liveness analysis
3693 if it exists, otherwise give up.
3695 2014-03-30 Kaz Kojima <kkojima@gcc.gnu.org>
3697 Backport from mainline
3698 2014-03-19 Kaz Kojima <kkojima@gcc.gnu.org>
3701 * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
3703 2014-03-26 Martin Jambor <mjambor@suse.cz>
3706 * ipa.c (symtab_remove_unreachable_nodes): Clear thunk and
3707 alias flags of nodes in the border.
3709 2014-03-26 Eric Botcazou <ebotcazou@adacore.com>
3711 PR rtl-optimization/60452
3712 * rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
3713 <case REG>: Return 1 for invalid offsets from the frame pointer.
3715 2014-03-24 Richard Biener <rguenther@suse.de>
3717 PR tree-optimization/60429
3718 * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Remove
3721 2014-03-23 Eric Botcazou <ebotcazou@adacore.com>
3723 PR rtl-optimization/60601
3724 * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
3726 * gcc.c (eval_spec_function): Initialize save_growing_value.
3728 2014-03-20 Jakub Jelinek <jakub@redhat.com>
3731 * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
3732 into CONST, put pic register as first operand of PLUS. Use
3733 gen_const_mem for both 32-bit and 64-bit PIC got loads.
3735 2014-03-20 Eric Botcazou <ebotcazou@adacore.com>
3737 * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
3738 around for store forwarding issue in the FPU on the UT699.
3739 * config/sparc/sparc.md (in_branch_delay): Return false for single FP
3740 loads and operations if -mfix-ut699 is specified.
3741 (divtf3_hq): Tweak attribute.
3742 (sqrttf2_hq): Likewise.
3744 2014-03-18 Kai Tietz <ktietz@redhat.com>
3746 PR rtl-optimization/56356
3747 * sdbout.c (sdbout_parms): Verify that parms'
3748 incoming argument is valid.
3749 (sdbout_reg_parms): Likewise.
3751 2014-03-18 Eric Botcazou <ebotcazou@adacore.com>
3753 * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
3754 proper constant for the store mode.
3756 2014-03-17 Mikael Pettersson <mikpelinux@gmail.com>
3757 Committed by Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3759 Backport from mainline:
3761 2013-06-20 Joern Rennecke <joern.rennecke@embecosm.com>
3763 PR rtl-optimization/57425
3764 PR rtl-optimization/57569
3765 * alias.c (write_dependence_p): Remove parameters mem_mode and
3766 canon_mem_addr. Add parameters x_mode, x_addr and x_canonicalized.
3767 Changed all callers.
3768 (canon_anti_dependence): Get comments and semantics in sync.
3769 Add parameter mem_canonicalized. Changed all callers.
3770 * rtl.h (canon_anti_dependence): Update prototype.
3772 2013-06-16 Joern Rennecke <joern.rennecke@embecosm.com>
3774 PR rtl-optimization/57425
3775 PR rtl-optimization/57569
3776 * alias.c (write_dependence_p): Add new parameters mem_mode,
3777 canon_mem_addr and mem_canonicalized. Change type of writep to bool.
3778 Changed all callers.
3779 (canon_anti_dependence): New function.
3780 * cse.c (check_dependence): Use canon_anti_dependence.
3781 * cselib.c (cselib_invalidate_mem): Likewise.
3782 * rtl.h (canon_anti_dependence): Declare.
3784 2014-03-17 Richard Biener <rguenther@suse.de>
3786 Backport from mainline
3787 2014-03-11 Richard Biener <rguenther@suse.de>
3789 PR tree-optimization/60429
3790 PR tree-optimization/60485
3791 * tree-ssa-structalias.c (set_union_with_increment): Properly
3792 take into account all fields that overlap the shifted vars.
3793 (do_sd_constraint): Likewise.
3794 (do_ds_constraint): Likewise.
3795 (get_constraint_for_ptr_offset): Likewise.
3797 2014-03-15 Eric Botcazou <ebotcazou@adacore.com>
3799 * config/sparc/sparc-protos.h (tls_call_delay): Delete.
3800 (eligible_for_call_delay): New prototype.
3801 * config/sparc/sparc.c (tls_call_delay): Rename into...
3802 (eligible_for_call_delay): ...this. Return false if the instruction
3803 cannot be put in the delay slot of a branch.
3804 (eligible_for_restore_insn): Simplify.
3805 (eligible_for_return_delay): Return false if the instruction cannot be
3806 put in the delay slot of a branch and simplify.
3807 (eligible_for_sibcall_delay): Return false if the instruction cannot be
3808 put in the delay slot of a branch.
3809 * config/sparc/sparc.md (fix_ut699): New attribute.
3810 (tls_call_delay): Delete.
3811 (in_call_delay): Reimplement.
3812 (eligible_for_sibcall_delay): Rename into...
3813 (in_sibcall_delay): ...this.
3814 (eligible_for_return_delay): Rename into...
3815 (in_return_delay): ...this.
3816 (in_branch_delay): Reimplement.
3817 (in_uncond_branch_delay): Delete.
3818 (in_annul_branch_delay): Delete.
3820 2014-03-14 Georg-Johann Lay <avr@gjlay.de>
3822 Backport from 2014-03-14 trunk r208562.
3825 * config/avr/avr.c (avr_set_current_function): Pass function name
3826 through default_strip_name_encoding before sanity checking instead
3827 of skipping the first char of the assembler name.
3829 2014-03-13 Georg-Johann Lay <avr@gjlay.de>
3831 Backport from 2014-03-13 trunk r208532.
3834 * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
3835 calls of avr_out_plus_1.
3837 2014-03-13 Joey Ye <joey.ye@arm.com>
3839 Backport from mainline
3840 2014-03-12 Thomas Preud'homme <thomas.preudhomme@arm.com>
3842 PR tree-optimization/60454
3843 * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
3845 2014-03-06 Matthias Klose <doko@ubuntu.com>
3847 * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
3848 MULTILIB_OSDIRNAMES is not defined.
3850 2014-03-06 Jakub Jelinek <jakub@redhat.com>
3852 PR tree-optimization/60276
3853 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Avoid
3854 a -Wsign-compare warning.
3856 * Makefile.in (tree-ssa-uninit.o): Depend on $(PARAMS_H).
3858 Backport from mainline
3859 2014-02-21 Jakub Jelinek <jakub@redhat.com>
3861 PR tree-optimization/56490
3862 * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
3863 * tree-ssa-uninit.c: Include params.h.
3864 (compute_control_dep_chain): Add num_calls argument, return false
3865 if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
3866 num_calls to recursive call.
3867 (find_predicates): Change dep_chain into normal array, add num_calls
3868 variable and adjust compute_control_dep_chain caller.
3869 (find_def_preds): Likewise.
3871 2014-02-13 Jakub Jelinek <jakub@redhat.com>
3874 * expr.c (compress_float_constant): If x is a hard register,
3875 extend into a pseudo and then move to x.
3877 2014-02-11 Richard Henderson <rth@redhat.com>
3878 Jakub Jelinek <jakub@redhat.com>
3881 * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
3882 around drhs if type conversion to lacc->type is not useless.
3884 2014-02-08 Jakub Jelinek <jakub@redhat.com>
3887 * ipa-cp.c (determine_versionability): Fail at -O0
3888 or __attribute__((optimize (0))) or -fno-ipa-cp functions.
3889 * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
3891 2014-02-06 Jakub Jelinek <jakub@redhat.com>
3894 * tree.h (opts_for_fn): New inline function.
3895 (opt_for_fn): Define.
3896 * config/i386/i386.c (ix86_function_regparm): Use
3897 opt_for_fn (decl, optimize) instead of optimize.
3899 2014-02-05 Jakub Jelinek <jakub@redhat.com>
3902 * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
3903 bb with no successors.
3905 2014-03-04 Richard Biener <rguenther@suse.de>
3907 PR tree-optimization/60382
3908 * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
3909 dead PHIs a reduction.
3911 2014-02-25 Richard Biener <rguenther@suse.de>
3913 Backport from mainline
3914 2014-02-21 Richard Biener <rguenther@suse.de>
3916 PR tree-optimization/60276
3917 * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
3918 (STMT_VINFO_MIN_NEG_DIST): New macro.
3919 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
3920 STMT_VINFO_MIN_NEG_DIST.
3921 * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
3922 made for negative dependence distances still hold.
3924 2014-02-25 Richard Biener <rguenther@suse.de>
3926 Backport from mainline
3927 2014-02-21 Richard Biener <rguenther@suse.de>
3930 * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
3931 DECL_INITIAL for globals not in the current function context.
3933 2014-02-20 Richard Biener <rguenther@suse.de>
3936 * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
3939 2014-02-14 Richard Biener <rguenther@suse.de>
3941 PR tree-optimization/60183
3942 * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating
3944 (tree_ssa_phiprop): Calculate and free post-dominators.
3946 2014-02-25 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3949 * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
3952 2014-02-24 John David Anglin <danglin@gcc.gnu.org>
3954 * config/pa/pa.c (pa_output_move_double): Don't valididate when
3955 adjusting offsetable addresses.
3957 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
3959 * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg names
3961 2014-02-23 Edgar E. Iglesias <edgar.iglesias@xilinx.com>
3963 * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
3966 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
3968 * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk
3969 and define TARGET_ASM_OUTPUT_MI_THUNK and
3970 TARGET_ASM_CAN_OUTPUT_MI_THUNK.
3972 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
3974 * config/microblaze/predicates.md: Add cmp_op predicate.
3975 * config/microblaze/microblaze.md: Add branch_compare instruction
3976 which uses cmp_op predicate and emits cmp insn before branch.
3977 * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
3978 to microblaze_expand_conditional_branch and consolidate logic.
3979 (microblaze_expand_conditional_branch): emit branch_compare
3980 insn instead of handling cmp op separate from branch insn.
3982 2014-02-21 Martin Jambor <mjambor@suse.cz>
3985 * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
3986 info when checking whether lattices are bottom.
3988 2014-02-21 Jakub Jelinek <jakub@redhat.com>
3990 * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
3991 mode for mask of V8SFmode permutation.
3993 2014-02-20 Richard Henderson <rth@redhat.com>
3996 * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
3997 on failure the store back into EXPECT. Always make a new pseudo for
4000 2014-02-20 Jakub Jelinek <jakub@redhat.com>
4003 * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
4004 gen_reg_rtx if d->testing_p.
4005 (expand_vec_perm_pshufb2, expand_vec_perm_even_odd_1,
4006 expand_vec_perm_broadcast_1): Return early if d->testing_p and
4007 we will certainly return true.
4009 2014-02-20 Richard Biener <rguenther@suse.de>
4011 * tree-cfg.c (replace_uses_by): Mark altered BBs before
4012 doing the substitution.
4014 2014-02-19 H.J. Lu <hongjiu.lu@intel.com>
4016 Backport from mainline
4017 2014-02-19 H.J. Lu <hongjiu.lu@intel.com>
4020 * config/i386/i386.c (construct_container): Remove TFmode check
4021 for X86_64_INTEGER_CLASS.
4023 2014-02-19 Uros Bizjak <ubizjak@gmail.com>
4025 Backport from mainline
4026 2014-02-19 Uros Bizjak <ubizjak@gmail.com>
4029 * config/i386/i386.c (type_natural_mode): Warn for ABI changes
4030 only when -Wpsabi is enabled.
4032 2014-02-19 Terry Guo <terry.guo@arm.com>
4034 Backport from mainline
4035 2014-02-08 Terry Guo <terry.guo@arm.com>
4037 * doc/invoke.texi: Document ARM -march=armv7e-m.
4039 2014-02-18 Kai Tietz <ktietz@redhat.com>
4041 Backport from mainline
4042 2014-02-18 Kai Tietz <ktietz@redhat.com>
4045 * config/i386/i386.c (ix86_expand_prologue): Use
4046 rax register as displacement for restoring %r10, %rax.
4047 Additional fix wrong offset for restoring both-registers.
4049 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
4051 * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
4052 assertion with conditional return.
4054 2014-02-18 Jakub Jelinek <jakub@redhat.com>
4055 Uros Bizjak <ubizjak@gmail.com>
4058 * config/i386/driver-i386.c (host_detect_local_cpu): If
4059 YMM state is not saved by the OS, also clear has_f16c. Move
4060 CPUID 0x80000001 handling before YMM state saving checking.
4062 2014-02-14 Roland McGrath <mcgrathr@google.com>
4064 * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
4065 * configure: Regenerated.
4066 * config.in: Regenerated.
4067 * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
4068 instead of ASM_SHORT.
4070 2014-02-13 Uros Bizjak <ubizjak@gmail.com>
4072 Backport from mainline
4073 2014-02-13 Uros Bizjak <ubizjak@gmail.com>
4075 * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
4076 operands[2], not operands[3].
4078 2014-02-13 Dominik Vogt <vogt@linux.vnet.ibm.com>
4080 * config/s390/s390.c (s390_asm_output_function_label): Fix crash
4081 caused by bad second argument to warning_at() with -mhotpatch and
4082 nested functions (e.g. with gfortran).
4084 2014-02-12 H.J. Lu <hongjiu.lu@intel.com>
4086 Backport from mainline
4087 2014-02-12 H.J. Lu <hongjiu.lu@intel.com>
4088 Uros Bizjak <ubizjak@gmail.com>
4091 * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
4093 2014-02-12 Eric Botcazou <ebotcazou@adacore.com>
4095 PR rtl-optimization/60116
4096 * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
4097 other_insn once the combination has been validated.
4099 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
4101 * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
4102 * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
4104 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
4106 * config/microblaze/microblaze.c: Extend mcpu version format
4108 2014-02-10 David Holsgrove <david.holsgrove@xilinx.com>
4110 * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
4112 2014-02-10 Richard Biener <rguenther@suse.de>
4114 Backport from mainline
4115 2014-01-30 Richard Biener <rguenther@suse.de>
4117 PR tree-optimization/59903
4118 * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
4121 2014-02-10 Richard Biener <rguenther@suse.de>
4123 PR tree-optimization/60115
4124 * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
4125 MEM_REF handling. Properly verify that the accesses are not
4126 out of the objects bound.
4128 2014-02-05 James Greenhalgh <james.greenhalgh@arm.com>
4130 Backport from mainline.
4131 2014-02-05 James Greenhalgh <james.greenhalgh@arm.com>
4134 * doc/invoke.texi (-march): Clarify documentation for ARM.
4138 2014-02-04 John David Anglin <danglin@gcc.gnu.org>
4141 * config/pa/pa.c (legitimize_tls_address): Return original address
4142 if not passed a SYMBOL_REF rtx.
4143 (hppa_legitimize_address): Call legitimize_tls_address for all TLS
4145 (pa_emit_move_sequence): Simplify TLS source operands.
4146 (pa_legitimate_constant_p): Reject all TLS constants.
4147 * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
4148 (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
4150 2014-02-04 Uros Bizjak <ubizjak@gmail.com>
4152 Backport from mainline
4153 2014-02-02 Uros Bizjak <ubizjak@gmail.com>
4156 * config/i386/i386.c (classify_argument): Fix handling of bit_offset
4157 when calculating size of integer atomic types.
4159 2014-02-02 Uros Bizjak <ubizjak@gmail.com>
4161 Backport from mainline
4162 2014-01-30 Jakub Jelinek <jakub@redhat.com>
4164 * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
4166 2014-01-31 Richard Henderson <rth@redhat.com>
4169 * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
4170 until after else_eh is processed.
4172 2014-01-30 David Holsgrove <david.holsgrove@xilinx.com>
4174 Backport from mainline
4175 * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
4176 comparison_operator with ordered_comparison_operator.
4178 2014-01-25 Walter Lee <walt@tilera.com>
4180 Backport from mainline
4181 2014-01-25 Walter Lee <walt@tilera.com>
4183 * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
4184 avoid clobbering a live register.
4186 2014-01-25 Walter Lee <walt@tilera.com>
4188 Backport from mainline
4189 2014-01-25 Walter Lee <walt@tilera.com>
4191 * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
4192 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
4193 * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
4194 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
4196 2014-01-25 Walter Lee <walt@tilera.com>
4198 Backport from mainline
4199 2014-01-25 Walter Lee <walt@tilera.com>
4201 * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
4202 insns before bundling.
4203 * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
4205 2014-01-25 Walter Lee <walt@tilera.com>
4207 Backport from mainline
4208 2014-01-25 Walter Lee <walt@tilera.com>
4210 * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
4211 PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
4212 * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
4214 2014-01-25 Walter Lee <walt@tilera.com>
4216 Backport from mainline
4217 2014-01-25 Walter Lee <walt@tilera.com>
4219 * config/tilepro/tilepro.md (ctzdi2): Use register_operand
4224 2014-01-25 Walter Lee <walt@tilera.com>
4226 Backport from mainline
4227 2014-01-25 Walter Lee <walt@tilera.com>
4229 * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
4230 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4232 2014-01-24 H.J. Lu <hongjiu.lu@intel.com>
4234 Backport from mainline
4235 2014-01-23 H.J. Lu <hongjiu.lu@intel.com>
4238 * config/i386/i386.md (pushsf splitter): Get stack adjustment
4239 from push operand if code of push isn't PRE_DEC.
4241 2014-01-23 David Holsgrove <david.holsgrove@xilinx.com>
4243 Backport from mainline.
4244 * config/microblaze/microblaze.md: Add trap insn and attribute
4246 2014-01-23 Marek Polacek <polacek@redhat.com>
4248 Backport from mainline
4249 2013-10-21 Marek Polacek <polacek@redhat.com>
4252 * fold-const.c (fold_range_test): Return 0 if the type is not
4255 2014-01-22 David Holsgrove <david.holsgrove@xilinx.com>
4257 * config/microblaze/microblaze.md: Correct bswaphi2 insn.
4259 2014-01-22 Uros Bizjak <ubizjak@gmail.com>
4261 Backport from mainline
4262 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
4264 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4265 for SImode_address_operand operands, having only a REG argument.
4267 2014-01-20 Jakub Jelinek <jakub@redhat.com>
4270 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4271 if operands[1] is a REG or ZERO_EXTEND of a REG.
4273 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
4274 H.J. Lu <hongjiu.lu@intel.com>
4277 * config/i386/i386.md (*lea<mode>): Zero-extend return register
4278 to DImode for zero-extended addresses.
4280 2014-01-21 Andrew Pinski <apinski@cavium.com>
4281 Steve Ellcey <sellcey@mips.com>
4284 * config/mips/mips.c (mips_print_operand): Check operand mode instead
4287 2014-01-21 Andrey Belevantsev <abel@ispras.ru>
4289 Backport from mainline
4290 2013-12-23 Andrey Belevantsev <abel@ispras.ru>
4292 PR rtl-optimization/57422
4293 * sel-sched.c (mark_unavailable_hard_regs): Fix typo when calling
4294 add_to_hard_reg_set.
4296 2014-01-20 Jakub Jelinek <jakub@redhat.com>
4299 * tree.h (fold_builtin_strcat): New prototype.
4300 * builtins.c (fold_builtin_strcat): No longer static. Add len
4301 argument, if non-NULL, don't call c_strlen. Optimize
4302 directly into __builtin_memcpy instead of __builtin_strcpy.
4303 (fold_builtin_2): Adjust fold_builtin_strcat caller.
4304 * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
4306 2014-01-20 Richard Biener <rguenther@suse.de>
4309 * builtins.c (fold_builtin_strcat): Remove case better handled
4310 by tree-ssa-strlen.c.
4312 2014-01-19 John David Anglin <danglin@gcc.gnu.org>
4314 * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
4315 long non-pic millicode calls.
4317 2014-01-17 John David Anglin <danglin@gcc.gnu.org>
4319 * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
4320 call to $$dyncall when TARGET_LONG_CALLS is true.
4322 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
4324 Backport from mainline
4325 2014-01-14 H.J. Lu <hongjiu.lu@intel.com>
4328 * config/i386/i386.c (type_natural_mode): Add a bool parameter
4329 to indicate if type is used for function return value. Warn
4330 ABI change if the vector mode isn't available for function
4332 (ix86_function_arg_advance): Pass false to type_natural_mode.
4333 (ix86_function_arg): Likewise.
4334 (ix86_gimplify_va_arg): Likewise.
4335 (function_arg_32): Don't warn ABI change.
4336 (ix86_function_value): Pass true to type_natural_mode.
4337 (ix86_return_in_memory): Likewise.
4338 (ix86_struct_value_rtx): Removed.
4339 (TARGET_STRUCT_VALUE_RTX): Likewise.
4341 2014-01-17 Charles Baylis <charles.baylis@linaro.org>
4343 Backport from mainline
4344 2013-12-19 Charles Baylis <charles.baylis@linaro.org>
4347 * config/arm/arm-ldmstm.ml: Use low_register_operand for Thumb
4349 * config/arm/ldmstm.md: Regenerate.
4351 2013-12-19 Charles Baylis <charles.baylis@linaro.org>
4354 * config/arm/predicates.md (arm_hard_general_register_operand):
4356 (arm_hard_register_operand): Remove.
4357 * config/arm/arm-ldmstm.ml: Use arm_hard_general_register_operand
4359 * config/arm/ldmstm.md: Regenerate.
4361 2013-12-19 Charles Baylis <charles.baylis@linaro.org>
4364 * config/arm/predicates.md (vfp_hard_register_operand): New predicate.
4365 * config/arm/arm.md (vfp_pop_multiple_with_writeback): Use
4366 vfp_hard_register_operand.
4368 2014-01-17 Kugan Vivekanandarajah <kuganv@linaro.org>
4370 Backport from mainline
4371 2014-01-15 Matthew Gretton-Dann <matthew.gretton-dann@linaro.org>
4372 Kugan Vivekanandarajah <kuganv@linaro.org>
4375 * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
4378 2014-01-17 Terry Guo <terry.guo@arm.com>
4381 * config/arm/arm.md (prefetch): Set insn type attribute to load1.
4383 2014-01-16 Jakub Jelinek <jakub@redhat.com>
4386 * config/i386/i386.c (ix86_expand_builtin): If target doesn't
4387 satisfy operand 0 predicate for gathers, use a new pseudo as
4390 2014-01-16 Richard Henderson <rth@redhat.com>
4393 * reginfo.c (global_regs_decl): Globalize.
4394 * rtl.h (global_regs_decl): Declare.
4395 * ira.c (do_reload): Diagnose frame_pointer_needed and it
4396 reserved via global_regs.
4398 2014-01-16 Peter Bergner <bergner@vnet.ibm.com>
4400 Backport from mainline
4401 2014-01-15 Uros Bizjak <ubizjak@gmail.com>
4403 * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
4404 VALID_AVX256_REG_OR_OI_MODE.
4406 2013-09-05 Peter Bergner <bergner@vnet.ibm.com>
4409 * reginfo.c (choose_hard_reg_mode): Scan through all mode classes
4410 looking for widest mode.
4412 2014-01-16 Marek Polacek <polacek@redhat.com>
4414 Backported from mainline
4415 2014-01-16 Marek Polacek <polacek@redhat.com>
4418 * gimple-low.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
4419 it is error_mark_node.
4421 2014-01-15 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
4424 * config/s390/s390.c (s390_preferred_reload_class): Don't return
4425 ADDR_REGS for invalid symrefs in non-PIC code.
4427 2014-01-14 Uros Bizjak <ubizjak@gmail.com>
4430 2014-01-08 Uros Bizjak <ubizjak@gmail.com>
4432 * config/i386/i386.c (ix86_data_alignment): Calculate max_align
4433 from prefetch_block tune setting.
4435 2014-01-13 Jakub Jelinek <jakub@redhat.com>
4437 Backported from mainline
4438 2014-01-10 Jakub Jelinek <jakub@redhat.com>
4440 PR tree-optimization/59745
4441 * tree-predcom.c (tree_predictive_commoning_loop): Call
4442 free_affine_expand_cache if giving up because components is NULL.
4444 2014-01-10 Yufeng Zhang <yufeng.zhang@arm.com>
4446 * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
4447 with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
4448 rtx is const0_rtx or not.
4450 2014-01-10 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
4452 * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
4453 extraction in good case.
4455 2014-01-10 Huacai Chen <chenhc@lemote.com>
4457 * config/mips/driver-native.c (host_detect_local_cpu): Handle new
4458 kernel strings for Loongson-2E/2F/3A.
4460 2014-01-10 Richard Biener <rguenther@suse.de>
4462 PR tree-optimization/59715
4463 * tree-flow.h (split_critical_edges): Declare.
4464 * tree-cfg.c (split_critical_edges): Export.
4465 * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4467 2014-01-09 Richard Sandiford <rdsandiford@googlemail.com>
4469 * config/mips/mips.h (ISA_HAS_WSBH): Define.
4470 * config/mips/mips.md (UNSPEC_WSBH, UNSPEC_DSBH, UNSPEC_DSHD): New
4472 (bswaphi2, bswapsi2, bswapdi2, wsbh, dsbh, dshd): New patterns.
4474 2014-01-09 Richard Sandiford <rdsandiford@googlemail.com>
4476 PR rtl-optimization/59137
4477 * reorg.c (steal_delay_list_from_target): Call update_block for
4479 (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
4481 2014-01-09 Richard Sandiford <rdsandiford@googlemail.com>
4484 2012-10-07 Richard Sandiford <rdsandiford@googlemail.com>
4486 * config/mips/mips.c (mips_truncated_op_cost): New function.
4487 (mips_rtx_costs): Adjust test for BADDU.
4488 * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
4490 2012-10-02 Richard Sandiford <rdsandiford@googlemail.com>
4492 * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
4493 (*baddu_si): ...this new pattern.
4495 2014-01-09 Richard Biener <rguenther@suse.de>
4497 Backport from mainline
4498 2013-11-18 Richard Biener <rguenther@suse.de>
4500 PR tree-optimization/59125
4501 PR tree-optimization/54570
4502 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): When inlining
4503 is not complete do not treat component-references with offset zero
4504 but different fields as equal.
4505 * tree-object-size.c: Include tree-phinodes.h and ssa-iterators.h.
4506 (compute_object_sizes): Apply TLC. Propagate the constant
4507 results into all uses and fold their stmts.
4508 * passes.def (pass_all_optimizations): Move pass_object_sizes
4509 after the first pass_forwprop and before pass_fre.
4511 2013-12-03 Jakub Jelinek <jakub@redhat.com>
4513 PR tree-optimization/59362
4514 * tree-object-size.c (object_sizes): Change into array of
4515 vec<unsigned HOST_WIDE_INT>.
4516 (compute_builtin_object_size): Check computed bitmap for
4517 non-NULL instead of object_sizes. Call safe_grow on object_sizes
4518 vector if new SSA_NAMEs appeared.
4519 (init_object_sizes): Check computed bitmap for non-NULL.
4520 Call safe_grow on object_sizes elements instead of initializing
4522 (fini_object_sizes): Call release on object_sizes elements, don't
4525 2014-01-09 Richard Earnshaw <rearnsha@arm.com>
4527 PR rtl-optimization/54300
4528 * regcprop.c (copyprop_hardreg_forward_1): Ensure any unused
4529 outputs in a single-set are killed from the value chains.
4531 2014-01-09 Jakub Jelinek <jakub@redhat.com>
4533 PR rtl-optimization/59724
4534 * ifcvt.c (cond_exec_process_if_block): Don't call
4535 flow_find_head_matching_sequence with 0 longest_match.
4536 * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4537 non-active insns if !stop_after.
4538 (try_head_merge_bb): Revert 2014-01-07 changes.
4540 2014-01-09 Hans-Peter Nilsson <hp@axis.com>
4542 Backport from mainline
4543 2013-12-23 Hans-Peter Nilsson <hp@axis.com>
4546 * config/cris/predicates.md (cris_nonsp_register_operand):
4547 New define_predicate.
4548 * config/cris/cris.md: Replace register_operand with
4549 cris_nonsp_register_operand for destinations in all
4550 define_splits where a register is set more than once.
4552 2014-01-08 H.J. Lu <hongjiu.lu@intel.com>
4554 Backport from mainline
4555 2013-12-25 H.J. Lu <hongjiu.lu@intel.com>
4558 * config/i386/i386.c (struct ptt): Add a field for processor name.
4559 (processor_target_table): Sync with processor_type. Add
4561 (cpu_names): Removed.
4562 (ix86_option_override_internal): Default x_ix86_tune_string
4563 to processor_target_table[TARGET_CPU_DEFAULT].name.
4564 (ix86_function_specific_print): Assert arch and tune <
4565 PROCESSOR_max. Use processor_target_table to print arch and
4567 * config/i386/i386.h (TARGET_CPU_DEFAULT): Default to
4568 PROCESSOR_GENERIC32.
4569 (target_cpu_default): Removed.
4570 (processor_type): Reordered.
4572 2014-01-08 Uros Bizjak <ubizjak@gmail.com>
4574 Backport from mainline
4575 2014-01-05 Uros Bizjak <ubizjak@gmail.com>
4577 * config/i386/i386.c (ix86_data_alignment): Calculate max_align
4578 from prefetch_block tune setting.
4579 (nocona_cost): Correct size of prefetch block to 64.
4581 2014-01-08 Martin Jambor <mjambor@suse.cz>
4584 * ipa-prop.c (ipa_compute_jump_functions): Bail out if not optimizing.
4585 (parm_preserved_before_stmt_p): Assume modification present when not
4588 2014-01-07 John David Anglin <danglin@gcc.gnu.org>
4591 * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
4592 for 14-bit register offsets when INT14_OK_STRICT is false.
4594 2014-01-07 Roland Stigge <stigge@antcom.de>
4595 Michael Meissner <meissner@linux.vnet.ibm.com>
4598 * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
4599 Only check TFmode for SPE constants. Don't check TImode or TDmode.
4601 2014-01-07 Jakub Jelinek <jakub@redhat.com>
4603 PR rtl-optimization/58668
4604 * cfgcleanup.c (flow_find_cross_jump): Don't count
4605 any jumps if dir_p is NULL. Remove p1 variable and make USE/CLOBBER
4606 check consistent with other places.
4607 (flow_find_head_matching_sequence): Don't count USE or CLOBBER insns.
4608 (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
4610 * ifcvt.c (count_bb_insns): Don't count USE or CLOBBER insns.
4612 2014-01-07 Mike Stump <mikestump@comcast.net>
4613 Jakub Jelinek <jakub@redhat.com>
4616 * tree.h (struct tree_optimization_option): Change optabs
4617 type from unsigned char * to void *.
4618 * optabs.c (init_tree_optimization_optabs): Adjust
4619 TREE_OPTIMIZATION_OPTABS initialization.
4621 2014-01-07 Jakub Jelinek <jakub@redhat.com>
4623 Backported from mainline
4624 2013-12-16 Jakub Jelinek <jakub@redhat.com>
4628 * gimple.h (walk_stmt_load_store_addr_fn): New typedef.
4629 (walk_stmt_load_store_addr_ops, walk_stmt_load_store_ops): Use it
4630 for callback params.
4631 * gimple.c (walk_stmt_load_store_ops): Likewise.
4632 (walk_stmt_load_store_addr_ops): Likewise. Adjust all callback
4633 calls to supply the gimple operand containing the base tree
4634 as an extra argument.
4635 * tree-ssa-ter.c (find_ssaname, find_ssaname_in_store): New helper
4637 (find_replaceable_in_bb): For calls or GIMPLE_ASM, only set
4638 same_root_var if USE is used somewhere in the stores of the stmt.
4639 * ipa-prop.c (visit_ref_for_mod_analysis): Remove name of the stmt
4640 argument and ATTRIBUTE_UNUSED, add another unnamed tree argument.
4641 * ipa-pure-const.c (check_load, check_store, check_ipa_load,
4642 check_ipa_store): Likewise.
4643 * gimple.c (gimple_ior_addresses_taken_1): Likewise.
4644 * ipa-split.c (test_nonssa_use, mark_nonssa_use): Likewise.
4645 (verify_non_ssa_vars, visit_bb): Adjust their callers.
4646 * cfgexpand.c (add_scope_conflicts_1): Use
4647 walk_stmt_load_store_addr_fn type for visit variable.
4648 (visit_op, visit_conflict): Remove name of the stmt
4649 argument and ATTRIBUTE_UNUSED, add another unnamed tree argument.
4650 * tree-sra.c (asm_visit_addr): Likewise. Remove name of the data
4651 argument and ATTRIBUTE_UNUSED.
4652 * cgraphbuild.c (mark_address, mark_load, mark_store): Add another
4653 unnamed tree argument.
4655 2014-01-03 Andreas Schwab <schwab@linux-m68k.org>
4657 * config/m68k/m68k.c (handle_move_double): Handle pushes with
4658 overlapping registers also for registers other than the stack
4661 2014-01-03 Jakub Jelinek <jakub@redhat.com>
4664 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
4667 2014-01-01 Jakub Jelinek <jakub@redhat.com>
4669 PR rtl-optimization/59647
4670 * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
4671 new_rtx into UNSIGNED_FLOAT rtxes.
4673 2013-12-28 Eric Botcazou <ebotcazou@adacore.com>
4675 * doc/invoke.texi (output file options): Document -fada-spec-parent.
4677 2013-12-26 Uros Bizjak <ubizjak@gmail.com>
4679 * config/i386/driver-i386.c (decode_caches_intel): Add missing entries.
4681 2013-12-20 Jakub Jelinek <jakub@redhat.com>
4684 Backported from mainline
4685 2013-08-19 Dehao Chen <dehao@google.com>
4687 * value-prof.c (gimple_ic): Fix the bug of adding EH edge.
4689 2013-12-19 James Greenhalgh <james.greenhalgh@arm.com>
4691 Backport from Mainline.
4692 2013-05-01 James Greenhalgh <james.greenhalgh@arm.com>
4694 * config/aarch64/aarch64-simd-builtins.def (cmhs): Rename to...
4696 (cmhi): Rename to...
4698 * config/aarch64/aarch64-simd.md
4699 (simd_mode): Add SF.
4700 (aarch64_vcond_internal): Use new names for unsigned comparison insns.
4701 (aarch64_cm<optab><mode>): Rewrite to not use UNSPECs.
4702 * config/aarch64/aarch64.md (*cstore<mode>_neg): Rename to...
4703 (cstore<mode>_neg): ...This.
4704 * config/aarch64/iterators.md
4706 (unspec): Remove UNSPEC_CM<EQ, LE, LT, GE, GT, HS, HI, TST>.
4708 (UCOMPARISONS): Likewise.
4709 (optab): Add missing comparisons.
4717 (V_cmp_result): Add DF, SF modes.
4718 (v_cmp_result): Likewise.
4721 * config/aarch64/predicates.md (aarch64_reg_or_fp_zero): New.
4723 Partial Backport from mainline.
4724 2013-05-01 James Greenhalgh <james.greenhalgh@arm.com>
4726 * config/aarch64/arm_neon.h
4727 (vc<eq, lt, le, gt, ge, tst><qsd>_<u><8,16,32,64>): Remap
4728 to builtins or C as appropriate.
4730 2013-12-19 Dominik Vogt <vogt@linux.vnet.ibm.com>
4731 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
4733 Backport from mainline
4734 2013-12-19 Dominik Vogt <vogt@linux.vnet.ibm.com>
4735 * config/s390/s390.c (s390_hotpatch_trampoline_halfwords_default): New
4737 (s390_hotpatch_trampoline_halfwords_max): New constant
4738 (s390_hotpatch_trampoline_halfwords): New static variable
4739 (get_hotpatch_attribute): New function
4740 (s390_handle_hotpatch_attribute): New function
4741 (s390_attribute_table): New target specific attribute table to implement
4742 the hotpatch attribute
4743 (s390_option_override): Parse hotpatch options
4744 (s390_function_num_hotpatch_trampoline_halfwords): New function
4745 (s390_can_inline_p): Implement target hook to
4746 suppress hotpatching for explicitly inlined functions
4747 (s390_asm_output_function_label): Generate hotpatch prologue
4748 (TARGET_ATTRIBUTE_TABLE): Define to implement target attribute table
4749 (TARGET_CAN_INLINE_P): Define to implement target hook
4750 * config/s390/s390.opt (mhotpatch): New options -mhotpatch, -mhotpatch=
4751 * config/s390/s390-protos.h (s390_asm_output_function_label): Add
4753 * config/s390/s390.h (ASM_OUTPUT_FUNCTION_LABEL): Target specific
4754 function label generation for hotpatching
4755 (FUNCTION_BOUNDARY): Align functions to eight bytes
4756 * doc/extend.texi: Document hotpatch attribute
4757 * doc/invoke.texi: Document -mhotpatch option
4759 2013-12-18 Eric Botcazou <ebotcazou@adacore.com>
4761 * config/arm/arm.c (arm_expand_epilogue_apcs_frame): Fix thinko.
4763 2013-12-12 Vladimir Makarov <vmakarov@redhat.com>
4766 * lra-coalesce.c (lra_coalesce): Invalidate inheritance pseudo
4767 values if necessary.
4769 2013-12-12 Jakub Jelinek <jakub@redhat.com>
4772 * gimplify.c (omp_check_private): Add copyprivate argument, if it
4773 is true, don't check omp_privatize_by_reference.
4774 (gimplify_scan_omp_clauses): For OMP_CLAUSE_COPYPRIVATE verify
4775 decl is private in outer context. Adjust omp_check_private caller.
4777 2013-12-10 Eric Botcazou <ebotcazou@adacore.com>
4779 PR rtl-optimization/58295
4780 * simplify-rtx.c (simplify_truncation): Restrict the distribution for
4781 WORD_REGISTER_OPERATIONS targets.
4783 2013-12-10 Kai Tietz <ktietz@redhat.com>
4786 * config/i386/i386.c (ix86_expand_prologue): Address saved
4787 registers stack-relative, not via frame-pointer.
4789 2013-12-09 Alan Modra <amodra@gmail.com>
4792 2013-12-05 Alan Modra <amodra@gmail.com>
4793 * configure.ac (BUILD_CXXFLAGS) Don't use ALL_CXXFLAGS for
4795 <recursive call for build != host>: Clear GMPINC. Don't bother
4797 * configure: Regenerate.
4799 2013-12-08 Uros Bizjak <ubizjak@gmail.com>
4801 Backport from mainline
4802 2013-12-06 Uros Bizjak <ubizjak@gmail.com>
4805 * config/i386/i386.c (type_natural_mode): Properly handle
4806 size 8 for !TARGET_64BIT.
4808 2013-12-07 Ralf Corsépius <ralf.corsepius@rtems.org>
4810 * config.gcc (microblaze*-*-rtems*): Add TARGET_BIG_ENDIAN_DEFAULT.
4812 2013-12-06 Jakub Jelinek <jakub@redhat.com>
4814 PR tree-optimization/59388
4815 * tree-ssa-reassoc.c (update_range_test): If op == range->exp,
4816 gimplify tem after stmt rather than before it.
4818 2013-12-06 Oleg Endo <olegendo@gcc.gnu.org>
4820 Backport from mainline
4821 2013-11-26 Oleg Endo <olegendo@gcc.gnu.org>
4825 * config/sh/sh.c (max_mov_insn_displacement, disp_addr_displacement):
4826 Prefix function names with 'sh_'. Make them non-static.
4827 * config/sh/sh-protos.h (sh_disp_addr_displacement,
4828 sh_max_mov_insn_displacement): Add declarations.
4829 * config/sh/constraints.md (Q): Reject QImode.
4830 (Sdd): Use match_code "mem".
4831 (Snd): Fix erroneous matching of non-memory operands.
4832 * config/sh/predicates.md (short_displacement_mem_operand): New
4834 (general_movsrc_operand): Disallow PC relative QImode loads.
4835 * config/sh/sh.md (*mov<mode>_reg_reg): Remove it.
4836 (*movqi, *movhi): Merge both insns into...
4837 (*mov<mode>): ... this new insn. Replace generic 'm' constraints with
4838 'Snd' and 'Sdd' constraints. Calculate insn length dynamically based
4839 on the operand types.
4841 2013-12-06 Richard Biener <rguenther@suse.de>
4843 Backport from mainline
4844 2013-11-29 Richard Biener <rguenther@suse.de>
4846 PR tree-optimization/59334
4847 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Fix bug
4850 2013-11-28 Richard Biener <rguenther@suse.de>
4852 PR tree-optimization/59330
4853 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Simplify
4854 and fix delayed marking of free calls not necessary.
4856 2013-12-06 Richard Biener <rguenther@suse.de>
4858 Backport from mainline
4859 2013-11-27 Richard Biener <rguenther@suse.de>
4861 PR tree-optimization/59288
4862 * tree-vect-loop.c (get_initial_def_for_induction): Do not
4863 re-analyze the PHI but use STMT_VINFO_LOOP_PHI_EVOLUTION_PART.
4865 2013-11-19 Richard Biener <rguenther@suse.de>
4867 PR tree-optimization/59164
4868 * tree-vect-loop.c (vect_analyze_loop_operations): Adjust
4869 check whether we can create an epilogue loop to reflect the
4870 cases where we create one.
4872 2013-09-05 Richard Biener <rguenther@suse.de>
4874 PR tree-optimization/58137
4875 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size):
4876 Do not create vectors of pointers.
4877 * tree-vect-loop.c (get_initial_def_for_induction): Use proper
4878 types for the components of the vector initializer.
4879 * tree-cfg.c (verify_gimple_assign_binary): Remove special-casing
4880 allowing pointer vectors with PLUS_EXPR/MINUS_EXPR.
4882 2013-12-06 Oleg Endo <olegendo@gcc.gnu.org>
4886 * config/sh/sh.md (*cbranch_t): Check that there are no labels between
4887 the s1 insn and the testing insn. Remove REG_DEAD note from s1 insn.
4889 2013-12-05 Richard Biener <rguenther@suse.de>
4891 Backport from mainline
4892 2013-11-19 Richard Biener <rguenther@suse.de>
4895 * tree-ssa-ter.c (find_replaceable_in_bb): Avoid forwarding
4896 loads into stmts that may clobber it.
4898 2013-12-04 Jakub Jelinek <jakub@redhat.com>
4900 PR rtl-optimization/58726
4901 * combine.c (force_to_mode): Fix comment typo. Don't destructively
4902 modify x for ROTATE, ROTATERT and IF_THEN_ELSE.
4904 2013-12-04 Jakub Jelinek <jakub@redhat.com>
4905 Uros Bizjak <ubizjak@gmail.com>
4908 * config/i386/i386.c (ix86_legitimate_combined_insn): If for
4909 !TARGET_AVX there is misaligned MEM operand with vector mode
4910 and get_attr_ssememalign is 0, return false.
4911 (ix86_expand_special_args_builtin): Add get_pointer_alignment
4912 computed alignment and for non-temporal loads/stores also
4913 at least GET_MODE_ALIGNMENT as MEM_ALIGN.
4914 * config/i386/sse.md
4915 (<sse>_loadu<ssemodesuffix><avxsizesuffix>,
4916 <sse>_storeu<ssemodesuffix><avxsizesuffix>,
4917 <sse2>_loaddqu<avxsizesuffix>,
4918 <sse2>_storedqu<avxsizesuffix>, <sse3>_lddqu<avxsizesuffix>,
4919 sse_vmrcpv4sf2, sse_vmrsqrtv4sf2, sse2_cvtdq2pd, sse_movhlps,
4920 sse_movlhps, sse_storehps, sse_loadhps, sse_loadlps,
4921 *vec_interleave_highv2df, *vec_interleave_lowv2df,
4922 *vec_extractv2df_1_sse, sse2_loadhpd, sse2_loadlpd, sse2_movsd,
4923 sse4_1_<code>v8qiv8hi2, sse4_1_<code>v4qiv4si2,
4924 sse4_1_<code>v4hiv4si2, sse4_1_<code>v2qiv2di2,
4925 sse4_1_<code>v2hiv2di2, sse4_1_<code>v2siv2di2, sse4_2_pcmpestr,
4926 *sse4_2_pcmpestr_unaligned, sse4_2_pcmpestri, sse4_2_pcmpestrm,
4927 sse4_2_pcmpestr_cconly, sse4_2_pcmpistr, *sse4_2_pcmpistr_unaligned,
4928 sse4_2_pcmpistri, sse4_2_pcmpistrm, sse4_2_pcmpistr_cconly): Add
4929 ssememalign attribute.
4930 * config/i386/i386.md (ssememalign): New define_attr.
4932 2013-12-03 Jakub Jelinek <jakub@redhat.com>
4935 * gimplify.c (nonlocal_vla_vars): New variable.
4936 (gimplify_var_or_parm_decl): Put VAR_DECLs for VLAs into
4937 nonlocal_vla_vars chain.
4938 (gimplify_body): Call declare_vars on nonlocal_vla_vars chain
4939 if outer_bind has DECL_INITIAL (current_function_decl) block.
4942 * optabs.c (emit_conditional_move): Save and restore
4943 pending_stack_adjust and stack_pointer_delta if cmove can't be used.
4945 2013-12-02 Jakub Jelinek <jakub@redhat.com>
4947 PR tree-optimization/59358
4948 * tree-vrp.c (union_ranges): To check for the partially
4949 overlapping ranges or adjacent ranges, also compare *vr0max
4952 2013-12-02 Richard Biener <rguenther@suse.de>
4954 PR tree-optimization/59139
4955 * tree-ssa-loop-niter.c (chain_of_csts_start): Properly match
4956 code in get_val_for.
4957 (get_val_for): Use gcc_checking_asserts.
4959 2013-11-27 Tom de Vries <tom@codesourcery.com>
4960 Marc Glisse <marc.glisse@inria.fr>
4963 * fold-const.c (fold_indirect_ref_1): Don't create out-of-bounds
4965 * gimplify.c (gimple_fold_indirect_ref): Same.
4967 2013-12-01 Eric Botcazou <ebotcazou@adacore.com>
4969 * config/i386/winnt.c (i386_pe_asm_named_section): Be prepared for an
4972 2013-12-01 Bernd Edlinger <bernd.edlinger@hotmail.de>
4974 * expr.c (emit_group_store): Fix off-by-one BITFIELD_END argument.
4976 2013-11-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4978 Backport from mainline
4979 2013-11-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4981 * config/arm/iterators.md (vrint_conds): New int attribute.
4982 * config/arm/vfp.md (<vrint_pattern><SDF:mode>2): Set conds attribute.
4983 (smax<mode>3): Likewise.
4984 (smin<mode>3): Likewise.
4986 2013-11-28 Uros Bizjak <ubizjak@gmail.com>
4988 Backport from mainline
4989 2013-11-27 Uros Bizjak <ubizjak@gmail.com>
4990 Ganesh Gopalasubramanian <Ganesh.Gopalasubramanian@amd.com>
4993 * gcc.target/i386/xop-frczX.c: New test.
4995 2013-11-28 Terry Guo <terry.guo@arm.com>
4997 Backport mainline r205391
4998 2013-11-26 Terry Guo <terry.guo@arm.com>
5000 * config/arm/arm.c (require_pic_register): Handle high pic base
5001 register for thumb-1.
5002 (arm_load_pic_register): Also initialize high pic base register.
5003 * doc/invoke.texi: Update documentation for option -mpic-register.
5005 2013-11-27 Jakub Jelinek <jakub@redhat.com>
5007 Backported from mainline
5008 2013-11-26 Jakub Jelinek <jakub@redhat.com>
5010 PR tree-optimization/59014
5011 * tree-vrp.c (register_edge_assert_for_1): Don't look
5012 through conversions from non-integral types or through
5013 narrowing conversions.
5015 2013-11-27 Eric Botcazou <ebotcazou@adacore.com>
5018 * expr.c (emit_group_store): Don't write past the end of the structure.
5019 (store_bit_field): Fix formatting.
5021 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
5023 Backport from mainline
5024 2013-09-17 Sebastian Huber <sebastian.huber@embedded-brains.de>
5026 * config/sparc/t-rtems: Add leon3 multilibs.
5028 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
5030 Backport from mainline
5031 2013-08-09 Eric Botcazou <ebotcazou@adacore.com>
5033 * configure.ac: Add GAS check for LEON instructions on SPARC.
5034 * configure: Regenerate.
5035 * config.in: Likewise.
5036 * config.gcc (with_cpu): Remove sparc-leon*-* and deal with LEON in the
5038 * config/sparc/sparc.opt (LEON, LEON3): New masks.
5039 * config/sparc/sparc.h (ASM_CPU32_DEFAULT_SPEC): Set to AS_LEON_FLAG
5041 (ASM_CPU_SPEC): Pass AS_LEON_FLAG if -mcpu=leon or -mcpu=leon3.
5042 (AS_LEON_FLAG): New macro.
5043 * config/sparc/sparc.c (sparc_option_override): Set MASK_LEON for leon
5044 and MASK_LEON3 for leon3 and unset them if HAVE_AS_LEON is not defined.
5045 Deal with LEON and LEON3 for the memory model.
5046 * config/sparc/sync.md (atomic_compare_and_swap<mode>): Enable if LEON3
5047 (atomic_compare_and_swap<mode>_1): Likewise.
5048 (*atomic_compare_and_swap<mode>_1): Likewise.
5050 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
5052 Backport from mainline
5053 2013-07-23 Eric Botcazou <ebotcazou@adacore.com>
5055 * doc/invoke.texi (SPARC Options): Document new leon3 processor value.
5057 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
5059 Backport from mainline
5060 2013-07-22 Eric Botcazou <ebotcazou@adacore.com>
5062 * config.gcc (sparc*-*-*): Accept leon3 processor.
5063 (sparc-leon*-*): Merge with sparc*-*-* and add leon3 support.
5064 * doc/invoke.texi (SPARC Options): Adjust -mfix-ut699 entry.
5065 * config/sparc/sparc-opts.h (enum processor_type): Add PROCESSOR_LEON3.
5066 * config/sparc/sparc.opt (enum processor_type): Add leon3.
5067 (mfix-ut699): Adjust comment.
5068 * config/sparc/sparc.h (TARGET_CPU_leon3): New define.
5069 (CPP_CPU32_DEFAULT_SPEC): Add leon3 support.
5070 (CPP_CPU_SPEC): Likewise.
5071 (ASM_CPU_SPEC): Likewise.
5072 * config/sparc/sparc.c (leon3_cost): New constant.
5073 (sparc_option_override): Add leon3 support.
5074 (mem_ref): New function.
5075 (sparc_gate_work_around_errata): Return true if -mfix-ut699 is enabled.
5076 (sparc_do_work_around_errata): Look into the instruction in the delay
5077 slot and adjust accordingly. Add fix for the data cache nullify issues
5078 of the UT699. Change insertion position for the NOP.
5079 * config/sparc/leon.md (leon_fpalu, leon_fpmds, write_buf): Delete.
5080 (leon3_load): New reservation.
5081 (leon_store): Bump latency to 2.
5082 (grfpu): New automaton.
5083 (grfpu_alu): New unit.
5084 (grfpu_ds): Likewise.
5085 (leon_fp_alu): Adjust.
5086 (leon_fp_mult): Delete.
5087 (leon_fp_div): Split into leon_fp_divs and leon_fp_divd.
5088 (leon_fp_sqrt): Split into leon_fp_sqrts and leon_fp_sqrtd.
5089 * config/sparc/sparc.md (cpu): Add leon3.
5090 * config/sparc/sync.md (atomic_exchangesi): Disable if -mfix-ut699.
5092 (atomic_test_and_set): Likewise.
5095 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
5097 Backport from mainline
5098 2013-04-10 Steven Bosscher <steven@gcc.gnu.org>
5100 * config/sparc/sparc.c: Include tree-pass.h.
5101 (TARGET_MACHINE_DEPENDENT_REORG): Do not redefine.
5102 (sparc_reorg): Rename to sparc_do_work_around_errata. Move to
5103 head of file. Change return type. Split off gate function.
5104 (sparc_gate_work_around_errata): New function.
5105 (pass_work_around_errata): New pass definition.
5106 (insert_pass_work_around_errata) New pass insert definition to
5107 insert pass_work_around_errata just after delayed-branch scheduling.
5108 (sparc_option_override): Insert the pass.
5109 * config/sparc/t-sparc (sparc.o): Add TREE_PASS_H dependence.
5111 2013-11-24 Sebastian Huber <sebastian.huber@embedded-brains.de>
5113 Backport from mainline
5114 2013-05-28 Eric Botcazou <ebotcazou@adacore.com>
5116 * doc/invoke.texi (SPARC Options): Document -mfix-ut699.
5117 * builtins.c (expand_builtin_mathfn) <BUILT_IN_SQRT>: Try to widen the
5118 mode if the instruction isn't available in the original mode.
5119 * config/sparc/sparc.opt (mfix-ut699): New option.
5120 * config/sparc/sparc.md (muldf3_extend): Disable if -mfix-ut699.
5121 (divdf3): Turn into expander.
5122 (divdf3_nofix): New insn.
5123 (divdf3_fix): Likewise.
5124 (divsf3): Disable if -mfix-ut699.
5125 (sqrtdf2): Turn into expander.
5126 (sqrtdf2_nofix): New insn.
5127 (sqrtdf2_fix): Likewise.
5128 (sqrtsf2): Disable if -mfix-ut699.
5130 2013-11-22 Eric Botcazou <ebotcazou@adacore.com>
5132 * print-rtl.c (print_rtx) <case MEM>: Output a space if no MEM_EXPR.
5134 2013-11-21 Zhenqiang Chen <zhenqiang.chen@linaro.org>
5137 Backport from mainline: r197467 and r198999.
5138 2013-04-03 Jeff Law <law@redhat.com>
5140 * Makefile.in (lra-constraints.o): Depend on $(OPTABS_H).
5141 (lra-eliminations.o): Likewise.
5143 2013-05-16 Jeff Law <law@redhat.com>
5145 * Makefile.in (tree-switch-conversion.o): Depend on $(OPTABS_H).
5147 2013-11-20 Eric Botcazou <ebotcazou@adacore.com>
5150 * config/sparc/sparc.c (sparc_fold_builtin) <case CODE_FOR_pdist_vis>:
5151 Make sure neg2_ovf is set before being used.
5153 2013-11-20 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5154 Dominik Vogt <vogt@linux.vnet.ibm.com>
5156 Backport from mainline
5157 * config/s390/s390.c (s390_canonicalize_comparison): Don't fold
5158 int comparisons with an out of range condition code.
5159 (s390_optimize_nonescaping_tx): Skip empty BBs.
5160 Generate the new tbegin RTX when removing the FPR clobbers (with
5162 (s390_expand_tbegin): Fix the retry loop counter. Copy CC to the
5163 result before doing the retry calculations.
5164 (s390_init_builtins): Make tbegin "returns_twice" and tabort
5166 * config/s390/s390.md (UNSPECV_TBEGIN_TDB): New constant used for
5167 the TDB setting part of an tbegin.
5168 ("tbegin_1", "tbegin_nofloat_1"): Add a set for the TDB.
5169 ("tx_assist"): Set unused argument to an immediate zero instead of
5170 loading zero into a GPR and pass it as argument.
5171 * config/s390/htmxlintrin.h (__TM_simple_begin, __TM_begin):
5172 Remove inline and related attributes.
5173 (__TM_nesting_depth, __TM_is_user_abort, __TM_is_named_user_abort)
5174 (__TM_is_illegal, __TM_is_footprint_exceeded)
5175 (__TM_is_nested_too_deep, __TM_is_conflict): Fix format value
5178 2013-11-19 Uros Bizjak <ubizjak@gmail.com>
5180 Backport from mainline
5181 2013-11-18 Uros Bizjak <ubizjak@gmail.com>
5183 * config/i386/i386.c (ix86_decompose_address): Use REG_P instead of
5184 ix86_address_subreg_operand. Move subreg checks to
5185 ix86_validate_address_register. Move address override check to
5186 ix86_legitimate_address_p.
5187 (ix86_validate_address_register): New function.
5188 (ix86_legitimate_address_p): Call ix86_validate_address_register
5189 to validate base and index registers. Add address override check
5190 from ix86_decompose_address.
5191 (ix86_decompose_address): Remove.
5193 Backport from mainline
5194 2013-11-17 Uros Bizjak <ubizjak@gmail.com>
5197 * config/i386/i386.c (ix86_address_subreg_operand): Do not
5198 reject non-integer subregs.
5199 (ix86_decompose_address): Do not reject invalid CONST_INT RTXes.
5200 Move check for invalid x32 constant addresses ...
5201 (ix86_legitimate_address_p): ... here.
5203 2013-11-19 Richard Biener <rguenther@suse.de>
5205 Backport from mainline
5206 2013-11-07 Richard Biener <rguenther@suse.de>
5208 * tree-dfa.c (get_ref_base_and_extent): Fix casting.
5210 2013-11-19 Richard Biener <rguenther@suse.de>
5212 PR tree-optimization/57517
5213 * tree-predcom.c (combinable_refs_p): Verify the combination
5214 is always executed when the refs are.
5216 2013-11-19 Richard Biener <rguenther@suse.de>
5218 Backport from mainline
5219 2013-11-05 Richard Biener <rguenther@suse.de>
5222 * tree-dfa.c (get_ref_base_and_extent): Merge common code
5223 in MEM_REF and TARGET_MEM_REF handling. Make sure to
5224 process trailing array detection before diving into the
5225 view-converted object (and possibly apply some extra offset).
5227 2013-11-18 Richard Biener <rguenther@suse.de>
5229 Backport from mainline
5230 2013-10-21 Richard Biener <rguenther@suse.de>
5232 PR tree-optimization/58794
5233 * fold-const.c (operand_equal_p): Compare FIELD_DECL operand
5234 of COMPONENT_REFs with OEP_CONSTANT_ADDRESS_OF left in place.
5236 2013-10-21 Richard Biener <rguenther@suse.de>
5239 * fold-const.c (fold_binary_loc): Fold ((T) (X /[ex] C)) * C
5240 to (T) X for sign-changing conversions (or no conversion).
5242 2013-11-06 Richard Biener <rguenther@suse.de>
5244 PR tree-optimization/58653
5245 * tree-predcom.c (ref_at_iteration): Rewrite to generate
5247 (prepare_initializers_chain): Adjust.
5249 PR tree-optimization/59047
5250 * tree-predcom.c (ref_at_iteration): Handle bitfield accesses
5253 2013-10-15 Richard Biener <rguenther@suse.de>
5255 PR tree-optimization/58143
5256 * tree-ssa-loop-im.c (arith_code_with_undefined_signed_overflow):
5258 (rewrite_to_defined_overflow): Likewise.
5259 (move_computations_dom_walker::before_dom): Rewrite stmts
5260 with undefined signed overflow that are not always executed
5261 into unsigned arithmetic.
5263 2013-11-14 Uros Bizjak <ubizjak@gmail.com>
5265 Backport from mainline
5266 2013-11-10 Uros Bizjak <ubizjak@gmail.com>
5268 * mode-switching.c (optimize_mode_switching): Mark block as
5269 nontransparent, if last_mode at block exit is different from no_mode.
5271 Backport from mainline
5272 2013-11-06 Uros Bizjak <ubizjak@gmail.com>
5275 * config/i386/i386.c (ix86_avx_u128_mode_needed): Require
5276 AVX_U128_DIRTY mode for call_insn RTXes that use AVX256 registers.
5277 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY mode for call_insn
5278 RTXes that return in AVX256 register.
5280 2013-11-14 Jakub Jelinek <jakub@redhat.com>
5281 Uros Bizjak <ubizjak@gmail.com>
5284 * config/i386/i386.md (*anddi_2): Only allow CCZmode if
5285 operands[2] satisfies_constraint_Z that might have bit 31 set.
5287 2013-11-12 H.J. Lu <hongjiu.lu@intel.com>
5289 Backported from mainline
5290 2013-11-12 H.J. Lu <hongjiu.lu@intel.com>
5293 * config/i386/i386.c (initial_ix86_tune_features): Set
5294 X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL and
5295 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL for m_HASWELL.
5297 2013-11-11 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5299 Backported from mainline
5300 2013-10-30 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5303 * config/arm/arm.c (arm_expand_epilogue_apcs_frame): Emit blockage
5305 2013-11-11 Jakub Jelinek <jakub@redhat.com>
5307 Backported from mainline
5308 2013-11-06 Jakub Jelinek <jakub@redhat.com>
5311 * expr.c (get_bit_range): Handle *offset == NULL_TREE.
5312 (expand_assignment): If *bitpos is negative, set *offset
5313 and adjust *bitpos, so that it is not negative.
5315 2013-11-05 Jakub Jelinek <jakub@redhat.com>
5317 PR rtl-optimization/58997
5318 * loop-iv.c (iv_subreg): For IV_UNKNOWN_EXTEND, expect
5319 get_iv_value to be in iv->mode rather than iv->extend_mode.
5320 (iv_extend): Likewise. Otherwise, if iv->extend != extend,
5321 use lowpart_subreg on get_iv_value before calling simplify_gen_unary.
5322 * loop-unswitch.c (may_unswitch_on): Make sure op[i] is in the right
5325 2013-11-10 Karlson2k <k2k@narod.ru>
5326 Kai Tietz <ktietz@redhat.com>
5330 * configure.ac: Adding for exported symbols check
5331 and for rdynamic-check executable-extension.
5332 * configure: Regenerated.
5334 2013-11-07 H.J. Lu <hongjiu.lu@intel.com>
5337 * config/i386/i386.md (push peepholer/splitter): Use Pmode
5338 with stack_pointer_rtx.
5340 2013-11-05 Uros Bizjak <ubizjak@gmail.com>
5342 * config/i386/t-rtems (MULTILIB_MATCHES): Fix option typos.
5344 2013-11-05 Jakub Jelinek <jakub@redhat.com>
5346 PR tree-optimization/58984
5347 * ipa-prop.c (ipa_load_from_parm_agg_1): Add SIZE_P argument,
5348 set *SIZE_P if non-NULL on success.
5349 (ipa_load_from_parm_agg, ipa_analyze_indirect_call_uses): Adjust
5351 (ipcp_transform_function): Likewise. Punt if size of access
5352 is different from TYPE_SIZE on v->value's type.
5354 2013-11-03 H.J. Lu <hongjiu.lu@intel.com>
5356 Backport from mainline
5357 2013-10-12 H.J. Lu <hongjiu.lu@intel.com>
5360 * config/i386/i386.c (ix86_copy_addr_to_reg): New function.
5361 (ix86_expand_movmem): Replace copy_addr_to_reg with
5362 ix86_copy_addr_to_reg.
5363 (ix86_expand_setmem): Likewise.
5365 2013-10-29 Uros Bizjak <ubizjak@gmail.com>
5367 Backport from mainline
5368 2013-08-08 Richard Sandiford <rdsandiford@googlemail.com>
5370 PR rtl-optimization/58079
5371 * combine.c (combine_simplify_rtx): Avoid using SUBST if
5372 simplify_comparison has widened a comparison with an integer.
5374 2013-10-29 Martin Jambor <mjambor@suse.cz>
5377 Backport from mainline
5378 2013-05-09 Martin Jambor <mjambor@suse.cz>
5381 * gimple-fold.c (canonicalize_constructor_val): Call
5382 cgraph_get_create_real_symbol_node instead of cgraph_get_create_node.
5384 Backport from mainline
5385 2013-03-16 Jan Hubicka <jh@suse.cz>
5387 * cgraph.h (cgraph_get_create_real_symbol_node): Declare.
5388 * cgraph.c (cgraph_get_create_real_symbol_node): New function.
5389 * cgrpahbuild.c: Use cgraph_get_create_real_symbol_node instead
5390 of cgraph_get_create_node.
5391 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5393 2013-10-28 Tom de Vries <tom@codesourcery.com>
5395 * cfgexpand.c (gimple_expand_cfg): Remove test for parm_birth_insn.
5396 Don't commit insertions after NOTE_INSN_FUNCTION_BEG.
5398 2013-10-26 Uros Bizjak <ubizjak@gmail.com>
5400 Backport from mainline
5401 2013-10-22 Uros Bizjak <ubizjak@gmail.com>
5404 * config/i386/i386.c (put_condition_code) <case GTU, case LEU>:
5405 Remove CCCmode handling.
5406 <case LTU>: Return 'c' suffix for CCCmode.
5407 <case GEU>: Return 'nc' suffix for CCCmode.
5408 (ix86_cc_mode) <case GTU, case LEU>: Do not generate overflow checks.
5409 * config/i386/i386.md (*sub<mode>3_cconly_overflow): Remove.
5410 (*sub<mode>3_cc_overflow): Ditto.
5411 (*subsi3_zext_cc_overflow): Ditto.
5413 2013-10-26 Uros Bizjak <ubizjak@gmail.com>
5415 Backport from mainline
5416 2013-10-19 Uros Bizjak <ubizjak@gmail.com>
5419 * config/i386/i386.c (ix86_function_value_regno): Add DX_REG,
5420 ST1_REG and XMM1_REG for 32bit and 64bit targets. Also add DI_REG
5421 and SI_REG for 64bit SYSV ABI targets.
5423 2013-08-25 Richard Henderson <rth@twiddle.net>
5426 * optabs.c (maybe_emit_atomic_exchange): Use create_input_operand
5427 instead of create_convert_operand_to.
5428 (maybe_emit_sync_lock_test_and_set): Likewise.
5429 (expand_atomic_compare_and_swap): Likewise.
5430 (maybe_emit_compare_and_swap_exchange_loop): Don't convert_modes.
5432 2013-10-25 Eric Botcazou <ebotcazou@adacore.com>
5434 PR rtl-optimization/58831
5435 * alias.c (init_alias_analysis): At the beginning of each iteration, set
5436 the reg_seen[N] bit if static_reg_base_value[N] is non-null.
5438 2013-10-25 Eric Botcazou <ebotcazou@adacore.com>
5440 * recog.c (search_ofs): New static variable moved from...
5441 (peep2_find_free_register): ...here.
5442 (peephole2_optimize): Initialize it.
5444 2013-10-24 David Edelsohn <dje.gcc@gmail.com>
5446 Backport from mainline
5447 2013-10-23 David Edelsohn <dje.gcc@gmail.com>
5450 * config/rs6000/rs6000.md (mulsi3_internal1 and splitter): Add
5451 TARGET_32BIT final condition.
5452 (mulsi3_internal2 and splitter): Same.
5454 2013-10-23 Tom de Vries <tom@codesourcery.com>
5456 PR tree-optimization/58805
5457 * tree-ssa-tail-merge.c (stmt_local_def): Add gimple_vdef check.
5459 2013-10-23 Richard Biener <rguenther@suse.de>
5461 Backport from mainline
5462 2013-06-24 Richard Biener <rguenther@suse.de>
5464 PR tree-optimization/57488
5465 * tree-ssa-pre.c (insert): Clear NEW sets before each iteration.
5467 2013-10-16 Ganesh Gopalasubramanian <Ganesh.Gopalasubramanian@amd.com>
5469 Backport from mainline
5470 2013-10-16 Ganesh Gopalasubramanian
5471 <Ganesh.Gopalasubramanian@amd.com>
5473 * config/i386/i386.c (ix86_option_override_internal): Enable FMA4
5476 2013-10-16 Jakub Jelinek <jakub@redhat.com>
5478 * BASE-VER: Set to 4.8.3.
5479 * DEV-PHASE: Set to prerelease.
5481 2013-10-16 Release Manager
5483 * GCC 4.8.2 released.
5485 2013-10-12 James Greenhalgh <james.greenhalgh@arm.com>
5487 Backport from mainline.
5488 2013-10-12 James Greenhalgh <james.greenhalgh@arm.com>
5490 * config/aarch64/arm_neon.h
5491 (vtbx<1,3>_<psu>8): Fix register constriants.
5493 2013-10-10 Jakub Jelinek <jakub@redhat.com>
5496 * stmt.c (expand_asm_operands): Add FALLTHRU_BB argument,
5497 if any labels are in FALLTHRU_BB, use a special label emitted
5498 immediately after the asm goto insn rather than label_rtx
5500 (expand_asm_stmt): Adjust caller.
5501 * cfgrtl.c (commit_one_edge_insertion): Force splitting of
5502 edge if the last insn in predecessor is a jump with single successor,
5503 but it isn't simplejump_p.
5505 2013-10-09 Jakub Jelinek <jakub@redhat.com>
5507 Backport from mainline
5508 2013-09-26 Richard Biener <rguenther@suse.de>
5510 PR tree-optimization/58539
5511 * tree-vect-loop.c (vect_create_epilog_for_reduction): Honor
5512 the fact that debug statements are not taking part in loop-closed
5515 2013-10-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5517 * config/s390/s390.c (s390_register_info): Make the call-saved FPR
5518 loop to work also for 31bit ABI.
5519 Save the stack pointer for frame_size > 0.
5521 2013-10-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5523 * config/s390/s390.md ("tbegin", "tbegin_nofloat", "tbegin_retry")
5524 ("tbegin_retry_nofloat", "tend", "tabort", "tx_assist"): Remove
5525 constraint letters from expanders.
5526 ("tbegin_retry", "tbegin_retry_nofloat"): Change predicate of the
5527 retry count to general_operand.
5528 ("tabort"): Give operand 0 a mode.
5529 ("tabort_1"): Add mode and constraint letter for operand 0.
5530 * doc/extend.texi: Fix protoype of __builtin_non_tx_store.
5532 2013-10-04 Marcus Shawcroft <marcus.shawcroft@arm.com>
5534 Backport from mainline.
5537 * config/aarch64/aarch64.md (*add_<shift>_<mode>)
5538 (*add_<shift>_si_uxtw,*add_mul_imm_<mode>)
5539 (*sub_<shift>_<mode>)
5540 (*sub_<shift>_si_uxtw,*sub_mul_imm_<mode>, *sub_mul_imm_si_uxtw):
5541 Remove k constraint.
5543 2013-10-02 John David Anglin <danglin@gcc.gnu.org>
5545 * config.gcc (hppa*64*-*-linux*): Don't add pa/t-linux to tmake_file.
5547 2013-10-01 Jakub Jelinek <jakub@redhat.com>
5548 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5551 * config/s390/s390.c (s390_chunkify_start): Handle tablejump_p first,
5552 continue when done, for other jumps look through PARALLEL
5555 2013-09-30 Jakub Jelinek <jakub@redhat.com>
5558 * fold-const.c (fold_ternary_loc): For A < 0 : <sign bit of A> : 0
5559 optimization, punt if sign_bit_p looked through any zero extension.
5561 2013-09-27 Paulo Matos <pmatos@broadcom.com>
5563 Backport from mainline.
5566 2013-03-27 Richard Biener <rguenther@suse.de>
5568 PR tree-optimization/56716
5569 * tree-ssa-structalias.c (perform_var_substitution): Adjust
5570 dumping for ref nodes.
5572 2013-09-27 Paulo Matos <pmatos@broadcom.com>
5574 Backport from mainline.
5576 2013-09-27 Paulo Matos <pmatos@broadcom.com>
5578 * gcc.dg/pr58463.c: New test.
5580 2013-09-23 Eric Botcazou <ebotcazou@adacore.com>
5582 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Recurse on copy
5583 assignment statements.
5585 2013-09-23 Alan Modra <amodra@gmail.com>
5588 * config/rs6000/rs6000.md (bswapdi2_64bit): Disable for volatile mems.
5590 2013-09-23 Alan Modra <amodra@gmail.com>
5592 * config/rs6000/predicates.md (add_cint_operand): New.
5593 (reg_or_add_cint_operand, small_toc_ref): Use add_cint_operand.
5594 * config/rs6000/rs6000.md (largetoc_high_plus): Restrict offset
5595 using add_cint_operand.
5596 (largetoc_high_plus_aix): Likewise.
5597 * config/rs6000/rs6000.c (toc_relative_expr_p): Use add_cint_operand.
5599 2013-09-20 John David Anglin <danglin@gcc.gnu.org>
5602 * config/pa/pa.c (pa_option_override): Disable auto increment and
5603 decrement instructions until reload is completed.
5605 * config/pa/pa.md: In "scc" insn patterns, change output template to
5606 handle const0_rtx in reg_or_0_operand operands.
5608 2013-09-19 Jakub Jelinek <jakub@redhat.com>
5610 * omp-low.c (expand_omp_sections): Always pass len - 1 to
5611 GOMP_sections_start, even if !exit_reachable.
5613 2013-09-18 Richard Earnshaw <rearnsha@arm.com>
5615 * arm.c (arm_expand_prologue): Validate architecture supports
5616 LDRD/STRD before accepting tuning preferences.
5617 (arm_expand_epilogue): Likewise.
5619 2013-09-18 Daniel Morris <danielm@ecoscentric.com>
5620 Paolo Carlini <paolo.carlini@oracle.com>
5623 * doc/implement-cxx.texi: Fix references to the C++ standards.
5625 2013-09-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5627 PR tree-optimization/58088
5628 * fold-const.c (mask_with_trailing_zeros): New function.
5629 (fold_binary_loc): Make sure we don't recurse infinitely
5630 when the X in (X & C1) | C2 is a tree of the form (Y * K1) & K2.
5631 Use mask_with_trailing_zeros where appropriate.
5633 2013-09-14 John David Anglin <danglin@gcc.gnu.org>
5636 * config/pa/pa.c (pa_expand_prologue): Change mode in gen_rtx_POST_INC
5639 2013-09-13 Christian Bruel <christian.bruel@st.com>
5642 * config/sh/sh.md (mov<mode>_reg_reg): Allow memory reloads.
5644 2013-09-11 Andi Kleen <ak@linux.intel.com>
5646 Backport from mainline
5647 * doc/extend.texi: Use __atomic_store_n instead of
5648 __atomic_store in HLE example.
5650 2013-09-11 Andi Kleen <ak@linux.intel.com>
5652 Backport from mainline
5653 * doc/extend.texi: Dont use __atomic_clear in HLE
5656 2013-09-11 Andi Kleen <ak@linux.intel.com>
5658 Backport from mainline
5659 * doc/extend.texi: Document that __atomic_clear and
5660 __atomic_test_and_set should only be used with bool.
5662 2013-09-11 Richard Biener <rguenther@suse.de>
5665 * passes.c (init_optimization_passes): Split critical edges
5666 before late uninit warning pass in the -Og pipeline.
5668 2013-09-11 Jakub Jelinek <jakub@redhat.com>
5670 PR tree-optimization/58385
5671 * fold-const.c (build_range_check): If both low and high are NULL,
5672 use omit_one_operand_loc to preserve exp side-effects.
5674 2013-09-10 Richard Earnshaw <rearnsha@arm.com>
5677 * arm/vfp.md (combine_vcvt_f32_<FCVTI32typename>): Fix pattern to
5678 support conditional execution.
5679 (combine_vcvt_f64_<FCVTI32typename>): Likewise.
5681 2013-09-10 Jakub Jelinek <jakub@redhat.com>
5683 PR rtl-optimization/58365
5684 * cfgcleanup.c (merge_memattrs): Also clear MEM_READONLY_P
5685 resp. MEM_NOTRAP_P if they differ, or set MEM_VOLATILE_P if
5688 2013-09-09 Jakub Jelinek <jakub@redhat.com>
5690 PR tree-optimization/58364
5691 * tree-ssa-reassoc.c (init_range_entry): For BIT_NOT_EXPR on
5692 BOOLEAN_TYPE, only invert in_p and continue with arg0 if
5693 the current range can't be an unconditional true or false.
5695 2013-09-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5698 Backport from mainline
5699 2013-04-30 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
5701 * explow.c (plus_constant): Pass "mode" to immed_double_int_const.
5702 Use gen_int_mode rather than GEN_INT.
5704 2013-09-09 Richard Biener <rguenther@suse.de>
5706 Backport from mainline
5707 2013-08-27 Richard Biener <rguenther@suse.de>
5709 PR tree-optimization/57521
5710 * tree-if-conv.c (if_convertible_bb_p): Verify that at least
5711 one edge is non-critical.
5712 (find_phi_replacement_condition): Make sure to use a non-critical
5713 edge. Cleanup and remove old bug workarounds.
5714 (bb_postdominates_preds): Remove.
5715 (if_convertible_loop_p_1): Do not compute post-dominators.
5716 (combine_blocks): Do not free post-dominators.
5717 (main_tree_if_conversion): Likewise.
5719 2013-09-09 Richard Biener <rguenther@suse.de>
5721 Backport from mainline
5722 2013-09-03 Richard Biener <rguenther@suse.de>
5725 * fold-const.c (negate_expr_p): Fix division case.
5726 (negate_expr): Likewise.
5728 2013-09-09 Richard Biener <rguenther@suse.de>
5730 Backport from mainline
5731 2013-08-29 Richard Biener <rguenther@suse.de>
5733 PR tree-optimization/57685
5734 * tree-vrp.c (register_edge_assert_for_1): Recurse only for
5735 single-use operands to avoid exponential complexity.
5737 2013-09-09 Richard Biener <rguenther@suse.de>
5739 Backport from mainline
5740 2013-08-30 Richard Biener <rguenther@suse.de>
5742 PR tree-optimization/58223
5743 * tree-loop-distribution.c (has_anti_dependence): Rename to ...
5744 (has_anti_or_output_dependence): ... this and adjust to also
5745 look for output dependences.
5746 (mark_nodes_having_upstream_mem_writes): Adjust.
5747 (rdg_flag_uses): Likewise.
5749 2013-09-03 Richard Biener <rguenther@suse.de>
5751 Backport from mainline
5752 2013-08-29 Richard Biener <rguenther@suse.de>
5754 PR tree-optimization/58246
5755 * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Properly
5756 handle the dominance check inside a basic-block.
5758 2013-09-03 Richard Biener <rguenther@suse.de>
5760 Backport from mainline
5761 2013-08-30 Richard Biener <rguenther@suse.de>
5763 PR tree-optimization/58228
5764 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Do not
5765 allow invariant loads in nested loop vectorization.
5767 2013-09-03 Richard Biener <rguenther@suse.de>
5769 Backport from mainline
5770 2013-08-30 Richard Biener <rguenther@suse.de>
5772 PR tree-optimization/58010
5773 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
5774 assert that we have a loop-closed PHI.
5776 2013-09-01 Uros Bizjak <ubizjak@gmail.com>
5778 Backport from mainline
5779 2013-08-31 Uros Bizjak <ubizjak@gmail.com>
5781 * config/alpha/alpha.c (alpha_emit_conditional_move): Update
5782 "cmp" RTX before signed_comparison_operator check to account
5785 2013-09-01 John David Anglin <danglin@gcc.gnu.org>
5787 * config/pa/pa.md: Allow "const 0" operand 1 in "scc" insns.
5789 2013-08-30 Jakub Jelinek <jakub@redhat.com>
5791 PR tree-optimization/58277
5792 * tree-ssa-strlen.c (strlen_enter_block): If do_invalidate gave up
5793 after seeing too many stmts with vdef in between dombb and current
5794 bb, invalidate everything.
5796 2013-08-29 Oleg Endo <olegendo@gcc.gnu.org>
5798 Backport from mainline
5799 2013-08-05 Oleg Endo <olegendo@gcc.gnu.org>
5802 * recog.h (rtx (*insn_gen_fn) (rtx, ...)): Replace typedef with new
5804 * expr.c (move_by_pieces_1, store_by_pieces_2): Replace argument
5805 rtx (*) (rtx, ...) with insn_gen_fn.
5806 * genoutput.c (output_insn_data): Cast gen_? function pointers to
5807 insn_gen_fn::stored_funcptr. Add initializer braces.
5809 Backport from mainline
5810 2013-08-07 Oleg Endo <olegendo@gcc.gnu.org>
5813 * config/rs6000/rs6000.c (gen_2arg_fn_t): Remove typedef.
5814 (rs6000_emit_swdiv_high_precision, rs6000_emit_swdiv_low_precision,
5815 rs6000_emit_swrsqrt): Don't cast result of GEN_FCN to gen_2arg_fn_t.
5817 2013-08-29 Jakub Jelinek <jakub@redhat.com>
5819 Backported from mainline
5820 2013-05-27 Richard Biener <rguenther@suse.de>
5823 PR tree-optimization/57417
5824 * tree-ssa-sccvn.c (vn_reference_fold_indirect): Fix test
5826 (set_ssa_val_to): Compare addresses using
5827 get_addr_base_and_unit_offset.
5829 PR tree-optimization/57396
5830 * tree-affine.c (double_int_constant_multiple_p): Properly
5831 return false for val == 0 and div != 0.
5833 PR tree-optimization/57343
5834 * tree-ssa-loop-niter.c (number_of_iterations_ne_max): Do not
5835 use multiple_of_p if not TYPE_OVERFLOW_UNDEFINED.
5836 (number_of_iterations_cond): Do not build the folded tree.
5838 2013-08-28 Jakub Jelinek <jakub@redhat.com>
5841 * omp-low.c (copy_var_decl): Copy over TREE_NO_WARNING flag.
5843 2013-08-28 Richard Biener <rguenther@suse.de>
5845 Backport from mainline
5846 2013-06-24 Richard Biener <rguenther@suse.de>
5849 * passes.c (init_optimization_passes): Move pass_fold_builtins
5850 and pass_dce earlier with -Og.
5852 2013-08-28 Uros Bizjak <ubizjak@gmail.com>
5854 Backport from mainline
5855 2013-08-27 H.J. Lu <hongjiu.lu@intel.com>
5857 * config/i386/driver-i386.c (host_detect_local_cpu): Update
5858 Haswell processor detection.
5860 Backport from mainline
5861 2013-08-27 Christian Widmer <shadow@umbrox.de>
5864 * config/i386/driver-i386.c (host_detect_local_cpu): Add detection
5865 of Ivy Bridge and Haswell processors. Assume core-avx2 for unknown
5866 AVX2 capable processors.
5868 2013-08-23 Jakub Jelinek <jakub@redhat.com>
5871 * config/i386/x86-64.h (TARGET_SECTION_TYPE_FLAGS): Define.
5872 * config/i386/i386.c (x86_64_elf_section_type_flags): New function.
5874 PR tree-optimization/58209
5875 * tree-tailcall.c (find_tail_calls): Give up for pointer result types
5876 if m or a is non-NULL.
5878 2013-08-21 Richard Earnshaw <rearnsha@arm.com>
5881 * arm.c (aapcs_vfp_allocate): Decompose the argument if the
5882 suggested mode for the assignment isn't compatible with the
5885 2013-08-20 Alan Modra <amodra@gmail.com>
5888 * config/rs6000/rs6000.c (rs6000_emit_prologue): Correct ool_adjust.
5889 (rs6000_emit_epilogue): Likewise.
5891 2013-08-19 Peter Bergner <bergner@vnet.ibm.com>
5892 Jakub Jelinek <jakub@redhat.com>
5894 Backport from mainline
5895 * config/rs6000/dfp.md (*negtd2_fpr): Handle non-overlapping
5896 destination and source operands.
5898 2013-08-18 Jakub Jelinek <jakub@redhat.com>
5900 PR tree-optimization/58006
5901 * tree-parloops.c (take_address_of): Don't ICE if get_name
5903 (eliminate_local_variables_stmt): Remove clobber stmts.
5905 2013-08-16 Jakub Jelinek <jakub@redhat.com>
5907 PR tree-optimization/58164
5908 * gimple.c (walk_stmt_load_store_addr_ops): For visit_addr
5909 walk gimple_goto_dest of GIMPLE_GOTO.
5911 PR tree-optimization/58165
5912 * tree-call-cdce.c (shrink_wrap_one_built_in_call): If
5913 bi_call must be the last stmt in a bb, don't split_block, instead
5914 use fallthru edge from it and give up if there is none.
5915 Release conds vector when returning early.
5917 2013-08-15 David Given <dg@cowlark.com>
5919 Backport from mainline
5920 2013-04-26 Vladimir Makarov <vmakarov@redhat.com>
5922 * lra-constraints.c (process_alt_operands): Use #if HAVE_ATTR_enable
5925 2013-08-14 Jakub Jelinek <jakub@redhat.com>
5927 PR tree-optimization/58145
5928 * tree-sra.c (build_ref_for_offset): If prev_base has
5929 TREE_THIS_VOLATILE or TREE_SIDE_EFFECTS, propagate it to MEM_REF.
5931 2013-08-14 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
5933 * config/s390/htmxlintrin.h: Add file missing from last commit.
5934 * config/s390/htmintrin.h: Likewise.
5935 * config/s390/s390intrin.h: Likewise.
5937 2013-08-14 Uros Bizjak <ubizjak@gmail.com>
5939 Backport from mainline
5940 2013-08-13 Uros Bizjak <ubizjak@gmail.com>
5942 * config/i386/sse.md (*sse2_maskmovdqu): Emit addr32 prefix
5943 when Pmode != word_mode. Add length_address attribute.
5944 (sse3_monitor_<mode>): Merge from sse3_monitor and
5945 sse3_monitor64_<mode> insn patterns. Emit addr32 prefix when
5946 Pmode != word_mode. Update insn length attribute.
5947 * config/i386/i386.c (ix86_option_override_internal): Update
5948 ix86_gen_monitor selection for merged sse3_monitor insn.
5950 2013-08-14 Jakub Jelinek <jakub@redhat.com>
5951 Alexandre Oliva <aoliva@redhat.com>
5954 * config/i386/i386.c (ix86_delegitimize_address): For CM_MEDIUM_PIC
5955 and CM_LARGE_PIC ix86_cmodel fall thru into the -m32 code, handle
5956 there also UNSPEC_PLTOFF.
5958 2013-08-13 Jakub Jelinek <jakub@redhat.com>
5961 * asan.c (instrument_strlen_call): Fix typo in comment.
5962 Use char * type even for the lhs of POINTER_PLUS_EXPR.
5964 2013-08-13 Vladimir Makarov <vmakarov@redhat.com>
5966 Backport from mainline
5967 2013-06-06 Vladimir Makarov <vmakarov@redhat.com>
5969 PR rtl-optimization/57459
5970 * lra-constraints.c (update_ebb_live_info): Fix typo for operand
5971 type when setting live regs.
5973 2013-08-13 Marek Polacek <polacek@redhat.com>
5974 Jakub Jelinek <jakub@redhat.com>
5976 PR tree-optimization/57980
5977 * tree-tailcall.c (process_assignment): Return false
5978 when not dealing with integers or floats.
5980 2013-08-12 Andrew Haley <aph@redhat.com>
5982 Backport from mainline:
5983 * 2013-07-11 Andreas Schwab <schwab@suse.de>
5985 * config/aarch64/aarch64-linux.h (CPP_SPEC): Define.
5987 2013-08-13 Uros Bizjak <ubizjak@gmail.com>
5989 Backport from mainline
5990 2013-08-12 Perez Read <netfirewall@gmail.com>
5993 * config/i386/i386.md (*movabs<mode>_1): Add <ptrsize> PTR before
5994 operand 0 for intel asm alternative.
5995 (*movabs<mode>_2): Ditto for operand 1.
5997 2013-08-09 Zhenqiang Chen <zhenqiang.chen@linaro.org>
5999 Backport from mainline:
6000 2013-08-09 Zhenqiang Chen <zhenqiang.chen@linaro.org>
6002 * config/arm/neon.md (vcond): Fix floating-point vector
6003 comparisons against 0.
6005 2013-08-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6007 Backport from mainline:
6008 2013-08-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6010 * config/arm/neon.md (movmisalign<mode>): Disable when we
6011 don't allow unaligned accesses.
6012 (*movmisalign<mode>_neon_store): Likewise.
6013 (*movmisalign<mode>_neon_load): Likewise.
6014 (*movmisalign<mode>_neon_store): Likewise.
6015 (*movmisalign<mode>_neon_load): Likewise.
6017 2013-08-06 Martin Jambor <mjambor@suse.cz>
6020 * gimple-ssa-strength-reduction.c (replace_ref): Make sure built
6021 MEM_REF has proper alignment information.
6023 2013-08-05 Richard Earnshaw <rearnsha@arm.com>
6025 PR rtl-optimization/57708
6026 * recog.c (peep2_find_free_register): Validate all regs in a
6029 2013-08-02 Eric Botcazou <ebotcazou@adacore.com>
6031 * config/sparc/sparc.c (sparc_emit_membar_for_model) <SMM_TSO>: Add
6032 the implied StoreLoad barrier for atomic operations if before.
6034 2013-08-02 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
6036 Backports from mainline:
6037 2013-06-27 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
6039 * config/s390/s390.c: Rename UNSPEC_CCU_TO_INT to
6040 UNSPEC_STRCMPCC_TO_INT and UNSPEC_CCZ_TO_INT to UNSPEC_CC_TO_INT.
6041 (struct machine_function): Add tbegin_p.
6042 (s390_canonicalize_comparison): Fold CC mode compares to
6043 conditional jump if possible.
6044 (s390_emit_jump): Return the emitted jump.
6045 (s390_branch_condition_mask, s390_branch_condition_mnemonic):
6046 Handle CCRAWmode compares.
6047 (s390_option_override): Default to -mhtm if available.
6048 (s390_reg_clobbered_rtx): Handle floating point regs as well.
6049 (s390_regs_ever_clobbered): Use s390_regs_ever_clobbered also for
6050 FPRs instead of df_regs_ever_live_p.
6051 (s390_optimize_nonescaping_tx): New function.
6052 (s390_init_frame_layout): Extend clobbered_regs array to cover
6054 (s390_emit_prologue): Call s390_optimize_nonescaping_tx.
6055 (s390_expand_tbegin): New function.
6056 (enum s390_builtin): New enum definition.
6057 (code_for_builtin): New array definition.
6058 (s390_init_builtins): New function.
6059 (s390_expand_builtin): New function.
6060 (TARGET_INIT_BUILTINS): Define.
6061 (TARGET_EXPAND_BUILTIN): Define.
6062 * common/config/s390/s390-common.c (processor_flags_table): Add PF_TX.
6063 * config/s390/predicates.md (s390_comparison): Handle CCRAWmode.
6064 (s390_alc_comparison): Likewise.
6065 * config/s390/s390-modes.def: Add CCRAWmode.
6066 * config/s390/s390.h (processor_flags): Add PF_TX.
6067 (TARGET_CPU_HTM): Define macro.
6068 (TARGET_HTM): Define macro.
6069 (TARGET_CPU_CPP_BUILTINS): Define __HTM__ for htm.
6070 * config/s390/s390.md: Rename UNSPEC_CCU_TO_INT to
6071 UNSPEC_STRCMPCC_TO_INT and UNSPEC_CCZ_TO_INT to UNSPEC_CC_TO_INT.
6072 (UNSPECV_TBEGIN, UNSPECV_TBEGINC, UNSPECV_TEND, UNSPECV_TABORT)
6073 (UNSPECV_ETND, UNSPECV_NTSTG, UNSPECV_PPA): New unspecv enum
6075 (TBEGIN_MASK, TBEGINC_MASK): New constants.
6076 ("*cc_to_int"): Move up.
6077 ("*mov<mode>cc", "*cjump_64", "*cjump_31"): Accept integer
6078 constants other than 0.
6079 ("*ccraw_to_int"): New insn and splitter definition.
6080 ("tbegin", "tbegin_nofloat", "tbegin_retry")
6081 ("tbegin_retry_nofloat", "tbeginc", "tend", "tabort")
6082 ("tx_assist"): New expander.
6083 ("tbegin_1", "tbegin_nofloat_1", "*tbeginc_1", "*tend_1")
6084 ("*tabort_1", "etnd", "ntstg", "*ppa"): New insn definition.
6085 * config/s390/s390.opt: Add -mhtm option.
6086 * config/s390/s390-protos.h (s390_emit_jump): Add return type.
6087 * config/s390/htmxlintrin.h: New file.
6088 * config/s390/htmintrin.h: New file.
6089 * config/s390/s390intrin.h: New file.
6090 * doc/extend.texi: Document htm builtins.
6091 * config.gcc: Add the new header files to extra_headers.
6093 2013-07-17 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
6095 * config/s390/s390.c: (s390_expand_builtin): Allow -mhtm to be
6096 enabled without -march=zEC12.
6097 * config/s390/s390.h (TARGET_HTM): Do not require EC12 machine
6100 2013-08-01 Ganesh Gopalasubramanian <Ganesh.Gopalasubramanian@amd.com>
6102 Backport from mainline
6103 2013-05-13 Ganesh Gopalasubramanian
6104 <Ganesh.Gopalasubramanian@amd.com>
6106 * config/i386/i386.c (processor_target_table): Modified default
6107 alignment values for AMD BD and BT architectures.
6109 2013-07-31 Sriraman Tallam <tmsriram@google.com>
6111 * config/i386/i386.c (dispatch_function_versions): Fix array
6112 indexing of function_version_info to match actual_versions.
6114 2013-07-31 Sebastian Huber <sebastian.huber@embedded-brains.de>
6116 * config.gcc (*-*-rtems*): Use __cxa_atexit by default.
6117 * config/rs6000/rtems.h (TARGET_LIBGCC_SDATA_SECTION): Define.
6119 2013-07-31 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
6121 Backport from mainline
6122 2013-03-27 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
6124 * config/s390/s390.h (TARGET_FLT_EVAL_METHOD): Define.
6126 2013-07-23 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
6128 * config/s390/linux-unwind.h: Use the proper dwarf to hard reg
6129 mapping for FPRs when creating the fallback framestate.
6131 2013-07-29 Dominik Vogt <vogt@linux.vnet.ibm.com>
6133 * config/s390/s390.md ("movcc"): Swap load and store instructions.
6135 2013-07-25 Terry Guo <terry.guo@arm.com>
6137 Backport from mainline:
6138 2013-07-25 Terry Guo <terry.guo@arm.com>
6140 * config/arm/arm.c (thumb1_size_rtx_costs): Assign proper cost for
6141 shift_add/shift_sub0/shift_sub1 RTXs.
6143 2013-07-22 Iain Sandoe <iain@codesourcery.com>
6145 Backport from mainline:
6146 2013-07-22 Uros Bizjak <ubizjak@gmail.com>
6148 * config/i386/i386.md (nonlocal_goto_receiver): Delete insn if
6149 it is not needed after split.
6151 2013-07-20 Iain Sandoe <iain@codesourcery.com>
6154 * config/i386/i386.c (output_set_got) [TARGET_MACHO]: Adjust to emit a
6155 second label for nonlocal goto receivers. Don't output pic base labels
6156 unless we're producing PIC; mark that action unreachable().
6157 (ix86_save_reg): If the function contains a nonlocal label, save the
6159 * config/darwin-protos.h (machopic_should_output_picbase_label): New.
6160 * gcc/config/darwin.c (emitted_pic_label_num): New GTY.
6161 (update_pic_label_number_if_needed): New.
6162 (machopic_output_function_base_name): Adjust for nonlocal receiver
6164 (machopic_should_output_picbase_label): New.
6165 * config/i386/i386.md (enum unspecv): UNSPECV_NLGR: New.
6166 (nonlocal_goto_receiver): New insn and split.
6168 2013-07-19 Wei Mi <wmi@google.com>
6170 Backport from mainline:
6171 2013-07-18 Vladimir Makarov <vmakarov@redhat.com>
6172 Wei Mi <wmi@google.com>
6174 PR rtl-optimization/57878
6175 * lra-assigns.c (assign_by_spills): Move non_reload_pseudos to the
6176 top. Promote lra_assert to gcc_assert.
6177 (reload_pseudo_compare_func): Check regs first for reload pseudos.
6179 2013-07-11 Georg-Johann Lay <avr@gjlay.de>
6181 Backport from 2013-07-19 trunk r201051.
6184 * config/avr/avr-fixed.md (round<mode>3_const): Turn expander to insn.
6185 * config/avr/avr.md (adjust_len): Add `round'.
6186 * config/avr/avr-protos.h (avr_out_round): New prototype.
6187 (avr_out_plus): Add `out_label' argument.
6188 * config/avr/avr.c (avr_out_plus_1): Add `out_label' argument.
6189 (avr_out_plus): Pass down `out_label' to avr_out_plus_1.
6190 Handle the case where `insn' is just a pattern.
6191 (avr_out_bitop): Handle the case where `insn' is just a pattern.
6192 (avr_out_round): New function.
6193 (avr_adjust_insn_length): Handle ADJUST_LEN_ROUND.
6195 2013-07-19 Kirill Yukhin <kirill.yukhin@intel.com>
6197 * config/i386/bmiintrin.h (_bextr_u32): New.
6198 (_bextr_u64): Ditto.
6203 (_blsmsk_u32): Ditto.
6204 (_blsmsk_u64): Ditto.
6205 (_tzcnt_u32): Ditto.
6206 (_tzcnt_u64): Ditto.
6208 2013-07-17 James Greenhalgh <james.greenhalgh@arm.com>
6210 Backport From mainline:
6211 2013-07-03 James Greenhalgh <james.greenhalgh@arm.com>
6213 * config/aarch64/aarch64-builtins.c
6214 (aarch64_simd_expand_builtin): Handle AARCH64_SIMD_STORE1.
6215 * config/aarch64/aarch64-simd-builtins.def (ld1): New.
6217 * config/aarch64/aarch64-simd.md
6218 (aarch64_ld1<VALL:mode>): New.
6219 (aarch64_st1<VALL:mode>): Likewise.
6220 * config/aarch64/arm_neon.h
6221 (vld1<q>_<fpsu><8, 16, 32, 64>): Convert to RTL builtins.
6223 2013-07-11 Georg-Johann Lay <avr@gjlay.de>
6225 Backport from 2013-07-11 trunk r200901.
6228 * config/avr/avr.c (avr_set_current_function): Sanity-check signal
6229 name seen by assembler/linker if available.
6231 2013-07-10 Georg-Johann Lay <avr@gjlay.de>
6233 Backport from 2013-07-10 trunk r200872.
6236 * config/avr/avr.c (avr_prologue_setup_frame): Trunk -size to mode
6239 2013-07-10 Georg-Johann Lay <avr@gjlay.de>
6241 Backport from 2013-07-10 trunk r200870.
6244 * config/avr/avr-mcus.def (atmega16hva, atmega16hva2, atmega16hvb)
6245 (atmega16m1, atmega16u4, atmega32a, atmega32c1, atmega32hvb)
6246 (atmega32m1, atmega32u4, atmega32u6, atmega64c1, atmega64m1):
6247 Remove duplicate devices.
6248 * config/avr/gen-avr-mmcu-texi.c (print_mcus): Fail on duplicate MCUs.
6249 * config/avr/t-multilib: Regenerate.
6250 * config/avr/avr-tables.opt: Regenerate.
6251 * doc/avr-mmcu.texi: Regenerate.
6253 2013-07-10 Georg-Johann Lay <avr@gjlay.de>
6256 * config/avr/avr.opt (Waddr-space-convert): Fix typo.
6258 2013-07-09 Joseph Myers <joseph@codesourcery.com>
6260 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Only
6261 adjust register size for TDmode and TFmode for VSX registers.
6263 2013-07-08 Kai Tietz <ktietz@redhat.com>
6265 Backport from mainline
6267 * config/i386/i386.c (TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P): Define as
6268 hook_bool_const_tree_true.
6270 2013-07-08 Uros Bizjak <ubizjak@gmail.com>
6272 Backport from mainline
6273 2013-07-07 Uros Bizjak <ubizjak@gmail.com>
6275 * config/i386/driver-i386.c (host_detect_local_cpu): Do not check
6276 signature_TM2_ebx, it interferes with signature_INTEL_ebx.
6278 Backport from mainline
6279 2013-07-06 Uros Bizjak <ubizjak@gmail.com>
6281 * config/i386/sse.md (sse_movlhps): Change alternative 3
6282 of operand 2 to "m".
6284 2013-07-08 Eric Botcazou <ebotcazou@adacore.com>
6286 * Makefile.in (tree-ssa-reassoc.o): Add dependency on $(PARAMS_H).
6288 2013-07-08 Jakub Jelinek <jakub@redhat.com>
6290 PR rtl-optimization/57829
6291 * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Ensure that
6292 mask bits outside of mode are just sign-extension from mode to HWI.
6294 2013-07-03 Jakub Jelinek <jakub@redhat.com>
6297 * config/i386/predicates.md (vsib_address_operand): Disallow
6298 SYMBOL_REF or LABEL_REF in parts.disp if TARGET_64BIT && flag_pic.
6300 2013-06-30 Terry Guo <terry.guo@arm.com>
6302 Backport from mainline
6303 2013-03-27 Bin Cheng <bin.cheng@arm.com>
6306 * config/arm/arm.c (thumb1_rtx_costs, thumb1_size_rtx_costs): Fix
6307 rtx costs for SET/ASHIFT/ASHIFTRT/LSHIFTRT/ROTATERT patterns with
6310 2013-06-28 Jakub Jelinek <jakub@redhat.com>
6313 * config/i386/i386.c (ix86_expand_builtin): If target == NULL
6314 and mode is VOIDmode, don't create a VOIDmode pseudo to copy result
6317 2013-06-27 Jakub Jelinek <jakub@redhat.com>
6320 * config/i386/i386.md (bmi_bextr_<mode>): Swap predicates and
6321 constraints of operand 1 and 2.
6324 * config/i386/i386.md (bmi2_bzhi_<mode>3): Swap AND arguments
6325 to match RTL canonicalization. Swap predicates and
6326 constraints of operand 1 and 2.
6328 * tree-vect-stmts.c (vectorizable_store): Move ptr_incr var
6329 decl before the loop, initialize to NULL.
6330 (vectorizable_load): Initialize ptr_incr to NULL.
6332 2013-06-24 Martin Jambor <mjambor@suse.cz>
6334 PR tree-optimization/57358
6335 * ipa-prop.c (parm_ref_data_preserved_p): Always return true when
6338 2013-06-24 Alan Modra <amodra@gmail.com>
6340 * config/rs6000/rs6000.c (vspltis_constant): Correct for little-endian.
6341 (gen_easy_altivec_constant): Likewise.
6342 * config/rs6000/predicates.md (easy_vector_constant_add_self,
6343 easy_vector_constant_msb): Likewise.
6345 2013-06-21 Uros Bizjak <ubizjak@gmail.com>
6347 Backport from mainline
6348 2013-06-20 Uros Bizjak <ubizjak@gmail.com>
6351 * config/i386/i386.c (construct_container): Report error if
6352 long double is used with disabled x87 float returns.
6354 2013-06-20 Wei Mi <wmi@google.com>
6356 Backport from mainline
6357 2013-06-19 Wei Mi <wmi@google.com>
6359 PR rtl-optimization/57518
6360 * ira.c (set_paradoxical_subreg): Set pdx_subregs[regno]
6361 if regno is used in paradoxical subreg.
6362 (update_equiv_regs): Check pdx_subregs[regno] before
6363 set a reg to be equivalent with a mem.
6366 2013-06-20 David Edelsohn <dje.gcc@gmail.com>
6368 Backport from mainline
6369 2013-06-19 David Edelsohn <dje.gcc@gmail.com>
6372 * collect2.c (collect_atexit): New.
6373 (collect_exit): Delete.
6374 (main): Register collect_atexit with atexit.
6375 (collect_wait): Change collect_exit to exit.
6377 * collect2.h (collect_exit): Delete.
6378 * tlink.c (do_tlink): Rename exit to ret. Change collect_exit to exit.
6380 2013-06-19 Matthias Klose <doko@ubuntu.com>
6383 * file-find.h (find_a_file): Add a mode parameter.
6384 * file-find.c (find_a_file): Likewise.
6385 * gcc-ar.c (main): Call find_a_file with R_OK for the plugin,
6386 with X_OK for the executables.
6387 * collect2.c (main): Call find_a_file with X_OK.
6389 2013-06-19 Igor Zamyatin <igor.zamyatin@intel.com>
6391 * doc/invoke.texi (core-avx2): Document.
6392 (atom): Updated with MOVBE.
6394 2013-06-19 Jakub Jelinek <jakub@redhat.com>
6397 * gcc-ar.c (main): If not CROSS_DIRECTORY_STRUCTURE, look for
6398 PERSONALITY in $PATH derived prefixes.
6400 2013-06-19 Paolo Carlini <paolo.carlini@oracle.com>
6403 * doc/cpp.texi [Standard Predefined Macros, __cplusplus]: Document
6404 that now in C++ the value is correct per the C++ standards.
6406 2013-06-19 Alan Modra <amodra@gmail.com>
6408 Apply mainline patches
6409 2013-06-13 Alan Modra <amodra@gmail.com>
6410 * config/rs6000/rs6000.h (LONG_DOUBLE_LARGE_FIRST): Define.
6411 * config/rs6000/rs6000.md (signbittf2): New insn.
6412 (extenddftf2_internal): Use LONG_DOUBLE_LARGE_FIRST.
6413 (abstf2_internal, cmptf_internal2): Likewise.
6414 * config/rs6000/spe.md (spe_abstf2_cmp, spe_abstf2_tst): Likewise.
6416 2013-06-11 Anton Blanchard <anton@samba.org>
6417 * config/rs6000/rs6000.c (rs6000_adjust_atomic_subword): Calculate
6418 correct shift value in little-endian mode.
6420 2013-06-07 Alan Modra <amodra@gmail.com>
6421 * config/rs6000/rs6000.c (setup_incoming_varargs): Round up
6424 2013-06-04 Alan Modra <amodra@gmail.com>
6425 * config/rs6000/rs6000.c (output_toc): Correct little-endian float
6428 2013-05-10 Alan Modra <amodra@gmail.com>
6429 * configure.ac (HAVE_AS_TLS): Swap powerpc64 and powerpc cases.
6430 (HAVE_LD_LARGE_TOC): Don't mention AIX in help text.
6431 * configure: Regenerate.
6433 2013-05-09 Alan Modra <amodra@gmail.com>
6434 * configure.ac (HAVE_AS_TLS): Enable tests for powerpcle and
6436 * configure: Regenerate.
6438 2013-05-07 Anton Blanchard <anton@samba.org>
6439 * configure.ac (HAVE_LD_LARGE_TOC): Use right linker emulation
6440 for powerpc64 little endian.
6441 * configure: Regenerate.
6443 2013-05-06 Alan Modra <amodra@gmail.com>
6444 * config/rs6000/linux.h (DEFAULT_ASM_ENDIAN): Define.
6445 (LINK_OS_LINUX_EMUL): Use ENDIAN_SELECT.
6446 * config/rs6000/linux64.h (DEFAULT_ASM_ENDIAN): Define.
6447 * config/rs6000/sysv4le.h (DEFAULT_ASM_ENDIAN): Define.
6448 (LINK_TARGET_SPEC): Use ENDIAN_SELECT.
6449 * config/rs6000/sysv4.h (DEFAULT_ASM_ENDIAN): Define as -mbig.
6451 2013-05-06 Alan Modra <amodra@gmail.com>
6452 * config/rs6000/sysv4.h (ENDIAN_SELECT): Define, extracted from
6453 (ASM_SPEC): ..here. Emit DEFAULT_ASM_ENDIAN too.
6454 (DEFAULT_ASM_ENDIAN): Define.
6455 (CC1_SPEC, LINK_TARGET_SPEC): Use ENDIAN_SELECT.
6456 * config/rs6000/linux64.h (ASM_SPEC32): Remove endian options.
6457 Update -K PIC clause from sysv4.h.
6458 (ASM_SPEC_COMMON): Use ENDIAN_SELECT.
6459 (LINK_OS_LINUX_EMUL32, LINK_OS_LINUX_EMUL64): Likewise.
6461 2013-05-06 Alan Modra <amodra@gmail.com>
6462 * config/rs6000/rs6000.md (bswapdi 2nd splitter): Don't swap words
6463 twice for little-endian.
6464 (ashrdi3_no_power, ashrdi3): Support little-endian.
6466 2013-04-25 Alan Modra <amodra@gmail.com>
6467 * config.gcc: Support little-endian powerpc-linux targets.
6468 * config/rs6000/linux.h (LINK_OS_LINUX_EMUL): Define.
6469 (LINK_OS_LINUX_SPEC): Define.
6470 * config/rs6000/linuxspe.h (TARGET_DEFAULT):
6471 Preserve MASK_LITTLE_ENDIAN.
6472 * config/rs6000/default64.h (TARGET_DEFAULT): Likewise.
6473 * config/rs6000/linuxaltivec.h (TARGET_DEFAULT): Likewise.
6474 * config/rs6000/linux64.h (OPTION_LITTLE_ENDIAN): Don't zero.
6475 (LINK_OS_LINUX_EMUL32, LINK_OS_LINUX_EMUL64): Define.
6476 (LINK_OS_LINUX_SPEC32, LINK_OS_LINUX_SPEC64): Use above.
6477 * config/rs6000/rs6000.c (output_toc): Don't use .tc for TARGET_ELF.
6478 Correct fp word order for little-endian. Don't shift toc entries
6479 smaller than a word for little-endian.
6480 * config/rs6000/rs6000.md (bswaphi2, bswapsi2 split): Comment.
6481 (bswapdi2 splits): Correct low-part subreg for little-endian.
6482 Remove wrong BYTES_BIG_ENDIAN tests, and rename vars to remove
6483 low/high where such is correct only for be.
6484 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Allow
6485 little-endian for -mcall-aixdesc.
6487 2013-06-12 Martin Jambor <mjambor@suse.cz>
6489 * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that param_index is
6490 within bounds at the beginning of the function.
6492 2013-06-12 Jakub Jelinek <jakub@redhat.com>
6494 PR tree-optimization/57537
6495 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): If
6496 vect_handle_widen_op_by_const, convert oprnd1 to half_type1.
6498 2013-06-10 Uros Bizjak <ubizjak@gmail.com>
6500 Backport from mainline
6501 2013-06-10 Uros Bizjak <ubizjak@gmail.com>
6503 * config/alpha/alpha.c (alpha_emit_xfloating_compare): Also use
6504 cmp_code to construct REG_EQUAL note.
6506 2013-06-10 Oleg Endo <olegendo@gcc.gnu.org>
6508 Backport from mainline
6509 2013-05-20 Oleg Endo <olegendo@gcc.gnu.org>
6512 * config/sh/sh.md (fmasf4): Remove empty constraints strings.
6513 (*fmasf4, *fmasf4_media): New insns.
6515 2013-06-09 Jakub Jelinek <jakub@redhat.com>
6518 * config/i386/i386.md (TARGET_READ_MODIFY_WRITE peepholes): Ensure
6519 that operands[2] doesn't overlap with operands[0].
6521 2013-06-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
6523 * recog.c (offsettable_address_addr_space_p): Fix calculation of
6524 address mode. Move pointer mode initialization to the same place.
6526 2013-06-07 Sofiane Naci <sofiane.naci@arm.com>
6528 Backport from mainline
6529 * config/aarch64/aarch64.md (*movdi_aarch64): Define "simd" attribute.
6531 2013-06-07 Uros Bizjak <ubizjak@gmail.com>
6533 Backport from mainline
6534 2013-06-05 Uros Bizjak <ubizjak@gmail.com>
6536 * config/alpha/alpha.c (alpha_emit_conditional_move): Swap all
6537 GE, GT, GEU and GTU compares, modulo DImode compares with zero.
6539 Backport from mainline
6540 2013-05-23 Uros Bizjak <ubizjak@gmail.com>
6543 * config/alpha/alpha.md (unspec): Add UNSPEC_XFLT_COMPARE.
6544 * config/alpha/alpha.c (alpha_emit_xfloating_compare): Construct
6545 REG_EQUAL note as UNSPEC_XFLT_COMPARE unspec.
6547 2013-06-04 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6549 Backport from mainline.
6550 2013-05-22 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6552 * config/rs6000/rs6000.h (MALLOC_ABI_ALIGNMENT): New #define.
6554 2013-06-03 James Greenhalgh <james.greenhalgh@arm.com>
6556 Backport from mainline.
6557 2013-04-25 James Greenhalgh <james.greenhalgh@arm.com>
6559 * config/aarch64/aarch64.c (aarch64_print_operand): Fix asm_fprintf
6560 format specifier in 'X' case.
6562 2013-05-31 Richard Henderson <rth@redhat.com>
6565 * config/i386/i386.c (ix86_seh_fixup_eh_fallthru): New.
6566 (ix86_reorg): Call it.
6568 2012-05-31 Jakub Jelinek <jakub@redhat.com>
6570 * BASE-VER: Set to 4.8.2.
6571 * DEV-PHASE: Set to prerelease.
6573 2013-05-31 Release Manager
6575 * GCC 4.8.1 released.
6577 2013-05-24 Greta Yorsh <Greta.Yorsh@arm.com>
6579 Backport from mainline
6580 2013-05-02 Greta Yorsh <Greta.Yorsh@arm.com>
6583 * config/arm/arm.c (arm_expand_epilogue): Check really_return before
6584 generating simple_return for naked functions.
6586 2013-05-24 Alexander Ivchenko <alexander.ivchenko@intel.com>
6589 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Check
6590 that index is not negative.
6592 2013-05-23 Martin Jambor <mjambor@suse.cz>
6595 * tree.h (contains_bitfld_component_ref_p): Declare.
6596 * tree-sra.c (contains_bitfld_comp_ref_p): Move...
6597 * tree.c (contains_bitfld_component_ref_p): ...here. Adjust its caller.
6598 * ipa-prop.c (determine_known_aggregate_parts): Check that LHS does
6599 not access a bit-field. Assert all final offsets are byte-aligned.
6601 2013-05-23 Richard Biener <rguenther@suse.de>
6603 PR rtl-optimization/57341
6604 * ira.c (validate_equiv_mem_from_store): Use anti_dependence
6605 instead of true_dependence.
6607 2013-05-23 Jakub Jelinek <jakub@redhat.com>
6610 * expmed.c (store_split_bit_field): If op0 is a REG or
6611 SUBREG of a REG, don't lower unit. Handle unit not being
6612 always BITS_PER_WORD.
6614 2013-05-22 Uros Bizjak <ubizjak@gmail.com>
6617 * config/i386/i386.md (*movti_internal_rex64): Emit movaps/movups
6618 for non-sse2 targets.
6619 (*movti_internal): Simplify mode attribute calculation.
6621 2013-05-22 Richard Biener <rguenther@suse.de>
6623 Backport from mainline
6624 2013-05-21 Richard Biener <rguenther@suse.de>
6626 PR tree-optimization/57318
6627 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Do not
6628 estimate stmts with side-effects as likely eliminated.
6630 2013-05-21 Richard Biener <rguenther@suse.de>
6632 PR tree-optimization/57330
6633 * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Properly
6634 preserve the call stmts fntype.
6636 2013-05-21 Richard Biener <rguenther@suse.de>
6638 PR tree-optimization/57303
6639 * tree-ssa-sink.c (statement_sink_location): Properly handle
6642 2013-05-21 Magnus Granberg <baldrick@free.fr>
6645 * Makefile.in (PLUGIN_HEADERS): Add $(TARGET_H).
6647 2013-05-21 Eric Botcazou <ebotcazou@adacore.com>
6649 Backport from mainline
6650 2013-05-14 Eric Botcazou <ebotcazou@adacore.com>
6652 * config/sparc/sp64-elf.h (CPP_SUBTARGET_SPEC): Delete.
6653 * config/sparc/openbsd64.h (CPP_SUBTARGET_SPEC): Likewise.
6655 2013-05-17 Jakub Jelinek <jakub@redhat.com>
6657 PR rtl-optimization/57281
6658 PR rtl-optimization/57300
6659 * config/i386/i386.md (extendsidi2_1 dead reg splitter): Remove.
6660 (extendsidi2_1 peephole2s): Add instead 2 new peephole2s, that undo
6661 what the other splitter did if the registers are dead.
6663 2013-05-17 Uros Bizjak <ubizjak@gmail.com>
6665 Backport from mainline
6666 2013-05-16 Uros Bizjak <ubizjak@gmail.com>
6668 * config/i386/driver-i386.c (host_detect_local_cpu): Determine
6669 cache parameters using detect_caches_amd also for CYRIX,
6670 NSC and TM2 signatures.
6672 2013-05-16 Uros Bizjak <ubizjak@gmail.com>
6673 Dzianis Kahanovich <mahatma@eu.by>
6677 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6678 VIA/Centaur processors and determine their cache parameters
6679 using detect_caches_amd.
6681 2013-05-15 Uros Bizjak <ubizjak@gmail.com>
6683 * config/i386/i386.c (ix86_option_override_internal): Update
6684 processor_alias_table for missing PTA_PRFCHW and PTA_FXSR flags. Add
6685 PTA_POPCNT to corei7 entry. Do not enable SSE prefetch on
6686 non-SSE 3dNow! targets. Enable TARGET_PRFCHW for TARGET_3DNOW targets.
6687 * config/i386/i386.md (prefetch): Enable for TARGET_PRFCHW instead
6689 (*prefetch_3dnow): Enable for TARGET_PRFCHW only.
6691 2013-05-17 Jakub Jelinek <jakub@redhat.com>
6693 * gcc.c (SANITIZER_SPEC): Reject -fsanitize=address -fsanitize=thread
6696 PR tree-optimization/57051
6697 * fold-const.c (const_binop) <case VEC_LSHIFT_EXPR,
6698 case VEC_RSHIFT_EXPR>: Fix BYTES_BIG_ENDIAN handling.
6700 2013-05-16 Jakub Jelinek <jakub@redhat.com>
6702 * omp-low.c (extract_omp_for_data): For collapsed loops,
6703 if at least one of the loops is known at compile time to
6704 iterate zero times, set count to 0.
6705 (expand_omp_regimplify_p): New function.
6706 (expand_omp_for_generic): For collapsed loops, if at least
6707 one of the loops isn't known to iterate at least once,
6708 add runtime check with setting count to 0.
6709 (expand_omp_for_static_nochunk, expand_omp_for_static_chunk):
6710 For unsigned types if it isn't known at compile time that
6711 the loop will iterate at least once, add runtime check to bypass
6712 the whole loop if initial condition isn't true.
6714 2013-05-14 Jakub Jelinek <jakub@redhat.com>
6717 * expr.c (expand_expr_real_2) <case WIDEN_MULT_EXPR>: Handle
6718 the case when both op0 and op1 have VOIDmode.
6720 2013-05-13 Jakub Jelinek <jakub@redhat.com>
6722 PR tree-optimization/57230
6723 * tree-ssa-strlen.c (handle_char_store): Add missing integer_zerop
6726 2013-05-12 Joern Rennecke <joern.rennecke@embecosm.com>
6728 * config/epiphany/epiphany.c (epiphany_init): Check size of
6729 NUM_MODES_FOR_MODE_SWITCHING.
6730 (epiphany_expand_prologue):
6731 Remove CONFIG_REGNUM initial value handling code.
6732 (epiphany_optimize_mode_switching): Handle EPIPHANY_MSW_ENTITY_CONFIG.
6733 (epiphany_mode_needed, epiphany_mode_entry_exit): Likewise.
6734 (emit_set_fp_mode, epiphany_mode_after): Likewise.
6735 (epiphany_mode_needed) <Handle EPIPHANY_MSW_ENTITY_AND>:
6736 Don't return 1 for FP_MODE_NONE.
6737 * config/epiphany/epiphany.h (NUM_MODES_FOR_MODE_SWITCHING):
6738 Add value for EPIPHANY_MSW_ENTITY_CONFIG.
6739 (EPIPHANY_MSW_ENTITY_CONFIG, EPIPHANY_MSW_ENTITY_NUM): Define.
6740 * config/epiphany/epiphany.md (save_config): New pattern.
6742 2013-05-10 Sebastian Huber <sebastian.huber@embedded-brains.de>
6744 * config/arm/t-rtems-eabi: Remove mthumb/march=armv7 multilib.
6745 Add mthumb/march=armv7-a multilib.
6746 Add mthumb/march=armv7-r multilib.
6747 Add mthumb/march=armv7-a/mfpu=neon/mfloat-abi=hard multilib.
6749 2013-05-10 Ralf Corsépius <ralf.corsepius@rtems.org>
6752 * config/v850/t-rtems: Add more multilibs.
6754 2013-05-10 Richard Biener <rguenther@suse.de>
6756 PR tree-optimization/57214
6757 * tree-ssa-loop-ivcanon.c (propagate_constants_for_unrolling): Do
6758 not propagate from SSA names that occur in abnormal PHI nodes.
6760 2013-05-10 Alan Modra <amodra@gmail.com>
6763 * varasm.c (default_elf_select_section): Move !DECL_P check..
6764 (get_named_section): ..to here before calling get_section_name.
6766 (default_section_type_flags): Add DECL_P check.
6767 * config/i386/winnt.c (i386_pe_section_type_flags): Likewise.
6768 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags): Likewise.
6770 2013-05-09 Joern Rennecke <joern.rennecke@embecosm.com>
6772 * config/epiphany/epiphany.c (epiphany_expand_prologue):
6773 When using gen_stack_adjust_str with a register offset, add a
6774 REG_FRAME_RELATED_EXPR note.
6776 2013-05-09 Martin Jambor <mjambor@suse.cz>
6779 * ipa-prop.h (ipa_agg_replacement_value): New flag by_ref.
6780 * ipa-cp.c (find_aggregate_values_for_callers_subset): Fill in the
6781 by_ref flag of ipa_agg_replacement_value structures.
6782 (known_aggs_to_agg_replacement_list): Likewise.
6783 * ipa-prop.c (write_agg_replacement_chain): Stream by_ref flag.
6784 (read_agg_replacement_chain): Likewise.
6785 (ipcp_transform_function): Also check that by_ref flags match.
6787 2013-05-08 Diego Novillo <dnovillo@google.com>
6793 2012-08-17 Diego Novillo <dnovillo@google.com>
6796 * configure.ac: Add libintl.h to AC_CHECK_HEADERS list.
6797 * config.in: Regenerate.
6798 * configure: Regenerate.
6799 * intl.h: Always include libintl.h if HAVE_LIBINTL_H is
6802 2013-05-08 Paolo Carlini <paolo.carlini@oracle.com>
6804 PR tree-optimization/57200
6805 * tree-ssa-loop-niter.c (do_warn_aggressive_loop_optimizations):
6806 Only call inform if the preceding warning_at returns true.
6808 2013-05-07 Jakub Jelinek <jakub@redhat.com>
6810 PR tree-optimization/57149
6811 * tree-ssa-uninit.c (uninit_undefined_value_p): New inline.
6812 (can_skip_redundant_opnd, compute_uninit_opnds_pos,
6813 collect_phi_def_edges, execute_late_warn_uninitialized): Use
6814 uninit_undefined_value_p instead of ssa_undefined_value_p.
6817 * expr.c (expand_expr_addr_expr_1): Handle COMPOUND_LITERAL_EXPR
6818 for modifier == EXPAND_INITIALIZER.
6820 2013-05-07 Richard Biener <rguenther@suse.de>
6822 Backport from mainline
6823 2013-05-06 Richard Biener <rguenther@suse.de>
6825 PR tree-optimization/57185
6826 * tree-parloops.c (add_field_for_reduction): Handle anonymous
6829 2013-04-19 Richard Biener <rguenther@suse.de>
6831 PR tree-optimization/57000
6832 * tree-ssa-reassoc.c (pass_reassoc): Add TODO_update_ssa_only_virtuals.
6834 2013-05-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6837 2013-05-03 Michael Meissner <meissner@linux.vnet.ibm.com>
6840 * config/rs6000/rs6000.h (HARD_REGNO_CALLER_SAVE_MODE): Use DFmode
6841 to save TFmode registers and DImode to save TImode registers for
6842 caller save operations.
6843 (HARD_REGNO_CALL_PART_CLOBBERED): TFmode and TDmode do not need to
6844 mark being partially clobbered since they only use the first
6847 * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): TFmode
6848 and TDmode only use the upper 64-bits of each VSX register.
6850 2013-05-06 Oleg Endo <olegendo@gcc.gnu.org>
6853 * config/sh/sh.md (tstsi_t_zero_extract_eq): Use QIHISIDI mode iterator.
6855 2013-05-06 Uros Bizjak <ubizjak@gmail.com>
6857 Backport from mainline
6858 2013-05-06 Uros Bizjak <ubizjak@gmail.com>
6861 * config/i386/i386.c (add_parameter_dependencies): Add dependence
6862 between "first_arg" and "insn", not "last" and "insn".
6864 2013-05-03 Jakub Jelinek <jakub@redhat.com>
6866 PR rtl-optimization/57130
6867 * combine.c (make_compound_operation) <case SUBREG>: Pass
6868 SET instead of COMPARE as in_code to the recursive call
6871 Backported from mainline
6872 2013-04-26 Jakub Jelinek <jakub@redhat.com>
6874 PR tree-optimization/57051
6875 * fold-const.c (const_binop): Handle VEC_LSHIFT_EXPR
6876 and VEC_RSHIFT_EXPR if shift count is a multiple of element
6879 2013-04-12 Marc Glisse <marc.glisse@inria.fr>
6881 * fold-const.c (fold_binary_loc): Call const_binop also for mixed
6882 vector-scalar operations.
6884 2013-05-03 Marek Polacek <polacek@redhat.com>
6886 Backport from mainline
6887 2013-04-25 Marek Polacek <polacek@redhat.com>
6889 PR tree-optimization/57066
6890 * builtins.c (fold_builtin_logb): Return +Inf for -Inf.
6892 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6894 Backport from mainline
6895 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6897 * lra-constraints.c (process_alt_operands): Add checking alt
6898 number to choose the best alternative.
6900 2013-05-01 Vladimir Makarov <vmakarov@redhat.com>
6903 * lra-constraints.c (best_small_class_operands_num): Remove.
6904 (process_alt_operands): Remove small_class_operands_num. Take
6905 small classes operands into losers and only if the operand is not
6906 matched. Modify debugging output.
6907 (curr_insn_transform): Remove best_small_class_operands_num.
6910 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6912 Backport from mainline
6913 2013-04-29 Vladimir Makarov <vmakarov@redhat.com>
6916 * lra-constraints.c (process_alt_operands): Discourage a bit more
6917 using memory for pseudos. Print cost dump for alternatives.
6918 Modify cost values for conflicts with early clobbers.
6919 (curr_insn_transform): Spill pseudos reassigned to NO_REGS.
6921 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6923 Backport from mainline
6924 2013-04-24 Vladimir Makarov <vmakarov@redhat.com>
6926 PR rtl-optimizations/57046
6927 * lra-constraints (split_reg): Set up lra_risky_transformations_p
6928 for multi-reg splits.
6930 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6932 Backport from mainline
6933 2013-04-22 Vladimir Makarov <vmakarov@redhat.com>
6936 * lra-eliminations.c (mark_not_eliminable): Prevent elimination of
6937 a set sp if no stack realignment.
6939 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6941 Backport from mainline
6942 2013-04-18 Vladimir Makarov <vmakarov@redhat.com>
6944 PR rtl-optimization/56999
6945 * lra-coalesce.c (coalescable_pseudo_p): Remove 2nd parameter and
6947 (lra_coalesce): Remove split_origin_bitmap and related code.
6948 * lra.c (lra): Coalesce after undoing inheritance. Recreate live
6949 ranges if necessary.
6951 2013-05-02 Vladimir Makarov <vmakarov@redhat.com>
6953 Backport from mainline
6954 2013-04-19 Vladimir Makarov <vmakarov@redhat.com>
6956 PR rtl-optimization/56847
6957 * lra-constraints.c (process_alt_operands): Discourage alternative
6958 with non-matche doffsettable memory constraint fro memory with
6961 2013-05-02 Ian Bolton <ian.bolton@arm.com>
6963 Backport from mainline
6964 2013-03-28 Ian Bolton <ian.bolton@arm.com>
6966 * config/aarch64/aarch64.md (aarch64_can_eliminate): Keep frame
6967 record only when desired or required.
6969 2013-04-30 Jakub Jelinek <jakub@redhat.com>
6971 PR tree-optimization/57104
6972 * tsan.c (instrument_expr): Don't instrument accesses to
6973 DECL_HARD_REGISTER VAR_DECLs.
6975 2013-04-30 Uros Bizjak <ubizjak@gmail.com>
6977 Backport from mainline
6978 2013-04-29 Uros Bizjak <ubizjak@gmail.com>
6981 * config/i386/i386.md (*zero_extendsidi2_rex64): Add "!" to m->?*y
6983 (*zero_extendsidi2): Ditto.
6985 Backport from mainline
6986 2013-04-29 Uros Bizjak <ubizjak@gmail.com>
6989 * config/i386/i386.c (ix86_expand_vec_perm): Validize constant memory.
6991 2013-04-29 Richard Biener <rguenther@suse.de>
6994 * tree-cfg.c (move_stmt_op): Fix condition under which to update
6996 (move_stmt_r): Remove redundant checking.
6998 2013-04-29 Christian Bruel <christian.bruel@st.com>
7001 * sh.md (tstsi_t_zero_extract_eq): Set mode for operand 0.
7003 2013-04-29 Jakub Jelinek <jakub@redhat.com>
7005 PR tree-optimization/57083
7006 * tree-vrp.c (extract_range_from_binary_expr_1): For LSHIFT_EXPR with
7007 non-singleton shift count range, zero extend low_bound for uns case.
7009 2013-04-28 Eric Botcazou <ebotcazou@gcc.gnu.org>
7011 * stor-layout.c (finalize_size_functions): Allocate a structure and
7012 reset cfun before dumping the functions.
7014 2013-04-27 Jakub Jelinek <jakub@redhat.com>
7017 * config/i386/i386.c (ix86_expand_mul_widen_evenodd): Don't
7018 use xop_pmacsdqh if uns_p.
7019 * config/i386/sse.md (xop_rotr<mode>3): Fix up computation of
7020 the immediate rotate count.
7022 2013-04-25 Jakub Jelinek <jakub@redhat.com>
7024 PR rtl-optimization/57003
7025 * regcprop.c (copyprop_hardreg_forward_1): If ksvd.ignore_set_reg,
7026 call note_stores with kill_clobbered_value callback again after
7027 killing regs_invalidated_by_call.
7029 2013-04-25 Ian Bolton <ian.bolton@arm.com>
7031 Backported from mainline.
7032 2013-03-22 Ian Bolton <ian.bolton@arm.com>
7034 * config/aarch64/aarch64.c (aarch64_print_operand): New
7035 format specifier for printing a constant in hex.
7036 * config/aarch64/aarch64.md (insv_imm<mode>): Use the X
7037 format specifier for printing second operand.
7039 2013-04-24 James Greenhalgh <james.greenhalgh@arm.com>
7041 Backported from mainline.
7042 2013-04-24 James Greenhalgh <james.greenhalgh@arm.com>
7044 * config/aarch64/arm_neon.h (vld1<q>_lane*): Fix constraints.
7045 (vld1<q>_dup_<sufp><8, 16, 32, 64>): Likewise.
7046 (vld1<q>_<sufp><8, 16, 32, 64>): Likewise.
7048 2013-04-24 Greta Yorsh <Greta.Yorsh@arm.com>
7050 Backported from mainline.
7052 * config/arm/arm.c (load_multiple_sequence): Require SP
7053 as base register for loads if SP is in the register list.
7055 2013-04-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7056 Steven Bosscher <steven@gcc.gnu.org>
7058 Backported from mainline.
7059 PR rtl-optimization/56605
7060 * loop-iv.c (implies_p): Handle equal RTXs and subregs.
7062 2013-04-22 Marek Polacek <polacek@redhat.com>
7064 Backported from mainline.
7065 2013-04-22 Marek Polacek <polacek@redhat.com>
7068 * tsan.c (instrument_expr): Don't instrument expression
7069 in case its size is zero.
7071 2013-04-22 Yufeng Zhang <yufeng.zhang@arm.com>
7073 Backported from mainline.
7074 2013-04-10 Yufeng Zhang <yufeng.zhang@arm.com>
7075 * config/aarch64/aarch64.c (aarch64_print_extension): New function.
7076 (aarch64_start_file): Use the new function.
7078 2013-04-18 Jakub Jelinek <jakub@redhat.com>
7080 PR tree-optimization/56984
7081 * tree-vrp.c (register_edge_assert_for_2): For (x >> M) < N
7082 and (x >> M) >= N don't register any assertion if N << M is the
7085 2013-04-17 David Edelsohn <dje.gcc@gmail.com>
7088 * config/rs6000/vsx.md (vsx_mov<mode>): Add j->r alternative.
7090 2013-04-15 Jakub Jelinek <jakub@redhat.com>
7092 PR tree-optimization/56962
7093 * gimple-ssa-strength-reduction.c (record_increment): Only set
7094 initializer if gimple_assign_rhs_code is {,POINTER_}PLUS_EXPR and
7095 either rhs1 or rhs2 is equal to c->base_expr.
7097 2013-04-15 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
7099 * emit-rtl.c (reset_all_used_flags): New function.
7100 (verify_rtl_sharing): Call reset_all_used_flags before and after
7101 performing the checks.
7103 2013-04-15 Eric Botcazou <ebotcazou@adacore.com>
7106 * config/sparc/sparc.c (enum sparc_mode_class): Add H_MODE value.
7107 (S_MODES): Set H_MODE bit.
7108 (SF_MODES): Set only S_MODE and SF_MODE bits.
7109 (DF_MODES): Set SF_MODES and only D_MODE and DF_MODE bits.
7110 (sparc_init_modes) <MODE_INT>: Set H_MODE bit for sub-word modes.
7111 <MODE_VECTOR_INT>: Do not set SF_MODE for sub-word modes.
7112 <MODE_FLOAT>: Likewise.
7114 2013-04-12 Vladimir Makarov <vmakarov@redhat.com>
7117 * config/i386/i386.c (ix86_hard_regno_mode_ok): Add
7118 lra_in_progress for return.
7120 2013-04-12 Jakub Jelinek <jakub@redhat.com>
7122 PR tree-optimization/56918
7123 PR tree-optimization/56920
7124 * fold-const.c (int_const_binop_1): Use op1.mul_with_sign (op2, ...)
7125 instead of op1 - op2. Pass 2 * TYPE_PRECISION (type) as second
7126 argument to rshift method.
7128 2013-04-12 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
7130 * ifcvt.c (end_ifcvt_sequence): Mark a and b for unsharing as
7133 2013-04-11 Marek Polacek <polacek@redhat.com>
7135 PR tree-optimization/48184
7136 * params.def (PARAM_ALIGN_THRESHOLD): Increase the minimum
7139 2013-04-11 James Greenhalgh <james.greenhalgh@arm.com>
7141 Backported from mainline.
7142 2013-04-11 James Greenhalgh <james.greenhalgh@arm.com>
7144 * config/aarch64/aarch64-simd.md (aarch64_vcond_internal): Fix
7145 floating-point vector comparisons against 0.
7147 2013-04-11 Jakub Jelinek <jakub@redhat.com>
7149 PR tree-optimization/56899
7150 * fold-const.c (extract_muldiv_1): Apply distributive law
7151 only if TYPE_OVERFLOW_WRAPS (ctype).
7153 2013-04-10 David S. Miller <davem@davemloft.net>
7155 * config/sparc/sparc.h (ASM_CPU_SPEC): Pass -Av8 if -mcpu=supersparc
7156 or -mcpu=hypersparc.
7158 2013-04-10 Jakub Jelinek <jakub@redhat.com>
7160 Backported from mainline
7161 2013-04-09 Jakub Jelinek <jakub@redhat.com>
7164 * omp-low.c (expand_omp_for_generic, expand_omp_for_static_nochunk,
7165 expand_omp_for_static_chunk): Use simple_p = true in
7166 force_gimple_operand_gsi calls when assigning to addressable decls.
7168 2013-04-09 Marek Polacek <polacek@redhat.com>
7170 PR tree-optimization/48762
7171 * params.def (PARAM_MAX_CSE_INSNS): Increase the minimum
7174 2013-04-08 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
7176 * config/s390/s390.c (s390_expand_insv): Only accept insertions
7179 2013-04-08 Marek Polacek <polacek@redhat.com>
7181 PR rtl-optimization/48182
7182 * params.def (PARAM_MIN_CROSSJUMP_INSNS): Increase the minimum
7185 2013-04-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
7188 * config/pa/pa.c (legitimize_pic_address): Before incrementing label
7189 nuses, make sure we have a label.
7191 2013-04-04 Ian Lance Taylor <iant@google.com>
7193 Backport from mainline:
7194 * doc/standards.texi (Standards): The Go frontend supports the Go
7195 1 language standard.
7197 2013-04-04 Marek Polacek <polacek@redhat.com>
7199 Backport from mainline:
7200 2013-04-04 Marek Polacek <polacek@redhat.com>
7202 PR tree-optimization/48186
7203 * predict.c (maybe_hot_frequency_p): Return false if
7204 HOT_BB_FREQUENCY_FRACTION is 0.
7205 (cgraph_maybe_hot_edge_p): Likewise.
7207 2013-04-04 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7209 Backport from mainline:
7210 2013-03-25 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7213 * config/arm/iterators.md (v_cmp_result): New mode attribute.
7214 * config/arm/neon.md (vcond<mode><mode>): Handle unordered cases.
7216 2013-04-04 Richard Biener <rguenther@suse.de>
7218 PR tree-optimization/56837
7219 * tree-loop-distribution.c (classify_partition): For non-zero
7220 values require that the value has the same precision as its
7221 mode to be useful as memset value.
7223 2013-04-03 Roland McGrath <mcgrathr@google.com>
7225 Backport from mainline:
7226 2013-03-26 Roland McGrath <mcgrathr@google.com>
7228 * config/arm/arm.c (arm_print_operand: case 'w'): Use fputs rather
7229 than fprintf with a non-constant, non-format string.
7231 2013-04-03 Marek Polacek <polacek@redhat.com>
7233 Backport from mainline:
7234 2013-04-03 Marek Polacek <polacek@redhat.com>
7237 * tsan.c (instrument_func_exit): Allow BUILT_IN_RETURN
7240 2013-04-03 Richard Biener <rguenther@suse.de>
7242 PR tree-optimization/56817
7243 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely):
7245 (tree_unroll_loops_completely_1): ... new function to manually
7246 walk the loop tree, properly defering outer loops of unrolled
7247 loops to later iterations.
7249 2013-04-02 Jakub Jelinek <jakub@redhat.com>
7251 PR rtl-optimization/56745
7252 * ifcvt.c (cond_exec_find_if_block): Don't try to optimize
7253 if then_bb has no successors and else_bb is EXIT_BLOCK_PTR.
7255 2013-04-02 Wei Mi <wmi@google.com>
7257 * config/i386/i386.c (ix86_rtx_costs): Set proper rtx cost for
7258 ashl<mode>3_mask, *<shift_insn><mode>3_mask and
7259 *<rotate_insn><mode>3_mask in i386.md.
7261 2013-04-01 Wei Mi <wmi@google.com>
7263 * config/i386/i386.md (*ashl<mode>3_mask): Rewrite as define_insn.
7264 Truncate operand 2 using %b asm operand modifier.
7265 (*<shift_insn><mode>3_mask): Ditto.
7266 (*<rotate_insn><mode>3_mask): Ditto.
7268 2013-04-01 Uros Bizjak <ubizjak@gmail.com>
7270 * config/i386/i386.md (*movsf_internal): Change type of
7271 alternatives 3,4 to imov.
7273 2013-03-29 Paolo Carlini <paolo.carlini@oracle.com>
7276 * doc/invoke.texi ([-fwhole-program]): Fix typo.
7278 2013-03-29 Kirill Yukhin <kirill.yukhin@intel.com>
7280 * config/i386/avx2intrin.h (_mm256_broadcastsi128_si256):
7281 Fix declaration name.
7283 2013-03-28 Gerald Pfeifer <gerald@pfeifer.com>
7285 * doc/invoke.texi (AVR Options): Tweak link for AVR-LibC user manual.
7286 * doc/extend.texi (Named Address Spaces): Ditto.
7287 (Variable Attributes): Ditto.
7289 2013-03-28 Eric Botcazou <ebotcazou@adacore.com>
7291 * toplev.c (process_options): Do not disable -fomit-frame-pointer on a
7292 general basis if unwind info is requested and ACCUMULATE_OUTGOING_ARGS
7295 2013-03-27 Walter Lee <walt@tilera.com>
7297 Backport from mainline:
7298 2013-03-27 Walter Lee <walt@tilera.com>
7300 * config/tilegx/tilegx.c (tilegx_expand_prologue): Avoid
7301 double-decrement of next_scratch_regno.
7303 2013-03-27 Walter Lee <walt@tilera.com>
7305 Backport from mainline:
7306 2013-03-27 Walter Lee <walt@tilera.com>
7308 * config/tilegx/tilegx.md (insn_v1mulu): Fix predicates on
7310 (insn_v1mulus): Ditto.
7311 (insn_v2muls): Ditto.
7313 2013-03-27 Walter Lee <walt@tilera.com>
7315 Backport from mainline:
7316 2013-03-27 Walter Lee <walt@tilera.com>
7318 * config/tilegx/tilegx.h (ASM_OUTPUT_ADDR_VEC_ELT): Delete
7320 (ASM_OUTPUT_ADDR_DIFF_ELT): Ditto.
7322 2013-03-27 Walter Lee <walt@tilera.com>
7324 Backport from mainline:
7325 2013-03-27 Walter Lee <walt@tilera.com>
7327 * config/tilegx/tilegx.md (*sibcall_insn): Fix type atribute for jr.
7328 (*sibcall_value): Ditto.
7330 2013-03-27 Walter Lee <walt@tilera.com>
7332 Backport from mainline:
7333 2013-03-27 Walter Lee <walt@tilera.com>
7335 * config/tilegx/tilegx.md (insn_mnz_<mode>): Replaced by ...
7336 (insn_mnz_v8qi): ... this ...
7337 (insn_mnz_v4hi): ... and this. Replace (const_int 0) with the
7339 (insn_v<n>mnz): Replaced by ...
7340 (insn_v1mnz): ... this ...
7341 (insn_v2mnz): ... and this. Replace (const_int 0) with the vector
7343 (insn_mz_<mode>): Replaced by ...
7344 (insn_mz_v8qi): ... this ...
7345 (insn_mz_v4hi): ... and this. Replace (const_int 0) with the
7347 (insn_v<n>mz): Replaced by ...
7348 (insn_v1mz): ... this ...
7349 (insn_v2mz): ... and this. Replace (const_int 0) with the vector
7352 2013-03-26 Eric Botcazou <ebotcazou@adacore.com>
7354 * doc/invoke.texi (SPARC options): Remove -mlittle-endian.
7356 2013-03-26 Sebastian Huber <sebastian.huber@embedded-brains.de>
7358 * config/rtems.opt: Add -pthread option.
7360 2013-03-26 Sofiane Naci <sofiane.naci@arm.com>
7362 * config/aarch64/aarch64.c (aarch64_classify_address): Support
7363 PC-relative load in SI modes and above only.
7365 2013-03-26 Walter Lee <walt@tilera.com>
7367 Backport from mainline:
7368 2013-03-26 Walter Lee <walt@tilera.com>
7370 * config/tilegx/tilegx.h (PROFILE_BEFORE_PROLOGUE): Define.
7371 * config/tilegx/tilepro.h (PROFILE_BEFORE_PROLOGUE): Define.
7373 2013-03-26 Walter Lee <walt@tilera.com>
7375 Backport from mainline:
7376 2013-03-25 Walter Lee <walt@tilera.com>
7378 * config/tilegx/tilegx-builtins.h (enum tilegx_builtin): Add
7379 TILEGX_INSN_SHUFFLEBYTES1.
7380 * config/tilegx/tilegx.c (tilegx_builtin_info): Add entry for
7382 (tilegx_builtins): Ditto.
7383 * config/tilegx/tilegx.md (insn_shufflebytes1): New pattern.
7385 2013-03-26 Walter Lee <walt@tilera.com>
7387 Backport from mainline:
7388 2013-03-25 Walter Lee <walt@tilera.com>
7390 * config/tilegx/tilegx.c (expand_set_cint64_one_inst): Inline
7391 tests for constraint J, K, N, P.
7393 2013-03-26 Walter Lee <walt@tilera.com>
7395 Backport from mainline:
7396 2013-03-25 Walter Lee <walt@tilera.com>
7398 * config/tilegx/tilegx.c (tilegx_asm_preferred_eh_data_format):
7399 Use indirect/pcrel encoding.
7400 * config/tilepro/tilepro.c (tilepro_asm_preferred_eh_data_format):
7403 2013-03-25 Richard Biener <rguenther@suse.de>
7406 * tree-eh.c (lower_eh_must_not_throw): Strip BLOCKs from the
7407 must-not-throw stmt location.
7409 2012-03-22 Jakub Jelinek <jakub@redhat.com>
7411 * BASE-VER: Set to 4.8.1.
7412 * DEV-PHASE: Set to prerelease.
7414 2013-03-22 Release Manager
7416 * GCC 4.8.0 released.
7418 2013-03-21 Walter Lee <walt@tilera.com>
7420 * config/tilegx/sync.md (atomic_test_and_set): New pattern.
7422 2013-03-21 Mark Wielaard <mjw@redhat.com>
7424 * dwarf2out.c (size_of_aranges): Skip DECL_IGNORED_P functions.
7426 2013-03-20 Richard Biener <rguenther@suse.de>
7428 PR tree-optimization/56661
7429 * tree-ssa-sccvn.c (visit_use): Only value-number calls if
7430 the result does not have to be distinct.
7432 2013-03-20 Jakub Jelinek <jakub@redhat.com>
7434 PR tree-optimization/56635
7435 * tree-ssa-phiopt.c (cond_if_else_store_replacement_1): Give up
7436 if lhs of then_assign and else_assign don't have compatible types.
7438 2013-03-17 Jakub Jelinek <jakub@redhat.com>
7441 * config/arm/arm.h (REG_CLASS_NAMES): Add "SFP_REG" and "AFP_REG"
7442 class names. Remove trailing comma after "ALL_REGS".
7444 2013-03-16 Jakub Jelinek <jakub@redhat.com>
7446 * DEV-PHASE: Set to prerelease.
7448 2013-03-14 Andi Kleen <ak@linux.intel.com>
7451 * doc/extend.texi: Document __ATOMIC_HLE_ACQUIRE,
7452 __ATOMIC_HLE_RELEASE. Document __builtin_ia32 TSX intrincs.
7453 Document _x* TSX intrinsics.
7455 2013-03-14 Edgar E. Iglesias <edgar.iglesias@xilinx.com>
7456 David Holsgrove <david.holsgrove@xilinx.com>
7458 * configure.ac: Add MicroBlaze TLS support detection.
7459 * configure: Regenerate.
7460 * config/microblaze/microblaze-protos.h
7461 (microblaze_cannot_force_const_mem, microblaze_tls_referenced_p,
7462 symbol_mentioned_p, label_mentioned_p): Add prototypes.
7463 * config/microblaze/microblaze.c (microblaze_address_type): Add
7464 ADDRESS_TLS and tls_reloc address types.
7465 (microblaze_address_info): Add tls_reloc.
7466 (TARGET_HAVE_TLS): Define.
7467 (get_tls_get_addr, microblaze_tls_symbol_p, microblaze_tls_operand_p_1,
7468 microblaze_tls_referenced_p, microblaze_cannot_force_const_mem,
7469 symbol_mentioned_p, label_mentioned_p, tls_mentioned_p,
7470 load_tls_operand, microblaze_call_tls_get_addr,
7471 microblaze_legitimize_tls_address): New functions.
7472 (microblaze_classify_unspec): Handle UNSPEC_TLS.
7473 (get_base_reg): Use microblaze_tls_symbol_p.
7474 (microblaze_classify_address): Handle TLS.
7475 (microblaze_legitimate_pic_operand): Use symbol_mentioned_p,
7476 label_mentioned_p and microblaze_tls_referenced_p.
7477 (microblaze_legitimize_address): Handle TLS.
7478 (microblaze_address_insns): Handle ADDRESS_TLS.
7479 (pic_address_needs_scratch): Handle TLS.
7480 (print_operand_address): Handle TLS.
7481 (microblaze_expand_prologue): Check TLS_NEEDS_GOT.
7482 (microblaze_expand_move): Handle TLS.
7483 (microblaze_legitimate_constant_p): Check
7484 microblaze_cannot_force_const_mem and microblaze_tls_symbol_p.
7485 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
7486 * config/microblaze/microblaze.h (TLS_NEEDS_GOT): Define
7487 (PIC_OFFSET_TABLE_REGNUM): Set.
7488 * config/microblaze/linux.h (TLS_NEEDS_GOT): Define.
7489 * config/microblaze/microblaze.md (UNSPEC_TLS): Define.
7490 (addsi3, movsi_internal2, movdf_internal): Update constraints
7491 * config/microblaze/predicates.md (arith_plus_operand): Define
7492 (move_operand): Redefine as move_src_operand, check
7493 microblaze_tls_referenced_p.
7495 2013-03-14 Ian Bolton <ian.bolton@arm.com>
7497 * config/aarch64/aarch64.md: (*and<mode>3nr_compare0): Use CC_NZ.
7498 (*and_<SHIFT:optab><mode>3nr_compare0): Likewise.
7500 2013-03-14 Ian Bolton <ian.bolton@arm.com>
7502 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return correct
7505 2013-03-14 Jakub Jelinek <jakub@redhat.com>
7507 PR tree-optimization/53265
7508 * common.opt (Waggressive-loop-optimizations): New option.
7509 * tree-ssa-loop-niter.c: Include tree-pass.h.
7510 (do_warn_aggressive_loop_optimizations): New function.
7511 (record_estimate): Call it. Don't add !is_exit bounds to loop->bounds
7512 if number_of_latch_executions returned constant.
7513 (estimate_numbers_of_iterations_loop): Call number_of_latch_executions
7514 early. If number_of_latch_executions returned constant, set
7515 nb_iterations_upper_bound back to it.
7516 * cfgloop.h (struct loop): Add warned_aggressive_loop_optimizations
7518 * Makefile.in (tree-ssa-loop-niter.o): Depend on $(TREE_PASS_H).
7519 * doc/invoke.texi (-Wno-aggressive-loop-optimizations): Document.
7521 * config/aarch64/t-aarch64-linux (MULTARCH_DIRNAME): Remove.
7522 (MULTILIB_OSDIRNAMES): Set.
7523 * genmultilib: If defaultosdirname doesn't start with :: , set
7524 defaultosdirname2 instead, clear it and emit two . multilib_raw
7525 entries instead of just one.
7527 2013-03-14 Kaz Kojima <kkojima@gcc.gnu.org>
7529 * config/sh/linux.h (TARGET_DEFAULT): Remove MASK_USERMODE.
7530 (SUBTARGET_OVERRIDE_OPTIONS): Set TARGET_USERMODE as default.
7531 * config/sh/netbsd-elf.h (TARGET_DEFAULT): Remove MASK_USERMODE.
7532 (SUBTARGET_OVERRIDE_OPTIONS): New.
7534 2013-03-13 Oleg Endo <olegendo@gcc.gnu.org>
7537 * config/sh/sh.opt (FPU_SINGLE_ONLY): New mask.
7538 (musermode): Convert to Var(TARGET_USERMODE).
7539 * config/sh/sh.h (SELECT_SH2A_SINGLE_ONLY, SELECT_SH4_SINGLE_ONLY,
7540 MASK_ARCH): Add MASK_FPU_SINGLE_ONLY.
7541 * config/sh/sh.c (sh_option_override): Use
7542 TARGET_FPU_DOUBLE || TARGET_FPU_SINGLE_ONLY for call-fp case.
7543 * config/sh/sh.md (udivsi3_i1, divsi3_i1): Remove ! TARGET_SH4
7545 (udivsi3_i4, divsi3_i4): Use TARGET_FPU_DOUBLE condition instead of
7547 (udivsi3_i4_single, divsi3_i4_single): Use
7548 TARGET_FPU_SINGLE_ONLY || TARGET_FPU_DOUBLE instead of TARGET_HARD_SH4.
7550 2013-03-13 Dave Korn <dave.korn.cygwin@....>
7552 * config/i386/cygwin.h (SHARED_LIBGCC_SPEC): Make shared libgcc the
7555 2013-03-13 Richard Biener <rguenther@suse.de>
7557 PR tree-optimization/56608
7558 * tree-vect-slp.c (vect_schedule_slp): Do not remove scalar
7559 calls when vectorizing basic-blocks.
7561 2013-03-13 Jakub Jelinek <jakub@redhat.com>
7564 * config.gcc: On arm, mips, sh and sparc add vxworks-dummy.h to
7567 2013-03-12 Jakub Jelinek <jakub@redhat.com>
7569 * doc/invoke.texi (-Waddr-space-convert): Move into the table earlier.
7571 2013-03-11 Jan Hubicka <jh@suse.cz>
7574 * lto-streamer-out.c (output_symbol_p): Skip references from
7575 constructors of external variables.
7577 2013-03-11 Jan Hubicka <jh@suse.cz>
7580 * valtrack.c (cleanup_auto_inc_dec): Unshare clobbers originating
7582 * emit-rtl.c (verify_rtx_sharing): Likewise.
7583 (copy_insn_1): Likewise.
7584 * rtl.c (copy_rtx): Likewise.
7586 2013-03-11 Georg-Johann Lay <avr@gjlay.de>
7589 * config/avr/avr.c (avr_print_operand): Add space after '%c' in
7590 output_operand_lossage message.
7592 2013-03-11 Richard Earnshaw <rearnsha@arm.com>
7595 * arm.c (shift_op): Validate RTL pattern on the fly.
7596 (arm_print_operand, case 'S'): Don't use shift_operator to validate
7599 2013-03-10 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
7602 * config/pa/pa.md (call_value): Check for calls to powf and direct to
7603 new call patterns that clobber %fr12.
7604 (call_val_powf, call_val_powf_pic, call_val_powf_64bit): New insn,
7605 split and postreload patterns.
7606 * config/pa/pa.c (pa_conditional_register_usage): Revert marking
7607 registers %fr12 and %fr12R as call used.
7609 2013-03-09 Steven Bosscher <steven@gcc.gnu.org>
7611 * dse.c (delete_dead_store_insn): Respect TDF_DETAILS.
7612 (canon_address, record_store, replace_read, check_mem_read_rtx,
7613 scan_insn, dse_step1, dse_step2_init, dse_step2_spill,
7614 dse_step4, dse_step5_nospill, dse_step5_spill, dse_step6,
7615 rest_of_handle_dse): Likewise.
7617 2013-03-09 Richard Sandiford <rdsandiford@googlemail.com>
7620 * tree.h (tree_optimization_option): Rename target_optabs to optabs.
7622 (TREE_OPTIMIZATION_OPTABS): Update after previous field change.
7623 (TREE_OPTIMIZATION_BASE_OPTABS): New macro.
7624 (save_optabs_if_changed): Replace with...
7625 (init_tree_optimization_optabs): ...this.
7626 * optabs.c (save_optabs_if_changed): Rename to...
7627 (init_tree_optimization_optabs): ...this. Take the optimization node
7628 as argument. Do nothing if the base optabs are already correct.
7629 Reuse the existing TREE_OPTIMIZATION_OPTABS memory if we need
7630 to recompute optabs.
7631 * function.h (function): Remove optabs field.
7632 * function.c (invoke_set_current_function_hook): Call
7633 init_tree_optimization_optabs. Use the result to initialize
7636 2013-02-27 Aldy Hernandez <aldyh@redhat.com>
7638 * trans-mem.c (expand_transaction): Do not set PR_INSTRUMENTEDCODE
7639 if GTMA_HAS_NO_INSTRUMENTATION.
7640 (generate_tm_state): Keep GTMA_HAS_NO_INSTRUMENTATION bit.
7641 (ipa_tm_transform_transaction): Set GTMA_HAS_NO_INSTRUMENTATION.
7642 * gimple.h (GTMA_HAS_NO_INSTRUMENTATION): Define.
7643 * gimple-pretty-print.c (dump_gimple_transaction): Handle
7644 GTMA_HAS_NO_INSTRUMENTATION.
7646 2013-03-08 Jakub Jelinek <jakub@redhat.com>
7648 * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Don't link against
7651 2013-03-08 Marek Polacek <polacek@redhat.com>
7652 Jakub Jelinek <jakub@redhat.com>
7654 PR tree-optimization/56478
7655 * predict.c (is_comparison_with_loop_invariant_p): Change the
7656 type of loop_step to tree.
7657 (predict_loops): Adjust.
7658 (predict_iv_comparison): Perform the computations on double_ints.
7660 2013-03-08 Richard Biener <rguenther@suse.de>
7662 PR tree-optimization/56570
7663 * tree-cfg.c (verify_expr_location_1): Verify locations for
7665 * tree-sra.c (create_access_replacement): Strip locations
7666 from DECL_DEBUG_EXPRs.
7668 2013-03-08 Richard Biener <rguenther@suse.de>
7670 * tree-inline.c (expand_call_inline): Do not associate
7671 a BLOCK with the location in BLOCK_SOURCE_LOCATION.
7672 * tree-cfg.c (verify_location): Verify BLOCK_SOURCE_LOCATION.
7674 2013-03-08 Richard Biener <rguenther@suse.de>
7676 * tree-ssa-ter.c (is_replaceable_p): Do not TER across location
7677 or block changes with -Og. Fix for location / block encoding
7678 changes and PHI arguments with locations.
7680 2013-03-07 Steven Bosscher <steven@gcc.gnu.org>
7682 * bitmap.c (struct bitmap_descriptor_d): Use unsigned HOST_WIDEST_INT
7684 (struct output_info): Likewise.
7685 (register_overhead): Remove bad gcc_assert.
7686 (bitmap_find_bit): If there is only a single bitmap element, do not
7687 count a miss as a search.
7688 (print_statistics): Update for counter type changes.
7689 (dump_bitmap_statistics): Likewise. Print headers such that they
7690 are properly lined up with the printed counters.
7692 2013-03-07 Jakub Jelinek <jakub@redhat.com>
7694 PR tree-optimization/56559
7695 * tree-ssa-reassoc.c (zero_one_operation): When looking at rhs2,
7696 check that it has only a single use.
7698 2013-03-07 Richard Biener <rguenther@suse.de>
7700 * doc/invoke.texi (fwhole-program): Discourage use in combination
7703 2013-03-06 Jakub Jelinek <jakub@redhat.com>
7705 * config/arm/t-arm (TM_H, OPTIONS_H_EXTRA): Add arm-cores.def.
7707 PR tree-optimization/56539
7708 * tree-tailcall.c (adjust_return_value_with_ops): Use GSI_SAME_STMT
7709 instead of GSI_CONTINUE_LINKING as last argument to
7710 force_gimple_operand_gsi. Adjust function comment.
7712 * config/aarch64/t-aarch64 (TM_H, OPTIONS_H_EXTRA): Add
7716 * expr.c (expand_cond_expr_using_cmove): When expanding cmove in
7717 promoted mode, convert the result back to the original mode.
7719 2013-03-06 Richard Biener <rguenther@suse.de>
7722 * tree-into-ssa.c (insert_phi_nodes_for): Add dumping.
7723 (insert_updated_phi_nodes_compare_uids): New function.
7724 (update_ssa): Sort symbols_to_rename after UID before
7725 traversing it to insert PHI nodes.
7727 2013-03-06 Richard Biener <rguenther@suse.de>
7730 * tree-vect-data-refs.c (vect_can_force_dr_alignment_p):
7731 Do not adjust alignment of DECL_IN_CONSTANT_POOL decls.
7734 2013-02-13 Richard Biener <rguenther@suse.de>
7737 * varasm.c (output_constant_def_1): Get the decl representing
7738 the constant as argument.
7739 (output_constant_def): Wrap output_constant_def_1.
7740 (make_decl_rtl): Use output_constant_def_1 with the decl
7741 representing the constant.
7742 (build_constant_desc): Optionally re-use a decl already
7743 representing the constant.
7744 (tree_output_constant_def): Adjust.
7746 2013-03-06 Joey Ye <joey.ye@arm.com>
7749 * gcc.c (convert_white_space): New function.
7750 (main): Handles white space in function name.
7752 2013-03-06 Oleg Endo <olegendo@gcc.gnu.org>
7755 * config/sh/sh.c (sh_option_override): Check for TARGET_DYNSHIFT
7756 instead of TARGET_SH2 for call-table case. Do not set sh_div_strategy
7757 to SH_DIV_CALL_TABLE for TARGET_SH2.
7758 * config.gcc (sh_multilibs): Add m2 and m2a to sh*-*-linux* multilib
7760 * doc/invoke.texi (SH options): Document mdiv= call-div1, call-fp,
7763 2013-03-05 Sterling Augustine <saugustine@google.com>
7764 Cary Coutant <ccoutant@google.com>
7767 * dwarf2out.c (resolve_addr): Don't call
7768 remove_loc_list_addr_table_entries a second time for the same
7771 2013-03-05 Jakub Jelinek <jakub@redhat.com>
7774 * cfgexpand.c (expand_debug_parm_decl): Call copy_rtx on incoming.
7775 (avoid_complex_debug_insns): New function.
7776 (expand_debug_locations): Call it.
7778 PR rtl-optimization/56484
7779 * ifcvt.c (noce_process_if_block): If else_bb is NULL, avoid extending
7780 lifetimes of hard registers on small register class machines.
7782 2013-03-05 David Holsgrove <david.holsgrove@xilinx.com>
7784 * config/microblaze/microblaze-protos.h: Rename
7785 microblaze_is_interrupt_handler to microblaze_is_interrupt_variant.
7786 * config/microblaze/microblaze.c (microblaze_attribute_table): Add
7788 (microblaze_fast_interrupt_function_p): New function.
7789 (microblaze_is_interrupt_handler): Rename to
7790 microblaze_is_interrupt_variant and add fast_interrupt check.
7791 (microblaze_must_save_register): Use microblaze_is_interrupt_variant.
7792 (save_restore_insns): Likewise.
7793 (compute_frame_size): Likewise.
7794 (microblaze_function_prologue): Add FAST_INTERRUPT_NAME.
7795 (microblaze_globalize_label): Likewise.
7796 * config/microblaze/microblaze.h: Define FAST_INTERRUPT_NAME.
7797 * config/microblaze/microblaze.md: Use wrapper
7798 microblaze_is_interrupt_variant.
7800 2013-03-05 Kai Tietz <ktietz@redhat.com>
7802 * sdbout.c (sdbout_one_type): Switch to current function's section
7803 supporting cold/hot.
7805 2013-03-05 David Holsgrove <david.holsgrove@xilinx.com>
7807 * doc/invoke.texi (MicroBlaze): Add -mbig-endian, -mlittle-endian,
7810 2013-03-05 Jakub Jelinek <jakub@redhat.com>
7813 * ggc-common.c (gt_pch_save): For ENABLE_VALGRIND_CHECKING,
7814 if VALGRIND_GET_VBITS is defined, temporarily make object
7815 memory all defined, and restore previous valgrind addressability
7816 and definability afterwards. Free this_object at the end.
7819 * lra.c (lra): Call lra_clear_live_ranges if live_p,
7820 right before calling lra_create_live_ranges, also call it
7821 when clearing live_p. Only call lra_clear_live_ranges
7822 at the end if live_p.
7825 * sched-deps.c (delete_dep_node): Free DEP_REPLACE.
7827 2013-03-05 Richard Biener <rguenther@suse.de>
7829 PR tree-optimization/56521
7830 * tree-ssa-sccvn.c (set_value_id_for_result): Always initialize
7833 2013-03-05 Steven Bosscher <steven@gcc.gnu.org>
7836 * except.h (remove_unreachable_eh_regions): New prototype.
7837 * except.c (remove_eh_handler_splicer): New function, split out
7838 of remove_eh_handler.
7839 (remove_eh_handler): Use remove_eh_handler_splicer. Add comment
7840 warning about running it on many EH regions one at a time.
7841 (remove_unreachable_eh_regions_worker): New function, walk the
7842 EH tree in depth-first order and remove non-marked regions.
7843 (remove_unreachable_eh_regions): New function.
7844 * tree-eh.c (mark_reachable_handlers): New function, split out
7845 from remove_unreachable_handlers.
7846 (remove_unreachable_handlers): Use mark_reachable_handlers and
7847 remove_unreachable_eh_regions.
7848 (remove_unreachable_handlers_no_lp): Use mark_reachable_handlers
7849 and remove_unreachable_eh_regions.
7851 2013-03-05 Richard Biener <rguenther@suse.de>
7854 * loop-init.c (fix_loop_structure): Remove loops in two stages,
7855 not freeing them until the end.
7857 2013-03-05 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
7859 * config/s390/s390.h: Define DWARF2_ASM_LINE_DEBUG_INFO.
7861 2013-03-05 Richard Biener <rguenther@suse.de>
7863 PR tree-optimization/56270
7864 * tree-vect-slp.c (vect_schedule_slp): Clear vectorized stmts
7865 of loads after scheduling an SLP instance.
7867 2013-03-05 Jakub Jelinek <jakub@redhat.com>
7869 * Makefile.in (dg_target_exps): Add aarch64.exp, epiphany.exp and
7871 (check_gcc_parallelize): Run guality.exp as a separate job from
7872 vect.exp with unsorted.exp and $(dg_target_exps) separately from
7873 struct-layout-1.exp with stackalign.exp.
7875 * alias.c (init_alias_analysis): Clear reg_known_equiv_p bitmap.
7878 * tree-vect-slp.c (vect_supported_load_permutation_p): Free
7879 load_index sbitmap even if some bit in it isn't set.
7882 * tree-ssa-loop-niter.c (bb_queue): Remove typedef.
7883 (discover_iteration_bound_by_body_walk): Change queues to
7884 vec<vec<basic_block> > and queue to vec<basic_block>. Fix up
7885 spelling in comment. Call safe_push on queues[bound_index] directly.
7886 Release queues[queue_index] in every iteration unconditionally.
7887 Release bounds vector.
7890 * tree-vect-stmts.c (free_stmt_vec_info_vec): Call
7891 free_stmt_vec_info on any left-over stmt_vec_info in the vector.
7892 * tree-vect-loop.c (vect_create_epilog_for_reduction): Release
7895 2013-03-05 Richard Biener <rguenther@suse.de>
7898 * tree-inline.c (remap_blocks_to_null): New function.
7899 (expand_call_inline): When expanding a call stmt without
7900 an associated block inline remap all callee blocks to NULL.
7902 2013-03-05 Jakub Jelinek <jakub@redhat.com>
7904 PR rtl-optimization/56494
7905 * simplify-rtx.c (simplify_truncation): If C is narrower than A,
7906 optimize (truncate:A (subreg:B (truncate:C X) 0)) into
7907 (subreg:A (truncate:C X) 0) instead of (truncate:A X).
7910 * sel-sched-ir.c (free_sched_pools): Release
7911 succs_info_pool.stack[succs_info_pool.max_top] vectors too
7912 if succs_info_pool.max_top isn't -1.
7915 * opts.c (opts_obstack, opts_concat): Moved to...
7916 * opts-common.c (opts_obstack, opts_concat): ... here.
7918 2013-03-04 Jakub Jelinek <jakub@redhat.com>
7921 * diagnostic.c (diagnostic_append_note): Save and restore old prefix.
7923 2013-03-04 Martin Jambor <mjambor@suse.cz>
7925 * tree-dfa.c (get_or_create_ssa_default_def): Use parameter fn in
7926 all appropriate places.
7928 2013-01-04 Eric Botcazou <ebotcazou@adacore.com>
7930 PR tree-optimization/56424
7931 * ipa-split.c (split_function): Do not set the RSO flag if result is
7932 not by reference and its type is a register type.
7934 2013-03-04 David Holsgrove <david.holsgrove@xilinx.com>
7936 * config/microblaze/microblaze.c (microblaze_valid_pic_const): New.
7937 (microblaze_legitimate_pic_operand): Likewise.
7938 * config/microblaze/microblaze.h (LEGITIMATE_PIC_OPERAND_P): Call
7939 new function microblaze_legitimate_pic_operand.
7940 * config/microblaze/microblaze-protos.h
7941 (microblaze_legitimate_pic_operand): Declare.
7943 2013-03-04 Edgar E. Iglesias <edgar.iglesias@gmail.com>
7945 * config/microblaze/predicates.md (call_insn_simple_operand):
7946 New predicate for supported rtx code types.
7947 * config/microblaze/microblaze.md (call_internal1): Use
7948 call_insn_simple_operand predicate.
7950 2013-03-04 Jakub Jelinek <jakub@redhat.com>
7953 * tree-loop-distribution.c (ldist_gen): Call partition_free after each
7954 partitions.ordered_remove.
7957 * tree-vect-stmts.c (vectorizable_conversion): Don't call
7958 vec_oprnds0.create (1) for modifier == NONE.
7961 * tree-vect-stmts.c (vectorizable_shift): Don't call create methods
7962 on vec_oprnds0 or vec_oprnds1 before loop, only call it on
7963 vec_oprnds1 right before pushing anything to it for
7967 * tree-vect-loop.c (destroy_loop_vec_info): For !clean_stmts, just
7968 set nbbs to 0 instead of having separate code path.
7969 (vect_analyze_loop_form): Call destroy_loop_vec_info with true
7970 instead of false as last argument if returning NULL.
7972 2013-03-03 Sandra Loosemore <sandra@codesourcery.com>
7974 * target.def (TARGET_OPTION_VALID_ATTRIBUTE_P): Update comments;
7975 the attribute is now called "target" instead of "option".
7976 (TARGET_OPTION_PRAGMA_PARSE): Likewise, for the pragma.
7977 * doc/tm.texi.in (Target Attributes): Likewise document the correct
7978 attribute/pragma name for TARGET_OPTION_VALID_P and
7979 TARGET_OPTION_PRAGMA_PARSE. Also copy-edit and correct markup.
7980 * doc/tm.texi: Regenerated.
7982 2013-03-02 David Holsgrove <david.holsgrove@xilinx.com>
7984 * config/microblaze/microblaze.c:
7985 Check mcpu, pcmp requirement and set TARGET_REORDER to 0 if not met.
7986 * config/microblaze/microblaze.h: Add -mxl-reorder to
7988 * config/microblaze/microblaze.md: New bswapsi2 and bswaphi2.
7989 instructions emitted if TARGET_REORDER.
7990 * config/microblaze/microblaze.opt: New option -mxl-reorder set to 1
7991 or 0 for -m/-mno case, but initialises as 2 to detect default use case
7994 2013-03-01 Xinliang David Li <davidxl@google.com>
7996 * tree-ssa-uninit.c (compute_control_dep_chain): Limit post-dom
7999 2013-03-01 Jakub Jelinek <jakub@redhat.com>
8002 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Release path
8003 vector even when returning true. Fix up function comment formatting.
8006 * ira-build.c (ira_loop_nodes_count): New variable.
8007 (create_loop_tree_nodes): Initialize it.
8008 (finish_loop_tree_nodes): Use it instead of looking at current_loops.
8011 * tree-vect-data-refs.c (vect_permute_store_chain): Avoid using copy
8012 method on dr_chain and result_chain.
8013 * tree-vect-stmts.c (vectorizable_store): Only call
8014 result_chain.create if j == 0.
8017 * tree-vect-stmts.c (vect_create_vectorized_promotion_stmts): Call
8018 vec_oprnds0->release (); rather than vec_oprnds0->truncate (0)
8019 before overwriting it.
8021 2013-03-01 Tobias Burnus <burnus@net-b.de>
8023 * doc/extended.texi (C Extensions): Change order in @menu
8025 (Other MIPS Built-in Functions): Move last MIPS entry before
8026 "picoChip Built-in Functions".
8027 (SH Built-in Functions): Move after RX Built-in Functions.
8028 * doc/gcc.texi (Introduction): Change order in @menu to match @node.
8029 * doc/md.texi (Constraints): Ditto.
8030 * gty.texi (Type Information): Ditto.
8031 (User-provided marking routines for template types): Make subsection.
8032 * doc/invoke.texi (AArch64 Options): Move before
8033 "Adapteva Epiphany Options".
8035 2013-02-28 Konstantin Serebryany <konstantin.s.serebryany@gmail.com>
8036 Jakub Jelinek <jakub@redhat.com>
8039 * asan.c (gate_asan): Lookup no_sanitize_address instead of
8040 no_address_safety_analysis attribute.
8041 * doc/extend.texi (no_address_safety_attribute): Rename to
8042 no_sanitize_address attribute, mention no_address_safety_analysis
8043 attribute as deprecated alias.
8045 2013-02-28 Jakub Jelinek <jakub@redhat.com>
8048 * tree-vectorizer.h (vect_get_slp_defs): Change 3rd argument
8049 type to vec<vec<tree> > *.
8050 * tree-vect-slp.c (vect_get_slp_defs): Likewise. Change vec_defs
8051 to be vec<tree> instead of vec<tree> *, set vec_defs
8052 to vNULL and call vec_defs.create (number_of_vects), adjust other
8054 * tree-vect-stmts.c (vect_get_vec_defs, vectorizable_call,
8055 vectorizable_condition): Adjust vect_get_slp_defs callers.
8057 2013-02-28 James Greenhalgh <james.greenhalgh@arm.com>
8059 * config/aarch64/aarch64.c
8060 (aarch64_float_const_representable): Remove unused variable.
8062 2013-02-28 James Greenhalgh <james.greenhalgh@arm.com>
8064 * config/aarch64/aarch64.c (aarch64_mangle_type): Make static.
8066 2013-02-28 James Greenhalgh <james.greenhalgh@arm.com>
8068 * config/aarch64/aarch64-builtins.c
8069 (aarch64_init_simd_builtins): Make static.
8071 2013-02-28 James Greenhalgh <james.greenhalgh@arm.com>
8073 * config/aarch64/aarch64.c
8074 (aarch64_simd_make_constant): Make static.
8076 2013-02-28 Martin Jambor <mjambor@suse.cz>
8078 * tree-sra.c (load_assign_lhs_subreplacements): Do not put replacements
8079 with no initialization to the RHS of debug statements.
8081 2013-02-28 Martin Jambor <mjambor@suse.cz>
8083 PR tree-optimization/56294
8084 * tree-sra.c (analyze_access_subtree): Create replacement declarations.
8086 (get_access_replacement): Do not call create_access_replacement.
8087 Assert a replacement exists.
8088 (get_repl_default_def_ssa_name): Create the replacement declaration
8091 2013-02-28 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8093 * config/arm/arm.c (arm_output_mi_thunk): Call final_start_function and
8096 2013-02-28 Marek Polacek <polacek@redhat.com>
8098 PR rtl-optimization/56466
8099 * loop-unroll.c (unroll_and_peel_loops): Call fix_loop_structure
8100 if we're changing a loop.
8101 (peel_loops_completely): Likewise.
8103 2013-02-28 Paolo Carlini <paolo.carlini@oracle.com>
8106 * doc/invoke.texi ([-Wctor-dtor-privacy]): Complete.
8108 2013-02-28 Georg-Johann Lay <avr@gjlay.de>
8111 * config/avr/avr.c (avr_init_builtins): Use 'n' instead of empty
8112 macro parameters with: FX_FTYPE_FX, FX_FTYPE_FX_INT, INT_FTYPE_FX,
8113 INTX_FTYPE_FX, FX_FTYPE_INTX.
8114 * config/avr/builtins.def: Adjust respective DEF_BUILTIN.
8116 2013-02-28 Georg-Johann Lay <avr@gjlay.de>
8118 * avr/avr-mcus.def (ata5272, ata5505, attiny1634, ata6285)
8119 (ata6286, atmega8a, atmega48pa, ata5790, ata5790n, ata5795)
8120 (atmega164pa, atmega165pa, atmega168pa, atmega16hva, atmega16hvb)
8121 (atmega16hvbrevb, atmega16m1, atmega16u4, atmega26hvg, atmega32a)
8122 (atmega32a, atmega3250pa, atmega3290pa, atmega32c1, atmega32m1)
8123 (atmega32u4, atmega32u6, atmega64a, atmega6490a, atmega6490p)
8124 (atmega64c1, atmega64m1, atmega64rfa2, atmega64rfr2, atmega32hvb)
8125 (atmega32hvbrevb, atmega16hva2, atmega48hvf, at90pwm161)
8126 (atmega128a, atmega1284, atmxt112sl, atmxt224, atmxt224e)
8127 (atmxt336s, atxmega16a4u, atxmega16c4, atxmega32a4u, atxmega32c4)
8128 (atxmega32e5, atxmega64a3u, atxmega64a4u, atxmega64b1, atxmega64b3)
8129 (atxmega64c3, atxmega64d4, atxmega128a3u, atxmega128b1)
8130 (atxmega128b3, atxmega128c3, atxmega128d4, atmxt540s, atmxt540sreva)
8131 (atxmega192a3u, atxmega192c3, atxmega256a3u, atxmega256c3)
8132 (atxmega384c3, atxmega384d3, atxmega128a4u): New AVR_MCU.
8133 (avrxmega6): Increase max flash segments from 5 to 6.
8134 * config/avr/t-multilib: Regenerate.
8135 * config/avr/avr-tables.opt: Regenerate.
8136 * doc/avr-mmcu.texi: Regenerate.
8138 2013-02-28 Georg-Johann Lay <avr@gjlay.de>
8140 * config/avr/avr.h (device_to_arch): Rename to device_to_ld.
8141 (avr_device_to_arch): Rename to avr_device_to_ld.
8142 (avr_device_to_as): New prototype.
8143 (EXTRA_SPEC_FUNCTIONS): Add device_to_as.
8144 (ASM_SPEC): Use device_to_as to get -mmcu= and -mno-skip-bug=.
8145 * config/avr/driver-avr.c (avr_device_to_as): New.
8146 (avr_device_to_arch): Rename to avr_device_to_ld.
8148 2013-02-27 Jakub Jelinek <jakub@redhat.com>
8151 * tree-vect-data-refs.c (vect_permute_load_chain): Avoid using copy
8152 method on dr_chain and result_chain.
8155 * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Call
8156 pointer_set_destroy on not_executed_last_iteration.
8159 * tree-vect-loop.c (vectorizable_reduction): Release vect_defs vector.
8162 * ipa-pure-const.c (propagate): Use FOR_EACH_FUNCTION instead of
8163 FOR_EACH_DEFINED_FUNCTION when freeing state.
8166 * df-scan.c (df_insn_delete): Use df_scan_free_mws_vec before
8168 (df_insn_rescan_debug_internal): Use df_scan_free_mws_vec before
8172 * ipa-cp.c (decide_whether_version_node): Call vec_free on
8173 known_aggs[i].items and release known_aggs vector.
8176 * ipa-reference.c (propagate): Free node_info even for alias nodes.
8178 2013-02-27 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8180 * config/microblaze/microblaze.c (microblaze_emit_compare):
8181 Use xor for EQ/NE comparisions.
8182 * config/microblaze/microblaze.md (cstoresf4): Add constraints
8183 (cbranchsf4): Adjust operator to comparison_operator.
8185 2013-02-27 Jakub Jelinek <jakub@redhat.com>
8188 * tree-flow.h (edge_var_map_vector): Change into va_heap, vl_embed
8190 * tree-ssa.c (redirect_edge_var_map_add): Use vec_safe_reserve and
8191 vec_safe_push, always update *slot.
8192 (redirect_edge_var_map_clear): Use vec_free.
8193 (redirect_edge_var_map_dup): Use vec_safe_copy and vec_safe_reserve.
8194 (free_var_map_entry): Use vec_free.
8195 * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Use
8196 FOR_EACH_VEC_SAFE_ELT instead of FOR_EACH_VEC_ELT.
8198 2013-02-27 Andrey Belevantsev <abel@ispras.ru>
8201 * sel-sched-ir.c (merge_expr): Also change vinsn of merged expr
8202 when the may_trap_p bit of the exprs being merged differs.
8203 Reorder tests for speculativeness in the logical and operator.
8205 2013-02-27 Jakub Jelinek <jakub@redhat.com>
8207 * incpath.c (add_standard_paths): Use reconcat instead of concat
8208 where appropriate and avoid leaking memory.
8210 * opts.h: Include obstack.h.
8211 (opts_concat): New prototype.
8212 (opts_obstack): New declaration.
8213 * opts.c (opts_concat): New function.
8214 (opts_obstack): New variable.
8215 (init_options_struct): Call gcc_init_obstack on opts_obstack.
8216 (finish_options): Use opts_concat instead of concat
8217 and XOBNEWVEC instead of XNEWVEC.
8218 * opts-common.c (generate_canonical_option, decode_cmdline_option,
8219 generate_option): Likewise.
8220 * Makefile.in (OPTS_H): Depend on $(OBSTACK_H).
8221 * lto-wrapper.c (main): Call gcc_init_obstack on opts_obstack.
8224 * stmt.c (expand_switch_as_decision_tree_p): If flag_pic
8225 and ASM_OUTPUT_ADDR_DIFF_ELT isn't defined, return true.
8227 2013-02-26 Jakub Jelinek <jakub@redhat.com>
8230 * lra-spills.c (lra_spill): Free spill_hard_reg at the end.
8232 2013-02-26 Joern Rennecke <joern.rennecke@embecosm.com>
8234 * config/arm/arm.c (const_ok_for_dimode_op): Back out last change.
8235 (arm_block_move_unaligned_straight): Likewise.
8236 (arm_adjust_block_mem): Likewise.
8238 2013-02-26 Joern Rennecke <joern.rennecke@embecosm.com>
8241 * config/lm32/lm32.c (gen_int_relational): Remove unused variables
8242 temp, cond and label.
8243 * config/lm32/lm32.md (ashlsi3): Remove unused variable one.
8246 * config/c6x/c6x.c (dbx_register_map): Change to unsigned.
8247 * config/c6x/c6x.h (dbx_register_map): Update declaration.
8250 * config/cr16/cr16-protos.h: Move end of RTX_CODE guard below end
8251 of prologue/epilogue functions.
8254 * config/tilegx/tilegx.c (tilegx_expand_prologue):
8255 Remove unused variable cfa_offset.
8256 * config/tilepro/tilepro.c (tilepro_expand_prologue): Likewise.
8259 * config/mn10300/mn10300.c (mn10300_expand_epilogue): Avoid offset
8260 type promotion to unsigned.
8263 * config/arm/arm.c (const_ok_for_dimode_op): Make code consistent
8264 for HOST_WIDE_INT of 32 bit / same size as int.
8265 (arm_block_move_unaligned_straight): Likewise.
8266 (arm_adjust_block_mem): Likewise.
8269 * config/mep/t-mep (mep-pragma.o): Use ALL_COMPILERFLAGS instead of
8272 2013-02-26 Marek Polacek <polacek@redhat.com>
8274 PR tree-optimization/56426
8275 * tree-ssa-loop.c (tree_ssa_loop_init): Always call
8278 2013-02-26 Richard Biener <rguenther@suse.de>
8281 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Remove
8282 unused variable loops.
8284 2013-02-26 Jakub Jelinek <jakub@redhat.com>
8286 PR tree-optimization/56448
8287 * fold-const.c (operand_equal_p) <case tcc_reference>: Don't look at
8288 TREE_SIDE_EFFECTS if flags contain OEP_CONSTANT_ADDRESS_OF.
8289 Clear OEP_CONSTANT_ADDRESS_OF from flags before recursing on second or
8290 later operands of the references, or even first operand for
8291 INDIRECT_REF, TARGET_MEM_REF or MEM_REF.
8293 PR tree-optimization/56443
8294 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): For
8295 overaligned types, pass TYPE_UNSIGNED (scalar_type) as second argument
8296 to type_for_mode langhook.
8298 2013-02-25 Matt Turner <mattst88@gmail.com>
8300 * doc/invoke.texi: Document r4700.
8302 2013-02-25 Richard Biener <rguenther@suse.de>
8304 PR tree-optimization/56175
8305 * tree-ssa-forwprop.c (hoist_conversion_for_bitop_p): New predicate,
8307 (simplify_bitwise_binary): ... here. Also guard the conversion
8308 of (type) X op CST to (type) (X op ((type-x) CST)) with it.
8310 2013-02-25 Catherine Moore <clm@codesourcery.com>
8313 2013-02-24 Catherine Moore <clm@codesourcery.com>
8314 Maciej W. Rozycki <macro@codesourcery.com>
8315 Tom de Vries <tom@codesourcery.com>
8316 Nathan Sidwell <nathan@codesourcery.com>
8317 Iain Sandoe <iain@codesourcery.com>
8318 Nathan Froyd <froydnj@codesourcery.com>
8319 Chao-ying Fu <fu@mips.com>
8321 * doc/extend.texi (micromips, nomicromips, nocompression):
8322 Document new function attributes.
8323 * doc/invoke.texi (minterlink-compressed, mmicromips,
8324 m14k, m14ke, m14kec): Document new options.
8325 (minterlink-mips16): Update documentation.
8326 * doc/md.texi (ZC, ZD): Document new constraints.
8327 * configure.ac (gcc_cv_as_micromips): Check if linker
8328 supports the .set micromips directive.
8329 * configure: Regenerate.
8330 * config.in: Regenerate.
8331 * config/mips/mips-tables.opt: Regenerate.
8332 * config/mips/micromips.md: New file.
8333 * constraints.md (ZC, AD): New constraints.
8334 * config/mips/predicates.md (movep_src_register): New predicate.
8335 (movep_src_operand): New predicate.
8336 (non_volatile_mem_operand): New predicate.
8337 * config/mips/mips.md (multimem): New type.
8338 (length): Differentiate between 17-bit and 18-bit branch offsets.
8339 (MOVEP1, MOVEP2): New mode iterator.
8340 (mov_<load>l): Use ZC constraint.
8341 (mov_<load>r): Likewise.
8342 (mov_<store>l): Likewise.
8343 (mov_<store>r): Likewise.
8344 (*branch_equality<mode>_inverted): Add microMIPS support.
8345 (*branch_equality<mode>): Likewise.
8346 (*jump_absolute): Likewise.
8347 (indirect_jump_<mode>): Likewise.
8348 (tablejump_<mode>): Likewise.
8349 (<optab>_internal): Likewise.
8350 (sibcall_internal): Likewise.
8351 (sibcall_value_internal): Likewise.
8352 (prefetch): Use constraint ZD.
8353 * config/mips/mips.opt (minterlink-compressed): New option.
8354 (minterlink-mips16): Now an alias for minterlink-compressed.
8355 (mmicromips): New option.
8356 * config/mips/sync.md (sync_compare_and_swap<mode>): Use ZR constraint.
8357 (compare_and_swap_12): Likewise.
8358 (sync_add<mode>): Likewise.
8359 (sync_<optab>_12): Likewise.
8360 (sync_old_<optab>_12): Likewise.
8361 (sync_new_<optab>_12): Likewise.
8362 (sync_nand_12): Likewise.
8363 (sync_old_nand_12): Likewise.
8364 (sync_new_nand_12): Likewise.
8365 (sync_sub<mode>): Likewise.
8366 (sync_old_add<mode>): Likewise.
8367 (sync_old_sub<mode>): Likewise.
8368 (sync_new_add<mode>): Likewise.
8369 (sync_new_sub<mode>): Likewise.
8370 (sync_<optab><mode>): Likewise.
8371 (sync_old_<optab><mode>): Likewise.
8372 (sync_new_<optab><mode>): Likewise.
8373 (sync_nand<mode>): Likewise.
8374 (sync_old_nand<mode>): Likewise.
8375 (sync_new_nand<mode>): Likewise.
8376 (sync_lock_test_and_set<mode>): Likewise.
8377 (test_and_set_12): Likewise.
8378 (atomic_compare_and_swap<mode>): Likewise.
8379 (atomic_exchange<mode>_llsc): Likewise.
8380 (atomic_fetch_add<mode>_llsc): Likewise.
8381 * config/mips/mips-cpus.def (m14kc, m14k): New processors.
8382 * config/mips/mips-protos.h (umips_output_save_restore): New prototype.
8383 (umips_save_restore_pattern_p): Likewise.
8384 (umips_load_store_pair_p): Likewise.
8385 (umips_output_load_store_pair): Likewise.
8386 (umips_movep_target_p): Likewise.
8387 (umips_12bit_offset_address_p): Likewise.
8388 * config/mips/mips.c (MIPS_MAX_FIRST_STEP): Update for microMIPS.
8389 (mips_base_mips16): Rename this...
8390 (mips_base_compression_flags): ...to this. Update all uses.
8391 (mips_attribute_table): Add micromips, nomicromips and nocompression.
8392 (mips_mips16_decl_p): Delete.
8393 (mips_nomips16_decl_p): Delete.
8394 (mips_get_compress_on_flags): New function.
8395 (mips_get_compress_off_flags): New function.
8396 (mips_get_compress_mode): New function.
8397 (mips_get_compress_on_name): New function.
8398 (mips_get_compress_off_name): New function.
8399 (mips_insert_attributes): Support multiple compression types.
8400 (mips_merge_decl_attributes): Likewise.
8401 (umips_12bit_offset_address_p): New function.
8402 (mips_start_function_definition): Emit .set micromips directive.
8403 (mips_call_may_need_jalx_p): New function.
8404 (mips_function_ok_for_sibcall): Add microMIPS support.
8405 (mips_print_operand_punctuation): Support short delay slots and
8407 (umips_swm_mask, umips_swm_encoding): New.
8408 (umips_build_save_restore): New function.
8409 (mips_for_each_saved_gpr_and_fpr): Add microMIPS support.
8410 (was_mips16_p): Remove.
8411 (old_compression_mode): New.
8412 (mips_set_compression_mode): New function.
8413 (mips_set_current_function): Add microMIPS support.
8414 (mips_option_override): Likewise.
8415 (umips_save_restore_pattern_p): New function.
8416 (umips_output_save_restore): New function.
8417 (umips_load_store_pair_p_1): New function.
8418 (umips_load_store_pair_p): New function.
8419 (umips_output_load_store_pair_1): New function.
8420 (umips_output_load_store_pair): New function.
8421 (umips_movep_target_p) New function.
8422 (mips_prepare_pch_save): Add microMIPS support.
8423 * config/mips/mips.h (TARGET_COMPRESSION): New.
8424 (TARGET_CPU_CPP_BUILTINS): Update macro
8425 to use new compression flags and to support microMIPS.
8426 (MIPS_ISA_LEVEL_SPEC): Add m14k processors.
8427 (MIPS_ARCH_FLOAT_SPEC): Likewise.
8428 (ISA_HAS_LWXS): Include TARGET_MICROMIPS.
8429 (ISA_HAS_LOAD_DELAY): Exclude TARGET_MICROMIPS.
8430 (ASM_SPEC): Support mmicromips and mno-micromips.
8431 (M16STORE_REG_P): New macro.
8432 (MIPS_CALL): Support TARGET_MICROMIPS.
8433 (MICROMIPS_J): New macro.
8434 (mips_base_mips16): Rename this...
8435 (mips_base_compression_flags): ...to this.
8436 (UMIPS_12BIT_OFFSET_P): New macro.
8437 * config/mips/t-sde: (MULTILIB_OPTIONS): Add microMIPS.
8438 (MULTILIB_DIRNAMES): Likewise.
8440 2013-02-25 Tom de Vries <tom@codesourcery.com>
8442 PR rtl-optimization/56131
8443 * insn-notes.def (INSN_NOTE_BASIC_BLOCK): Update comment.
8444 * cfgrtl.c (delete_insn): Don't reorder NOTE_INSN_DELETED_LABEL and
8445 NOTE_INSN_BASIC_BLOCK if BLOCK_FOR_INSN == NULL.
8447 2013-02-25 Tobias Burnus <burnus@net-b.de>
8449 * doc/invoke.texi (-fsanitize=): Move from optimization
8450 to debugging options.
8452 2013-02-25 Andrey Belevantsev <abel@ispras.ru>
8454 * sched-deps.c (sched_analyze_insn): Fix typo in comment.
8456 2013-02-25 Andrey Belevantsev <abel@ispras.ru>
8457 Alexander Monakov <amonakov@ispras.ru>
8460 * sched-deps.c (sched_analyze_insn): When reg_pending_barrier,
8461 flush pending lists also on non-jumps. Adjust comment.
8463 2013-02-24 Catherine Moore <clm@codesourcery.com>
8464 Maciej W. Rozycki <macro@codesourcery.com>
8465 Tom de Vries <tom@codesourcery.com>
8466 Nathan Sidwell <nathan@codesourcery.com>
8467 Iain Sandoe <iain@codesourcery.com>
8468 Nathan Froyd <froydnj@codesourcery.com>
8469 Chao-ying Fu <fu@mips.com>
8471 * doc/extend.texi (micromips, nomicromips, nocompression):
8472 Document new function attributes.
8473 * doc/invoke.texi (minterlink-compressed, mmicromips,
8474 m14k, m14ke, m14kec): Document new options.
8475 (minterlink-mips16): Update documentation.
8476 * doc/md.texi (ZC, ZD): Document new constraints.
8477 * configure.ac (gcc_cv_as_micromips): Check if linker
8478 supports the .set micromips directive.
8479 * configure: Regenerate.
8480 * config.in: Regenerate.
8481 * config/mips/mips-tables.opt: Regenerate.
8482 * config/mips/micromips.md: New file.
8483 * constraints.md (ZC, AD): New constraints.
8484 * config/mips/predicates.md (movep_src_register): New predicate.
8485 (movep_src_operand): New predicate.
8486 (non_volatile_mem_operand): New predicate.
8487 * config/mips/mips.md (multimem): New type.
8488 (length): Differentiate between 17-bit and 18-bit branch offsets.
8489 (MOVEP1, MOVEP2): New mode iterator.
8490 (mov_<load>l): Use ZC constraint.
8491 (mov_<load>r): Likewise.
8492 (mov_<store>l): Likewise.
8493 (mov_<store>r): Likewise.
8494 (*branch_equality<mode>_inverted): Add microMIPS support.
8495 (*branch_equality<mode>): Likewise.
8496 (*jump_absolute): Likewise.
8497 (indirect_jump_<mode>): Likewise.
8498 (tablejump_<mode>): Likewise.
8499 (<optab>_internal): Likewise.
8500 (sibcall_internal): Likewise.
8501 (sibcall_value_internal): Likewise.
8502 (prefetch): Use constraint ZD.
8503 * config/mips/mips.opt (minterlink-compressed): New option.
8504 (minterlink-mips16): Now an alias for minterlink-compressed.
8505 (mmicromips): New option.
8506 * config/mips/sync.md (sync_compare_and_swap<mode>): Use ZR constraint.
8507 (compare_and_swap_12): Likewise.
8508 (sync_add<mode>): Likewise.
8509 (sync_<optab>_12): Likewise.
8510 (sync_old_<optab>_12): Likewise.
8511 (sync_new_<optab>_12): Likewise.
8512 (sync_nand_12): Likewise.
8513 (sync_old_nand_12): Likewise.
8514 (sync_new_nand_12): Likewise.
8515 (sync_sub<mode>): Likewise.
8516 (sync_old_add<mode>): Likewise.
8517 (sync_old_sub<mode>): Likewise.
8518 (sync_new_add<mode>): Likewise.
8519 (sync_new_sub<mode>): Likewise.
8520 (sync_<optab><mode>): Likewise.
8521 (sync_old_<optab><mode>): Likewise.
8522 (sync_new_<optab><mode>): Likewise.
8523 (sync_nand<mode>): Likewise.
8524 (sync_old_nand<mode>): Likewise.
8525 (sync_new_nand<mode>): Likewise.
8526 (sync_lock_test_and_set<mode>): Likewise.
8527 (test_and_set_12): Likewise.
8528 (atomic_compare_and_swap<mode>): Likewise.
8529 (atomic_exchange<mode>_llsc): Likewise.
8530 (atomic_fetch_add<mode>_llsc): Likewise.
8531 * config/mips/mips-cpus.def (m14kc, m14k): New processors.
8532 * config/mips/mips-protos.h (umips_output_save_restore): New prototype.
8533 (umips_save_restore_pattern_p): Likewise.
8534 (umips_load_store_pair_p): Likewise.
8535 (umips_output_load_store_pair): Likewise.
8536 (umips_movep_target_p): Likewise.
8537 (umips_12bit_offset_address_p): Likewise.
8538 * config/mips/mips.c (MIPS_MAX_FIRST_STEP): Update for microMIPS.
8539 (mips_base_mips16): Rename this...
8540 (mips_base_compression_flags): ...to this. Update all uses.
8541 (mips_attribute_table): Add micromips, nomicromips and nocompression.
8542 (mips_mips16_decl_p): Delete.
8543 (mips_nomips16_decl_p): Delete.
8544 (mips_get_compress_on_flags): New function.
8545 (mips_get_compress_off_flags): New function.
8546 (mips_get_compress_mode): New function.
8547 (mips_get_compress_on_name): New function.
8548 (mips_get_compress_off_name): New function.
8549 (mips_insert_attributes): Support multiple compression types.
8550 (mips_merge_decl_attributes): Likewise.
8551 (umips_12bit_offset_address_p): New function.
8552 (mips_start_function_definition): Emit .set micromips directive.
8553 (mips_call_may_need_jalx_p): New function.
8554 (mips_function_ok_for_sibcall): Add microMIPS support.
8555 (mips_print_operand_punctuation): Support short delay slots and
8557 (umips_swm_mask, umips_swm_encoding): New.
8558 (umips_build_save_restore): New function.
8559 (mips_for_each_saved_gpr_and_fpr): Add microMIPS support.
8560 (was_mips16_p): Remove.
8561 (old_compression_mode): New.
8562 (mips_set_compression_mode): New function.
8563 (mips_set_current_function): Add microMIPS support.
8564 (mips_option_override): Likewise.
8565 (umips_save_restore_pattern_p): New function.
8566 (umips_output_save_restore): New function.
8567 (umips_load_store_pair_p_1): New function.
8568 (umips_load_store_pair_p): New function.
8569 (umips_output_load_store_pair_1): New function.
8570 (umips_output_load_store_pair): New function.
8571 (umips_movep_target_p) New function.
8572 (mips_prepare_pch_save): Add microMIPS support.
8573 * config/mips/mips.h (TARGET_COMPRESSION): New.
8574 (TARGET_CPU_CPP_BUILTINS): Update macro
8575 to use new compression flags and to support microMIPS.
8576 (MIPS_ISA_LEVEL_SPEC): Add m14k processors.
8577 (MIPS_ARCH_FLOAT_SPEC): Likewise.
8578 (ISA_HAS_LWXS): Include TARGET_MICROMIPS.
8579 (ISA_HAS_LOAD_DELAY): Exclude TARGET_MICROMIPS.
8580 (ASM_SPEC): Support mmicromips and mno-micromips.
8581 (M16STORE_REG_P): New macro.
8582 (MIPS_CALL): Support TARGET_MICROMIPS.
8583 (MICROMIPS_J): New macro.
8584 (mips_base_mips16): Rename this...
8585 (mips_base_compression_flags): ...to this.
8586 (UMIPS_12BIT_OFFSET_P): New macro.
8587 * config/mips/t-sde: (MULTILIB_OPTIONS): Add microMIPS.
8588 (MULTILIB_DIRNAMES): Likewise.
8590 2013-02-24 Jakub Jelinek <jakub@redhat.com>
8593 * target-globals.c (save_target_globals): For init_reg_sets and
8594 target_reinit remporarily set this_fn_optabs to this_target_optabs.
8596 2013-02-22 James Grennahlgh <james.greenhalgh@arm.com>
8598 * config/aarch64/aarch64-simd-builtins.def: Add copyright header.
8599 * config/aarch64/t-aarch64
8600 (aarch64-builtins.o): Depend on aarch64-simd-builtins.def.
8602 2013-02-22 Vladimir Makarov <vmakarov@redhat.com>
8605 * lra-constraints.c (process_alt_operands): Reload operand
8606 conflicting with earlier clobber only if no more other conflicting
8609 2013-02-22 Jakub Jelinek <jakub@redhat.com>
8612 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Link in libasan_preinit.o
8613 if not linking a shared library.
8615 2013-02-22 Seth LaForge <sethml@google.com>
8617 * config.gcc (arm*-*-eabi*): Treat arm*eb as big-endian.
8619 2013-02-22 Greta Yorsh <Greta.Yorsh@arm.com>
8621 * config/arm/arm.md (split for extendsidi): Update condition.
8622 (zero_extend<mode>di2,extend<mode>di2): Add an alternative.
8623 * config/arm/iterators.md (qhs_extenddi_cstr): Likewise.
8624 (qhs_zextenddi_cstr): Likewise.
8626 2013-02-21 Jakub Jelinek <jakub@redhat.com>
8629 * expmed.c (EXACT_POWER_OF_2_OR_ZERO_P): Do subtraction in uhwi, to
8630 avoid signed wrapping.
8631 (expand_mult): Handle properly multiplication by
8632 ((dword_type) -1) << (BITS_PER_WORD - 1). Improve multiplication by
8633 ((dword_type) 1) << (BITS_PER_WORD - 1). Avoid undefined behavior
8634 in the compiler if coeff is HOST_WIDE_INT_MIN.
8635 (expand_divmod): Don't make ext_op1 static, change it's type to uhwi.
8636 Avoid undefined behavior in -INTVAL (op1).
8638 PR rtl-optimization/50339
8639 * lower-subreg.h (struct lower_subreg_choices): Add splitting_ashiftrt
8641 * lower-subreg.c (compute_splitting_shift): Handle ASHIFTRT.
8642 (compute_costs): Call compute_splitting_shift also for ASHIFTRT
8643 into splitting_ashiftrt field.
8644 (find_decomposable_shift_zext, resolve_shift_zext): Handle also
8646 (dump_choices): Fix up printing LSHIFTRT choices, print ASHIFTRT
8649 2013-02-20 Aldy Hernandez <aldyh@redhat.com>
8652 * trans-mem.c (execute_tm_mark): Do not expand transactions that
8653 are sure to go irrevocable.
8655 2013-02-21 Hans-Peter Nilsson <hp@axis.com>
8657 * doc/rtl.texi (vec_concat, vec_duplicate): Mention that
8658 scalars are valid operands.
8660 2013-02-21 Martin Jambor <mjambor@suse.cz>
8662 PR tree-optimization/56310
8663 * ipa-cp.c (agg_replacements_to_vector): New parameter index, copy
8664 only matching indices and non-negative final offsets.
8665 (intersect_aggregates_with_edge): Pass src_idx to
8666 agg_replacements_to_vector. Pass src_idx insstead of index to
8667 intersect_with_agg_replacements.
8669 2013-02-21 Martin Jambor <mjambor@suse.cz>
8671 * ipa-cp.c (good_cloning_opportunity_p): Dump the real threshold
8672 instead of hard-wired defaults.
8674 2013-02-21 Maciej W. Rozycki <macro@codesourcery.com>
8676 * doc/invoke.texi (MIPS Options): Update documentation of the
8677 floating-point multiply-accumulate instruction restrictions.
8679 2013-02-21 Kostya Serebryany <kcc@google.com>
8681 * config/i386/i386.c (ix86_asan_shadow_offset): Use 0x7fff8000 as
8682 asan_shadow_offset on x86_64 linux.
8684 2013-02-21 Richard Biener <rguenther@suse.de>
8686 PR tree-optimization/56415
8688 2013-02-11 Richard Biener <rguenther@suse.de>
8690 PR tree-optimization/56273
8691 * tree-vrp.c (simplify_cond_using_ranges): Disable for the
8694 2013-02-21 Jakub Jelinek <jakub@redhat.com>
8697 * doc/invoke.texi (-fdump-rtl-pro_and_epilogue): Use @item
8701 * expr.c (expand_expr_real_1) <case TARGET_MEM_REF, MEM_REF>: Don't
8702 use movmisalign or extract_bit_field for EXPAND_MEMORY modifier.
8704 2013-02-20 Jan Hubicka <jh@suse.cz>
8706 PR tree-optimization/56265
8707 * ipa-prop.c (ipa_make_edge_direct_to_target): Fixup callgraph when
8708 target is referenced for first time.
8710 2013-02-20 Richard Biener <rguenther@suse.de>
8712 * tree-call-cdce.c (tree_call_cdce): Do not remove unused locals.
8713 * tree-ssa-forwprop.c (ssa_forward_propagate_and_combine): Likewise.
8714 * tree-ssa-dce.c (perform_tree_ssa_dce): Likewise.
8715 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Do
8716 not return anything.
8717 (rename_ssa_copies): Do not remove unused locals.
8718 * tree-ssa-ccp.c (do_ssa_ccp): Likewise.
8719 * tree-ssanames.c (pass_release_ssa_names): Remove unused locals first.
8720 * passes.c (execute_function_todo): Do not schedule unused locals
8721 removal if cleanup_tree_cfg did something.
8722 * tree-ssa-live.c (remove_unused_locals): Dump statistics
8723 about the number of removed locals.
8725 2013-02-20 Richard Biener <rguenther@suse.de>
8727 PR tree-optimization/56398
8728 * tree-vect-loop-manip.c (adjust_debug_stmts): Skip
8731 2013-02-20 Martin Jambor <mjambor@suse.cz>
8733 PR tree-optimization/55334
8734 * ipa-cp.c (initialize_node_lattices): Disable IPA-CP through and to
8735 restricted pointers to arrays.
8737 2013-02-20 Richard Biener <rguenther@suse.de>
8738 Jakub Jelinek <jakub@redhat.com>
8740 PR tree-optimization/56396
8741 * tree-ssa-ccp.c (n_const_val): New static variable.
8742 (get_value): Return NULL for SSA names we don't have a lattice
8744 (ccp_initialize): Initialize n_const_val.
8745 * tree-ssa-copy.c (n_copy_of): New static variable.
8746 (init_copy_prop): Initialize n_copy_of.
8747 (get_value): Return NULL_TREE for SSA names we don't have a
8750 2013-02-20 Martin Jambor <mjambor@suse.cz>
8752 * ipa-cp.c (initialize_node_lattices): Fix dumping condition.
8754 2013-02-20 Richard Biener <rguenther@suse.de>
8756 * genpreds.c (write_lookup_constraint): Do not compare first
8757 letter of the constraint again.
8759 2013-02-20 Richard Biener <rguenther@suse.de>
8761 * tree-ssa-loop-ivopts.c (alloc_use_cost_map): Use bitmap_count_bits
8763 (get_use_iv_cost): Terminate hashtable walk when coming across
8766 2013-02-20 Igor Zamyatin <igor.zamyatin@intel.com>
8768 * config/i386/i386.c (initial_ix86_tune_features): Turn on fp
8769 reassociation for avx2 targets.
8771 2012-02-19 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8773 * config/microblaze/microblaze.c: microblaze_has_clz = 0
8774 Add version check for v8.10.a to enable microblaze_has_clz
8775 * config/microblaze/microblaze.h: Add TARGET_HAS_CLZ as combined
8776 version and TARGET_PATTERN_COMPARE check
8777 * config/microblaze/microblaze.md: New clzsi2 instruction
8779 2012-02-19 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8781 * config/microblaze/microblaze.md (call_value_intern): Check symbol is
8782 function before branching.
8784 2012-02-19 Andrey Belevantsev <abel@ispras.ru>
8786 * sel-sched-dump.c (dump_insn_rtx_flags): Explicitly set
8788 (dump_insn_rtx_1): Pass PATTERN (insn) to str_pattern_slim.
8790 2012-02-19 Andrey Belevantsev <abel@ispras.ru>
8793 * sel-sched.c: Include ira.h.
8794 (implicit_clobber_conflict_p): New function.
8795 (moveup_expr): Use it.
8796 * Makefile.in (sel-sched.o): Depend on ira.h.
8798 2013-02-19 Richard Biener <rguenther@suse.de>
8800 PR tree-optimization/56384
8801 * tree-ssa-sccvn.h (struct vn_phi_s): Add type member.
8802 (vn_hash_type): Split out from ...
8803 (vn_hash_constant_with_type): ... here.
8804 * tree-ssa-sccvn.c (vn_phi_compute_hash): Use vn_hash_type.
8805 (vn_phi_eq): Compare types from vn_phi_s structure.
8806 (vn_phi_lookup): Populate vn_phi_s type.
8807 (vn_phi_insert): Likewise.
8809 2013-02-19 Jakub Jelinek <jakub@redhat.com>
8811 PR tree-optimization/56350
8812 * tree-vect-loop.c (vectorizable_reduction): If orig_stmt, return false
8813 if haven't found reduction or nested cycle operand, rather than
8814 asserting we must find it.
8816 PR tree-optimization/56381
8817 * tree-ssa-pre.c (create_expression_by_pieces): Fix up last argument
8820 2013-02-18 Aldy Hernandez <aldyh@redhat.com>
8821 Jakub Jelinek <jakub@redhat.com>
8824 * genopinit.c (raw_optab_handler): Use this_fn_optabs.
8825 (swap_optab_enable): Same.
8826 (init_all_optabs): Use argument instead of global.
8827 * tree.h (struct tree_optimization_option): New field target_optabs.
8828 * expr.h (init_all_optabs): Add argument to prototype.
8829 (TREE_OPTIMIZATION_OPTABS): New.
8830 (save_optabs_if_changed): Protoize.
8831 * optabs.h: Declare this_fn_optabs.
8832 * optabs.c (save_optabs_if_changed): New.
8833 Declare this_fn_optabs.
8834 (init_optabs): Add argument to init_all_optabs() call.
8835 * function.c (invoke_set_current_function_hook): Handle per
8837 * function.h (struct function): New field optabs.
8838 * config/mips/mips.c (mips_set_mips16_mode): Handle when
8839 optimization_current_node has changed.
8840 * target-globals.h (save_target_globals_default_opts): Protoize.
8841 * target-globals.c (save_target_globals_default_opts): New.
8843 2013-02-18 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
8846 * config/pa/pa.c (pa_conditional_register_usage): On HP-UX, mark
8847 registers %fr12 and %fr12R as call used.
8850 * config/pa/predicates.md (base14_operand): Except for BLKmode, QImode
8851 and HImode, require all displacements to be an integer multiple of
8853 * config/pa/pa.c (pa_legitimate_address_p): For REG+BASE addresses,
8854 only allow QImode and HImode when reload is in progress and strict is
8855 true. Likewise for symbolic addresses. Use base14_operand to check
8856 displacements in REG+BASE addresses.
8858 2013-02-18 Richard Biener <rguenther@suse.de>
8860 PR tree-optimization/56366
8861 * tree-vect-loop.c (get_initial_def_for_induction): Properly
8862 handle sign-conversion of outer-loop initial induction value.
8864 2013-02-18 Richard Biener <rguenther@suse.de>
8867 * cfghooks.c (merge_blocks): If we merge a latch into another
8868 block adjust references to it.
8869 * cfgloop.c (flow_loops_find): Reset latch before recomputing it.
8870 (verify_loop_structure): Verify that a recorded latch is in fact
8873 2013-02-18 Richard Biener <rguenther@suse.de>
8875 PR tree-optimization/56321
8876 * tree-ssa-reassoc.c (propagate_op_to_single_use): Properly
8877 order SSA name release and virtual operand unlinking.
8879 2013-02-17 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8881 * config/microblaze/microblaze.md (save_stack_block): Define.
8882 (restore_stack_block): Likewise.
8884 2013-02-16 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8886 * config/microblaze/linux.h (TARGET_SUPPORTS_PIC): Define as 1.
8887 * config/microblaze/microblaze.h (TARGET_SUPPORTS_PIC): Define as 1.
8888 * config/microblaze/microblaze.c (microblaze_option_override):
8889 Bail out early for PIC modes when target does not support PIC.
8891 2013-02-16 Edgar E. Iglesias <edgar.iglesias@gmail.com>
8893 * config/microblaze/microblaze.c (microblaze_asm_trampoline_template):
8894 Replace with a microblaze version.
8895 (microblaze_trampoline_init): Adapt for microblaze.
8896 * config/microblaze/microblaze.h (TRAMPOLINE_SIZE): Adapt for
8899 2013-02-16 Jakub Jelinek <jakub@redhat.com>
8900 Dodji Seketeli <dodji@redhat.com>
8903 * asan.c (get_mem_refs_of_builtin_call): White space and style
8905 (instrument_mem_region_access): Do not forget to always put
8906 instrumentation of the of 'base' and 'base + len' in a "if (len !=
8907 0) statement, even for cases where either 'base' or 'base + len'
8908 are not instrumented -- because they have been previously
8909 instrumented. Simplify the logic by putting all the statements
8910 instrument 'base + len' inside a sequence, and then insert that
8911 sequence right before the current insertion point. Then, to
8912 instrument 'base + len', just get an iterator on that statement.
8913 And do not forget to update the pointer to iterator the function
8914 received as argument.
8916 2013-02-15 Vladimir Makarov <vmakarov@redhat.com>
8918 PR rtl-optimization/56348
8919 * lra-assigns.c (reload_pseudo_compare_func): Prefer bigger pseudos.
8921 2013-02-15 Steven Bosscher <steven@gcc.gnu.org>
8923 * graph.c (start_graph_dump): Print dumpfile base as digraph label.
8924 (clean_graph_dump_file): Pass base to start_graph_dump.
8926 2013-02-14 Richard Henderson <rth@redhat.com>
8929 * lower-subreg.c (simple_move): Check dest mode instead of src mode.
8931 2013-02-14 Steven Bosscher <steven@gcc.gnu.org>
8933 * collect2-aix.h: Define F_LOADONLY.
8935 2013-02-14 Richard Biener <rguenther@suse.de>
8938 * varasm.c (output_constant_def_1): Get the decl representing
8939 the constant as argument.
8940 (output_constant_def): Wrap output_constant_def_1.
8941 (make_decl_rtl): Use output_constant_def_1 with the decl
8942 representing the constant.
8943 (build_constant_desc): Optionally re-use a decl already
8944 representing the constant.
8945 (tree_output_constant_def): Adjust.
8947 2013-02-14 Dodji Seketeli <dodji@redhat.com>
8950 * asan.c (instrument_builtin_call): Really put the length of the
8951 second source argument into src1_len.
8953 2013-02-13 Jakub Jelinek <jakub@redhat.com>
8955 * asan.c (create_cond_insert_point): Add create_then_fallthru_edge
8956 argument. If it is false, don't create edge from then_bb to
8958 (insert_if_then_before_iter): Pass true to it.
8959 (build_check_stmt): Pass false to it.
8960 (transform_statements): Flush hash table only on extended basic
8961 block boundaries, rather than at the beginning of every bb.
8962 Don't flush hash table on nonfreeing_call_p calls.
8963 * tree-flow.h (nonfreeing_call_p): New prototype.
8964 * tree-ssa-phiopt.c (nonfreeing_call_p): No longer static.
8966 2013-02-13 David S. Miller <davem@davemloft.net>
8968 * expmed.c (expand_shift_1): Only strip scalar integer subregs.
8970 2013-02-13 Vladimir Makarov <vmakarov@redhat.com>
8973 * ira.c (max_regno_before_ira): Move from ...
8975 (fix_reg_equiv_init): Use max_regno_before_ira instead of
8978 2013-02-13 Jakub Jelinek <jakub@redhat.com>
8980 * config/i386/i386.c (ix86_asan_shadow_offset): Revert last change.
8982 2013-02-13 Richard Biener <rguenther@suse.de>
8985 * gimple-streamer-out.c (output_gimple_stmt): Undo wrapping
8986 globals in MEM_REFs.
8988 2013-02-13 Richard Biener <rguenther@suse.de>
8990 * loop-init.c (loop_optimizer_init): Clear loop state when
8991 re-initializing preserved loops.
8992 * loop-unswitch.c (unswitch_single_loop): Return whether
8993 we unswitched the loop. Do not verify loop state here.
8994 (unswitch_loops): When we unswitched a loop discover new
8997 2013-02-13 Kostya Serebryany <kcc@google.com>
8999 * config/i386/i386.c: Use 0x7fff8000 as asan_shadow_offset on x86_64
9001 * sanitizer.def: Rename __asan_init to __asan_init_v1.
9003 2013-02-12 Dodji Seketeli <dodji@redhat.com>
9005 Avoid instrumenting duplicated memory access in the same basic block
9006 * Makefile.in (asan.o): Add new dependency on hash-table.h
9007 * asan.c (struct asan_mem_ref, struct mem_ref_hasher): New types.
9008 (asan_mem_ref_init, asan_mem_ref_get_end, get_mem_ref_hash_table)
9009 (has_stmt_been_instrumented_p, empty_mem_ref_hash_table)
9010 (free_mem_ref_resources, has_mem_ref_been_instrumented)
9011 (has_stmt_been_instrumented_p, update_mem_ref_hash_table)
9012 (get_mem_ref_of_assignment): New functions.
9013 (get_mem_refs_of_builtin_call): Extract from
9014 instrument_builtin_call and tweak a little bit to make it fit with
9016 (instrument_builtin_call): Use the new
9017 get_mem_refs_of_builtin_call. Use gimple_call_builtin_p instead
9018 of is_gimple_builtin_call.
9019 (instrument_derefs, instrument_mem_region_access): Insert the
9020 instrumented memory reference into the hash table.
9021 (maybe_instrument_assignment): Renamed instrument_assignment into
9022 this, and change it to advance the iterator when instrumentation
9023 actually happened and return true in that case. This makes it
9024 homogeneous with maybe_instrument_assignment, and thus give a
9025 chance to callers to be more 'regular'.
9026 (transform_statements): Clear the memory reference hash table
9027 whenever we enter a new BB, when we cross a function call, or when
9028 we are done transforming statements. Use
9029 maybe_instrument_assignment instead of instrumentation. No more
9030 need to special case maybe_instrument_assignment and advance the
9031 iterator after calling it; it's now handled just like
9032 maybe_instrument_call. Update comment.
9034 2013-02-13 Richard Biener <rguenther@suse.de>
9036 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc):
9037 Fix loop discovery code.
9039 2013-02-12 Vladimir Makarov <vmakarov@redhat.com>
9042 * lra-constraints.c (process_alt_operands): Match early clobber
9043 operand with itself. Check conflicts with earlyclobber only if
9044 the operand is not reloaded. Prefer to reload conflicting operand
9045 if earlyclobber and matching operands are the same.
9047 2013-02-12 Richard Biener <rguenther@suse.de>
9050 * lto-streamer-out.c (write_symbol): Do not output symbols
9051 for hard register variables.
9053 2013-02-12 Georg-Johann Lay <avr@gjlay.de>
9056 * config/avr/avr-dimode.md (umulsidi3, mulsidi3): New expanders.
9057 (umulsidi3_insn, mulsidi3_insn): New insns.
9059 2013-02-12 Christophe Lyon <christophe.lyon@linaro.org>
9061 * config/arm/arm-protos.h (struct cpu_vec_costs): New struct type.
9062 (struct tune_params): Add vec_costs field.
9063 * config/arm/arm.c (arm_builtin_vectorization_cost)
9064 (arm_add_stmt_cost): New functions.
9065 (TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST)
9066 (TARGET_VECTORIZE_ADD_STMT_COST): Define.
9067 (arm_default_vec_cost): New struct of type cpu_vec_costs.
9068 (arm_slowmul_tune, arm_fastmul_tune, arm_strongarm_tune)
9069 (arm_xscale_tune, arm_9e_tune, arm_v6t2_tune, arm_cortex_tune)
9070 (arm_cortex_a15_tune, arm_cortex_a5_tune, arm_cortex_a9_tune)
9071 (arm_v6m_tune, arm_fa726te_tune): Define new vec_costs field.
9073 2013-02-12 Richard Biener <rguenther@suse.de>
9076 * gimple-streamer-in.c (input_gimple_stmt): Strip MEM_REFs off
9077 decls again if possible.
9079 2013-02-12 Richard Biener <rguenther@suse.de>
9082 * tree-ssa.c (verify_ssa_name): Fix check, move
9083 SSA_NAME_IN_FREE_LIST check up.
9085 2013-02-12 Jakub Jelinek <jakub@redhat.com>
9086 Steven Bosscher <steven@gcc.gnu.org>
9088 PR rtl-optimization/56151
9089 * optabs.c (add_equal_note): Don't return 0 if target is a MEM,
9090 equal to op0 or op1, and last_insn pattern is CODE operation
9091 with MEM dest and one of the operands matches that MEM.
9093 2013-02-11 Sriraman Tallam <tmsriramgoogle.com>
9095 * doc/extend.texi: Document Function Multiversioning and "default"
9096 parameter string to target attribute.
9097 * config/i386/i386.c (get_builtin_code_for_version): Return 0 if
9098 target attribute parameter is "default".
9099 (ix86_compare_version_priority): Remove checks for target attribute.
9100 (ix86_mangle_function_version_assembler_name): Change error to sorry.
9101 Remove check for target attribute equal to NULL. Add assert.
9102 (ix86_generate_version_dispatcher_body): Change error to sorry.
9104 2013-02-11 Iain Sandoe <iain@codesourcery.com>
9105 Jack Howarth <howarth@bromo.med.uc.edu>
9106 Patrick Marlier <patrick.marlier@gmail.com>
9109 * config/darwin.h: Replace ENDFILE_SPEC with TM_DESTRUCTOR and
9110 define ENDFILE_SPEC as TM_DESTRUCTOR.
9111 * config/i386/darwin.h (ENDFILE_SPEC): Use TM_DESTRUCTOR.
9113 2013-02-11 Alexander Potapenko <glider@google.com>
9114 Jack Howarth <howarth@bromo.med.uc.edu>
9115 Jakub Jelinek <jakub@redhat.com>
9118 * config/darwin.c (cdtor_record): Rename ctor_record.
9119 (sort_cdtor_records): Rename sort_ctor_records.
9120 (finalize_dtors): New routine to sort destructors by
9121 priority before use in assemble_integer.
9122 (machopic_asm_out_destructor): Use finalize_dtors if needed.
9124 2013-02-11 Uros Bizjak <ubizjak@gmail.com>
9126 PR rtl-optimization/56275
9127 * simplify-rtx.c (avoid_constant_pool_reference): Check that
9128 offset is non-negative and less than cmode size before
9129 calling simplify_subreg.
9131 2013-02-11 Richard Biener <rguenther@suse.de>
9133 PR tree-optimization/56264
9134 * cfgloop.h (fix_loop_structure): Adjust prototype.
9135 * loop-init.c (fix_loop_structure): Return the number of
9136 newly discovered loops.
9137 * tree-cfgcleanup.c (repair_loop_structures): When new loops
9138 are discovered, do a full loop-closed SSA rewrite.
9140 2013-02-11 Richard Biener <rguenther@suse.de>
9142 PR tree-optimization/56273
9143 * tree-vrp.c (simplify_cond_using_ranges): Disable for the
9145 (check_array_ref): Fix missing newline in dumps.
9146 (search_for_addr_array): Likewise.
9148 2013-02-09 David Edelsohn <dje.gcc@gmail.com>
9150 * config/rs6000/aix61.h (OS_MISSING_ALTIVEC): Undefine.
9152 2013-02-09 Jakub Jelinek <jakub@redhat.com>
9155 * config/rs6000/rs6000.h (ASSEMBLER_DIALECT): Define.
9157 2013-02-08 Vladimir Makarov <vmakarov@redhat.com>
9159 PR rtl-optimization/56246
9160 * lra-constraints.c (simplify_operand_subreg): Try to reuse
9162 * lra.c (lra): Clear lra_optional_reload_pseudos only when all
9163 constraints are satisfied.
9165 2013-02-08 Jeff Law <law@redhat.com>
9168 * emit-rtl.c (reg_is_parm_p): New function.
9169 * regs.h (reg_is_parm_p): New prototype.
9170 * ira-conflicts.c (ira_build_conflicts): Allow parameters in
9171 callee-clobbered registers.
9173 2013-02-08 Michael Meissner <meissner@linux.vnet.ibm.com>
9176 * config/rs6000/rs6000.c (rs6000_builtin_vectorized_libmass):
9177 If there is no implicit builtin declaration, just return NULL.
9179 2013-02-08 Uros Bizjak <ubizjak@gmail.com>
9181 * config/i386/sse.md (FMAMODEM): New mode iterator.
9182 (fma<mode>4, fms<mode>4, fnma<mode>4, fnms<mode>4): Use FMAMODEM
9183 mode iterator. Do not use TARGET_SSE_MATH in insn constraint.
9185 2013-02-08 Uros Bizjak <ubizjak@gmail.com>
9187 * config/i386/gnu-user.h (TARGET_CAN_SPLIT_STACK): Define only
9188 when HAVE_GAS_CFI_PERSONALITY_DIRECTIVE is set.
9189 * config/i386/gnu-user64.h (TARGET_CAN_SPLIT_STACK): Ditto.
9191 2013-02-08 Edgar E. Iglesias <edgar.iglesias@gmail.com>
9193 * config.gcc (microblaze*-linux*): Add TARGET_BIG_ENDIAN_DEFAULT.
9194 (microblaze*-*-elf): Likewise.
9195 * config/microblaze/linux.h: Add -mbig-endian / -mlittle-endian to
9197 * config/microblaze/microblaze-c.c: Add builtin defines for
9198 _LITTLE_ENDIAN and _BIG_ENDIAN.
9199 * config/microblaze/microblaze.h: Add TARGET_ENDIAN_DEFAULT and
9200 add to TARGET_DEFAULT flags.
9201 Expand ASM_SPEC and LINK_SPEC.
9202 Update BYTES_BIG_ENDIAN and WORDS_BIG_ENDIAN.
9203 * config/microblaze/microblaze.md: Update extendsidi2 and
9204 movdi_internal instructions to use low-order / high-order reg
9206 * config/microblaze/microblaze.opt: Add mbig-endian and mlittle-endian
9207 options and inversemask / mask of LITTLE_ENDIAN.
9208 * config/microblaze/t-microblaze: Expand multilib options to
9209 include mlittle-endian (le) and update exceptions patterns.
9211 2013-02-08 Jakub Jelinek <jakub@redhat.com>
9213 PR rtl-optimization/56195
9214 * lra-constraints.c (get_reload_reg): Don't reuse regs
9215 if they have smaller mode than requested, if they have
9216 wider mode than requested, try to return a SUBREG.
9218 PR tree-optimization/56250
9219 * fold-const.c (extract_muldiv_1) <case NEGATE_EXPR>: Don't optimize
9220 if type is unsigned and code isn't MULT_EXPR.
9222 2013-02-08 Georg-Johann Lay <avr@gjlay.de>
9224 PR tree-optimization/56064
9225 * fixed-value.c (fixed_from_double_int): Sign/zero extend payload
9226 bits according to mode.
9227 * fixed-value.h (fixed_from_double_int)
9228 (const_fixed_from_double_int): Adjust comments.
9230 2013-02-08 Richard Biener <rguenther@suse.de>
9233 * lto-streamer.h (struct data_in): Remove current_file, current_line
9234 and current_col members.
9235 * lto-streamer-out.c (lto_output_location): Stream changed bits
9236 en-block for efficiency.
9237 * lto-streamer-in.c (clear_line_info): Remove.
9238 (lto_input_location): Cache current file, line and column
9239 globally via local statics. Read changed bits en-block.
9240 (input_function): Do not call clear_line_info.
9241 (lto_read_body): Likewise.
9242 (lto_input_toplevel_asms): Likewise.
9244 2013-02-08 Michael Matz <matz@suse.de>
9246 PR tree-optimization/52448
9247 * tree-ssa-phiopt.c (struct name_to_bb): Add phase member.
9248 (nt_call_phase): New static.
9249 (add_or_mark_expr): Only mark accesses with newer phase than any
9251 (nonfreeing_call_p): New.
9252 (nt_init_block): Update nt_call_phase, mark blocks as visited.
9253 (nt_fini_block): Keep blocks marked as visited.
9254 (get_non_trapping): Initialize nt_call_phase, and reset aux pointer.
9256 2013-02-08 Richard Biener <rguenther@suse.de>
9258 * ira.c (ira): Free broken dominator information.
9260 2013-02-08 Uros Bizjak <ubizjak@gmail.com>
9262 * config/i386/i386.c (ix86_spill_class): Use INTEGER_CLASS_P macro.
9264 2013-02-08 Marek Polacek <polacek@redhat.com>
9266 * cfgloop.c (verify_loop_structure): Add more checking of headers.
9268 2013-02-08 Richard Biener <rguenther@suse.de>
9271 * cfgloop.h (flow_loops_find): Adjust.
9272 (bb_loop_header_p): Declare.
9273 * cfgloop.c (bb_loop_header_p): New function split out from ...
9274 (flow_loops_find): ... here. Adjust function signature,
9275 support incremental loop structure update.
9276 (verify_loop_structure): Cleanup. Verify a loop is a loop.
9277 * cfgloopmanip.c (fix_loop_structure): Move ...
9278 * loop-init.c (fix_loop_structure): ... here.
9279 (apply_loop_flags): Split out from ...
9280 (loop_optimizer_init): ... here.
9281 (fix_loop_structure): Use apply_loop_flags. Use flow_loops_find
9282 in incremental mode, only remove dead loops here.
9284 2013-02-08 Georg-Johann Lay <avr@gjlay.de>
9287 * config/avr/avr.md (unspec) <UNSPEC_ROUND>: Add.
9288 * config/avr/avr-fixed.md (ALL4QA, ALL124QA): New mode iterators.
9289 (round<mode>3, round<mode>3_const): New expanders for fixed-mode.
9290 (*round<mode>3.libgcc): New insns for fixed-modes.
9291 * config/avr/builtins.def (ABSxx): Use a non-NULL LIBNAME.
9292 (ROUNDxx, COUNTLSxx, BITSxx, xxBITS): New DEF_BUILTINs.
9293 (ROUNDFX, COUNTLSFX, ABSFX): New DEF_BUILTINs.
9294 * config/avr/stdfix.h (absFX, bitsFX, FXbits): Remove inline
9295 implementations. Define to __builtin_avr_absFX,
9296 __builtin_avr_bitsFX, __builtin_avr_FXbits, respectively.
9297 (roundFX, countlsFX): Define to __builtin_avr_roundFX,
9298 __builtin_avr_countlsFX, respectively.
9299 * config/avr/avr-c.c (target.h): Include it.
9300 (enum avr_builtin_id): New enum.
9301 (avr_resolve_overloaded_builtin): New static function.
9302 (avr_register_target_pragmas): Use it to set
9303 targetm.resolve_overloaded_builtin.
9304 * config/avr/avr.c (avr_init_builtins): Supply myriads of local
9305 tree nodes used by DEF_BUILTIN.
9306 (avr_expand_builtin) <AVR_BUILTIN_ROUNDxx>: Sanity-check them.
9307 (avr_fold_builtin) <AVR_BUILTIN_BITSxx>: Fold to VIEW_COVERT_EXPR.
9308 <AVR_BUILTIN_xxBITS>: Same.
9310 2013-02-08 Richard Biener <rguenther@suse.de>
9312 * cfgloop.c (verify_loop_structure): Properly handle
9313 a loop exiting to another loop header.
9314 * ira-int.h (ira_loops): Remove.
9315 * ira.c (ira_loops): Remove.
9316 (ira): Use loop_optimizer_init and loop_optimizer_finalize.
9317 (do_reload): Use loop_optimizer_finalize.
9318 * ira-build.c (create_loop_tree_nodes): Use get_loops and
9319 number_of_loops to access the loop tree.
9320 (more_one_region_p): Likewise.
9321 (finish_loop_tree_nodes): Likewise.
9322 (rebuild_regno_allocno_maps): Likewise.
9323 (mark_loops_for_removal): Likewise.
9324 (mark_all_loops_for_removal): Likewise.
9325 (remove_unnecessary_regions): Likewise.
9326 (ira_build): Likewise.
9327 * ira-emit.c (setup_entered_from_non_parent_p): Likewise.
9329 2013-02-08 Richard Biener <rguenther@suse.de>
9331 * Makefile.in (tree-tailcall.o): Add $(CFGLOOP_H) dependency.
9332 * ipa-pure-const.c (analyze_function): Avoid calling
9333 mark_irreducible_loops twice.
9334 * tree-tailcall.c (tree_optimize_tail_calls_1): Mark loops for fixup.
9336 2013-02-07 David S. Miller <davem@davemloft.net>
9338 * dwarf2out.c (based_loc_descr): Perform leaf register remapping
9340 * var-tracking.c (vt_add_function_parameter): Test the presence of
9341 HAVE_window_save properly and do not remap argument registers when
9342 we have a leaf function.
9344 2013-02-07 Uros Bizjak <ubizjak@gmail.com>
9347 * ggc-page.c (ggc_print_statistics): Use HOST_LONG_LONG_FORMAT
9349 * config/i386/i386.c (ix86_print_operand): Ditto.
9351 2013-02-07 Vladimir Makarov <vmakarov@redhat.com>
9353 * lra-constraints.c (process_alt_operands): Fix recently added comment.
9355 2013-02-07 Vladimir Makarov <vmakarov@redhat.com>
9357 PR rtl-optimization/56225
9358 * lra-constraints.c (process_alt_operands): Check that reload hard
9359 reg can hold value for strict_low_part.
9361 2013-02-07 Jakub Jelinek <jakub@redhat.com>
9364 * dwarf2out.c (dwarf2_debug_hooks): Set end_function hook to
9365 dwarf2out_end_function.
9366 (in_first_function_p, maybe_at_text_label_p,
9367 first_loclabel_num_not_at_text_label): New variables.
9368 (dwarf2out_var_location): In the first function find out
9369 lowest loclabel_num N where .LVLN is known not to be equal to .Ltext0.
9370 (find_empty_loc_ranges_at_text_label, dwarf2out_end_function): New
9373 2013-02-07 Eric Botcazou <ebotcazou@adacore.com>
9375 PR rtl-optimization/56178
9376 * cse.c (cse_insn): Do not create a REG_EQUAL note if the source is a
9377 SUBREG of a register. Tidy up related block of code.
9378 * fwprop.c (forward_propagate_and_simplify): Do not create a REG_EQUAL
9379 note if the source is a register or a SUBREG of a register.
9381 2013-02-07 Jakub Jelinek <jakub@redhat.com>
9384 * config/rs6000/rs6000.md (ptrm): New mode attr.
9385 (call_indirect_aix<ptrsize>, call_indirect_aix<ptrsize>_nor11,
9386 call_value_indirect_aix<pttrsize>,
9387 call_value_indirect_aix<pttrsize>_nor11): Use <ptrm> instead of
9390 2013-02-07 Michael Haubenwallner <michael.haubenwallner@salomon.at>
9392 * collect2.c (main): Set aix64_flag for -G and -bsvr4 too, disable
9393 if -bnortl. Convert to strcmp and strncmp.
9395 2013-02-07 Alan Modra <amodra@gmail.com>
9398 * config/rs6000/rs6000.c (mem_operand_gpr): Check that LO_SUM
9399 addresses won't wrap when offsetting.
9400 (rs6000_secondary_reload): Provide secondary reloads needed for
9401 wrapping LO_SUM addresses.
9403 2013-02-06 Thomas Schwinge <thomas@codesourcery.com>
9405 * config/gnu.h (GNU_USER_TARGET_OS_CPP_BUILTINS): Never define
9406 MACH, just __MACH__.
9408 2013-02-06 Richard Biener <rguenther@suse.de>
9410 * tracer.c (tracer): Mark loops with LOOPS_NEED_FIXUP
9411 instead of calling fix_loop_structure.
9413 2013-02-06 Jakub Jelinek <jakub@redhat.com>
9416 * omp-low.c (use_pointer_for_field): Return false if
9417 lower_send_shared_vars doesn't generate any copy-out code.
9419 2013-02-06 Tom de Vries <tom@codesourcery.com>
9421 PR rtl-optimization/56131
9422 * cfgrtl.c (delete_insn): Use NOTE_BASIC_BLOCK instead of BLOCK_FOR_INSN
9423 to get the bb of a NOTE_INSN_BASIC_BLOCK. Handle the case that the bb
9424 of the label is NULL. Add comment.
9426 2013-02-05 Jakub Jelinek <jakub@redhat.com>
9428 * tree.h (struct tree_decl_with_vis): Remove thread_local field.
9431 * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Define.
9432 (STATIC_LIBTSAN_LIBS): Likewise.
9433 * gcc.c (ADD_STATIC_LIBTSAN_LIBS, LIBTSAN_EARLY_SPEC): Define.
9434 (LIBTSAN_SPEC): Add ADD_STATIC_LIBTSAN_LIBS, if LIBTSAN_EARLY_SPEC
9435 is defined, don't add anything else beyond that.
9436 (SANITIZER_EARLY_SPEC, SANITIZER_SPEC): Define.
9437 (LINK_COMMAND_SPEC): Use them.
9439 PR tree-optimization/56205
9440 * tree-stdarg.c (check_all_va_list_escapes): Return true if
9441 there are any PHI nodes that set non-va_list_escape_vars SSA_NAME
9442 and some va_list_escape_vars SSA_NAME appears in some PHI argument.
9444 2013-02-05 Richard Biener <rguenther@suse.de>
9446 PR tree-optimization/53342
9447 PR tree-optimization/53185
9448 * tree-vectorizer.h (vect_check_strided_load): Remove.
9449 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Do
9450 not disallow peeling for vectorized strided loads.
9451 (vect_check_strided_load): Make static and simplify.
9452 (vect_analyze_data_refs): Adjust.
9453 * tree-vect-stmts.c (vectorizable_load): Handle peeled loops
9454 correctly when vectorizing strided loads.
9456 2013-02-05 Richard Biener <rguenther@suse.de>
9458 * doc/install.texi: Refer to ISL, not PPL.
9460 2013-02-05 Jan Hubicka <jh@suse.cz>
9462 PR tree-optimization/55789
9463 * params.def (PARAM_EARLY_INLINER_MAX_ITERATIONS): Drop to 1.
9465 2013-02-05 Jan Hubicka <jh@suse.cz>
9467 PR tree-optimization/55789
9468 * cgraphclones.c (cgraph_remove_node_and_inline_clones): Remove
9469 the dead call anyway.
9471 2013-02-05 Eric Botcazou <ebotcazou@adacore.com>
9474 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Add missing guard.
9476 2013-02-04 Alexander Potapenko <glider@google.com>
9477 Jack Howarth <howarth@bromo.med.uc.edu>
9478 Jakub Jelinek <jakub@redhat.com>
9481 * config/darwin.c (sort_ctor_records): Stabilized qsort
9482 on constructor priority by using original position.
9483 (finalize_ctors): New routine to sort constructors by
9484 priority before use in assemble_integer.
9485 (machopic_asm_out_constructor): Use finalize_ctors if needed.
9487 2013-02-04 Jakub Jelinek <jakub@redhat.com>
9490 * config/i386/winnt.c (i386_pe_assemble_visibility): Don't warn
9491 about visibility on artificial decls.
9492 * config/sol2.c (solaris_assemble_visibility): Likewise.
9494 2013-02-04 Kai Tietz <ktietz@redhat.com>
9497 * config/i386/i386.c (function_value_ms_64): Add additional valtype
9498 argument and improve checking of return-argument types for 16-byte
9500 (ix86_function_value_1): Add additional valtype argument on call
9501 of function_value_64.
9502 (return_in_memory_ms_64): Sync 16-byte sized mode handling with
9503 handling infunction_value_64 function.
9505 2013-02-04 Matthew Gretton-Dann <matthew.gretton-dann@linaro.org>
9507 * reload.c (subst_reloads): Fix DEBUG_RELOAD build issue.
9509 2013-02-04 Richard Biener <rguenther@suse.de>
9511 PR tree-optimization/56188
9512 * tree-ssa-structalias.c (label_visit): Consider case with
9513 initially non-empty points-to set.
9514 (perform_var_substitution): Dump node mapping and clean up.
9516 2013-02-04 Richard Guenther <rguenther@suse.de>
9519 * lto-symtab.c (lto_symtab_merge_decls_1): Make non-builtin
9520 node prevail as last resort.
9521 (lto_symtab_merge_decls): Remove guard on LTRANS here.
9522 (lto_symtab_prevailing_decl): Builtins are their own prevailing decl.
9524 2013-02-04 Richard Biener <rguenther@suse.de>
9526 PR tree-optimization/56113
9527 * tree-ssa-structalias.c (equiv_class_lookup, equiv_class_add):
9529 (equiv_class_lookup_or_add): ... this.
9530 (label_visit): Adjust and fix error in previous patch.
9531 (perform_var_substitution): Adjust.
9533 2013-02-03 Oleg Endo <olegendo@gcc.gnu.org>
9535 * config/sh/divtab.c: Fix formatting and comments throughout the file.
9536 * config/sh/sh4-300.md: Likewise.
9537 * config/sh/sh4a.md: Likewise.
9538 * config/sh/constraints.md: Likewise.
9539 * config/sh/sh.md: Likewise.
9540 * config/sh/netbsd-elf.h: Likewise.
9541 * config/sh/predicates.md: Likewise.
9542 * config/sh/sh-protos.h: Likewise.
9543 * config/sh/ushmedia.h: Likewise.
9544 * config/sh/linux.h: Likewise.
9545 * config/sh/sh.c: Likewise.
9546 * config/sh/superh.h: Likewise.
9547 * config/sh/elf.h: Likewise.
9548 * config/sh/sh4.md: Likewise.
9549 * config/sh/sh.h: Likewise.
9551 2013-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
9553 * config/pa/constraints.md: Adjust unused letters. Change "T"
9554 constraint to match_test floating_point_store_memory_operand().
9555 * config/pa/predicates.md (reg_plus_base_memory_operand): New.
9556 (base14_operand): New.
9557 (floating_point_store_memory_operand): New.
9558 (integer_store_memory_operand): Revise to use base14_operand and
9559 reg_plus_base_memory_operand.
9560 (move_dest_operand): Allow symbolic_memory_operands.
9561 (symbolic_memory_operand): Check for LO_SOM.
9562 (symbolic_operand): Change default case to break.
9563 * config/pa/pa.md: Remove unamed DFmode and SFmode patterns to force
9564 CONST_DOUBLE values to be reloaded by putting them into memory when
9565 the destination is a floating point register.
9566 (movdf): Remove code to handle CONST_DOUBLE.
9568 (reload_indf_r1): New.
9569 (reload_insf_r1): New.
9570 Consistently use "Q" and "T" constraints with integer and floating
9571 point move instructions, respectively.
9572 (movdi): Remove FAIL.
9573 Change predicate for source operand unamed DImode move from
9574 general_operand to move_src_operand.
9575 (umulsidi3): Change predicate for destination operand to
9577 Likewise for similar unamed patterns.
9578 * config/pa/pa-protos.h (pa_legitimize_reload_address): Declare.
9579 * config/pa/pa.c (pa_symbolic_expression_p): Remove extra parenthesis.
9580 (hppa_legitimize_address): Simplify mask calculation.
9581 (pa_emit_move_sequence): Revised handling of secondary reloads from
9582 REG+D addresses for floating point loads and stores. Directly handle
9583 loading CONST0_RTX (mode) to a floating point register.
9584 (pa_secondary_reload): Handle reloading DF and SFmode constant values
9585 to floating point registers. Don't restrict secondary reloads to
9586 floating point registers to integer modes. Revise some comments and
9588 (TARGET_LEGITIMATE_ADDRESS_P): Define.
9589 (pa_legitimate_address_p): New.
9590 (pa_legitimize_reload_address): New.
9591 * config/pa/pa.h (STRICT_REG_OK_FOR_INDEX_P): New.
9592 (STRICT_REG_OK_FOR_BASE_P): New.
9593 (GO_IF_LEGITIMATE_ADDRESS): Delete. Update some related comments.
9594 (LEGITIMIZE_RELOAD_ADDRESS): Revise to use pa_legitimize_reload_address.
9596 2013-02-03 David Edelsohn <dje.gcc@gmail.com>
9597 Andrew Dixie <andrewd@gentrack.com>
9599 * collect2.c (GCC_CHECK_HDR): Do not scan objects with F_LOADONLY
9602 2013-02-03 Richard Sandiford <rdsandiford@googlemail.com>
9604 * expmed.c (extract_bit_field_1): Pass the full width of the
9605 structure to get_best_reg_extraction_insn.
9607 2013-02-01 David Edelsohn <dje.gcc@gmail.com>
9610 * configure.ac (use_cxa_atexit): Add AIX.
9611 * configure: Regenerate.
9613 * config/rs6000/aix61.h (STARTFILE_SPEC): Add crtcxa.o.
9615 2013-02-01 Jakub Jelinek <jakub@redhat.com>
9618 * final.c (need_profile_function): New variable.
9619 (final_start_function): Drop ATTRIBUTE_UNUSED from first argument.
9620 If first of NOTE_INSN_BASIC_BLOCK or NOTE_INSN_FUNCTION_BEG
9621 is only preceeded by NOTE_INSN_VAR_LOCATION or NOTE_INSN_DELETED
9622 notes, targetm.asm_out.function_prologue doesn't emit anything,
9623 HAVE_prologue and profiler should be emitted before prologue,
9624 set need_profile_function instead of emitting it.
9625 (final_scan_insn): If need_profile_function, emit
9626 profile_function on the first NOTE_INSN_BASIC_BLOCK or
9627 NOTE_INSN_FUNCTION_BEG note.
9629 2013-02-01 Richard Henderson <rth@redhat.com>
9631 * config/rs6000/rs6000.md (smulditi3): New.
9634 * config/alpha/alpha.md (umulditi3): New.
9636 2013-02-01 David Edelsohn <dje.gcc@gmail.com>
9638 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_COMMON): Use floor_log2.
9639 (ASM_OUTPUT_ALIGNED_LOCAL): New.
9641 2013-02-01 Richard Biener <rguenther@suse.de>
9643 PR tree-optimization/56113
9644 * tree-ssa-structalias.c (label_visit): Reduce work for
9645 single-predecessor nodes.
9647 2013-02-01 Eric Botcazou <ebotcazou@adacore.com>
9649 * fold-const.c (make_range_step) <TRUTH_NOT_EXPR>: Bail out if the
9650 range isn't testing for zero.
9652 2013-01-31 Steven Bosscher <steven@gcc.gnu.org>
9655 * fwprop.c (fwprop_init): Set up loops without CFG modifications.
9657 2013-01-31 Hiroyuki Ono <hiroyuki.ono.jc@renesas.com>
9658 Nick Clifton <nickc@redhat.com>
9660 * config/v850/constraints.md (Q): Define as a memory constraint.
9661 * config/v850/predicates.md (label_ref_operand): New predicate.
9662 (e3v5_shift_operand): New predicate.
9663 (ior_operator): New predicate.
9664 * config/v850/t-v850: Add e3v5 multilib.
9665 * config/v850/v850-protos.h (v850_adjust_insn_length): Prototype.
9666 (v850_gen_movdi): Prototype.
9667 * config/v850/v850.c: Add support for e3v5 architecture.
9668 Rename all uses of TARGET_V850E || TARGET_V850E2_ALL to
9670 (construct_save_jarl): Add e3v5 long JARL support.
9671 (v850_adjust_insn_length): New function. Adjust length of call
9672 insns when using e3v5 instructions.
9673 (v850_gen_movdi): New function: Generate instructions to move a
9675 * config/v850/v850.h (TARGET_CPU_v850e3v5): Define.
9676 (CPP_SPEC): Define __v850e3v5__ as appropriate.
9677 (TARGET_USE_FPU): Enable for e3v5.
9678 (CONST_OK_FOR_W): New macro.
9679 (ADJUST_INSN_LENGTH): Define.
9680 * config/v850/v850.md (UNSPEC_LOOP): Define.
9681 (attr cpu): Add v850e3v5.
9682 Rename all uses of TARGET_V850E2 to TARGET_V850E2V3_UP.
9683 (movdi): New pattern.
9684 (movdi_internal): New pattern.
9685 (cbranchsf4): Conditionalize on TARGET_USE_FPU.
9686 (cbranchdf4): Conditionalize on TARGET_USE_FPU.
9687 (cstoresf4): Likewise.
9688 (cstoredf4): Likewise.
9689 (insv): New pattern.
9690 (rotlso3_a): New pattern.
9691 (rotlsi3_b): New pattern
9692 (rotlsi3_v850e3v5): New pattern.
9693 (doloop_begin): New pattern.
9694 (fix_loop_counter): New pattern.
9695 (doloop_end): New pattern.
9696 (branch_normal): Add e3v5 long branch support.
9697 (branch_invert): Likewise.
9698 (branch_z_normal): Likewise.
9699 (branch_z_invert): Likewise.
9700 (branch_nz_normal): Likewise.
9701 (branch_nz_invert): Likewise.
9702 (call_internal_short): Add e3v5 register-indirect JARL support.
9703 (call_internal_long): Likewise.
9704 (call_value_internal_short): Likewise.
9705 (call_value_internal_long): Likewise.
9706 * config/v850/v850.opt (mv850e3v5, mv850e2v4): New options.
9707 (mloop): New option.
9708 * config.gcc: Add support for configuring v840e3v5 target.
9709 * doc/invoke.texi: Document new v850 specific command line options.
9711 2013-01-31 Paul Koning <ni1d@arrl.net>
9715 * dwarf2out.c (prune_unused_types_mark): Mark all of parent's
9716 children if parent is a class.
9717 (prune_unused_types_prune): Don't add DW_AT_declaration.
9719 2013-01-31 Richard Biener <rguenther@suse.de>
9721 PR tree-optimization/56157
9722 * tree-vect-slp.c (vect_get_slp_defs): More thoroughly try to
9723 match up operand with SLP child.
9725 2013-01-31 Jason Merrill <jason@redhat.com>
9728 * dwarf2out.c (gen_struct_or_union_type_die): Always schedule template
9729 parameters the first time.
9730 (gen_scheduled_generic_parms_dies): Check completeness here.
9732 2013-01-31 Richard Biener <rguenther@suse.de>
9735 * common.opt (faggressive-loop-optimizations): New flag,
9737 * doc/invoke.texi (faggressive-loop-optimizations): Document.
9738 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Guard
9739 infer_loop_bounds_from_undefined by it.
9741 2013-01-31 Richard Biener <rguenther@suse.de>
9743 PR tree-optimization/56150
9744 * tree-ssa-loop-manip.c (find_uses_to_rename_stmt): Do not
9745 visit virtual operands.
9746 (find_uses_to_rename_bb): Likewise.
9748 2013-01-31 Richard Biener <rguenther@suse.de>
9750 PR tree-optimization/56150
9751 * tree-ssa-tail-merge.c (gimple_equal_p): Properly handle
9752 mixed store non-store stmts.
9754 2013-01-30 Jakub Jelinek <jakub@redhat.com>
9757 * gcc.c (LIBASAN_SPEC): Define just to ADD_STATIC_LIBASAN_LIBS if
9758 LIBASAN_EARLY_SPEC is defined.
9759 (LIBASAN_EARLY_SPEC): Define to empty string if not already defined.
9760 (LINK_COMMAND_SPEC): Add LIBASAN_EARLY_SPEC for -fsanitize=address,
9762 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Define.
9765 * config/i386/i386.c (ix86_valid_target_attribute_inner_p): Diagnose
9766 invalid args instead of ICEing on it.
9767 (ix86_valid_target_attribute_tree): Return error_mark_node if
9768 ix86_valid_target_attribute_inner_p failed.
9769 (ix86_valid_target_attribute_p): Return false only if
9770 ix86_valid_target_attribute_tree returned error_mark_node. Allow
9771 target("default") attribute.
9772 (sorted_attr_string): Change argument from const char * to tree,
9773 merge in all target attribute arguments rather than just one.
9774 Formatting fix. Use XNEWVEC instead of xmalloc and XDELETEVEC
9775 instead of free. Avoid using strcat.
9776 (ix86_mangle_function_version_assembler_name): Mangle
9777 target("default") as if no target attribute is present. Adjust
9778 sorted_attr_string caller. Avoid leaking memory. Use XNEWVEC
9779 instead of xmalloc and XDELETEVEC instead of free.
9780 (ix86_function_versions): Don't return true if one of the decls
9781 doesn't have target attribute. If they don't and one of the decls
9782 is DECL_FUNCTION_VERSIONED, report an error. Adjust
9783 sorted_attr_string caller. Use XDELETEVEC instead of free.
9784 (ix86_supports_function_versions): Remove.
9785 (make_name): Fix up formatting.
9786 (make_dispatcher_decl): Remove resolver_name and its initialization.
9787 Avoid leaking memory.
9788 (is_function_default_version): Return true if there is
9789 target("default") attribute rather than no target attribute at all.
9790 (make_resolver_func): Avoid leaking memory.
9791 (ix86_generate_version_dispatcher_body): Likewise.
9792 (TARGET_OPTION_SUPPORTS_FUNCTION_VERSIONS): Remove.
9793 * target.def (supports_function_versions): Remove.
9794 * doc/tm.texi.in (SUPPORTS_FUNCTION_VERSIONS): Remove.
9795 * doc/tm.texi: Regenerated.
9797 2013-01-30 Vladimir Makarov <vmakarov@redhat.com>
9799 PR rtl-optimization/56144
9800 * lra-constraints.c (get_reload_reg): Don't reuse reload pseudo
9801 for values with side effects.
9803 2013-01-30 Richard Biener <rguenther@suse.de>
9805 * sparseset.h (sparseset_bit_p): Use gcc_checking_assert.
9806 (sparseset_pop): Likewise.
9807 * cfganal.c (compute_idf): Likewise. Increase work-stack size
9808 to be able to use quick_push in the worker loop.
9810 2013-01-30 Marek Polacek <polacek@redhat.com>
9812 * cfgcleanup.c (cleanup_cfg): Don't mark affected BBs.
9814 2013-01-30 Richard Biener <rguenther@suse.de>
9817 * lto-symtab.c (lto_symtab_merge_decls_1): Guard DECL_BUILT_IN check.
9819 2013-01-30 Georg-Johann Lay <avr@gjlay.de>
9821 PR tree-optimization/56064
9822 * fixed-value.c (fixed_from_double_int): New function.
9823 * fixed-value.h (fixed_from_double_int): New prototype.
9824 (const_fixed_from_double_int): New static inline function.
9825 * fold-const.c (native_interpret_fixed): New static function.
9826 (native_interpret_expr) <FIXED_POINT_TYPE>: Use it.
9827 (can_native_interpret_type_p) <FIXED_POINT_TYPE>: Return true.
9828 (native_encode_fixed): New static function.
9829 (native_encode_expr) <FIXED_CST>: Use it.
9830 (native_interpret_int): Move double_int worker code to...
9831 * double-int.c (double_int::from_buffer): ...this new static method.
9832 * double-int.h (double_int::from_buffer): Prototype it.
9834 2013-01-30 Richard Biener <rguenther@suse.de>
9836 * tree-ssa-structalias.c (final_solutions, final_solutions_obstack):
9837 New pointer-map and obstack.
9838 (init_alias_vars): Allocate pointer-map and obstack.
9839 (delete_points_to_sets): Free them.
9840 (find_what_var_points_to): Cache result.
9841 (find_what_p_points_to): Adjust for changed interface of
9842 find_what_var_points_to.
9843 (compute_points_to_sets): Likewise.
9844 (ipa_pta_execute): Likewise.
9846 2013-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9848 * configure.ac (HAVE_AS_SPARC_NOBITS): New test.
9849 * configure: Regenerate.
9850 * config.in: Regenerate.
9851 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section): Emit
9852 #nobits/#progbits if supported.
9854 2013-01-29 Oleg Endo <olegendo@gcc.gnu.org>
9857 * config/sh/sh.md (bclr_m2a, bset_m2a, bst_m2a, bld_m2a, bldsign_m2a,
9858 bld_reg, *bld_regqi, band_m2a, bandreg_m2a, bor_m2a, borreg_m2a,
9859 bxor_m2a, bxorreg_m2a): Add satisfies_constraint_K03 condition.
9861 2013-01-29 Greta Yorsh <Greta.Yorsh@arm.com>
9863 * config/arm/cortex-a7.md (cortex_a7_neon, cortex_a7_all): Remove.
9864 (cortex_a7_idiv): Use cortex_a7_both instead of cortex_a7_all.
9866 2013-01-29 Greta Yorsh <Greta.Yorsh@arm.com>
9868 * config/arm/arm.c (cortexa7_younger): Return true for TYPE_CALL.
9869 * config/arm/cortex-a7.md (cortex_a7_call): Update required units.
9871 2013-01-29 Greta Yorsh <Greta.Yorsh@arm.com>
9873 * config/arm/arm-protos.h (arm_mac_accumulator_is_result): New
9875 * config/arm/arm.c (arm_mac_accumulator_is_result): New function.
9876 * config/arm/cortex-a7.md: New bypasses using
9877 arm_mac_accumulator_is_result.
9879 2013-01-29 Greta Yorsh <Greta.Yorsh@arm.com>
9881 * config/arm/cortex-a7.md (cortex_a7_neon_mul): New reservation.
9882 (cortex_a7_neon_mla): Likewise.
9883 (cortex_a7_fpfmad): New reservation.
9884 (cortex_a7_fpmacs): Use ffmas and update required units.
9885 (cortex_a7_fpmuld): Update required units and latency.
9886 (cortex_a7_fpmacd): Likewise.
9887 (cortex_a7_fdivs, cortex_a7_fdivd): Likewise.
9888 (cortex_a7_neon). Likewise.
9889 (bypass) Update participating units.
9891 2013-01-29 Greta Yorsh <Greta.Yorsh@arm.com>
9893 * config/arm/arm.md (type): Add ffmas and ffmad to "type" attribute.
9894 * config/arm/vfp.md (fma,fmsub,fnmsub,fnmadd): Change type
9896 * config/arm/vfp11.md (vfp_farith): Use ffmas.
9897 (vfp_fmul): Use ffmad.
9898 * config/arm/cortex-r4f.md (cortex_r4_fmacs): Use ffmas.
9899 (cortex_r4_fmacd): Use ffmad.
9900 * config/arm/cortex-m4-fpu.md (cortex_m4_fmacs): Use ffmas.
9901 * config/arm/cortex-a9.md (cortex_a9_fmacs): Use ffmas.
9902 (cortex_a9_fmacd): Use ffmad.
9903 * config/arm/cortex-a8-neon.md (cortex_a8_vfp_macs): Use ffmas.
9904 (cortex_a8_vfp_macd): Use ffmad.
9905 * config/arm/cortex-a5.md (cortex_a5_fpmacs): Use ffmas.
9906 (cortex_a5_fpmacd): Use ffmad.
9907 * config/arm/cortex-a15-neon.md (cortex_a15_vfp_macs) Use ffmas.
9908 (cortex_a15_vfp_macd): Use ffmad.
9909 * config/arm/arm1020e.md (v10_fmul): Use ffmas and ffmad.
9911 2013-01-29 Jason Merrill <jason@redhat.com>
9914 * varasm.c (default_assemble_visibility): Don't warn about
9915 visibility on artificial decls.
9917 2013-01-29 Richard Biener <rguenther@suse.de>
9919 PR tree-optimization/56113
9920 * tree-ssa-structalias.c (equiv_class_lookup): Also return
9922 (label_visit): Free duplicate bitmaps and record the leader instead.
9923 (perform_var_substitution): Adjust.
9925 2013-01-29 Richard Biener <rguenther@suse.de>
9927 PR tree-optimization/55270
9928 * tree-ssa-dom.c (eliminate_degenerate_phis): If we changed
9929 the CFG, schedule loops for fixup.
9931 2013-01-29 Nick Clifton <nickc@redhat.com>
9933 * config/rl78/rl78.c (rl78_regno_mode_code_ok_for_base_p): Allow
9936 2013-01-28 Leif Ekblad <leif@rdos.net>
9938 * config.gcc (i[34567]86-*-rdos*, x86_64-*-rdos*): New targets.
9939 * config/i386/i386.h (TARGET_RDOS): New macro.
9940 (DEFAULT_LARGE_SECTION_THRESHOLD): New macro.
9941 * config/i386/i386.c (ix86_option_override_internal): For 64bit
9942 TARGET_RDOS, set ix86_cmodel to CM_MEDIUM_PIC and flag_pic to 1.
9943 * config/i386/i386.opt (mlarge-data-threshold): Initialize to
9944 DEFAULT_LARGE_SECTION_THRESHOLD.
9945 * config/i386/i386.md (R14_REG, R15_REG): New constants.
9946 * config/i386/rdos.h: New file.
9947 * config/i386/rdos64.h: New file.
9949 2013-01-28 Bernd Schmidt <bernds@codesourcery.com>
9952 * reload.c (find_valid_class_1): Use in_hard_reg_set_p instead of
9955 2013-01-28 Jakub Jelinek <jakub@redhat.com>
9957 PR rtl-optimization/56117
9958 * sched-deps.c (sched_analyze_2) <case PREFETCH>: For use_cselib
9959 call cselib_lookup_from_insn on the MEM before calling
9960 add_insn_mem_dependence.
9962 2013-01-28 Richard Biener <rguenther@suse.de>
9964 * tree-inline.c (remap_gimple_stmt): Do not assing a BLOCK
9965 to a stmt that didn't have one.
9966 (copy_phis_for_bb): Likewise for PHI arguments.
9967 (copy_debug_stmt): Likewise for debug stmts.
9969 2013-01-28 Richard Biener <rguenther@suse.de>
9971 PR tree-optimization/56034
9972 * tree-loop-distribution.c (enum partition_kind): Add PKIND_REDUCTION.
9973 (partition_builtin_p): Adjust.
9974 (generate_code_for_partition): Handle PKIND_REDUCTION. Assert
9975 it is the last partition.
9976 (rdg_flag_uses): Check SSA_NAME_IS_DEFAULT_DEF before looking
9977 up the vertex for the definition.
9978 (classify_partition): Classify whether a partition is a
9979 PKIND_REDUCTION, thus has uses outside of the loop.
9980 (ldist_gen): Inherit PKIND_REDUCTION when merging partitions.
9981 Merge all PKIND_REDUCTION partitions into the last partition.
9982 (tree_loop_distribution): Seed partitions from reductions as well.
9984 2013-01-28 Jakub Jelinek <jakub@redhat.com>
9986 PR tree-optimization/56125
9987 * tree-ssa-math-opts.c (gimple_expand_builtin_pow): Don't optimize
9988 pow(x,c) into sqrt(x) * powi(x, n/2) or
9989 1.0 / (sqrt(x) * powi(x, abs(n/2))) if c is an integer or when
9990 optimizing for size.
9991 Don't optimize pow(x,c) into powi(x, n/3) * powi(cbrt(x), n%3) or
9992 1.0 / (powi(x, abs(n)/3) * powi(cbrt(x), abs(n)%3)) if 2c is an
9995 PR tree-optimization/56094
9996 * gimplify.c (force_gimple_operand_1): Temporarily set input_location
9997 to UNKNOWN_LOCATION while gimplifying expr.
9999 2013-01-27 Uros Bizjak <ubizjak@gmail.com>
10002 * config/i386/i386.md (*movabs<mode>_1): Add square brackets around
10003 operand 0 in movabs insn template for -masm=intel asm alternative.
10004 (*movabs<mode>_2): Ditto for operand 1.
10006 2013-01-26 David Holsgrove <david.holsgrove@xilinx.com>
10009 * config.gcc (microblaze*-linux*): Add tmake_file to allow building
10012 2013-01-26 Edgar E. Iglesias <edgar.iglesias@gmail.com>
10014 * config.gcc (microblaze*-*-*): Rename microblaze*-*-elf, update
10017 2013-01-25 Naveen H.S <Naveen.Hurugalawadi@caviumnetworks.com>
10019 * config/aarch64/aarch64.c (TARGET_FIXED_CONDITION_CODE_REGS):
10020 Undef to avoid warning.
10022 2013-01-25 Michael Haubenwallner <michael.haubenwallner@salomon.at>
10024 * configure.ac (gcc_cv_ld_static_dynamic): Define for AIX native ld.
10025 * configure: Regenerate.
10027 2013-01-25 Jakub Jelinek <jakub@redhat.com>
10029 PR tree-optimization/56098
10030 * tree-ssa-phiopt.c (nt_init_block): Don't call add_or_mark_expr
10031 for stmts with volatile ops.
10032 (cond_store_replacement): Don't optimize if assign has volatile ops.
10033 (cond_if_else_store_replacement_1): Don't optimize if either
10034 then_assign or else_assign have volatile ops.
10035 (hoist_adjacent_loads): Don't optimize if either def1 or def2 have
10038 2013-01-25 Georg-Johann Lay <avr@gjlay.de>
10040 * doc/invoke.texi (AVR Built-in Macros): Document __XMEGA__.
10042 2013-01-25 Georg-Johann Lay <avr@gjlay.de>
10044 * doc/extend.texi (Example of asm with clobbered asm reg): Fix
10045 missing ':' in asm example.
10047 2013-01-25 Tejas Belagod <tejas.belagod@arm.com>
10049 * config/aarch64/aarch64-simd-builtins.def: Separate sq<r>dmulh_lane
10050 entries into lane and laneq entries.
10051 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_lane<mode>):
10052 Remove AdvSIMD scalar modes.
10053 (aarch64_sq<r>dmulh_laneq<mode>): New.
10054 (aarch64_sq<r>dmulh_lane<mode>): New RTL pattern for Scalar AdvSIMD
10056 * config/aarch64/arm_neon.h: Fix all the vq<r>dmulh_lane* intrinsics'
10057 builtin implementations to relfect changes in RTL in aarch64-simd.md.
10058 * config/aarch64/iterators.md (VCOND): New.
10061 2013-01-25 Georg-Johann Lay <avr@gjlay.de>
10064 * config/avr/builtins.def (DEF_BUILTIN): Add LIBNAME argument.
10065 Add NULL LIBNAME argument to existing definitions.
10066 (ABSHR, ABSR, ABSLR, ABSLLR, ABSHK, ABSK, ABSLK, ABSLLK): New.
10067 * config/avr/avr-c.c (DEF_BUILTIN): Add LIBNAME argument.
10068 * config/avr/avr.c (DEF_BUILTIN): Same.
10069 (avr_init_builtins): Pass down LIBNAME to add_builtin_function.
10070 (avr_expand_builtin): Expand to a vanilla call if a libgcc
10071 implementation is available (DECL_ASSEMBLER_NAME is set).
10072 (avr_fold_absfx): New static function.
10073 (avr_fold_builtin): Use it to handle: AVR_BUILTIN_ABSHR,
10074 AVR_BUILTIN_ABSR, AVR_BUILTIN_ABSLR, AVR_BUILTIN_ABSLLR,
10075 AVR_BUILTIN_ABSHK, AVR_BUILTIN_ABSK, AVR_BUILTIN_ABSLK,
10076 AVR_BUILTIN_ABSLLK.
10077 * config/avr/stdfix.h (abshr, absr, abslr, absllr)
10078 (abshk, absk, abslk, absllk): Provide as static inline functions.
10080 2013-01-25 Marek Polacek <polacek@redhat.com>
10082 PR tree-optimization/56035
10083 * cfgloopmanip.c (fix_loop_structure): Remove redundant condition.
10085 2012-01-24 Uros Bizjak <ubizjak@gmail.com>
10087 * config/i386/i386.md (*movti_internal_rex64): Add (o,e) alternative.
10088 (*movtf_internal_rex64): Add (!o,C) alternative
10089 (*movxf_internal_rex64): Ditto.
10090 (*movdf_internal_rex64): Add (?r,C) and (?m,C) alternatives.
10092 2013-01-24 Shenghou Ma <minux.ma@gmail.com>
10094 * doc/invoke.texi: fix typo.
10095 * doc/objc.texi: fix typo.
10097 2013-01-24 Richard Sandiford <rdsandiford@googlemail.com>
10099 * config/mips/mips.md (*and<mode>3_mips16): Use the "W" constraint
10100 for the first two alternatives.
10102 2013-01-24 Diego Novillo <dnovillo@google.com>
10104 * Makefile.in (GGC): Remove. Replace all instances with ggc-page.o.
10105 (ggc-zone.o): Remove.
10106 * configure.ac: Remove option --with-gc.
10107 * configure: Re-generate.
10108 * doc/install.texi: Remove documentation for --with-gc.
10109 * gengtype.c (write_enum_defn): Remove. Update all users.
10110 (write_Types_process_field): Remove generation of gt_e_* argument.
10111 (output_type_enum): Remove. Update all users.
10112 (write_enum_defn): Remove. Update all users.
10113 (enum alloc_zone): Remove. Update all users.
10114 (write_splay_tree_allocator_def): Remove generation of gt_e_* argument.
10115 * ggc-common.c (ggc_splay_alloc): Remove first argument.
10116 Update all callers.
10117 (struct ptr_data): Remove field TYPE. Update all users.
10118 (gt_pch_note_object): Remove argument TYPE. Update all users.
10119 * ggc-internal.h (ggc_pch_alloc_object): Remove last argument.
10121 * ggc-none.c (ggc_alloc_typed_stat): Remove.
10122 (struct alloc_zone): Remove.
10123 (ggc_internal_alloc_zone_stat): Remove.
10124 (ggc_internal_cleared_alloc_zone_stat): Remove.
10125 * ggc-page.c (ggc_alloc_typed_stat): Remove.
10126 (ggc_pch_count_object): Remove last argument. Update all users.
10127 (ggc_pch_alloc_object): Remove last argument. Update all users.
10128 (struct alloc_zone): Remove.
10129 * ggc-zone.c: Remove.
10130 * ggc.h (gt_pch_note_object): Remove last argument. Update all users.
10131 (struct alloc_zone): Remove.
10132 (ggc_alloc_typed_stat): Remove.
10133 (ggc_alloc_typed): Remove.
10134 (ggc_splay_alloc): Remove first argument.
10135 (rtl_zone): Remove. Update all users.
10136 (tree_zone): Remove. Update all users.
10137 (tree_id_zone): Remove. Update all users.
10138 (ggc_internal_zone_alloc_stat): Remove. Update all users.
10139 (ggc_internal_zone_cleared_alloc_stat): Remove. Update all users.
10140 (ggc_internal_zone_vec_alloc_stat): Remove. Update all users.
10141 * tree-ssanames.c: Remove references to zone allocator in comments.
10143 2013-01-24 Georg-Johann Lay <avr@gjlay.de>
10145 * config/avr/avr.c (avr_out_fract): Make register numbers that
10146 might be outside of source operand signed.
10148 2013-01-24 Uros Bizjak <ubizjak@gmail.com>
10150 * config/i386/constraints.md (Yf): New constraint.
10151 * config/i386/i386.md (*movdf_internal_rex64): Use Yf*f instead
10152 of f constraint to conditionaly disable x87 register preferences.
10153 (*movdf_internal): Ditto.
10154 (*movsf_internal): Ditto.
10156 2013-01-24 Steven Bosscher <steven@gcc.gnu.org>
10158 PR inline-asm/55934
10159 * lra-assigns.c (assign_by_spills): Throw away the pattern of asms
10160 that have operands with impossible constraints.
10161 Add a FIXME for a speed-up opportunity.
10162 * lra-constraints.c (process_alt_operands): Verify that a class
10163 selected from constraints on asms is valid for the operand mode.
10164 (curr_insn_transform): Remove incorrect comment.
10166 2013-01-23 David Edelsohn <dje.gcc@gmail.com>
10168 * config/rs6000/rs6000.c (rs6000_delegitimize_address): Check that
10169 TOC operand is a valid symbol ref in the constant pool.
10171 2013-01-23 Edgar E. Iglesias <edgar.iglesias@gmail.com>
10173 * config/microblaze/linux.h: Add TARGET_OS_CPP_BUILTINS
10175 2013-01-23 Georg-Johann Lay <avr@gjlay.de>
10178 * config/avr/stdfix.h: New file.
10179 * t-avr (stdfix-gcc.h): New rule to build it.
10180 (EXTRA_HEADERS): Set it to install stdfix.h, stdfix-gcc.h.
10182 2013-01-23 Kostya Serebryany <kcc@google.com>
10184 * config/darwin.h: remove dependency on
10185 CoreFoundation (asan on Mac OS).
10187 2013-01-23 Jakub Jelinek <jakub@redhat.com>
10190 * config/arm/arm.md (cbranchdi4, cstoredi4): Use s_register_operand
10191 instead of cmpdi_operand for first comparison operand.
10192 Don't assert that comparison operands aren't both constants.
10194 2013-01-22 Jonathan Wakely <jwakely.gcc@gmail.com>
10196 * doc/install.texi (Downloading the Source): Update references to
10197 downloading separate components.
10199 2013-01-22 Jonathan Wakely <jwakely.gcc@gmail.com>
10201 * doc/extend.texi (__int128): Improve grammar.
10203 2013-01-22 Uros Bizjak <ubizjak@gmail.com>
10206 * config/i386/i386.md (*movti_internal_rex64): Change (o,riF)
10207 alternative to (o,r).
10208 (*movdi_internal_rex64): Remove (!o,n) alternative.
10209 (DImode immediate->memory splitter): Remove.
10210 (DImode immediate->memory peephole2): Remove.
10211 (movtf): Enable for TARGET_64BIT || TARGET_SSE.
10212 (*movtf_internal_rex64): Rename from *movtf_internal. Change (!o,F*r)
10213 alternative to (!o,*r).
10214 (*movtf_internal_sse): New pattern.
10215 (*movxf_internal_rex64): New pattern.
10216 (*movxf_internal): Disable for TARGET_64BIT.
10217 (*movdf_internal_rex64): Remove (!o,F) alternative.
10219 2013-01-22 Jakub Jelinek <jakub@redhat.com>
10221 PR middle-end/56074
10222 * dumpfile.c (dump_loc): Only print loc if LOCATION_LOCUS (loc)
10223 isn't UNKNOWN_LOCATION nor BUILTINS_LOCATION.
10224 * tree-vect-loop-manip.c (find_loop_location): Also ignore
10225 stmt locations where LOCATION_LOCUS of the stmt location is
10226 UNKNOWN_LOCATION or BUILTINS_LOCATION.
10229 * config/i386/i386.md (UNSPEC_STOS): New.
10230 (strset_singleop, *strsetdi_rex_1, *strsetsi_1, *strsethi_1,
10231 *strsetqi_1): Add UNSPEC_STOS.
10233 2013-01-22 Paolo Carlini <paolo.carlini@oracle.com>
10236 * doc/invoke.texi: Remove left over -Wsynth example.
10238 2013-01-21 Jakub Jelinek <jakub@redhat.com>
10240 PR tree-optimization/56051
10241 * fold-const.c (fold_binary_loc): Don't fold
10242 X < (cast) (1 << Y) into (X >> Y) != 0 if cast is either
10243 a narrowing conversion, or widening conversion from signed
10246 2013-01-21 Uros Bizjak <ubizjak@gmail.com>
10248 PR rtl-optimization/56023
10249 * haifa-sched.c (fix_inter_tick): Do not update ticks of instructions,
10250 dependent on debug instruction.
10252 2013-01-21 Martin Jambor <mjambor@suse.cz>
10254 PR middle-end/56022
10255 * function.c (allocate_struct_function): Call
10256 invoke_set_current_function_hook earlier.
10258 2013-01-21 Jakub Jelinek <jakub@redhat.com>
10260 * reload1.c (init_reload): Only initialize reload_obstack
10261 during the first call.
10263 2013-01-21 Marek Polacek <polacek@redhat.com>
10265 * cfgloop.c (verify_loop_structure): Fix up grammar.
10267 2013-01-21 Yi-Hsiu Hsu <ahsu@marvell.com>
10269 * config/arm/marvell-pj4.md (pj4_shift_conds, pj4_alu_shift,
10270 pj4_alu_shift_conds, pj4_shift): Handle simple_alu_shift.
10272 2013-01-21 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10275 * config/arm/marvell-pj4.md: Update copyright year.
10276 Fix up use of alu to alu_reg and simple_alu_imm.
10278 2013-01-21 Uros Bizjak <ubizjak@gmail.com>
10280 * config/i386/i386.md (enabled): Do not disable fma4 for TARGET_FMA.
10282 2013-01-20 Vladimir Makarov <vmakarov@redhat.com>
10285 * lra-constraints.c (curr_insn_transform): Don't reuse original
10286 insn for secondary memory move when memory mode should be different.
10288 2013-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
10290 * config/pa/pa.md (atomic_loaddi, atomic_loaddi_1, atomic_storedi,
10291 atomic_storedi_1): New patterns.
10293 2013-01-20 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
10295 btver2 pipeline descriptions.
10296 * config/i386/i386.c: Enable CPU_BTVER2 to use btver2 pipeline
10298 * config/i386/i386.md (btver2_decode): New type attributes.
10299 * config/i386/sse.md (btver2_decode, btver2_sse_attr): New
10301 * config/i386/btver2.md: New file describing btver2 pipelines.
10303 2013-01-19 Andrew Pinski <apinski@cavium.com>
10305 PR tree-optimization/52631
10306 * tree-ssa-sccvn (visit_use): Before looking up the original
10307 statement, try looking up the simplified expression.
10309 2013-01-19 Anthony Green <green@moxielogic.com>
10311 * config/moxie/moxie.c (moxie_expand_prologue): Set
10312 current_function_static_stack_size.
10314 2013-01-18 Jakub Jelinek <jakub@redhat.com>
10316 PR tree-optimization/56029
10317 * tree-phinodes.c (reserve_phi_args_for_new_edge): Set
10318 gimple_phi_arg_location for the new arg to UNKNOWN_LOCATION.
10320 2013-01-18 Sharad Singhai <singhai@google.com>
10322 PR tree-optimization/55995
10323 * dumpfile.c (dump_loc): Print location only if available.
10324 * tree-vectorizer.c (increase_alignment): Intialize vect_location.
10326 2013-01-18 Vladimir Makarov <vmakarov@redhat.com>
10329 * lra-constraints.c (curr_insn_transform): Reuse original insn for
10330 secondary memory move.
10331 (inherit_reload_reg): Use rclass instead of cl for
10332 check_secondary_memory_needed_p.
10334 2013-01-18 Jakub Jelinek <jakub@redhat.com>
10336 PR middle-end/56015
10337 * expr.c (expand_expr_real_2) <case COMPLEX_EXPR>: Handle
10338 the case where writing real complex part of target modifies op1.
10340 2013-01-18 James Greenhalgh <james.greenhalgh@arm.com>
10342 * config/aarch64/aarch64-simd.md
10343 (aarch64_vcond_internal<mode>): Handle unordered cases.
10344 * config/aarch64/iterators.md (v_cmp_result): New.
10346 2013-01-18 Yi-Hsiu Hsu <ahsu@marvell.com>
10347 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10349 * config/arm/marvell-pj4.md: New file.
10350 * config/arm/arm.c (arm_issue_rate): Add marvell_pj4.
10351 * config/arm/arm.md (generic_sched): Add marvell_pj4.
10352 (generic_vfp): Likewise.
10353 * config/arm/arm-cores.def: Add marvell-pj4.
10354 * config/arm/arm-tune.md: Regenerate.
10355 * config/arm/arm-tables.opt: Regenerate.
10356 * config/arm/bpabi.h (BE8_LINK_SPEC): Add marvell_pj4.
10357 * doc/invoke.texi: Document marvell-pj4.
10359 2013-01-18 Tejas Belagod <tejas.belagod@arm.com>
10361 * config/aarch64/arm_neon.h: Map scalar types to standard types.
10363 2013-01-18 Alexandre Oliva <aoliva@redhat.com>
10368 * var-tracking.c (negative_power_of_two_p): New.
10369 (global_get_addr_cache, local_get_addr_cache): New.
10370 (get_addr_from_global_cache, get_addr_from_local_cache): New.
10371 (vt_canonicalize_addr): Rewrite using the above. Adjust the
10373 (vt_stack_offset_p): Remove.
10374 (vt_canon_true_dep): Always canonicalize loc's address.
10375 (clobber_overlapping_mems): Make sure we have a MEM.
10376 (local_get_addr_clear_given_value): New.
10377 (val_reset): Clear local cached entries.
10378 (compute_bb_dataflow): Create and release the local cache.
10379 Disable duplicate MEMs clobbering.
10380 (emit_notes_in_bb): Clobber MEMs likewise.
10381 (vt_emit_notes): Create and release the local cache.
10382 (vt_initialize, vt_finalize): Create and release the global
10383 cache, respectively.
10384 * alias.c (rtx_equal_for_memref_p): Compare operands of ENTRY_VALUEs.
10386 2013-01-18 Alexandre Oliva <aoliva@redhat.com>
10388 PR libmudflap/53359
10389 * tree-mudflap.c (mudflap_finish_file): Skip deferred decls
10390 not found in the symtab.
10392 2013-01-18 Alexandre Oliva <aoliva@redhat.com>
10395 PR rtl-optimization/55547
10396 PR rtl-optimization/53827
10399 * alias.c (offset_overlap_p): New, factored out of...
10400 (memrefs_conflict_p): ... this. Use absolute sizes. Retain
10401 the conservative special case for symbolic constants. Don't
10402 adjust zero sizes on alignment.
10404 2013-01-18 Bernd Schmidt <bernds@codesourcery.com>
10406 PR rtl-optimization/52573
10407 * regrename.c (build_def_use): Ignore REG_DEAD notes if there is a
10408 REG_UNUSED for the same register.
10410 2013-01-17 Richard Biener <rguenther@suse.de>
10411 Marek Polacek <polacek@redhat.com>
10413 PR rtl-optimization/55833
10414 * loop-unswitch.c (unswitch_loops): Move loop verification...
10415 (unswitch_single_loop): ...here. Call mark_irreducible_loops.
10416 * cfgloopmanip.c (fix_loop_placement): Add IRRED_INVALIDATED parameter.
10417 Set it to true when we're removing a loop from hierarchy tree in
10418 an irreducible region.
10419 (fix_bb_placements): Adjust caller.
10420 (fix_loop_placements): Likewise.
10422 2013-01-17 Georg-Johann Lay <avr@gjlay.de>
10424 * config/avr/builtins.def (DEF_BUILTIN): Factor out
10425 "__builtin_avr_" from NAME, turn NAME to an uppercase identifier.
10426 Factor out 'CODE_FOR_' from ICODE, use 'nothing' instead of '-1'.
10427 Remove ID. Adjust comments.
10428 * config/avr/avr-c.c (avr_builtin_name): Remove.
10429 (avr_cpu_cpp_builtins): Use DEF_BUILTIN instead of for-loop.
10430 * config/avr/avr.c (avr_tolower): New static function.
10431 (DEF_BUILTIN): Remove parameter ID. Prefix ICODE by 'CODE_FOR_'.
10432 Stringify NAME, prefix it with "__builtin_avr_" and lowercase it.
10433 (avr_expand_builtin): Assert insn_code != CODE_FOR_nothing for
10436 2013-01-17 Jan Hubicka <jh@suse.cz>
10438 PR tree-optimization/55273
10439 * loop-iv.c (iv_number_of_iterations): Consider zero iteration case.
10441 2013-01-17 Uros Bizjak <ubizjak@gmail.com>
10444 * config/i386/sync.md (atomic_store<mode>): Always generate SWImode
10445 store through atomic_store<mode>_1.
10446 (atomic_store<mode>_1): Macroize insn using SWI mode iterator.
10448 2013-01-17 Martin Jambor <mjambor@suse.cz>
10450 PR tree-optimizations/55264
10451 * ipa-inline-transform.c (can_remove_node_now_p_1): Never return true
10452 for virtual methods.
10453 * ipa.c (symtab_remove_unreachable_nodes): Never return true for
10454 virtual methods before inlining is over.
10455 * cgraph.h (cgraph_only_called_directly_or_aliased_p): Return false for
10457 * cgraphclones.c (cgraph_create_virtual_clone): Mark clones as
10460 2013-01-16 Vladimir Makarov <vmakarov@redhat.com>
10462 PR rtl-optimization/56005
10463 * sched-deps.c (sched_analyze_2): Check deps->readonly for adding
10464 pending reads for prefetch.
10466 2013-01-16 Ian Bolton <ian.bolton@arm.com>
10468 * config/aarch64/aarch64.md
10469 (*cstoresi_neg_uxtw): New pattern.
10470 (*cmovsi_insn_uxtw): New pattern.
10471 (*<optab>si3_uxtw): New pattern.
10472 (*<LOGICAL:optab>_<SHIFT:optab>si3_uxtw): New pattern.
10473 (*<optab>si3_insn_uxtw): New pattern.
10474 (*bswapsi2_uxtw): New pattern.
10476 2013-01-16 Richard Biener <rguenther@suse.de>
10478 * tree-inline.c (tree_function_versioning): Remove set but
10479 never used variable.
10481 2013-01-16 Richard Biener <rguenther@suse.de>
10483 PR tree-optimization/55964
10484 * tree-flow.h (rename_variables_in_loop): Remove.
10485 (rename_variables_in_bb): Likewise.
10486 * tree-loop-distribution.c (update_phis_for_loop_copy): Remove.
10487 (copy_loop_before): Adjust and delete update-ssa status.
10488 * tree-vect-loop-manip.c (rename_variables_in_bb): Make static.
10489 (rename_variables_in_bb): Likewise. Properly walk over predecessors.
10490 (rename_variables_in_loop): Remove.
10491 (slpeel_update_phis_for_duplicate_loop): Likewise.
10492 (slpeel_tree_duplicate_loop_to_edge_cfg): Handle nested loops,
10493 use available cfg machinery instead of duplicating it.
10494 Update PHI nodes and perform poor-mans SSA update here.
10495 (slpeel_tree_peel_loop_to_edge): Adjust.
10497 2013-01-16 Richard Biener <rguenther@suse.de>
10499 PR tree-optimization/54767
10500 PR tree-optimization/53465
10501 * tree-vrp.c (vrp_meet_1): Revert original fix for PR53465.
10502 (vrp_visit_phi_node): For PHI arguments coming via backedges
10503 drop all symbolical range information.
10504 (execute_vrp): Compute backedges.
10506 2013-01-16 Richard Biener <rguenther@suse.de>
10508 * doc/install.texi: Update CLooG and ISL requirements to
10511 2013-01-16 Christian Bruel <christian.bruel@st.com>
10514 * config/sh/sh.c (sh_expand_prologue): Postpone new_stack mem symbol.
10515 (broken_move): Handle UNSPECV_SP_SWITCH_B.
10516 * config/sh/sh.md (sp_switch_1): Use set (reg:SI SP_REG).
10518 2013-01-16 DJ Delorie <dj@redhat.com>
10520 * config/sh/sh.md (UNSPECV_SP_SWITCH_B): New.
10521 (UNSPECV_SP_SWITCH_E): New.
10522 (sp_switch_1): Change to an unspec.
10523 (sp_switch_2): Change to an unspec. Don't use post-inc when we
10526 2013-01-16 Uros Bizjak <ubizjak@gmail.com>
10528 * emit-rtl.c (need_atomic_barrier_p): Mask memory model argument
10529 with MEMMODEL_MASK before comparing with MEMMODEL_* memory types.
10530 * optabs.c (maybe_emit_sync_lock_test_and_set): Ditto.
10531 (expand_mem_thread_fence): Ditto.
10532 (expand_mem_signal_fence): Ditto.
10533 (expand_atomic_load): Ditto.
10534 (expand_atomic_store): Ditto.
10536 2013-01-16 Alexandre Oliva <aoliva@redhat.com>
10538 PR rtl-optimization/55547
10539 PR rtl-optimization/53827
10542 * alias.c (memrefs_conflict_p): Set sizes to negative after
10545 2013-01-15 Jakub Jelinek <jakub@redhat.com>
10548 * function.c (thread_prologue_and_epilogue_insns): Always
10549 add crtl->drap_reg to set_up_by_prologue.set, even if
10550 stack_realign_drap is false.
10552 2013-01-15 Jan-Benedict Glaw <jbglaw@lug-owl.de>
10554 * config/vax/vax.md (add<mode>3, sub<mode>3, mul<mode>3, div<mode>3,
10555 and<mode>3, *and<mode>_const_int, ior<mode>3, xor<mode>3, ashrsi3,
10556 *call): Fix indention.
10558 2013-01-15 Tom de Vries <tom@codesourcery.com>
10561 * optabs.c (widen_operand): Use gen_lowpart instead of gen_rtx_SUBREG.
10564 2013-01-15 Vladimir Makarov <vmakarov@redhat.com>
10566 PR rtl-optimization/55153
10567 * sched-deps.c (sched_analyze_2): Add pending reads for prefetch.
10569 2013-01-15 Martin Jambor <mjambor@suse.cz>
10571 PR tree-optimization/55920
10572 * tree-sra.c (analyze_access_subtree): Do not mark non-removable
10573 accesses as grp_to_be_debug_replaced.
10575 2013-01-15 Jakub Jelinek <jakub@redhat.com>
10577 PR tree-optimization/55920
10578 * tree-sra.c (sra_modify_assign): If for lacc->grp_to_be_debug_replaced
10579 there is non-useless type conversion needed from debug rhs to lhs,
10580 use build_debug_ref_for_model and/or VIEW_CONVERT_EXPR.
10582 2013-01-15 Joseph Myers <joseph@codesourcery.com>
10583 Mikael Pettersson <mikpe@it.uu.se>
10586 * config/arm/arm.h (ADDR_VEC_ALIGN): Align SImode jump tables for
10588 (ASM_OUTPUT_CASE_LABEL): Remove.
10589 (ASM_OUTPUT_BEFORE_CASE_LABEL): Define to empty.
10590 * final.c (shorten_branches): Update alignment of labels before
10591 jump tables if CASE_VECTOR_SHORTEN_MODE.
10593 2013-01-15 Richard Biener <rguenther@suse.de>
10596 * system.h: Do not include gmp.h for building host tools.
10598 2013-01-15 Richard Biener <rguenther@suse.de>
10600 PR middle-end/55882
10601 * emit-rtl.c (set_mem_attributes_minus_bitpos): Correctly
10602 account for bitpos when computing alignment.
10604 2013-01-15 Vladimir Yakovlev <vladimir.b.yakovlev@intel.com>
10606 * config/i386/i386-c.c (ix86_target_macros_internal): New case.
10607 (ix86_target_macros_internal): Likewise.
10609 * config/i386/i386.c (m_CORE2I7): Removed.
10610 (m_CORE_HASWELL): New macro.
10611 (m_CORE_ALL): Likewise.
10612 (initial_ix86_tune_features): m_CORE2I7 is replaced by m_CORE_ALL.
10613 (initial_ix86_arch_features): Likewise.
10614 (processor_target_table): Initializations for Core avx2.
10615 (cpu_names): New names "core-avx2".
10616 (ix86_option_override_internal): Changed PROCESSOR_COREI7 by
10617 PROCESSOR_CORE_HASWELL.
10618 (ix86_issue_rate): New case.
10619 (ia32_multipass_dfa_lookahead): Likewise.
10620 (ix86_sched_init_global): Likewise.
10622 * config/i386/i386.h (TARGET_HASWELL): New macro.
10623 (target_cpu_default): New TARGET_CPU_DEFAULT_haswell.
10624 (processor_type): New PROCESSOR_HASWELL.
10626 2013-01-15 Jakub Jelinek <jakub@redhat.com>
10628 PR tree-optimization/55955
10629 * tree-vect-loop.c (vectorizable_reduction): Give up early on
10630 *SHIFT_EXPR and *ROTATE_EXPR codes.
10632 PR tree-optimization/48766
10633 * opts.c (common_handle_option): For -fwrapv disable -ftrapv, for
10634 -ftrapv disable -fwrapv.
10636 2013-01-14 Georg-Johann Lay <avr@gjlay.de>
10639 * config/avr/avr-c.c (avr_cpu_cpp_builtins): Define __FLASH
10640 etc. to 1 and not to __flash.
10641 Use LL suffix for __INT24_MAX__ with -mint8.
10642 Use ULL suffix for __UINT24_MAX__ with -mint8.
10644 2013-01-14 Georg-Johann Lay <avr@gjlay.de>
10646 * config/avr/avr-arch.h
10647 (struct base_arch_s): Use typedef avr_arch_t instead.
10648 (struct arch_info_s): Use typedef avr_arch_info_t instead.
10649 (struct mcu_type_s): Use typedef avr_mcu_t instead.
10650 * config/avr/avr.c: Same.
10651 * config/avr/avr-devices.c: Same.
10652 * config/avr/driver-avr.c: Same.
10653 * config/avr/gen-avr-mmcu-texi.c: Same.
10654 * config/avr/avr-mcus.def: Adjust comment.
10656 2013-01-14 Tejas Belagod <tejas.belagod@arm.com>
10658 * config/aarch64/aarch64-simd.md (*aarch64_simd_ld1r<mode>): New.
10659 * config/aarch64/iterators.md (VALLDI): New.
10661 2013-01-14 Uros Bizjak <ubizjak@gmail.com>
10662 Andi Kleen <ak@linux.intel.com>
10665 * config/i386/sync.md (atomic_store<mode>_1): New pattern.
10666 (atomic_store<mode>): Call atomic_store<mode>_1 for IX86_HLE_RELEASE
10669 2013-01-14 Georg-Johann Lay <avr@gjlay.de>
10671 * config/avr/avr-stdint.h: Remove trailing blanks.
10672 * config/avr/avr-log.h: Same.
10673 * config/avr/avr-arch.h: Same.
10674 * config/avr/avr-devices.c: Same.
10675 * config/avr/avr-dimode.md: Same.
10676 * config/avr/predicates.md: Same.
10677 * config/avr/avr-c.c: Same. And fix typo.
10679 * config/avr/avr-protos.h: Same. And:
10680 (function_arg_regno_p): Rename to avr_function_arg_regno_p.
10681 (init_cumulative_args): Rename to avr_init_cumulative_args.
10682 (expand_prologue): Rename to avr_expand_prologue.
10683 (expand_epilogue): Rename to avr_expand_epilogue.
10684 (adjust_insn_length): Rename to avr_adjust_insn_length.
10685 (notice_update_cc): Rename to avr_notice_update_cc.
10686 (final_prescan_insn): Rename to avr_final_prescan_insn.
10687 * config/avr/avr.c: Same.
10688 * config/avr/avr.h: Same.
10689 * config/avr/avr.md: Remove trailing blanks.
10690 (prologue): Use avr_expand_prologue.
10691 (epilogue, sibcall_epilogue): Use avr_expand_epilogue.
10693 2013-01-14 Richard Biener <rguenther@suse.de>
10695 * tree-cfg.c (verify_expr_location, verify_expr_location_1,
10696 verify_location, collect_subblocks): New functions.
10697 (verify_gimple_in_cfg): Verify that locations only reference
10698 BLOCKs in the functions BLOCK tree.
10700 2013-01-14 Richard Biener <rguenther@suse.de>
10702 * tree-cfgcleanup.c (remove_forwarder_block): Unshare propagated
10704 * graphite-sese-to-poly.c (insert_out_of_ssa_copy): Properly
10706 (insert_out_of_ssa_copy_on_edge): Likewise.
10707 (rewrite_close_phi_out_of_ssa): Likewise.
10708 * tree-ssa.c (insert_debug_temp_for_var_def): Properly unshare
10710 * tree-ssa-pre.c (insert_into_preds_of_block): Properly unshare
10711 propagated constants.
10712 * tree-cfg.c (tree_node_can_be_shared): Handled component-refs
10715 2013-01-14 Georg-Johann Lay <avr@gjlay.de>
10717 * config/avr/avr-modes.def: Add GPL copyright notice.
10719 2013-01-13 Uros Bizjak <ubizjak@gmail.com>
10721 * config/i386/sync.md (mem_thread_fence): Mask operands[0] with
10722 MEMMODEL_MASK to determine memory model.
10723 (atomic_store<mode>): Ditto from operands[2].
10724 * config/i386/i386.c (ix86_memmodel_check): Declare "strong" as bool.
10726 2013-01-13 Jakub Jelinek <jakub@redhat.com>
10729 * gimple-fold.c (get_symbol_constant_value): Call unshare_expr.
10730 (fold_gimple_assign): Don't call unshare_expr here.
10731 (fold_ctor_reference): Call unshare_expr.
10733 2013-01-13 Terry Guo <terry.guo@arm.com>
10735 * Makefile.in (s-mlib): New argument MULTILIB_REUSE.
10736 * doc/fragments.texi: Document MULTILIB_REUSE.
10737 * gcc.c (multilib_reuse): New internal spec.
10738 (set_multilib_dir): Also search multilib from multilib_reuse.
10739 * genmultilib (tmpmultilib3): Refactor code.
10740 (tmpmultilib4): Ditto.
10741 (multilib_reuse): New multilib argument.
10743 2013-01-13 Richard Sandiford <rdsandiford@googlemail.com>
10745 * Makefile.in: Update copyright.
10747 2013-01-12 Tom de Vries <tom@codesourcery.com>
10749 PR middle-end/55890
10750 * calls.c (expand_call): Check if arg_nr is valid.
10752 2013-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10754 * doc/extend.texi (X86 Built-in Functions): Add whitespace in
10755 __builtin_ia32_paddb256 and __builtin_ia32_pavgb256
10756 documentation. Add missing '__' in front of
10757 __builtin_ia32_packssdw256.
10759 2013-01-11 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
10762 * config/s390/s390.c (s390_preferred_reload_class): Do not return
10763 NO_REGS for larl operands.
10764 (s390_reload_larl_operand): Use s390_load_address instead of
10767 2013-01-11 Richard Biener <rguenther@suse.de>
10769 * tree-cfg.c (verify_node_sharing_1): Split out from ...
10770 (verify_node_sharing): ... here.
10771 (verify_gimple_in_cfg): Use verify_node_sharing_1 for walk_tree.
10773 2013-01-11 Eric Botcazou <ebotcazou@adacore.com>
10775 * configure.ac (Tree checking): Set TREECHECKING to yes if enabled.
10776 Substitute TREECHECKING.
10777 * configure: Regenerate.
10778 * Makefile.in (TREECHECKING): New.
10780 2013-01-11 Richard Guenther <rguenther@suse.de>
10782 PR tree-optimization/44061
10783 * tree-vrp.c (extract_range_basic): Compute zero as
10784 value-range for __builtin_constant_p of function parameters.
10786 2013-01-10 Richard Sandiford <rdsandiford@googlemail.com>
10788 Update copyright years.
10790 2013-01-10 Vladimir Makarov <vmakarov@redhat.com>
10792 PR rtl-optimization/55672
10793 * lra-eliminations.c (mark_not_eliminable): Permit addition with
10794 const to be eliminable.
10796 2013-01-10 David Edelsohn <dje.gcc@gmail.com>
10798 * configure.ac (HAVE_AS_TLS): Add check for powerpc-ibm-aix.
10799 * configure: Regenerate.
10801 2013-01-10 Richard Biener <rguenther@suse.de>
10803 * builtins.c (expand_builtin_init_trampoline): Use set_mem_attributes.
10805 2013-01-10 Richard Biener <rguenther@suse.de>
10808 * tree-into-ssa.c (rewrite_add_phi_arguments): Do not set
10809 locations for virtual PHI arguments.
10810 (rewrite_update_phi_arguments): Likewise.
10812 2013-01-10 Joel Sherrill <joel.sherrill@OARcorp.com>
10814 * config/v850/rtems.h (ASM_SPEC): Pass -m8byte-align and -mgcc-abi
10817 2013-01-10 Jakub Jelinek <jakub@redhat.com>
10819 PR tree-optimization/55921
10820 * tree-complex.c (expand_complex_asm): New function.
10821 (expand_complex_operations_1): Call it for GIMPLE_ASM.
10823 2013-01-10 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
10826 * config/s390/s390.c (s390_symref_operand_p)
10827 (s390_loadrelative_operand_p): Merge the two functions.
10828 (s390_check_qrst_address, print_operand_address): Add parameters
10829 to s390_loadrelative_operand_p invokation.
10830 (s390_check_symref_alignment): Use s390_loadrelative_operand_p.
10831 (s390_reload_larl_operand, s390_secondary_reload): Use
10832 s390_loadrelative_operand_p instead of s390_symref_operand_p.
10833 (legitimize_pic_address): Handle @GOTENT and @PLT + addend.
10835 2013-01-09 Mike Stump <mikestump@comcast.net>
10837 * dse.c (record_store): Remove unnecessary assert.
10839 2013-01-09 Jan Hubicka <jh@suse.cz>
10841 PR tree-optimization/55569
10842 * cfgloopmanip.c (scale_loop_profile): Make ITERATION_BOUND gcov_type.
10843 * cfgloop.h (scale_loop_profile): Likewise.
10845 2013-01-09 Jan Hubicka <jh@suse.cz>
10848 * ipa-inline.c (ipa_inline): Remove extern inlines and virtual
10850 * cgraphclones.c (cgraph_clone_node): Cpoy also LTO file data.
10852 2013-01-09 Richard Sandiford <rdsandiford@googlemail.com>
10854 PR middle-end/55114
10855 * expr.h (maybe_emit_group_store): Declare.
10856 * expr.c (maybe_emit_group_store): New function.
10857 * builtins.c (expand_builtin_int_roundingfn): Call it.
10858 (expand_builtin_int_roundingfn_2): Likewise.
10860 2013-01-09 Vladimir Makarov <vmakarov@redhat.com>
10862 PR rtl-optimization/55829
10863 * lra-constraints.c (match_reload): Add code for absent output.
10864 (curr_insn_transform): Add code for reloads of matched inputs
10867 2013-01-09 Uros Bizjak <ubizjak@gmail.com>
10869 * config/i386/sse.md (*vec_interleave_highv2df): Change mode
10870 attribute of movddup insn to DF.
10871 (*vec_interleave_lowv2df): Ditto.
10872 (vec_dupv2df): Ditto.
10874 2013-01-09 Jan Hubicka <jh@suse.cz>
10876 PR tree-optimiation/55875
10877 * tree-ssa-loop-niter.c (number_of_iterations_cond): Add
10878 EVERY_ITERATION parameter.
10879 (number_of_iterations_exit): Check if exit is executed every iteration.
10880 (idx_infer_loop_bounds): Similarly here.
10881 (n_of_executions_at_most): Simplify
10882 to only test for cases where statement is dominated by the
10883 particular bound; handle correctly the "postdominance" test.
10884 (scev_probably_wraps_p): Use max loop iterations info
10885 as a global bound first.
10887 2013-01-09 Nguyen Duy Dat <dat.nguyen.yn@rvc.renesas.com>
10888 Nick Clifton <nickc@redhat.com>
10890 * config/v850/v850.md (cbranchsf4): New pattern.
10891 (cstoresf4): New pattern.
10892 (cbranchdf4): New pattern.
10893 (cstoredf4): New pattern.
10894 (movsicc): Disallow floating point comparisons.
10895 (cmpsf_le_insn): Fix order of operators.
10896 (cmpsf_lt_insn): Likewise.
10897 (cmpsf_eq_insn): Likewise.
10898 (cmpdf_le_insn): Likewise.
10899 (cmpdf_lt_insn): Likewise.
10900 (cmpdf_eq_insn): Likewise.
10901 (cmpsf_ge_insn): Use LE comparison.
10902 (cmpdf_ge_insn): Likewise.
10903 (cmpsf_gt_insn): Use LT comparison.
10904 (cmpdf_gt_insn): Likewise.
10905 (cmpsf_ne_insn): Delete pattern.
10906 (cmpdf_ne_insn): Delete pattern.
10907 * config/v850/v850.c (v850_gen_float_compare): Use
10908 gen_cmpdf_eq_insn for NE comparison.
10909 (v850_float_z_comparison_operator)
10910 (v850_float_nz_comparison_operator): Move from here ...
10911 * config/v850/predicates.md: ... to here. Move GT and GE
10912 comparisons into v850_float_z_comparison_operator.
10913 * config/v850/v850-protos.h (v850_float_z_comparison_operator):
10915 (v850_float_nz_comparison_operator): Likewise.
10917 2013-01-09 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
10919 * config/pa/pa.c (pa_emit_move_sequence): Replace calls to gen_insv
10920 with calls to gen_insvsi/gen_insvdi.
10922 2013-01-09 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
10924 * config/i386/i386.c (initial_ix86_tune_features): Set up
10925 X86_TUNE_AVX128_OPTIMAL for m_BTVER2.
10927 2013-01-09 Steven Bosscher <steven@gcc.gnu.org>
10928 Jakub Jelinek <jakub@redhat.com>
10930 PR tree-optimization/48189
10931 * predict.c (predict_loops): If max is 0, don't call compare_tree_int.
10932 If nitercst is 0, don't predict the exit edge.
10934 2013-01-08 Naveen H.S <Naveen.Hurugalawadi@caviumnetworks.com>
10936 * config/aarch64/aarch64.c (aarch64_print_operand): Replace %r
10937 in asm_fprintf with reg_names.
10938 (aarch64_print_operand_address): Likewise.
10939 (aarch64_return_addr): Likewise.
10940 * config/aarch64/aarch64.h (ASM_FPRINTF_EXTENSIONS): Remove.
10942 2013-01-08 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
10944 * config/pa/pa.h (VAL_U6_BITS_P): Define.
10945 (INT_U6_BITS): Likewise.
10946 * config/pa/predicates.md (uint6_operand): New predicate.
10947 (shift5_operand, shift6_operand): Likewise.
10948 * config/pa/pa.md (lshrsi3, rotrsi3): Use shift5_operand instead of
10950 (lshrdi3): Use shift6_operand.
10951 (shrpsi4, shrpdi4): New insn patterns.
10952 (extzv): Delete expander.
10953 (extzvsi, extzvdi): New expanders. Use uint5_operand and uint6_operand
10954 predicates in unamed zero extract patterns. Tighten common constraint.
10955 (extv): Delete expander.
10956 (extvsi, extvdi): New expanders. Use uint5_operand and uint6_operand
10957 predicates in unamed sign extract patterns. Tighten common constraint.
10958 (insv): Delete expander.
10959 (insvsi, insvdi): New expanders. Use uint5_operand and uint6_operand
10960 predicates in unamed insert patterns. Tighten common constraint.
10961 Change uint32_operand predicate to uint6_operand predicate in unamed
10962 DImode pattern to insert constant values of type 1...1xxxx.
10964 2013-01-04 Jan Hubicka <jh@suse.cz>
10966 PR tree-optimization/55823
10967 * ipa-prop.c (update_indirect_edges_after_inlining): Fix ordering
10970 2013-01-08 Jakub Jelinek <jakub@redhat.com>
10971 Uros Bizjak <ubizjak@gmail.com>
10973 PR rtl-optimization/55845
10974 * df-problems.c (can_move_insns_across): Stop scanning at
10975 volatile_insn_p source instruction or give up if
10976 across_from .. across_to range contains any volatile_insn_p
10979 2013-01-08 Tejas Belagod <tejas.belagod@arm.com>
10981 * config/aarch64/aarch64-simd.md (vec_init<mode>): New.
10982 * config/aarch64/aarch64-protos.h (aarch64_expand_vector_init):
10984 * config/aarch64/aarch64.c (aarch64_simd_dup_constant,
10985 aarch64_simd_make_constant, aarch64_expand_vector_init): New.
10987 2013-01-08 Jakub Jelinek <jakub@redhat.com>
10990 * asan.c (asan_clear_shadow): New function.
10991 (asan_emit_stack_protection): Use it.
10993 2013-01-08 Tejas Belagod <tejas.belagod@arm.com>
10995 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_<su>mult_lo_<mode>,
10996 aarch64_simd_vec_<su>mult_hi_<mode>): Separate instruction and operand
10997 with tab instead of space.
10999 2013-01-08 Nick Clifton <nickc@redhat.com>
11001 * config/rl78/rl78.c (rl78_expand_prologue): Always select
11002 register bank 0 at the start of an interrupt handler.
11003 * config/rl78/rl78.md (mulsi3_g13): Correct values for MDBL and
11006 2013-01-08 James Greenhalgh <james.greenhalgh@arm.com>
11008 * config/aarch64/aarch64-simd.md
11009 (aarch64_simd_bsl<mode>_internal): Add floating-point modes.
11010 (aarch64_simd_bsl): Likewise.
11011 (aarch64_vcond_internal<mode>): Likewise.
11012 (vcond<mode><mode>): Likewise.
11013 (aarch64_cm<cmp><mode>): Fix constraints, add new modes.
11014 * config/aarch64/iterators.md (V_cmp_result): Add V2DF.
11016 2013-01-08 James Greenhalgh <james.greenhalgh@arm.com>
11018 * config/aarch64/aarch64-builtins.c
11019 (aarch64_builtin_vectorized_function): Handle sqrt, sqrtf.
11021 2013-01-08 Martin Jambor <mjambor@suse.cz>
11024 * tree-sra.c (analyze_access_subtree): Return true also after
11025 potentially creating a debug-only replacement.
11027 2013-01-08 Jakub Jelinek <jakub@redhat.com>
11029 PR middle-end/55890
11030 * tree-ssa-ccp.c (evaluate_stmt): Use gimple_call_builtin_p.
11032 PR tree-optimization/54120
11033 * tree-vrp.c (range_fits_type_p): Don't allow
11034 src_precision < precision from signed vr to unsigned_p
11035 if vr->min or vr->max is negative.
11036 (simplify_float_conversion_using_ranges): Test can_float_p
11037 against CODE_FOR_nothing.
11039 2013-01-08 Jakub Jelinek <jakub@redhat.com>
11040 Richard Biener <rguenther@suse.de>
11042 PR middle-end/55851
11043 * fold-const.c (int_binop_types_match_p): Allow all INTEGRAL_TYPE_P
11044 types instead of just INTEGER_TYPE types.
11046 2013-01-07 Mark Kettenis <kettenis@openbsd.org>
11048 * config/i386/openbsdelf.h (LIBGCC2_HAS_TF_MODE, LIBGCC2_TF_CEXT,
11051 2013-01-07 Steve Ellcey <sellcey@mips.com>
11054 * config/mips/mips.opt: Change mad to mmad to match documentation.
11056 2013-01-07 Georg-Johann Lay <avr@gjlay.de>
11059 * doc/extend.texi (AVR Named Address Spaces): __memx goes into
11060 .progmemx.data now.
11062 2013-01-07 Georg-Johann Lay <avr@gjlay.de>
11065 * config/avr/avr.h (ADDR_SPACE_COUNT): New enum.
11066 (avr_addrspace_t): Add .section_name field.
11067 * config/avr/avr.c (progmem_section): Use ADDR_SPACE_COUNT as
11069 (avr_addrspace): Same. Initialize .section_name. Remove last
11070 NULL entry. Put __memx into .progmemx.data.
11071 (progmem_section_prefix): Remove.
11072 (avr_asm_init_sections): No need to initialize progmem_section.
11073 (avr_asm_named_section): Use avr_addrspace[].section_name to get
11074 section name prefix.
11075 (avr_asm_select_section): Ditto. And use get_unnamed_section to
11076 retrieve the progmem section.
11077 * avr-c.c (avr_cpu_cpp_builtins): Use ADDR_SPACE_COUNT as loop
11078 boundary to run over avr_addrspace[].
11079 (avr_register_target_pragmas): Ditto.
11081 2013-01-06 Jakub Jelinek <jakub@redhat.com>
11083 * varasm.c (output_constant_def_contents): For asan_protect_global
11084 protected strings, adjust DECL_ALIGN if needed, before testing for
11086 (place_block_symbol): Adjust size for asan protected STRING_CSTs if
11087 TREE_CONSTANT_POOL_ADDRESS_P. Increase alignment for asan protected
11089 (output_object_block): For asan protected decls, emit asan padding
11090 after their contents.
11091 * asan.c (asan_protect_global): Don't check TREE_ASM_WRITTEN here.
11092 (asan_finish_file): Test it here instead.
11094 2013-01-07 Nick Clifton <nickc@redhat.com>
11095 Matthias Klose <doko@debian.org>
11096 Doug Kwan <dougkwan@google.com>
11097 H.J. Lu <hongjiu.lu@intel.com>
11100 * collect2.c (main): Support -fuse-ld=bfd and -fuse-ld=gold.
11102 * common.opt: Add fuse-ld=bfd and fuse-ld=gold.
11104 * gcc.c (LINK_COMMAND_SPEC): Pass -fuse-ld=* to collect2.
11106 * opts.c (comman_handle_option): Ignore -fuse-ld=bfd and -fuse-ld=gold.
11108 * doc/invoke.texi: Document -fuse-ld=bfd and -fuse-ld=gold.
11110 2013-01-07 Georg-Johann Lay <avr@gjlay.de>
11113 * doc/install.texi (Cross-Compiler-Specific Options): Document
11116 2013-01-07 Tejas Belagod <tejas.belagod@arm.com>
11118 * config/aarch64/arm_neon.h (vmovn_high_is16, vmovn_high_s32,
11119 vmovn_high_s64, vmovn_high_u16, vmovn_high_u32, vmovn_high_u64,
11120 vqmovn_high_s16, vqmovn_high_s32, vqmovn_high_s64, vqmovn_high_u16,
11121 vqmovn_high_u32, vqmovn_high_u64, vqmovun_high_s16, vqmovun_high_s32,
11122 vqmovun_high_s64): Fix source operand number and update copyright.
11124 2013-01-07 Richard Biener <rguenther@suse.de>
11126 PR middle-end/55890
11127 * gimple.h (gimple_call_builtin_p): New overload.
11128 * gimple.c (validate_call): New function.
11129 (gimple_call_builtin_p): Likewise.
11130 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
11131 Use gimple_call_builtin_p.
11132 (find_func_clobbers): Likewise.
11133 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
11134 (strlen_optimize_stmt): Likewise.
11136 2013-01-07 James Greenhalgh <james.greenhalgh@arm.com>
11138 * config/aarch64/arm_neon.h (vld1_dup_*): Make argument const.
11139 (vld1q_dup_*): Likewise.
11140 (vld1_*): Likewise.
11141 (vld1q_*): Likewise.
11142 (vld1_lane_*): Likewise.
11143 (vld1q_lane_*): Likewise.
11145 2013-01-07 Richard Biener <rguenther@suse.de>
11147 * lto-streamer.h (LTO_minor_version): Bump to 2.
11149 2013-01-07 James Greenhalgh <james.greenhalgh@arm.com>
11151 * config/aarch64/aarch64-protos.h
11152 (aarch64_const_double_zero_rtx_p): Rename to...
11153 (aarch64_float_const_zero_rtx_p): ...this.
11154 (aarch64_float_const_representable_p): New.
11155 (aarch64_output_simd_mov_immediate): Likewise.
11156 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>): Refactor
11157 move immediate case.
11158 * config/aarch64/aarch64.c
11159 (aarch64_const_double_zero_rtx_p): Rename to...
11160 (aarch64_float_const_zero_rtx_p): ...this.
11161 (aarch64_print_operand): Allow printing of new constants.
11162 (aarch64_valid_floating_const): New.
11163 (aarch64_legitimate_constant_p): Check for valid floating-point
11165 (aarch64_simd_valid_immediate): Likewise.
11166 (aarch64_vect_float_const_representable_p): New.
11167 (aarch64_float_const_representable_p): Likewise.
11168 (aarch64_simd_imm_zero_p): Also allow for floating-point 0.0.
11169 (aarch64_output_simd_mov_immediate): New.
11170 * config/aarch64/aarch64.md (*movsf_aarch64): Add new alternative.
11171 (*movdf_aarch64): Likewise.
11172 * config/aarch64/constraints.md (Ufc): New.
11173 (Y): call aarch64_float_const_zero_rtx.
11174 * config/aarch64/predicates.md (aarch64_fp_compare_operand): New.
11176 2013-01-07 Richard Biener <rguenther@suse.de>
11178 PR tree-optimization/55888
11179 PR tree-optimization/55862
11180 * tree-ssa-pre.c (phi_translate_1): Revert previous change.
11181 (valid_in_sets): Check if a NAME has a leader in AVAIL_OUT,
11182 not if it is contained therein.
11184 2013-01-07 Georg-Johann Lay <avr@gjlay.de>
11186 * config/avr/t-avr: Typo.
11188 2013-01-07 Georg-Johann Lay <avr@gjlay.de>
11191 * config/avr/t-avr: Don't automatically rebuild
11192 $(srcdir)/config/avr/t-multilib
11193 $(srcdir)/config/avr/avr-tables.opt
11194 $(srcdir)/doc/avr-mmcu.texi
11195 (avr-mcus): New phony target to build them on request.
11196 (s-avr-mlib, s-avr-mmcu-texi): Remove.
11197 * avr/avr-mcus.def: Adjust comments.
11199 2013-01-07 Uros Bizjak <ubizjak@gmail.com>
11201 * config/i386/i386.c (DEFAULT_PCC_STRUCT_RETURN): Remove.
11203 2013-01-06 Richard Sandiford <rdsandiford@googlemail.com>
11205 * file-find.c, file-find.h, realmpfr.c: Add FSF as copyright holder.
11207 2013-01-06 Richard Sandiford <rdsandiford@googlemail.com>
11209 * config/tilepro/gen-mul-tables.cc: Put copyright on one line.
11211 2013-01-05 David Edelsohn <dje.gcc@gmail.com>
11213 * config/rs6000/aix53.h (LIB_SPEC): Add -lpthreads when compiling
11214 to generate profiling.
11215 * config/rs6000/aix64.h (LIB_SPEC): Same.
11217 2013-01-04 Andrew Pinski <apinski@cavium.com>
11219 * config/aarch64/aarch64.c (aarch64_fixed_condition_code_regs):
11221 (TARGET_FIXED_CONDITION_CODE_REGS): Define.
11223 2013-01-04 Uros Bizjak <ubizjak@gmail.com>
11225 * config/i386/i386.c (ix86_legitimize_address): Call convert_to_mode
11227 (ix86_expand_move): Ditto.
11228 (ix86_zero_extend_to_Pmode): Ditto.
11229 (ix86_expand_call): Ditto.
11230 (ix86_expand_special_args_builtin): Ditto.
11231 (ix86_expand_builtin): Ditto.
11233 2013-01-04 Richard Biener <rguenther@suse.de>
11235 PR tree-optimization/55862
11236 * tree-ssa-pre.c (phi_translate_1): Valueize SSA names after
11237 translating them through PHI nodes.
11239 2013-01-04 Martin Jambor <mjambor@suse.cz>
11241 PR tree-optimization/55755
11242 * tree-sra.c (sra_modify_assign): Do not check that an access has no
11243 children when trying to avoid producing a VIEW_CONVERT_EXPR.
11245 2013-01-04 Marek Polacek <polacek@redhat.com>
11247 PR middle-end/55859
11248 * opts.c (default_options_optimization): Clarify error message.
11250 2013-01-04 Richard Biener <rguenther@suse.de>
11252 PR middle-end/55863
11253 * fold-const.c (split_tree): Undo -X - 1 to ~X folding for
11256 2013-01-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
11259 * config/pa/pa.md (movsi): Revert previous change.
11260 * config/pa/pa.c (pa_legitimate_constant_p): Reject all TLS symbol
11263 2013-01-03 Richard Henderson <rth@redhat.com>
11265 * config/i386/i386.c (ix86_expand_move): Always assign to op1
11266 after eliminating TLS symbols.
11268 2013-01-03 Marc Glisse <marc.glisse@inria.fr>
11271 * graphite-interchange.c (pdr_stride_in_loop): Use gmp_fprintf.
11272 * graphite-poly.c (debug_gmp_value): Likewise.
11274 2013-01-03 Uros Bizjak <ubizjak@gmail.com>
11277 * config/i386/i386-c.c (ix86_target_macros_internal): Depending on
11278 selected code model, define __code_mode_small__, __code_model_medium__,
11279 __code_model_large__, __code_model_32__ or __code_model_kernel__.
11280 * config/i386/cpuid.h (__cpuid, __cpuid_count) [__i386__]: Prefix
11281 xchg temporary register with %k. Declare temporary register as
11283 [__x86_64__]: For medium and large code models, preserve %rbx register.
11285 2013-01-03 Richard Biener <rguenther@suse.de>
11287 * tree-data-ref.c (dump_conflict_function): Use less vertical spacing.
11288 (dump_subscript): Adjust.
11289 (finalize_ddr_dependent): Do not dump redundant info.
11290 (analyze_siv_subscript): Adjust.
11291 (subscript_dependence_tester): Likewise.
11292 (compute_affine_dependence): Likewise.
11294 2013-01-03 Richard Biener <rguenther@suse.de>
11297 2013-01-03 Richard Biener <rguenther@suse.de>
11299 PR tree-optimization/55857
11300 * tree-vect-stmts.c (vectorizable_load): Do not setup
11301 re-alignment for invariant loads.
11303 2013-01-02 Richard Biener <rguenther@suse.de>
11305 * tree-vect-stmts.c (vectorizable_load): When vectorizing an
11306 invariant load do not generate a vector load from the scalar location.
11308 2013-01-03 Richard Biener <rguenther@suse.de>
11310 * tree-vect-loop.c (vect_analyze_loop_form): Clarify reason
11311 for not vectorizing.
11312 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Do
11313 not build INDIRECT_REFs, call get_name once only.
11314 (vect_create_data_ref_ptr): Likewise. Dump base object kind
11315 based on DR_BASE_OBJECT, not DR_BASE_ADDRESS.
11317 2013-01-03 Richard Biener <rguenther@suse.de>
11319 PR tree-optimization/55857
11320 * tree-vect-stmts.c (vectorizable_load): Do not setup
11321 re-alignment for invariant loads.
11323 2013-01-03 Richard Biener <rguenther@suse.de>
11326 * lto-symtab.c (lto_symtab_merge_decls_1): As last resort, always
11327 prefer a built-in decl.
11329 2013-01-03 Jakub Jelinek <jakub@redhat.com>
11331 * gcc.c (process_command): Update copyright notice dates.
11332 * gcov.c (print_version): Likewise.
11333 * gcov-dump.c (print_version): Likewise.
11335 PR rtl-optimization/55838
11336 * loop-iv.c (iv_number_of_iterations): Call lowpart_subreg on
11337 iv0.step, iv1.step and step.
11339 2013-01-03 Jakub Jelinek <jakub@redhat.com>
11340 Marc Glisse <marc.glisse@inria.fr>
11342 PR tree-optimization/55832
11343 * fold-const.c (fold_binary_loc): For ABS_EXPR<x> >= 0 and
11344 ABS_EXPR<x> < 0 folding use constant_boolean_node instead of
11345 integer_{one,zero}_node.
11347 2013-01-03 Jakub Jelinek <jakub@redhat.com>
11350 * params.def (PARAM_MAX_VARTRACK_REVERSE_OP_SIZE): New param.
11351 * var-tracking.c (reverse_op): Don't add reverse ops to
11352 VALUEs that have already
11353 PARAM_VALUE (PARAM_MAX_VARTRACK_REVERSE_OP_SIZE) or longer locs list.
11355 2013-01-02 Gerald Pfeifer <gerald@pfeifer.com>
11357 * doc/contrib.texi: Note years as release manager for Mark Mitchell.
11359 2013-01-02 Teresa Johnson <tejohnson@google.com>
11361 * dumpfile.c (dump_loc): Print filename with location.
11362 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Use
11363 new location_t parameter to emit complete unroll message with
11364 new dump framework.
11365 (canonicalize_loop_induction_variables): Compute loops location
11366 and pass to try_unroll_loop_completely.
11367 * loop-unroll.c (report_unroll_peel): New function.
11368 (peel_loops_completely): Use new dump format with location
11369 for main dumpfile message, and invoke report_unroll_peel on success.
11370 (decide_unrolling_and_peeling): Ditto.
11371 (decide_peel_once_rolling): Remove old dumpfile message subsumed
11372 by report_unroll_peel.
11373 (decide_peel_completely): Ditto.
11374 (decide_unroll_constant_iterations): Ditto.
11375 (decide_unroll_runtime_iterations): Ditto.
11376 (decide_peel_simple): Ditto.
11377 (decide_unroll_stupid): Ditto.
11378 * cfgloop.c (get_loop_location): New function.
11379 * cfgloop.h (get_loop_location): Declare.
11381 2013-01-02 Sriraman Tallam <tmsriram@google.com>
11383 * config/i386/i386.c (fold_builtin_cpu): Remove unnecessary checks for
11386 2013-01-02 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
11388 PR middle-end/55198
11389 * expr.c (expand_expr_real_1): Don't use bitfield extraction for non
11390 BLKmode objects when EXPAND_MEMORY is specified.
11392 2013-01-02 Sriraman Tallam <tmsriram@google.com>
11394 * config/i386/i386.c (ix86_get_function_versions_dispatcher): Fix bug
11396 (fold_builtin_cpu): Do not share cpu model decls across statements.
11398 2013-01-02 Jason Merrill <jason@redhat.com>
11401 * tree.c (build_array_type_1): Revert earlier change.
11403 2013-01-02 Yufeng Zhang <yufeng.zhang@arm.com>
11405 * config/aarch64/aarch64-cores.def: Add entries for "cortex-a53" and
11407 * config/aarch64/aarch64-tune.md: Re-generate.
11409 2013-01-02 Richard Biener <rguenther@suse.de>
11411 * tree-vect-stmts.c (vectorizable_load): When vectorizing an
11412 invariant load do not generate a vector load from the scalar location.
11414 2013-01-02 Richard Biener <rguenther@suse.de>
11417 * configure.ac: Add $GMPINC to CFLAGS/CXXFLAGS.
11418 * configure: Regenerate.
11420 2013-01-02 Richard Sandiford <rdsandiford@googlemail.com>
11422 * builtins.c (expand_builtin_mathfn, expand_builtin_mathfn_2)
11423 (expand_builtin_mathfn_ternary, expand_builtin_mathfn_3)
11424 (expand_builtin_int_roundingfn_2): Keep the original target around
11425 for the fallback case.
11427 2013-01-02 Richard Sandiford <rdsandiford@googlemail.com>
11429 * tree-vrp.c (range_fits_type_p): Require the MSB of the double_int
11430 to be clear for sign changes.
11432 2013-01-01 Jan Hubicka <jh@suse.cz>
11434 * ipa-inline-analysis.c: Fix formatting.
11436 2013-01-01 Jakub Jelinek <jakub@redhat.com>
11438 PR tree-optimization/55831
11439 * tree-vect-loop.c (get_initial_def_for_induction): Use
11440 gsi_after_labels instead of gsi_start_bb.
11442 Copyright (C) 2013 Free Software Foundation, Inc.
11444 Copying and distribution of this file, with or without modification,
11445 are permitted in any medium without royalty provided the copyright
11446 notice and this notice are preserved.