gcc.git
8 years agore PR target/69894 (dependency of gcc-plugin.h not installed on aarch64-linux-gnu)
Jakub Jelinek [Mon, 22 Feb 2016 19:55:47 +0000 (20:55 +0100)]
re PR target/69894 (dependency of gcc-plugin.h not installed on aarch64-linux-gnu)

PR target/69894
PR target/69895
* config/m68k/t-opts (OPTIONS_H_EXTRA): Add m68k-microarchs.def
and m68k-devices.def.
* config/c6x/t-c6x (OPTIONS_H_EXTRA): Add c6x-isas.def.
* config/aarch64/t-aarch64 (OPTIONS_H_EXTRA): Add aarch64-arches.def.

From-SVN: r233611

8 years agoMAINTAINERS: add myself to libcpp and diagnostic messages
David Malcolm [Mon, 22 Feb 2016 17:35:45 +0000 (17:35 +0000)]
MAINTAINERS: add myself to libcpp and diagnostic messages

2016-02-22  David Malcolm  <dmalcolm@redhat.com>

* MAINTAINERS (libcpp): Add myself.
(diagnostic messages): Likewise.

From-SVN: r233608

8 years agonvptx.c (nvptx_gen_shuffle): Add support for QImode and HImode registers.
Cesar Philippidis [Mon, 22 Feb 2016 16:28:25 +0000 (08:28 -0800)]
nvptx.c (nvptx_gen_shuffle): Add support for QImode and HImode registers.

gcc/
* config/nvptx/nvptx.c (nvptx_gen_shuffle): Add support for QImode
and HImode registers.

libgomp/
* testsuite/libgomp.oacc-c-c++-common/vprop.c: New test.

From-SVN: r233607

8 years agore PR tree-optimization/69882 (Excessive reduction statements generated by SLP)
Richard Biener [Mon, 22 Feb 2016 14:53:17 +0000 (14:53 +0000)]
re PR tree-optimization/69882 (Excessive reduction statements generated by SLP)

2016-02-22  Richard Biener  <rguenther@suse.de>

PR tree-optimization/69882
* tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Properly
preserve permutations present because of gaps.
(vect_supported_load_permutation_p): Always continue checking
permutations after vect_attempt_slp_rearrange_stmts.

* gfortran.dg/vect/pr69882.f90: New testcase.

From-SVN: r233605

8 years agotree-vect-loop.c (vect_estimate_min_profitable_iters): Dump min_profitable_estimate...
Bin Cheng [Mon, 22 Feb 2016 14:38:10 +0000 (14:38 +0000)]
tree-vect-loop.c (vect_estimate_min_profitable_iters): Dump min_profitable_estimate, rather than min_profitable_iters.

* tree-vect-loop.c (vect_estimate_min_profitable_iters): Dump
min_profitable_estimate, rather than min_profitable_iters.

From-SVN: r233604

8 years agore PR bootstrap/69885 (ICE in maybe_legitimize_operand, at optabs.c:6903 on m68k...
Jakub Jelinek [Mon, 22 Feb 2016 14:22:07 +0000 (15:22 +0100)]
re PR bootstrap/69885 (ICE in maybe_legitimize_operand, at optabs.c:6903 on m68k-linux-gnu)

PR target/69885
* config/m68k/m68k.md (ashldi3, ashrdi3, lshrdi3): Use
SImode for last match_operand.

* gcc.dg/pr69885.c: New test.

From-SVN: r233603

8 years agoHSA: fix emission of clrsb{l,ll} builtins
Martin Liska [Mon, 22 Feb 2016 14:14:51 +0000 (15:14 +0100)]
HSA: fix emission of clrsb{l,ll} builtins

* hsa-gen.c (gen_hsa_clrsb): In case of zero value,
return bitsize - 1 as the return value.

From-SVN: r233602

8 years agore PR rtl-optimization/69806 ([SH] Combine doesn't see constant)
Oleg Endo [Mon, 22 Feb 2016 13:33:31 +0000 (13:33 +0000)]
re PR rtl-optimization/69806 ([SH] Combine doesn't see constant)

gcc/
PR target/69806
PR target/54089
* config/sh/sh.c (sh_lshrsi_clobbers_t_reg_p, sh_dynamicalize_shift_p):
Handle negative shift counts.
* config/sh/sh.md (ashlsi3, lshrsi3_n, lshrsi3_n_clobbers_t): Don't use
force_reg on the shift constant.
(lshrsi3): Likewise.  Expand into lshrsi3_n* instead of lshrsi3_d.
(lshrsi3_d): Handle negative shift counts.

gcc/testsuite/
PR target/69806
PR target/54089
* gcc.target/sh/pr54089-10.c: New.

From-SVN: r233601

8 years agoAdd dot-fn to gdbhooks.py
Richard Biener [Mon, 22 Feb 2016 10:31:41 +0000 (10:31 +0000)]
Add dot-fn to gdbhooks.py

2016-02-22  Richard Biener  <rguenther@suse.de>
    Tom de Vries  <tom@codesourcery.com>

* graph.c: Include dumpfile.h.
(print_graph_cfg): Split into three overloads.
* gdbhooks.py (class DotFn): Add and instantiate, adding command dot-fn.

Co-Authored-By: Tom de Vries <tom@codesourcery.com>
From-SVN: r233600

8 years agoAdd dump-fn to gdbhooks.py
Tom de Vries [Mon, 22 Feb 2016 10:31:32 +0000 (10:31 +0000)]
Add dump-fn to gdbhooks.py

2016-02-22  Tom de Vries  <tom@codesourcery.com>

* gdbhooks.py (class DumpFn): Add and instantiate, adding command
dump-fn.

From-SVN: r233599

8 years agore PR middle-end/37448 (cannot compile big function)
Richard Biener [Mon, 22 Feb 2016 09:32:35 +0000 (09:32 +0000)]
re PR middle-end/37448 (cannot compile big function)

2016-02-22  Richard Biener  <rguenther@suse.de>

PR ipa/37448
* ipa-inline-transform.c (inline_call): When not updating
overall summaries adjust self size by the growth estimate.
* ipa-inline.c (inline_to_all_callers_1): Add to the callers
hash-set, do not update overall summaries here.  Renamed from ...
(inline_to_all_callers): ... this which is now wrapping the
above and performing delayed overall summary update.
(early_inline_small_functions): Delay updating of the overall
summary.

From-SVN: r233598

8 years agoDaily bump.
GCC Administrator [Mon, 22 Feb 2016 00:16:13 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r233597

8 years agoProperly initialize variable in tree-chkp.c (chkp_mark_invalid_bounds_walker)
Markus Trippelsdorf [Sun, 21 Feb 2016 11:25:31 +0000 (11:25 +0000)]
Properly initialize variable in tree-chkp.c (chkp_mark_invalid_bounds_walker)

* tree-chkp.c (chkp_mark_invalid_bounds_walker): Initialize
variable.

From-SVN: r233593

8 years agoDaily bump.
GCC Administrator [Sun, 21 Feb 2016 00:16:17 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r233592

8 years agore PR fortran/69423 (Invalid optimization with deferred-length character)
Paul Thomas [Sat, 20 Feb 2016 18:26:59 +0000 (18:26 +0000)]
re PR fortran/69423 (Invalid optimization with deferred-length character)

2016-02-20  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/69423
* trans-decl.c (create_function_arglist): Deferred character
length functions, with and without declared results, address
the passed reference type as '.result' and the local string
length as '..result'.
(gfc_null_and_pass_deferred_len): Helper function to null and
return deferred string lengths, as needed.
(gfc_trans_deferred_vars): Call it, thereby reducing repeated
code, add call for deferred arrays and reroute pointer function
results. Avoid using 'tmp' for anything other that a temporary
tree by introducing 'type_of_array' for the arrayspec type.

2016-02-20  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/69423
* gfortran.dg/deferred_character_15.f90 : New test.

From-SVN: r233589

8 years ago[multiple changes]
Dominique d'Humieres [Sat, 20 Feb 2016 14:10:55 +0000 (15:10 +0100)]
[multiple changes]

2016-02-20  Dominique d'Humieres  <dominiq@lps.ens.fr>

PR fortran/57365
gfortran.dg/allocate_with_source_18.f03: New test.

2016-02-20  Harald Anlauf  <anlauf@gmx.de>

PR fortran/52531
gfortran.dg/gomp/pr52531.f90: New test.

From-SVN: r233588

8 years agoFix whitespace in libgomp/ChangeLog
Tom de Vries [Sat, 20 Feb 2016 11:42:54 +0000 (11:42 +0000)]
Fix whitespace in libgomp/ChangeLog

From-SVN: r233587

8 years agoFix whitespace in gcc/testsuite/ChangeLog
Tom de Vries [Sat, 20 Feb 2016 11:42:46 +0000 (11:42 +0000)]
Fix whitespace in gcc/testsuite/ChangeLog

From-SVN: r233586

8 years agoFix whitespace in gcc/ChangeLog
Tom de Vries [Sat, 20 Feb 2016 11:42:35 +0000 (11:42 +0000)]
Fix whitespace in gcc/ChangeLog

From-SVN: r233585

8 years agoFix whitespace in ChangeLog
Tom de Vries [Sat, 20 Feb 2016 11:42:25 +0000 (11:42 +0000)]
Fix whitespace in ChangeLog

From-SVN: r233584

8 years agoRemove stray log entry in gcc/ChangeLog-2015
Tom de Vries [Sat, 20 Feb 2016 11:42:16 +0000 (11:42 +0000)]
Remove stray log entry in gcc/ChangeLog-2015

From-SVN: r233583

8 years agoFix whitespace in MAINTAINERS
Tom de Vries [Sat, 20 Feb 2016 11:27:08 +0000 (11:27 +0000)]
Fix whitespace in MAINTAINERS

2016-02-20  Tom de Vries  <tom@codesourcery.com>

* MAINTAINERS: Fix whitespace.

From-SVN: r233582

8 years agore PR c++/69743 (function overload with variadic arguments - template instantiation...
Jason Merrill [Sat, 20 Feb 2016 04:31:16 +0000 (23:31 -0500)]
re PR c++/69743 (function overload with variadic arguments - template instantiation depth exceeds maximum (gcc4, clang - no problem))

PR c++/69743
* call.c (remaining_arguments): No longer static.
* cp-tree.h: Declare it.
* pt.c (more_specialized_fn): Use it.

From-SVN: r233579

8 years agoDaily bump.
GCC Administrator [Sat, 20 Feb 2016 00:16:15 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r233578

8 years agore PR c++/69865 (-trigraphs option broken)
Bernd Edlinger [Fri, 19 Feb 2016 22:22:04 +0000 (22:22 +0000)]
re PR c++/69865 (-trigraphs option broken)

gcc/c-family/ChangeLog:
2016-02-19  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR c++/69865
        * c-opts.c (c_common_post_options): Move call to set_std_cxx14 from
        here...
        (c_common_init_options): ...to here.
        (set_std_cxx98): Initialize flag_isoc94 and flag_isoc99.

gcc/testsuite/ChangeLog:
2016-02-19  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR c++/69865
        * c-c++-common/Wshift-negative-value-6.c: Add -std=c++11 before
        -std=c++03.

From-SVN: r233574

8 years agore PR driver/69805 (ICE in greater_than_spec_func, at gcc.c:9722)
Jakub Jelinek [Fri, 19 Feb 2016 22:18:38 +0000 (23:18 +0100)]
re PR driver/69805 (ICE in greater_than_spec_func, at gcc.c:9722)

PR driver/69805
* gcc.c (LINK_COMMAND_SPEC, GOMP_SELF_SPECS): Use
:%* in %:gt() argument.
(greater_than_spec_func): Adjust for expecting only numbers,
if there are more than two numbers, compare the last two.

* testsuite/libgomp.c/pr69805.c: New test.

From-SVN: r233573

8 years agoMake-lang.in: Invoke gperf with -L C++.
Jakub Jelinek [Fri, 19 Feb 2016 22:13:29 +0000 (23:13 +0100)]
Make-lang.in: Invoke gperf with -L C++.

2016-02-19  Jakub Jelinek  <jakub@redhat.com>
            Bernd Edlinger  <bernd.edlinger@hotmail.de>

        * Make-lang.in: Invoke gperf with -L C++.
        * cfns.gperf: Remove prototypes for hash and libc_name_p
        inlines.
        * cfns.h: Regenerated.
        * except.c (nothrow_libfn_p): Adjust.

Co-Authored-By: Bernd Edlinger <bernd.edlinger@hotmail.de>
From-SVN: r233572

8 years agore PR c++/69826 (problem with cilkplus pragma and preprocessor variable)
Jakub Jelinek [Fri, 19 Feb 2016 22:12:54 +0000 (23:12 +0100)]
re PR c++/69826 (problem with cilkplus pragma and preprocessor variable)

PR c++/69826
* c-pragma.c (c_pp_lookup_pragma): Handle PRAGMA_CILK_GRAINSIZE.
(init_pragma): Register PRAGMA_CILK_GRAINSIZE even for
flag_preprocess_only.

* c-c++-common/cilk-plus/CK/pr69826-1.c: New test.
* c-c++-common/cilk-plus/CK/pr69826-2.c: New test.

From-SVN: r233571

8 years agoinvoke.texi (C++ Dialect Options): Clarify interaction of -Wnarrowing with -std.
Jonathan Wakely [Fri, 19 Feb 2016 19:32:06 +0000 (19:32 +0000)]
invoke.texi (C++ Dialect Options): Clarify interaction of -Wnarrowing with -std.

* doc/invoke.texi (C++ Dialect Options): Clarify interaction of
-Wnarrowing with -std.

From-SVN: r233569

8 years agore PR c++/69850 (unnecessary -Wnonnull-compare warning)
Jakub Jelinek [Fri, 19 Feb 2016 19:16:31 +0000 (20:16 +0100)]
re PR c++/69850 (unnecessary -Wnonnull-compare warning)

PR c++/69850
* rtti.c (ifnonnull): Set TREE_NO_WARNING on the condition, use
NE_EXPR instead of EQ_EXPR and swap last two arguments on COND_EXPR.

* g++.dg/warn/Wnonnull-compare-4.C: New test.

From-SVN: r233568

8 years agoTweak comment further.
Jakub Jelinek [Fri, 19 Feb 2016 19:13:20 +0000 (20:13 +0100)]
Tweak comment further.

From-SVN: r233567

8 years agore PR c++/69851 (ICE: in assign_temp, at function.c:961)
Jakub Jelinek [Fri, 19 Feb 2016 19:11:58 +0000 (20:11 +0100)]
re PR c++/69851 (ICE:  in assign_temp, at function.c:961)

PR c++/69851
* expr.c (store_field): Don't use bit-field path if exp is
COMPONENT_REF with TREE_ADDRESSABLE type, where TYPE_SIZE is
different from bitsize, but DECL_SIZE of FIELD_DECL is bitsize
and the assignment can be performed by bitwise copy.  Formatting
fix.

* g++.dg/torture/pr69851.C: New test.

From-SVN: r233566

8 years agoPR testsuite/69573 - FAIL: gcc.dg/pr61053.c (test for excess errors)
Martin Sebor [Fri, 19 Feb 2016 18:37:40 +0000 (18:37 +0000)]
PR testsuite/69573 - FAIL: gcc.dg/pr61053.c (test for excess errors)

Tested on x86_64 by running make check-c with
RUNTESTFLAGS='--target_board=unix/-m32/-m128bit-long-double dg.exp=pr61053.c'
gcc/testsuite/ChangeLog:
2016-01-30  Martin Sebor  <msebor@redhat.com>

PR testsuite/69573
* gcc.dg/pr61053.c: Tweaked test to avoid bogus failures.

From-SVN: r233564

8 years agoAdjust fix PR c++/68948
Patrick Palka [Fri, 19 Feb 2016 17:04:29 +0000 (17:04 +0000)]
Adjust fix PR c++/68948

gcc/cp/ChangeLog:

PR c++/68948
* pt.c (tsubst_baselink): Don't diagnose an invalid constructor
call here.
* semantics.c (finish_call_expr): Don't assume a constructor
call is dependent if only the "this" pointer is dependent.  When
building a constructor call, always use a dummy object.

From-SVN: r233563

8 years agore PR middle-end/69838 (Lra deletes EH_REGION)
Jakub Jelinek [Fri, 19 Feb 2016 16:25:58 +0000 (17:25 +0100)]
re PR middle-end/69838 (Lra deletes EH_REGION)

PR middle-end/69838
* lra.c (lra_process_new_insns): If non-call exceptions are enabled,
call copy_reg_eh_region_note_forward on before and/or after sequences
and remove note from insn if it no longer can throw.

From-SVN: r233562

8 years agore PR c++/69850 (unnecessary -Wnonnull-compare warning)
Jakub Jelinek [Fri, 19 Feb 2016 16:02:51 +0000 (17:02 +0100)]
re PR c++/69850 (unnecessary -Wnonnull-compare warning)

PR c++/69850
* init.c (build_vec_delete_1): Set TREE_NO_WARNING on the NE_EXPR
condition.
* cp-gimplify.c (cp_fold): Propagate TREE_NO_WARNING from binary
operators if folding preserved the binop, just with different
arguments.

* g++.dg/warn/Wnonnull-compare-2.C: New test.
* g++.dg/warn/Wnonnull-compare-3.C: New test.

From-SVN: r233561

8 years agore PR c++/67767 (-Wsuggest-attribute=noreturn suggests noreturn for function which...
Jakub Jelinek [Fri, 19 Feb 2016 16:01:50 +0000 (17:01 +0100)]
re PR c++/67767 (-Wsuggest-attribute=noreturn suggests noreturn for function which already has noreturn and cold.)

PR c++/67767
* parser.c (cp_parser_std_attribute_spec_seq): Don't assume
attr_spec is always single element chain, chain all the attributes
properly together in the right order.

* g++.dg/cpp0x/pr67767.C: New test.

From-SVN: r233560

8 years ago[AArch64] Skip gcc.target/aarch64/assembler_arch_1.c if assembler does not support it
Kyrylo Tkachov [Fri, 19 Feb 2016 14:20:27 +0000 (14:20 +0000)]
[AArch64] Skip gcc.target/aarch64/assembler_arch_1.c if assembler does not support it

* lib/target-supports.exp: Define aarch64_asm_FUNC_ok checks
for fp, simd, crypto, crc, lse.
* doc/sourcebuild.texi (AArch64-specific attributes): Document the
above.
* gcc.target/aarch64/assembler_arch_1.c: Add aarch64_asm_lse_ok
effective target check.

From-SVN: r233559

8 years agore PR tree-optimization/69820 (Test miscompiled with -O3 option)
Jakub Jelinek [Fri, 19 Feb 2016 13:42:38 +0000 (14:42 +0100)]
re PR tree-optimization/69820 (Test miscompiled with -O3 option)

PR target/69820
* config/i386/sse.md (VI_512): Only include V64QImode and V32HImode
if TARGET_AVX512BW.

* gcc.target/i386/pr69820.c: New test.

From-SVN: r233558

8 years agoS/390: z13 Add missing commutative operand markers.
Andreas Krebbel [Fri, 19 Feb 2016 10:39:15 +0000 (10:39 +0000)]
S/390: z13 Add missing commutative operand markers.

gcc/ChangeLog:

* config/s390/vector.md: Add missing commutative operand markers
to the patterns which qualify for one.
* config/s390/vx-builtins.md: Likewise.

From-SVN: r233556

8 years agoS/390: Add single element vector types to iterators.
Andreas Krebbel [Fri, 19 Feb 2016 10:38:22 +0000 (10:38 +0000)]
S/390: Add single element vector types to iterators.

gcc/ChangeLog:

* config/s390/vector.md (VI, VI_QHS): Add single element vector
types to mode iterators.
(vec_double): ... and mode attribute.
* config/s390/vx-builtins.md (non_vec_int): Likewise.

From-SVN: r233555

8 years agoS/390: z13 Change predicates of 128 bit add sub.
Andreas Krebbel [Fri, 19 Feb 2016 10:37:19 +0000 (10:37 +0000)]
S/390: z13 Change predicates of 128 bit add sub.

So far usage of 128 bit add/sub instruction was rejected if the second
operand was a constant because the predicate rejected this.

gcc/testsuite/ChangeLog:

* gcc.target/s390/vector/int128-1.c: New test.

gcc/ChangeLog:

* config/s390/vector.md ("<ti*>add<mode>3", "<ti*>sub<mode>3"):
Change the predicate of op2 from nonimmediate to general and let
reload fix it if necessary.

From-SVN: r233554

8 years agoS/390: Add vec_sub_u128 to vecintrin.h
Andreas Krebbel [Fri, 19 Feb 2016 10:35:48 +0000 (10:35 +0000)]
S/390: Add vec_sub_u128 to vecintrin.h

This adds a missing macro to the vecintrin.h header file.

gcc/ChangeLog:

* config/s390/vecintrin.h (vec_sub_u128): Define missing macro.

From-SVN: r233553

8 years agoS/390: z13 fix mode in vcond expansion
Andreas Krebbel [Fri, 19 Feb 2016 10:34:12 +0000 (10:34 +0000)]
S/390: z13 fix mode in vcond expansion

For floating point vector compares the target mode is an integer mode
which accidently was used as register mode when forcing the compare
operands into regs.

gcc/ChangeLog:

* config/s390/s390.c (s390_expand_vcond): Use the compare operand
mode.

gcc/testsuite/ChangeLog:

* gcc.target/s390/vector/vec-vcond-1.c: New test.

From-SVN: r233552

8 years agoS/390: Adjust movstr-1.c testcase to work with the z13 stpcpy implementation.
Andreas Krebbel [Fri, 19 Feb 2016 10:32:38 +0000 (10:32 +0000)]
S/390: Adjust movstr-1.c testcase to work with the z13 stpcpy implementation.

2016-02-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>

* gcc.target/s390/md/movstr-1.c: Allow also the z13 strings
instruction pattern name to prevent the testcase from failing with
-march=z13.

From-SVN: r233551

8 years agoS/390: z13 inline stpcpy implementation.
Andreas Krebbel [Fri, 19 Feb 2016 10:31:05 +0000 (10:31 +0000)]
S/390: z13 inline stpcpy implementation.

A handwritten loop for stpcpy using the new z13 vector instructions
appears to be much faster than the millicoded instruction.  However,
the implementation is much longer and therefore will only be enabled
when optimization for speed.

gcc/testsuite/ChangeLog:

* gcc.target/s390/md/movstr-2.c: New test.

gcc/ChangeLog:

* config/s390/s390-protos.h: Add s390_expand_vec_movstr prototype.
* config/s390/s390.c (s390_expand_vec_movstr): New function.
* config/s390/s390.md ("movstr<P:mode>"): Call
s390_expand_vec_movstr.

From-SVN: r233550

8 years agoS/390: z13 lcbb fix address operand.
Andreas Krebbel [Fri, 19 Feb 2016 10:28:45 +0000 (10:28 +0000)]
S/390: z13 lcbb fix address operand.

gcc/ChangeLog:

2016-02-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>

* config/s390/s390.md: Add missing output modifier for operand 1
to print it as address properly.

From-SVN: r233549

8 years agoS/390: Add IBM z13 pipeline description
Andreas Krebbel [Fri, 19 Feb 2016 10:27:12 +0000 (10:27 +0000)]
S/390: Add IBM z13 pipeline description

This patch adds proper support for the -mtune=z13 option by adding a
z13 pipeline description.  As started with zEC12 we mostly make use of
the sched reorder hooks to implement a grouping strategy.  However,
this time we also keep an eye at the instruction mix provided in the
out of order window to allow the hardware to exploit the different
units.

gcc/ChangeLog:

* config/s390/2827.md: Rename ooo_* insn attributes to zEC12_*.
* config/s390/2964.md: New file.
* config/s390/s390.c (s390_get_sched_attrmask): Use the right set
of insn grouping attributes depending on the CPU level.
(s390_get_unit_mask): New function.
(s390_sched_score): Remove the OOO from the scheduling macros.
Add loop to calculate a score for the instruction mix.
(s390_sched_reorder): Likewise plus improve debug output.
(s390_sched_variable_issue): Rename macros as above.  Calculate
the unit distances after actually scheduling an insn.  Improve
debug output.
(s390_sched_init): Clear last_scheduled_unit_distance array.
* config/s390/s390.md: Include 2964.md.

From-SVN: r233548

8 years agoS/390: Add __morestack test.
Marcin Kościelnicki [Fri, 19 Feb 2016 10:20:23 +0000 (10:20 +0000)]
S/390: Add __morestack test.

gcc/testsuite/ChangeLog:

* gcc.target/s390/morestack.c: New test.

From-SVN: r233547

8 years agoS/390: PR 69625: Add test case
Dominik Vogt [Fri, 19 Feb 2016 10:15:43 +0000 (10:15 +0000)]
S/390: PR 69625: Add test case

gcc/testsuite/ChangeLog

PR target/69625
* gcc.target/s390/pr69625.c: Add test case.

From-SVN: r233546

8 years agore PR target/69671 (FAIL: gcc.target/i386/avx512vl-vpmovqb-1.c scan-assembler-times...
Jakub Jelinek [Fri, 19 Feb 2016 07:56:36 +0000 (08:56 +0100)]
re PR target/69671 (FAIL: gcc.target/i386/avx512vl-vpmovqb-1.c scan-assembler-times vpmovqb[ \\t]+[^{\n]*%ymm[0-9]+[^\n]*%xmm[0-9]+{%k[1-7]}{z}(?)

PR target/69671
* config/i386/sse.md (*<floatsuffix>floatv2div2sf2_mask_1,
*avx512vl_<code>v2div2qi2_mask_1, *avx512vl_<code><mode>v4qi2_mask_1,
*avx512vl_<code><mode>v8qi2_mask_1, *avx512vl_<code><mode>v4hi2_mask_1,
*avx512vl_<code>v2div2hi2_mask_1, *avx512vl_<code>v2div2si2_mask_1,
*avx512f_<code>v8div16qi2_mask_1): New insns.

From-SVN: r233545

8 years agomangle.c (maybe_check_abi_tags): Add for_decl parm.
Jason Merrill [Fri, 19 Feb 2016 06:25:05 +0000 (01:25 -0500)]
mangle.c (maybe_check_abi_tags): Add for_decl parm.

* mangle.c (maybe_check_abi_tags): Add for_decl parm.  Call
mangle_decl.
(mangle_decl): Call maybe_check_abi_tags for function scope.
(mangle_guard_variable): Call maybe_check_abi_tags here.
(write_guarded_var_name): Not here.

From-SVN: r233544

8 years agoDaily bump.
GCC Administrator [Fri, 19 Feb 2016 00:16:16 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r233543

8 years agore PR testsuite/68580 (FAIL: c-c++-common/tsan/pr65400-1.c -O0 execution test)
Bernd Edlinger [Thu, 18 Feb 2016 22:27:09 +0000 (22:27 +0000)]
re PR testsuite/68580 (FAIL: c-c++-common/tsan/pr65400-1.c -O0 execution test)

PR testsuite/68580
* c-c++-common/tsan/pr65400-1.c (v, q, o): Make 8-byte aligned.

From-SVN: r233538

8 years agore PR bootstrap/68404 (PGO/LTO bootstrap failure on ppc64le)
Michael Meissner [Thu, 18 Feb 2016 19:36:39 +0000 (19:36 +0000)]
re PR bootstrap/68404 (PGO/LTO bootstrap failure on ppc64le)

[gcc]
2016-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/68404
* config/rs6000/predicates.md (fusion_gpr_addis): Revert
2016-02-09 change.

* config/rs6000/rs6000.md (fusion_gpr_load_<mode>): Remove
earlyclobber from target.  Use wF constraint for fused memory
address.
(fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.

[gcc/testsuites]
2016-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/68404
* gcc.target/powerpc/fusion.c: Revert the 2016-02-09 change.
* gcc.target/powerpc/fusion3.c: Likewise.

From-SVN: r233532

8 years agoDo not emit red stack zones for a fn with no_sanitize_address
Jakub Jelinek [Thu, 18 Feb 2016 17:15:25 +0000 (18:15 +0100)]
Do not emit red stack zones for a fn with no_sanitize_address

PR sanitizer/69863
* cfgexpand.c (asan_sanitize_stack_p): New function.
(partition_stack_vars): Use the function.
(expand_stack_vars): Likewise.
(defer_stack_allocation): Likewise.
(expand_used_vars): Likewise.

Co-Authored-By: Martin Liska <mliska@suse.cz>
From-SVN: r233524

8 years agore PR middle-end/69553 (Optimizations O1/O2 makes std::array value incorrect when...
Richard Biener [Thu, 18 Feb 2016 14:34:59 +0000 (14:34 +0000)]
re PR middle-end/69553 (Optimizations O1/O2 makes std::array value incorrect when passed to function)

2016-02-18  Richard Biener  <rguenther@suse.de>

PR middle-end/69553
* fold-const.c (operand_equal_p): Properly compare offsets for
IMAGPART_EXPR and ARRAY_REF.

* g++.dg/torture/pr69553.C: New testcase.

From-SVN: r233520

8 years agoFix PR number in changelog entries.
Nick Clifton [Thu, 18 Feb 2016 13:23:06 +0000 (13:23 +0000)]
Fix PR number in changelog entries.

From-SVN: r233519

8 years agoPR target/62554
Nick Clifton [Thu, 18 Feb 2016 13:00:07 +0000 (13:00 +0000)]
PR target/62554
        PR target/69610
gcc     * config/arm/arm.c (arm_option_override_internal): Disable
        interworking if the target does not support thumb instructions.
        (arm_reload_in_hi): Handle the case where a register to register
        move needs reloading because there is no simple pattern to handle
        it.
        (arm_reload_out_hi): Likewise.

tests   * gcc.target/arm/pr62554.c: New test.
        * gcc.target/arm/pr69610-1.c: New test.
        * gcc.target/arm/pr69610-2.c: New test.

From-SVN: r233518

8 years agore PR middle-end/69854 (ICE: tree check: expected class 'constant', have 'binary...
Richard Biener [Thu, 18 Feb 2016 08:43:58 +0000 (08:43 +0000)]
re PR middle-end/69854 (ICE: tree check: expected class 'constant', have 'binary' (plus_expr) in generic_simplify_65, at generic-match.c:3110)

2016-02-18  Richard Biener  <rguenther@suse.de>

PR middle-end/69854
* match.pd: Don't use fold_binary or fold_unary for folding
constants.

* gcc.dg/torture/pr69854.c: New testcase.

From-SVN: r233516

8 years agolibgo: Update to final Go 1.6 release.
Ian Lance Taylor [Thu, 18 Feb 2016 05:56:46 +0000 (05:56 +0000)]
libgo: Update to final Go 1.6 release.

    Reviewed-on: https://go-review.googlesource.com/19592

From-SVN: r233515

8 years agore PR c++/65985 (compiler segfault with assert() in constexpr constructor body)
Jason Merrill [Thu, 18 Feb 2016 05:08:09 +0000 (00:08 -0500)]
re PR c++/65985 (compiler segfault with assert() in constexpr constructor body)

PR c++/65985
* constexpr.c (build_constexpr_constructor_member_initializers):
Handle an additional STATEMENT_LIST.

From-SVN: r233514

8 years agore PR c++/68585 (c++14 code accepted by 4.9 not accepted by 5 and 6)
Jason Merrill [Thu, 18 Feb 2016 05:08:02 +0000 (00:08 -0500)]
re PR c++/68585 (c++14 code accepted by 4.9 not accepted by 5 and 6)

PR c++/68585
* constexpr.c (cxx_eval_bare_aggregate): Fix 'changed' detection.

From-SVN: r233513

8 years agore PR c++/68679 (gcc-5.2.1 ICE in C++11 anon union of structs with template fns,...
Jason Merrill [Thu, 18 Feb 2016 05:07:55 +0000 (00:07 -0500)]
re PR c++/68679 (gcc-5.2.1 ICE in C++11 anon union of structs with template fns, OK in gcc <= 4.9.3)

PR c++/68679
* decl2.c (reset_type_linkage_2): Look through member templates.

From-SVN: r233512

8 years agoDaily bump.
GCC Administrator [Thu, 18 Feb 2016 00:16:12 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r233511

8 years agore PR c++/69850 (unnecessary -Wnonnull-compare warning)
Jakub Jelinek [Wed, 17 Feb 2016 22:27:24 +0000 (23:27 +0100)]
re PR c++/69850 (unnecessary -Wnonnull-compare warning)

PR c++/69850
* gimplify.c (gimplify_cond_expr): Call gimple_set_no_warning
on the cond_stmt from TREE_NO_WARNING on COND_EXPR_COND.
* gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Don't
warn on gimple_no_warning_p statements.

* init.c (build_delete): Set TREE_NO_WARNING on ifexp.

* g++.dg/warn/Wnonnull-compare-1.C: New test.

From-SVN: r233508

8 years agore PR c++/69842 (Parameter deduction in polymorphic lambdas)
Jason Merrill [Wed, 17 Feb 2016 20:45:15 +0000 (15:45 -0500)]
re PR c++/69842 (Parameter deduction in polymorphic lambdas)

PR c++/69842
* method.c (forward_parm): Split out from...
(add_one_base_init): ...here.
* lambda.c (maybe_add_lambda_conv_op): Use it.

From-SVN: r233506

8 years agoxtensa: fix libgcc build with --text-section-literals
Max Filippov [Wed, 17 Feb 2016 20:21:48 +0000 (20:21 +0000)]
xtensa: fix libgcc build with --text-section-literals

Functions __muldf3_aux, __divdf3_aux, __mulsf3_aux and __divsf3_aux
don't start with leaf_entry, so they need explicit .literal_position,
otherwise libgcc build fails in the presence of --text-section-literals.

2016-02-17  Max Filippov  <jcmvbkbc@gmail.com>
libgcc/
* config/xtensa/ieee754-df.S (__muldf3_aux, __divdf3_aux): Add
.literal_position before the function.
* config/xtensa/ieee754-sf.S (__mulsf3_aux, __divsf3_aux):
Likewise.

From-SVN: r233505

8 years agoextend.texi (C++ Attributes): Correct description of warn_unused type attribute.
Jonathan Wakely [Wed, 17 Feb 2016 17:36:22 +0000 (17:36 +0000)]
extend.texi (C++ Attributes): Correct description of warn_unused type attribute.

* doc/extend.texi (C++ Attributes): Correct description of
warn_unused type attribute.

From-SVN: r233502

8 years agore PR libfortran/69651 ([6 Regession] Usage of unitialized pointer io/list_read.c)
Jerry DeLisle [Wed, 17 Feb 2016 16:48:57 +0000 (16:48 +0000)]
re PR libfortran/69651 ([6 Regession] Usage of unitialized pointer io/list_read.c)

2016-02-17  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR libgfortran/69651
* io/list_read.c (push_char4): Fix the pointer usage for xrealloc.

From-SVN: r233500

8 years agoaltivec.md (*altivec_lvxl_<mode>_internal): Output correct instruction.
Bill Schmidt [Wed, 17 Feb 2016 16:23:55 +0000 (16:23 +0000)]
altivec.md (*altivec_lvxl_<mode>_internal): Output correct instruction.

[gcc]

2016-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* config/rs6000/altivec.md (*altivec_lvxl_<mode>_internal): Output
correct instruction.

[gcc/testsuite]

2012-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* gcc.target/powerpc/vec-cg.c: New test.

From-SVN: r233499

8 years agore PR rtl-optimization/69609 (block reordering consumes an inordinate amount of time...
Richard Biener [Wed, 17 Feb 2016 14:57:58 +0000 (14:57 +0000)]
re PR rtl-optimization/69609 (block reordering consumes an inordinate amount of time, REE consumes much memory)

2016-02-17  Richard Biener  <rguenther@suse.de>

PR rtl-optimization/69609
* bb-reorder.c (struct bbro_basic_block_data): Add priority member.
(find_traces_1_round): When ending a trace update cached priority
of successors.
(bb_to_key): Use cached priority when available.
(copy_bb): Initialize cached priority.
(reorder_basic_blocks_software_trace_cache): Likewise.

From-SVN: r233498

8 years agore PR testsuite/69586 (FAIL: gcc.dg/uninit-21.c for target defaulting to short enum)
Richard Biener [Wed, 17 Feb 2016 14:51:27 +0000 (14:51 +0000)]
re PR testsuite/69586 (FAIL: gcc.dg/uninit-21.c for target defaulting to short enum)

2016-02-17  Richard Biener  <rguenther@suse.de>

PR testsuite/69586
* gcc.dg/uninit-21.c: Add -fno-short-enums.

From-SVN: r233497

8 years ago[AArch64] PR target/69161: Don't use special predicate for CCmode comparisons in...
Kyrylo Tkachov [Wed, 17 Feb 2016 13:42:35 +0000 (13:42 +0000)]
[AArch64] PR target/69161: Don't use special predicate for CCmode comparisons in expressions that require matching modes

PR target/69161
* config/aarch64/predicates.md (aarch64_comparison_operator_mode):
New predicate.
(aarch64_comparison_operator): Break overly long line into two.
(aarch64_comparison_operation): Likewise.
* config/aarch64/aarch64.md (cstorecc4): Use
aarch64_comparison_operator_mode instead of
aarch64_comparison_operator.
(cstore<mode>4): Likewise.
(aarch64_cstore<mode>): Likewise.
(*cstoresi_insn_uxtw): Likewise.
(cstore<mode>_neg): Likewise.
(*cstoresi_neg_uxtw): Likewise.

* gcc.c-torture/compile/pr69161.c: New test.

From-SVN: r233496

8 years ago[ARM] PR target/69161: Don't ignore mode when matching comparison operator in cstore...
Kyrylo Tkachov [Wed, 17 Feb 2016 13:39:30 +0000 (13:39 +0000)]
[ARM] PR target/69161: Don't ignore mode when matching comparison operator in cstore-like patterns

PR target/69161
* config/arm/predicates.md (arm_comparison_operator_mode):
New predicate.
* config/arm/arm.md (*mov_scc): Use arm_comparison_operator_mode
instead of arm_comparison_operator.
(*mov_negscc): Likewise.
(*mov_notscc): Likewise.
* config/arm/thumb2.md (*thumb2_mov_scc): Likewise.
(*thumb2_mov_negscc): Likewise.
(*thumb2_mov_negscc_strict_it): Likewise.
(*thumb2_mov_notscc): Likewise.
(*thumb2_mov_notscc_strict_it): Likewise.

From-SVN: r233495

8 years agore PR target/69532 (FAIL: gcc.target/arm/{vect-,}fmaxmin.c execution test on armv7)
David Sherwood [Wed, 17 Feb 2016 13:16:31 +0000 (13:16 +0000)]
re PR target/69532 (FAIL: gcc.target/arm/{vect-,}fmaxmin.c execution test on armv7)

2016-02-17  David Sherwood  <david.sherwood@arm.com>

    gcc/testsuite:
        PR target/69532
        * gcc.target/arm/fmaxmin.c: Change effective target to arm_v8_neon_hw.
        * gcc.target/arm/vect-fmaxmin.c: Likewise.

From-SVN: r233494

8 years agoAs part of the final cleanup of aarch64_internal_mov_immediate a return was accidenta...
Wilco Dijkstra [Wed, 17 Feb 2016 11:52:35 +0000 (11:52 +0000)]
As part of the final cleanup of aarch64_internal_mov_immediate a return was accidentally removed.

As part of the final cleanup of aarch64_internal_mov_immediate a return was
accidentally removed.  This causes the 2-instruction case to fallthrough
into the general case even when it found a match.  An example immediate is
0xcccccccccccccccd which was using 2 instructions in GCC5 but now requires 4.
Adding the return fixes the regressions.

2016-02-17  Wilco Dijkstra  <wdijkstr@arm.com>

    gcc/
* config/aarch64/aarch64.c (aarch64_internal_mov_immediate):
Add missing return.

From-SVN: r233490

8 years ago* gcc.dg/ifcvt-4.c: Skip on Visium.
Eric Botcazou [Wed, 17 Feb 2016 11:20:06 +0000 (11:20 +0000)]
* gcc.dg/ifcvt-4.c: Skip on Visium.

From-SVN: r233489

8 years agovisium.c (machine_libfunc_index): New enum.
Eric Botcazou [Wed, 17 Feb 2016 11:17:15 +0000 (11:17 +0000)]
visium.c (machine_libfunc_index): New enum.

* config/visium/visium.c (machine_libfunc_index): New enum.
(machine_libfuncs): New structure.
(visium_libfuncs): New static variable.
(TARGET_INIT_LIBFUNCS): Define to...
(visium_init_libfuncs): ...this.  New function.
(expand_block_move_4): Use the appropriate libfunc.
(expand_block_move_2): Likewise.
(expand_block_move_1): Likewise.
(expand_block_set_4): Likewise.
(expand_block_set_2): Likewise.
(expand_block_set_1): Likewise.
(visium_trampoline_init): Likewise.

From-SVN: r233488

8 years agomsp430.c (msp430_mcu_data): Sync with data from TI's devices.csv file as of March...
Nick Clifton [Wed, 17 Feb 2016 09:53:42 +0000 (09:53 +0000)]
msp430.c (msp430_mcu_data): Sync with data from TI's devices.csv file as of March 2016.

* config/msp430/msp430.c (msp430_mcu_data): Sync with data from
TI's devices.csv file as of March 2016.

From-SVN: r233487

8 years agoexp_ch4.adb (Expand_N_Indexed_Component): Active synchronization if the prefix denote...
Eric Botcazou [Wed, 17 Feb 2016 09:21:58 +0000 (09:21 +0000)]
exp_ch4.adb (Expand_N_Indexed_Component): Active synchronization if the prefix denotes an entity which Has_Atomic_Components.

* exp_ch4.adb (Expand_N_Indexed_Component): Active synchronization if
the prefix denotes an entity which Has_Atomic_Components.
* gcc-interface/trans.c (node_is_atomic): Return true if the prefix
denotes an entity which Has_Atomic_Components.

From-SVN: r233485

8 years agoutils2.c (gnat_protect_expr): Make a SAVE_EXPR only for fat pointer or scalar types.
Eric Botcazou [Wed, 17 Feb 2016 09:08:09 +0000 (09:08 +0000)]
utils2.c (gnat_protect_expr): Make a SAVE_EXPR only for fat pointer or scalar types.

* gcc-interface/utils2.c (gnat_protect_expr): Make a SAVE_EXPR only
for fat pointer or scalar types.

From-SVN: r233484

8 years agore PR libstdc++/69794 (std::regex_search match failure with regex object with flags...
Tim Shen [Wed, 17 Feb 2016 03:33:02 +0000 (03:33 +0000)]
re PR libstdc++/69794 (std::regex_search match failure with regex object with flags grep|icase)

2016-02-16  Tim Shen  <timshen@google.com>

PR libstdc++/69794
* include/bits/regex_scanner.h: Add different special character
sets for grep and egrep regex.
* include/bits/regex_scanner.tcc: Use _M_spec_char more uniformly.
* testsuite/28_regex/regression.cc: Add new testcase.

From-SVN: r233482

8 years agore PR c++/10200 (Weird clash with same names in different scopes)
Jason Merrill [Wed, 17 Feb 2016 00:50:23 +0000 (19:50 -0500)]
re PR c++/10200 (Weird clash with same names in different scopes)

PR c++/10200
PR c++/69753
* call.c, cp-tree.h, name-lookup.c, pt.c, search.c, semantics.c,
tree.c, typeck2.c: Revert earlier changes.
* parser.c (cp_parser_lookup_name): Ignore namespace-scope
non-type templates after -> or .

From-SVN: r233481

8 years agoDaily bump.
GCC Administrator [Wed, 17 Feb 2016 00:16:12 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r233480

8 years agore PR target/48344 (powerpc ICE with -fstack-limit-register=r2)
Kelvin Nilsen [Tue, 16 Feb 2016 23:12:19 +0000 (23:12 +0000)]
re PR target/48344 (powerpc ICE with -fstack-limit-register=r2)

[gcc]

2016-02-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>

PR Target/48344
* opts-global.c (handle_common_deferred_options): Introduce and
initialize two global variables to remember command-line options
specifying a stack-limiting register.
* opts.h: Add extern declarations of the two new global variables.
* emit-rtl.c (init_emit_once): Initialize the stack_limit_rtx
variable based on the values of the two new global variables.

[gcc/testsuite]

2016-02-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>

PR Target/48344
* gcc.target/powerpc/pr48344-1.c: New test.

From-SVN: r233477

8 years agore PR fortran/69742 (ICE with -O3 and ASSOCIATE containing repeated expression)
Thomas Koenig [Tue, 16 Feb 2016 21:10:00 +0000 (21:10 +0000)]
re PR fortran/69742 (ICE with -O3 and ASSOCIATE containing repeated expression)

2015-02-16  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/69742
* frontend-passes.c (cfe-expr_0):  Don't register functions
from within an ASSOCIATE statement.

2015-02-16  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/69742
* gfortran.dg/associate_21.f90:  New test.

From-SVN: r233474

8 years agore PR c/69835 (-Wnonnull diagnoses parameter comparisons with NULL even when those...
Jakub Jelinek [Tue, 16 Feb 2016 20:46:17 +0000 (21:46 +0100)]
re PR c/69835 (-Wnonnull diagnoses parameter comparisons with NULL even when those could have changed)

PR c/69835
* common.opt (Wnonnull-compare): New warning.
* doc/invoke.texi (-Wnonnull): Remove text about comparison
of arguments against NULL.
(-Wnonnull-compare): Document.
* Makefile.in (OBJS): Add gimple-ssa-nonnull-compare.o.
* tree-pass.h (make_pass_warn_nonnull_compare): Declare.
* passes.def (pass_warn_nonnull_compare): Add.
* gimple-ssa-nonnull-compare.c: New file.
c-family/
* c.opt (Wnonnull-compare): Enable for -Wall.
c/
* c-typeck.c (build_binary_op): Revert 2015-09-09 change.
cp/
* typeck.c (cp_build_binary_op): Revert 2015-09-09 change.
testsuite/
* c-c++-common/nonnull-1.c: Use -Wnonnull-compare instead of
-Wnonnull in dg-options.
* c-c++-common/nonnull-2.c: New test.

From-SVN: r233472

8 years agore PR c++/69657 (abs() not inlined after including math.h)
Jason Merrill [Tue, 16 Feb 2016 19:01:49 +0000 (14:01 -0500)]
re PR c++/69657 (abs() not inlined after including math.h)

PR c++/69657
* name-lookup.c (lookup_qualified_name): Add find_hidden parm.
(set_decl_namespace): Pass it.  Complain about finding a hidden friend.
* name-lookup.h: Adjust.

From-SVN: r233470

8 years agogigi.h (maybe_debug_type): New inline function.
Eric Botcazou [Tue, 16 Feb 2016 18:01:04 +0000 (18:01 +0000)]
gigi.h (maybe_debug_type): New inline function.

* gcc-interface/gigi.h (maybe_debug_type): New inline function.
* gcc-interface/misc.c (gnat_get_array_descr_info): Use it.
Call maybe_character_value on the array bounds.  Get to the base type
of the index type and call maybe_debug_type on it.
* gcc-interface/utils.c (finish_character_type): Add special treatment

From-SVN: r233469

8 years agomisc.c (gnat_enum_underlying_base_type): New function.
Eric Botcazou [Tue, 16 Feb 2016 17:51:56 +0000 (17:51 +0000)]
misc.c (gnat_enum_underlying_base_type): New function.

* gcc-interface/misc.c (gnat_enum_underlying_base_type): New function.
(LANG_HOOKS_ENUM_UNDERLYING_BASE_TYPE): Define to above.

From-SVN: r233467

8 years ago[AArch64] Remove AARCH64_EXTRA_TUNE_RECIP_SQRT from Cortex-A57 tuning
James Greenhalgh [Tue, 16 Feb 2016 16:04:42 +0000 (16:04 +0000)]
[AArch64] Remove AARCH64_EXTRA_TUNE_RECIP_SQRT from Cortex-A57 tuning

gcc/

* config/aarch64/aarch64.c (cortexa57_tunings): Remove
AARCH64_EXTRA_TUNE_RECIP_SQRT.

From-SVN: r233463

8 years ago[Patch AArch64] Use software sqrt expansion always for -mlow-precision-recip-sqrt
James Greenhalgh [Tue, 16 Feb 2016 16:03:27 +0000 (16:03 +0000)]
[Patch AArch64] Use software sqrt expansion always for -mlow-precision-recip-sqrt

gcc/

* config/aarch64/aarch64.c (use_rsqrt_p): Always use software
reciprocal sqrt for -mlow-precision-recip-sqrt.

From-SVN: r233462

8 years ago[Patch AArch64] GCC 6 regression in vector performance. - Fix vector initialization...
James Greenhalgh [Tue, 16 Feb 2016 16:02:09 +0000 (16:02 +0000)]
[Patch AArch64] GCC 6 regression in vector performance. - Fix vector initialization to happen with lane load instructions.

gcc/

* config/aarch64/aarch64.c (aarch64_expand_vector_init): Refactor,
always use lane loads to construct non-constant vectors.

gcc/testsuite/

* gcc.target/aarch64/vector_initialization_nostack.c: New.

Co-Authored-By: Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
From-SVN: r233461

8 years ago[Patch AArch64] Restrict 16-bit sqrdml{sa}h instructions to FP_LO_REGS
James Greenhalgh [Tue, 16 Feb 2016 15:59:51 +0000 (15:59 +0000)]
[Patch AArch64] Restrict 16-bit sqrdml{sa}h instructions to FP_LO_REGS

gcc/

* config/aarch64/aarch64.md
(arch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Fix register
constraints for operand 3.
(aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.

From-SVN: r233460

8 years agore PR tree-optimization/69801 (gcc ICE on valid code on x86_64-linux-gnu in "operand...
Jakub Jelinek [Tue, 16 Feb 2016 15:34:07 +0000 (16:34 +0100)]
re PR tree-optimization/69801 (gcc ICE  on valid code on x86_64-linux-gnu in "operand_equal_p")

PR middle-end/69801
* gcc.dg/pr69801.c: Add empty dg-options.

From-SVN: r233459

8 years agore PR c/64748 (OpenACC: "is not a variable" error with deviceptr())
James Norris [Tue, 16 Feb 2016 15:29:04 +0000 (15:29 +0000)]
re PR c/64748 (OpenACC: "is not a variable" error with deviceptr())

PR c/64748

gcc/c/
* c-parser.c (c_parser_oacc_data_clause_deviceptr): Allow parms.

gcc/cp/
* parser.c (cp_parser_oacc_data_clause_deviceptr): Remove checking.
* semantics.c (finish_omp_clauses): Add deviceptr checking.

gcc/testsuite/
* c-c++-common/goacc/deviceptr-1.c: Add tests.
* g++.dg/goacc/deviceptr-1.c: New file.

From-SVN: r233458

8 years agore PR tree-optimization/69820 (Test miscompiled with -O3 option)
Jakub Jelinek [Tue, 16 Feb 2016 15:16:48 +0000 (16:16 +0100)]
re PR tree-optimization/69820 (Test miscompiled with -O3 option)

PR tree-optimization/69820
* tree-vect-patterns.c (type_conversion_p): Return false if
*orig_type is unsigned single precision or boolean.
(vect_recog_dot_prod_pattern, vect_recog_widen_mult_pattern):
Formatting fix.

* gcc.dg/vect/pr69820.c: New test.

Co-Authored-By: Richard Biener <rguenther@suse.de>
From-SVN: r233457

8 years agore PR rtl-optimization/69764 (ICE on x86_64-linux-gnu at -O0 (in decompose, at rtl...
Jakub Jelinek [Tue, 16 Feb 2016 15:15:40 +0000 (16:15 +0100)]
re PR rtl-optimization/69764 (ICE on x86_64-linux-gnu at -O0 (in decompose, at rtl.h:2107))

PR rtl-optimization/69764
PR rtl-optimization/69771
* optabs.c (expand_binop): Ensure for shift optabs invalid CONST_INT
op1 is valid for GET_MODE_INNER (mode) and force it into a reg.

From-SVN: r233456

8 years agore PR tree-optimization/69776 (Wrong optimization with aliasing)
Richard Biener [Tue, 16 Feb 2016 15:00:45 +0000 (15:00 +0000)]
re PR tree-optimization/69776 (Wrong optimization with aliasing)

2016-02-16  Richard Biener  <rguenther@suse.de>

PR tree-optimization/69776
* tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Get alias
sets from caller.
(indirect_refs_may_alias_p): Likewise.
(refs_may_alias_p_1): Pass alias sets as from ao_ref.
* tree-ssa-sccvn.c (vn_reference_lookup): Also adjust vr alias-set
according to tbaa_p.
* tree-ssa-dom.c (lookup_avail_expr): Add tbaa_p flag.
(optimize_stmt): For redundant store discovery do not allow tbaa.

* gcc.dg/torture/pr69776-2.c: New testcase.

From-SVN: r233453