linux-unwind.h (pa32_read_access_ok): New function.

* config/pa/linux-unwind.h (pa32_read_access_ok): New function.
(pa32_fallback_frame_state): Use pa32_read_access_ok to check if
memory read accesses are ok.

From-SVN: r216716

mips.c (mips_at_reg_p): Delete.

gcc/
* config/mips/mips.c (mips_at_reg_p): Delete.
(mips_need_noat_wrapper_p): Use FOR_EACH_SUBRTX.

From-SVN: r216715

mips.c (mips_record_lo_sum): Replace with...

gcc/
* config/mips/mips.c (mips_record_lo_sum): Replace with...
(mips_record_lo_sums): ...this new function.
(mips_reorg_process_insns): Update accordingly.

From-SVN: r216714

mips.c (mips_sim_insn): Update comment.

gcc/
* config/mips/mips.c (mips_sim_insn): Update comment.
(mips_sim_wait_regs_2): Delete.
(mips_sim_wait_regs_1): Use FOR_EACH_SUBRTX_VAR.

From-SVN: r216713

mips.c (r10k_needs_protection_p_call): Take a const_rtx and return a bool.

gcc/
* config/mips/mips.c (r10k_needs_protection_p_call): Take a const_rtx
and return a bool.  Iterate over all subrtxes here.
(r10k_needs_protection_p): Update accordingly.

From-SVN: r216712

mips.c (r10k_needs_protection_p_1): Take an rtx rather than an rtx pointer.

gcc/
* config/mips/mips.c (r10k_needs_protection_p_1): Take an rtx
rather than an rtx pointer.  Change type of insn from "void *"
to its real type.  Return bool rather than int.  Iterate over
all subrtxes here.
(r10k_needs_protection_p_store): Update accordingly.
(r10k_needs_protection_p): Likewise.

From-SVN: r216711

mips.c (mips16_rewrite_pool_refs_info): Delete.

gcc/
* config/mips/mips.c (mips16_rewrite_pool_refs_info): Delete.
(mips16_rewrite_pool_refs): Take the insn and constant pool as
parameters.  Iterate over the instruction's pattern and return void.
(mips16_lay_out_constants): Update accordingly.

From-SVN: r216710

mips.c (mips_kernel_reg_p): Replace with...

gcc/
* config/mips/mips.c (mips_kernel_reg_p): Replace with...
(mips_refers_to_kernel_reg_p): ...this new function.
(mips_expand_prologue): Update accordingly.

From-SVN: r216709

mips.c (mips_rewrite_small_data_1): Take the context as a parameter instead of the containing MEM.

gcc/
* config/mips/mips.c (mips_rewrite_small_data_1): Take the context
as a parameter instead of the containing MEM.  Iterate over all
subrtxes.  Don't return a value.
(mips_rewrite_small_data): Update call accordingly.

From-SVN: r216708

mips.c: Include rtl-iter.h.

gcc/
* config/mips/mips.c: Include rtl-iter.h.
(mips_small_data_pattern_1): Take an rtx rather than an rtx pointer.
Take the context as a parameter instead of the containing MEM.
Iterate over all subrtxes.
(mips_small_data_pattern_p): Update call accordingly.

From-SVN: r216707

mep.c (mep_mul_hilo_bypass_1): Delete.

gcc/
* config/mep/mep.c (mep_mul_hilo_bypass_1): Delete.
(mep_mul_hilo_bypass_p): Use FOR_EACH_SUBRTX.

From-SVN: r216706

mep.c (mep_store_find_set): Take a const_rtx and return a bool.

gcc/
* config/mep/mep.c (mep_store_find_set): Take a const_rtx and
return a bool.  Replace "void *" with specific type.  Iterate
over all subrtxes.
(mep_store_data_bypass_1): Update calls accordingly.

From-SVN: r216705

mep.c: Include rtl-iter.h.

gcc/
* config/mep/mep.c: Include rtl-iter.h.
(global_reg_mentioned_p_1): Take a const_rtx and return a bool.
(xtensa_tls_referenced_p): Return a bool.  Use FOR_EACH_SUBRTX.

From-SVN: r216704

xtensa.c: Include rtl-iter.h.

gcc/
* config/xtensa/xtensa.c: Include rtl-iter.h.
(xtensa_tls_referenced_p_1): Delete.
(xtensa_tls_referenced_p): Use FOR_EACH_SUBRTX.

From-SVN: r216703

sh.c (sh_contains_memref_p_1): Delete.

gcc/
* config/sh/sh.c (sh_contains_memref_p_1): Delete.
(sh_contains_memref_p): Use FOR_EACH_SUBRTX.

From-SVN: r216702

sh-protos.h (shmedia_cleanup_truncate): Take an rtx as argument and return the number of changes.

gcc/
* config/sh/sh-protos.h (shmedia_cleanup_truncate): Take an
rtx as argument and return the number of changes.
* config/sh/sh.c: Include rtl-iter.h.
(shmedia_cleanup_truncate): Take an rtx as argument and iterate
over all subrtxes.  Return the number of changes made.
* config/sh/sh.md: Update caller accordingly.

From-SVN: r216701

m68k.c (m68k_tls_reference_p_1): Delete.

gcc/
* config/m68k/m68k.c (m68k_tls_reference_p_1): Delete.
(m68k_tls_reference_p): Use FOR_EACH_SUBRTX_VAR.

From-SVN: r216700

m68k.c: Include rtl-iter.h.

gcc/
* config/m68k/m68k.c: Include rtl-iter.h.
(m68k_final_prescan_insn_1): Delete.
(m68k_final_prescan_insn): Use FOR_EACH_SUBRTX_VAR.

From-SVN: r216699

Daily bump.

From-SVN: r216696

re PR tree-optimization/63641 (Invalid shift leads to incorrect code on 32-bit system)

PR tree-optimization/63641
* tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Set high
to low + prec - 1 - clz (mask) instead of low + prec - clz (mask).

* gcc.c-torture/execute/pr63641.c: New test.

From-SVN: r216693

status_cxx2011.xml: Update unordered container specific behavior.

2014-10-25  François Dumont  <fdumont@gcc.gnu.org>

* doc/xml/manual/status_cxx2011.xml: Update unordered container
specific behavior.

From-SVN: r216692

Add forgotten "configure: regenerate" stanza to last CL entry.

From-SVN: r216690

re PR rtl-optimization/63615 (FAIL: gcc.target/i386/addr-sel-1.c)

PR rtl-optimization/63615
* simplify-rtx.c (simplify_plus_minus): Set "canonicalized" on
decomposing PLUS or MINUS if operands are not placed adjacent
in the "ops" array.

From-SVN: r216689

Only allow e500 double in SPE_SIMD_REGNO_P registers.

rs6000_hard_regno_nregs_internal allows SPE vectors in single
registers satisfying SPE_SIMD_REGNO_P (i.e. register numbers 0 to
31).  However, the corresponding test for e500 double treats all
registers as being able to store a 64-bit value, rather than just
those GPRs.

Logically this inconsistency is wrong; in addition, it causes problems
unwinding from signal handlers.  linux-unwind.h uses
ARG_POINTER_REGNUM as a place to store the return address from a
signal handler, but this logic in rs6000_hard_regno_nregs_internal
results in that being considered an 8-byte register, resulting in
assertion failures.
(<https://gcc.gnu.org/ml/gcc-patches/2014-09/msg02625.html> first
needs to be applied for unwinding to work in general on e500.)  This
patch makes rs6000_hard_regno_nregs_internal handle the e500 double
case consistently with SPE vectors.

Tested with no regressions with cross to powerpc-linux-gnuspe (given
the aforementioned patch applied).  Failures of signal handling
unwinding tests such as gcc.dg/cleanup-{8,9,10,11}.c are fixed by this
patch.

* config/rs6000/rs6000.c (rs6000_hard_regno_nregs_internal): Do
not allow e500 double in registers not satisyfing
SPE_SIMD_REGNO_P.

From-SVN: r216688

Optimize powerpc*-*-linux* 32-bit classic hard/soft float hardfp/soft-fp use.

Continuing the cleanups of libgcc soft-fp configuration for
powerpc*-*-linux* in preparation for implementing
TARGET_ATOMIC_ASSIGN_EXPAND_FENV for soft-float and e500, this patch
optimizes the choice of which functions to build for the 32-bit
classic hard-float and soft-float cases.  (e500 will be dealt with in
a separate patch which will need to add new features to t-hardfp and
t-softfp; this patch keeps the status quo for e500.)

For hard-float, while the functions in question are part of the libgcc
ABI there is no need for them to contain software floating point code:
no newly built code should use them, and if anything does use them
it's most efficient (space and speed) for them to pass straight
through to floating-point hardware instructions; this case is made to
use t-hardfp to achieve that.  For soft-float, direct use of soft-fp
functions for operations involving DImode or unsigned integers is more
efficient than using the libgcc2.c versions of those operations to
convert to operations on other types (which then end up calling
soft-fp functions for those other types, possibly more than once);
this case is thus stopped from using t-softfp-excl.  (A future patch
will stop the e500 cases from using t-softfp-excl as well.)

Tested with no regressions for crosses to powerpc-linux-gnu (soft
float and classic hard float); also checked that the same set of
symbols and versions is exported from shared libgcc before and after
the patch.

* configure.ac (ppc_fp_type): Set variable on powerpc*-*-linux*.
* configure: Regenerate.
* config.host (powerpc*-*-linux*): Use $ppc_fp_type to determine
additions to tmake_file.  Use t-hardfp-sfdf and t-hardfp instead
of soft-fp for 32-bit classic hard float.  Do not use
t-softfp-excl for soft float.

From-SVN: r216687

Daily bump.

From-SVN: r216686

dwarf2out.c (declare_in_namespace): Only emit external declarations in the local scope once.

* dwarf2out.c (declare_in_namespace): Only emit external
declarations in the local scope once.

From-SVN: r216683

C++11 explicitly forbids macros for bool, true and false.

gcc:
* ginclude/stdbool.h: Do not define bool, true or false in C++11.

libstdc++-v3:
* testsuite/18_support/headers/cstdbool/macros.cc: New.

From-SVN: r216679

check.c (check_co_collective): Reject coindexed A args.

2014-10-24  Tobias Burnus  <burnus@net-b.de>

gcc/fortran
        * check.c (check_co_collective): Reject coindexed A args.
        (gfc_check_co_reduce): Add OPERATOR checks.
        * gfortran.texi (_gfortran_caf_co_broadcast,
        * _gfortran_caf_co_max,
        _gfortran_caf_co_min, _gfortran_caf_co_sum,
        _gfortran_caf_co_reduce): Add ABI documentation.
        * intrinsic.texi (CO_REDUCE): Document intrinsic.
        (DPROD): Returns double not single precision.
        * trans-decl.c (gfor_fndecl_co_reduce): New global var.
        (gfc_build_builtin_function_decls): Init it.
        * trans.h (gfor_fndecl_co_reduce): Declare it.
        * trans-intrinsic.c (conv_co_collective,
        gfc_conv_intrinsic_subroutine): Handle CO_REDUCE.

gcc/testsuite/
        * gfortran.dg/coarray_collectives_9.f90: Remove dg-error.
        * gfortran.dg/coarray_collectives_13.f90: New.
        * gfortran.dg/coarray_collectives_14.f90: New.
        * gfortran.dg/coarray_collectives_15.f90: New.
        * gfortran.dg/coarray_collectives_16.f90: New.

From-SVN: r216678

Fix error in earlier ChangeLog entry: s/define/undefine/.

From-SVN: r216676

[ARM] gnu11 cleanup for aapcs testcases

gcc/testsuite/
   * gcc.target/arm/aapcs/abitest.h: Declare memcpy.

From-SVN: r216675

re PR c/56980 (C pretty-printer does not handle well pointer to typedef of struct)

PR c/56980
* c-pretty-print.c (c_pretty_printer::simple_type_specifier): Don't
print "struct"/"union"/"enum" for typedefed names.

* gcc.dg/pr56980.c: New test.

From-SVN: r216674

fix up changelog format for r216671.

From-SVN: r216673

[PATCH 2/2] [AARCH64,NEON] Convert arm_neon.h to use new builtins for vld[234](q?)_lane_*

2014-10-24  Charles Baylis  <charles.baylis@linaro.org>

* config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Rewrite using builtins,
update uses to use new macro arguments.
(__LD3_LANE_FUNC): Likewise.
(__LD4_LANE_FUNC): Likewise.

From-SVN: r216672

[PATCH 1/2] [AARCH64,NEON] Add patterns + builtins for vld[234](q?)_lane_* intrinsics

2014-10-24  Charles Baylis  <charles.baylis@linaro.org>
* config/aarch64/aarch64-builtins.c
(aarch64_types_loadstruct_lane_qualifiers): Define.
* config/aarch64/aarch64-simd-builtins.def (ld2_lane, ld3_lane,
ld4_lane): New builtins.
* config/aarch64/aarch64-simd.md (aarch64_vec_load_lanesoi_lane<mode>):
New pattern.
(aarch64_vec_load_lanesci_lane<mode>): Likewise.
(aarch64_vec_load_lanesxi_lane<mode>): Likewise.
(aarch64_ld2_lane<mode>): New expand.
(aarch64_ld3_lane<mode>): Likewise.
(aarch64_ld4_lane<mode>): Likewise.
* config/aarch64/aarch64.md (define_c_enum "unspec"): Add
UNSPEC_LD2_LANE, UNSPEC_LD3_LANE, UNSPEC_LD4_LANE.

From-SVN: r216671

Fix bad email last commit

From-SVN: r216670

configure.ac (build_configargs): Don't share config.cache between build subdirs.

* configure.ac (build_configargs): Don't share config.cache between
build subdirs.

From-SVN: r216669

avr-protos.h (avr_out_sign_extend): New.

* avr-protos.h (avr_out_sign_extend): New.
* avr.c (avr_adjust_insn_length) [ADJUST_LEN_SEXT]: Handle.
(avr_out_sign_extend): New function.
* avr.md (extendqihi2, extendqipsi2, extendqisi2, extendhipsi2)
(extendhisi2, extendpsisi2): Use it.
(adjust_len) [sext]: New.

From-SVN: r216668

atomic_base.h: Avoid including <stdbool.h>.

2014-10-24  Paolo Carlini  <paolo.carlini@oracle.com>

* include/bits/atomic_base.h: Avoid including <stdbool.h>.
* include/std/atomic: When __cplusplus < 201103L skip the rest of
the header.
* testsuite/29_atomics/headers/atomic/std_c++0x_neg.cc: Adjust.

From-SVN: r216667

config.gcc (sparc*-*-*): Accept mcpu=leon3v7 processor.

2014-10-24  Daniel Hellstrom  <daniel@gaisler.com>

* config.gcc (sparc*-*-*): Accept mcpu=leon3v7 processor.
* doc/invoke.texi (SPARC options): Add mcpu=leon3v7 comment.
* config/sparc/leon.md (leon3_load, leon_store, leon_fp_*): Handle
leon3v7 as leon3.
* config/sparc/sparc-opts.h (enum processor_type): Add LEON3V7.
* config/sparc/sparc.c (sparc_option_override): Add leon3v7 support.
* config/sparc/sparc.h (TARGET_CPU_leon3v7): New define.
* config/sparc/sparc.md (cpu): Add leon3v7.
* config/sparc/sparc.opt (enum processor_type): Add leon3v7.

From-SVN: r216666

[ARM] revert changes on check_effective_target_arm_*_ok

  gcc/testsuite/
    * lib/target-supports.exp
    (check_effective_target_arm_crypto_ok_nocache): Remove declaration for
    vaeseq_u8.
    (check_effective_target_arm_neon_fp16_ok_nocache): Remove declaration for
    vcvt_f16_f32.
    (check_effective_target_arm_neonv2_ok_nocache): Remove declaration for
    vfma_f32.

From-SVN: r216663

re PR tree-optimization/63595 (Segmentation faults inside kernel)

Fix for PR63595

* gcc.dg/ipa/pr63595.c: New test.
* ipa-icf.c (sem_function::compare_phi_node): PHI result comparison

From-SVN: r216662

vuzp.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vuzp.c: New file.
* gcc.target/aarch64/advsimd-intrinsics/vzip.c: Likewise.

From-SVN: r216661

vmul.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vmul.c: New file.

From-SVN: r216660

vldX_lane.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vldX_lane.c: New file.

From-SVN: r216659

vldX.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vldX.c: New file.

From-SVN: r216658

vld1_dup.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vld1_dup.c: New file.

From-SVN: r216657

vdup-vmov.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vdup-vmov.c: New file.

From-SVN: r216656

vclz.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vclz.c: New file.

From-SVN: r216655

vbsl.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vbsl.c: New file.

From-SVN: r216654

vaddw.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vaddw.c: New file.

From-SVN: r216653

vaddl.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vaddl.c: New file.

From-SVN: r216652

vaddhn.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vaddhn.c: New file.

From-SVN: r216651

vabdl.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vabdl.c: New file.

From-SVN: r216650

vabd.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vabd.c: New file.

From-SVN: r216649

vabal.c: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/vabal.c: New file.

From-SVN: r216648

binary_sat_op.inc: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/binary_sat_op.inc: New
file.
* gcc.target/aarch64/advsimd-intrinsics/vqadd.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vqsub.c: Likewise.

From-SVN: r216647

unary_sat_op.inc: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/unary_sat_op.inc: New
file.
* gcc.target/aarch64/advsimd-intrinsics/vqabs.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vqneg.c: Likewise.

From-SVN: r216646

cmp_fp_op.inc: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/cmp_fp_op.inc: New file.
* gcc.target/aarch64/advsimd-intrinsics/vcage.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vcagt.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vcale.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vcalt.c: Likewise.

From-SVN: r216645

cmp_op.inc: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/cmp_op.inc: New file.
* gcc.target/aarch64/advsimd-intrinsics/vceq.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vcge.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vcgt.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vcle.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vclt.c: Likewise.

From-SVN: r216644

binary_op.inc: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/binary_op.inc: New file.
* gcc.target/aarch64/advsimd-intrinsics/vadd.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vand.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vbic.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/veor.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vorn.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vorr.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vsub.c: Likewise.

From-SVN: r216643

unary_op.inc: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/unary_op.inc: New file.
* gcc.target/aarch64/advsimd-intrinsics/vabs.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vneg.c: Likewise.

From-SVN: r216642

Add new files missing from previous commit.

From-SVN: r216641

README.advsimd-intrinsics: New file.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/arm/README.advsimd-intrinsics: New file.
* gcc.target/aarch64/advsimd-intrinsics/README: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/arm-neon-ref.h: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/compute-ref-data.h:
Likewise.
* gcc.target/aarch64/advsimd-intrinsics/advsimd-intrinsics.exp:
Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vaba.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vld1.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vshl.c: Likewise.

From-SVN: r216640

[AArch64] LINK_SPEC changes for Cortex-A53 erratum 835769 workaround

     * config/aarch64/aarch64-elf-raw.h (CA53_ERR_835769_SPEC): Define.
     (LINK_SPEC): Include CA53_ERR_835769_SPEC.
     * config/aarch64/aarch64-linux.h (CA53_ERR_835769_SPEC): Define.
     (LINK_SPEC): Include CA53_ERR_835769_SPEC.

From-SVN: r216639

wrapper.exp ({tool}_maybe_build_wrapper): Clear wrap_compile_flags before setting it.

2014-10-24  Christophe Lyon  <christophe.lyon@linaro.org>

* lib/wrapper.exp ({tool}_maybe_build_wrapper): Clear
wrap_compile_flags before setting it.

From-SVN: r216638

[AArch64] Cleanup logic around aarch64_final_prescan

* config/aarch64/aarch64.h (ADJUST_INSN_LENGTH): Wrap definition in
do while (0).
* config/aarch64/aarch64.c (is_mem_p): Delete.
(is_memory_op): Rename to...
(has_memory_op): ... This.  Use FOR_EACH_SUBRTX.
(dep_between_memop_and_curr): Assert that the input is a SET.
(aarch64_madd_needs_nop): Add comment.  Do not call
dep_between_memop_and_curr on NULL body.
(aarch64_final_prescan_insn): Add comment.
Include rtl-iter.h.

From-SVN: r216637

Makefile.in (BUILD_CPPLIB): Move $(LIBINTL) $(LIBICONV) to genmatch BUILD_LIBS instead.

2014-10-24  Richard Biener  <rguenther@suse.de>

* Makefile.in (BUILD_CPPLIB): Move $(LIBINTL) $(LIBICONV)
to genmatch BUILD_LIBS instead.

From-SVN: r216632

genmatch.c (expr::gen_transform): Use fold_buildN_loc and build_call_expr_loc.

2014-10-24  Richard Biener  <rguenther@suse.de>

* genmatch.c (expr::gen_transform): Use fold_buildN_loc
and build_call_expr_loc.
(dt_simplify::gen): Drop non_lvalue for GIMPLE, use
non_lvalue_loc to build it for GENERIC.
(decision_tree::gen_generic): Add location argument to
generic_simplify prototype.
(capture_info): New class.
(capture_info::capture_info): New constructor.
(capture_info::walk_match): New method.
(capture_info::walk_result): New method.
(capture_info::walk_c_expr): New method.
(dt_simplify::gen): Handle preserving side-effects for
GENERIC code generation.
(decision_tree::gen_generic): Do not reject operands
with TREE_SIDE_EFFECTS.
* generic-match.h: New file.
* generic-match-head.c: Include generic-match.h, not gimple-match.h.
* match.pd: Add some constant folding patterns from fold-const.c.
* fold-const.c: Include generic-match.h.
(fold_unary_loc): Dispatch to generic_simplify.
(fold_ternary_loc): Likewise.
(fold_binary_loc): Likewise.  Remove patterns now implemented
by generic_simplify.
* gimple-fold.c (replace_stmt_with_simplification): New function.
(fold_stmt_1): Add valueize parameter, dispatch to gimple_simplify.
(no_follow_ssa_edges): New function.
(fold_stmt): New overload with valueization hook.  Use
no_follow_ssa_edges for the overload without hook.
(fold_stmt_inplace): Likewise.
* gimple-fold.h (no_follow_ssa_edges): Declare.

From-SVN: r216631

re PR target/63173 (performance problem with simd intrinsics vld2_dup_* on aarch64-none-elf)

        PR target/63173
        * config/aarch64/arm_neon.h (__LD2R_FUNC): Remove macro.
        (__LD3R_FUNC): Ditto.
        (__LD4R_FUNC): Ditto.
        (vld2_dup_s8, vld2_dup_s16, vld2_dup_s32, vld2_dup_f32, vld2_dup_f64,
         vld2_dup_u8, vld2_dup_u16, vld2_dup_u32, vld2_dup_p8, vld2_dup_p16
         vld2_dup_s64, vld2_dup_u64, vld2q_dup_s8, vld2q_dup_p8,
         vld2q_dup_s16, vld2q_dup_p16, vld2q_dup_s32, vld2q_dup_s64,
         vld2q_dup_u8, vld2q_dup_u16, vld2q_dup_u32, vld2q_dup_u64
         vld2q_dup_f32, vld2q_dup_f64): Rewrite using builtin functions.
        (vld3_dup_s64, vld3_dup_u64, vld3_dup_f64, vld3_dup_s8
         vld3_dup_p8, vld3_dup_s16, vld3_dup_p16, vld3_dup_s32
         vld3_dup_u8, vld3_dup_u16, vld3_dup_u32, vld3_dup_f32
         vld3q_dup_s8, vld3q_dup_p8, vld3q_dup_s16, vld3q_dup_p16
         vld3q_dup_s32, vld3q_dup_s64, vld3q_dup_u8, vld3q_dup_u16
         vld3q_dup_u32, vld3q_dup_u64, vld3q_dup_f32, vld3q_dup_f64): Likewise.
        (vld4_dup_s64, vld4_dup_u64, vld4_dup_f64, vld4_dup_s8
         vld4_dup_p8, vld4_dup_s16, vld4_dup_p16, vld4_dup_s32
         vld4_dup_u8, vld4_dup_u16, vld4_dup_u32, vld4_dup_f32
         vld4q_dup_s8, vld4q_dup_p8, vld4q_dup_s16, vld4q_dup_p16
         vld4q_dup_s32, vld4q_dup_s64, vld4q_dup_u8, vld4q_dup_u16
         vld4q_dup_u32, vld4q_dup_u64, vld4q_dup_f32, vld4q_dup_f64): Likewise.
        * config/aarch64/aarch64.md (define_c_enum "unspec"): Add
        UNSPEC_LD2_DUP, UNSPEC_LD3_DUP, UNSPEC_LD4_DUP.
        * config/aarch64/aarch64-simd-builtins.def (ld2r, ld3r, ld4r): New
        builtins.
        * config/aarch64/aarch64-simd.md (aarch64_simd_ld2r<mode>): New pattern.
        (aarch64_simd_ld3r<mode>): Likewise.
        (aarch64_simd_ld4r<mode>): Likewise.
        (aarch64_ld2r<mode>): New expand.
        (aarch64_ld3r<mode>): Likewise.
        (aarch64_ld4r<mode>): Likewise.

Co-Authored-By: Jiji Jiang <jiangjiji@huawei.com>
From-SVN: r216630

Fix gcc.dg/pr63594-1.c execution on Solaris/x86

        * gcc.dg/pr63594-1.c: Apply -mno-mmx to all i?86-*-* and x86_64-*-*
        targets.
        * gcc.dg/pr63594-2.c: Likewise.

From-SVN: r216629

980217-1.c (main): Fix implicit int.

* gcc.target/alpha/980217-1.c (main): Fix implicit int.
* gcc.target/alpha/pr19518.c (_mm_setzero_si64): New function.

From-SVN: r216627

MAINTAINERS (write-after-approval): Add myself.

2014-10-24  Daniel Hellstrom  <daniel@gaisler.com>

* MAINTAINERS (write-after-approval): Add myself.

From-SVN: r216626

Handle SCRATCH in decompose_address

* rtlanal.c (get_base_term): Handle SCRATCH.

From-SVN: r216624

Disable max_issue when scheduling for register pressure

* haifa-sched.c (sched_init): Disable max_issue when scheduling for
register pressure.

From-SVN: r216623

Remove cached_first_cycle_multipass_dfa_lookahead and cached_issue_rate

* haifa-sched.c (cached_first_cycle_multipass_dfa_lookahead,)
(cached_issue_rate): Remove.  Use dfa_lookahead and issue_rate instead.
(max_issue, choose_ready, sched_init): Update.

From-SVN: r216622

Improve scheduler dumps of ready list

* sched-int.h (struct _haifa_insn_data:last_rfs_win): New field.
* haifa-sched.c (INSN_LAST_RFS_WIN): New access macro.
(rfs_result): Set INSN_LAST_RFS_WIN.  Update signature.
(rank_for_schedule): Update calls to rfs_result to pass new parameters.
(print_rank_for_schedule_stats): Print out elements of ready list that
ended up on their respective places due to each of the sorting
heuristics.
(ready_sort): Update.
(debug_ready_list_1): Improve printout for SCHED_PRESSURE_MODEL.
(schedule_block): Update.

From-SVN: r216621

Account for prologue spills in reg_pressure scheduling

* haifa-sched.c (sched_class_regs_num, call_used_regs_num): New static
arrays.  Use sched_class_regs_num instead of ira_class_hard_regs_num.
(print_curr_reg_pressure, setup_insn_reg_pressure_info,)
(model_update_pressure, model_spill_cost): Use sched_class_regs_num.
(model_start_schedule): Update.
(sched_pressure_start_bb): New static function.  Calculate
sched_class_regs_num.
(schedule_block): Use it.
(alloc_global_sched_pressure_data): Calculate call_used_regs_num.

From-SVN: r216620

Makefile.in (BUILD_CPPLIB): When in stage2+ use the host library and make sure to pull in the required...

2014-10-24  Richard Biener  <rguenther@suse.de>

* Makefile.in (BUILD_CPPLIB): When in stage2+ use the
host library and make sure to pull in the required libintl
and libiconv dependencies.

From-SVN: r216619

fold-const.c (fold_binary_loc): Fix copy-and-pasto.

2014-10-24  Richard Biener  <rguenther@suse.de>

* fold-const.c (fold_binary_loc): Fix copy-and-pasto.

From-SVN: r216617

Fix bootstrap/PR63632

r216566 (r216568 for 4.9 branch) added %{fno-lto} to LINK_COMMAND_SPEC.
However the linker doesn't understand -fno-lto and errors out.
This causes an LTO/PGO bootstrap failure, because -fno-lto is used
during STAGEprofile.
Fixed by filtering out -fno-lto in collect2.c.

From-SVN: r216613

ipa-icf.c (sem_item_optimizer::parse_nonsingleton_classes): Guard division by zero in dumps.

* ipa-icf.c (sem_item_optimizer::parse_nonsingleton_classes): Guard
division by zero in dumps.
(sem_item_optimizer::merge_classes): Ditto.

From-SVN: r216612

compiler: Use MPC library for complex numbers.

* go-gcc.cc (Gcc_backend::complex_constant_expression): Take one
mpc_t parameter instead of two mpfr_t parameters.

From-SVN: r216611

compiler: Simplify making integer expressions.

Instead of always needing an mpz_t, add helper functions to
create an integer functions from signed or unsigned long
values.

From-SVN: r216610

Daily bump.

From-SVN: r216609

pa.c (pa_can_combine_p): Fix typo in last change.

* config/pa/pa.c (pa_can_combine_p): Fix typo in last change.

From-SVN: r216606

* tree-vrp.c (extract_range_from_assert): Fix typo in comment.

From-SVN: r216605

* config/mep/mep.h (TARGET_HAS_F_SETLKW): Don't define.

From-SVN: r216604

internal.h (backtrace_atomic_load_pointer): Fix to return void *.

* internal.h (backtrace_atomic_load_pointer) [no atomic or sync]:
Fix to return void *.

From-SVN: r216603

configure.ac: Disable the Go frontend on systems where it is known to not work.

* configure.ac: Disable the Go frontend on systems where it is known
to not work.
* configure: Regenerate.

From-SVN: r216602

complexity.cc: Add missing test variable.

2014-10-23  François Dumont  <fdumont@gcc.gnu.org>

* testsuite/25_algorithms/make_heap/complexity.cc: Add missing test
variable.
* testsuite/25_algorithms/sort_heap/complexity.cc: Likewise and use
log2. Now require cmath support.
* testsuite/25_algorithms/pop_heap/complexity.cc: Likewise and require
normal mode.
* testsuite/25_algorithms/push_heap/complexity.cc: Likewise.

From-SVN: r216601

re PR debug/63623 (Lots of functions get -fvar-tracking silently turned off unnecessarily)

PR debug/63623
* var-tracking.c (stack_adjust_offset_pre_post_cb): New function.
(stack_adjust_offset_pre_post): Use it through for_each_inc_dec,
instead of only handling autoinc in dest if it is a MEM.
(vt_stack_adjustments): Fix up formatting.

From-SVN: r216600

re PR c++/63619 (warning: deleting ‘void*’ is undefined has no -W flag)

PR c++/63619

gcc/cp:
PR c++/63619
* decl2.c (delete_sanity): Use OPT_Wdelete_incomplete in warning.

gcc/testsuite:
PR c++/63619
* g++.dg/warn/Wdelete-incomplete-3.C: New.
* g++.dg/warn/Wdelete-incomplete-4.C: New.

From-SVN: r216599

msp430.c (msp430_print_operand): 'x' modifier is independend of -mlarge.

* config/msp430/msp430.c (msp430_print_operand): 'x' modifier is
independend of -mlarge.
* config/msp430/constraints.md (Ys): Update comment.

From-SVN: r216598

status_cxx2011.xml: Update status.

* doc/xml/manual/status_cxx2011.xml: Update status.
* doc/xml/manual/status_cxx2014.xml: Update TS status.

From-SVN: r216597

re PR target/63534 (Bootstrap failure on x86_64/i686-linux)

PR target/63534
PR target/63618
gcc/
* cse.c (delete_trivially_dead_insns): Consider PIC register is used
while it is pseudo.
* dse.c (deletable_insn_p): Likewise.
gcc/testsuite/
* gcc.target/i386/pr63618.c: New test.

From-SVN: r216596

re PR c/63626 (Warning "inline function declared but never defined" output twice)

PR c/63626
* c-decl.c (pop_scope): Don't print warning in external_scope.

* gcc.dg/pr63626.c: New test.

From-SVN: r216595

c-ubsan.c (ubsan_instrument_shift): Perform the MINUS_EXPR in unsigned type.

* c-ubsan.c (ubsan_instrument_shift): Perform the MINUS_EXPR
in unsigned type.

* c-c++-common/ubsan/undefined-2.c: New test.

From-SVN: r216593

avr.c: Fix GNU coding rules and typos.

* config/avr/avr.c: Fix GNU coding rules and typos.
* config/avr/avr.h: Dito.
* config/avr/avr-c.c: Dito.
* config/avr/avr.md: Dito.

From-SVN: r216592

AVX. Fix block absq emit for non AVX-512 targets.

gcc/
* config/i386/sse.md (define_mode_iterator VI1248_AVX512VL_AVX512BW):
New.
(define_insn "*abs<mode>2"): Use VI1248_AVX512VL_AVX512BW mode
iterator.
(define_expand "abs<mode>2"): Ditto.

gcc/testsuite/
* gcc.target/i386/pr63600.c: New.

From-SVN: r216591