On AArch64 the UXTB and UXTH instructions are aliases of UBFM,
which does a shift as part of its operation. An AND immediate is a
simpler operation, and might be faster on some implementations, so
it is better to emit this this instead of UBFM.

Benchmarking showed no difference on implementations where UBFM has
the same performance as AND, and minor speedups across several
benchmarks on an implementation where UBFM is slower than AND.

Bootstrapped and tested on aarch64-none-elf.

    gcc/
* config/aarch64/aarch64.md
(zero_extend<SHORT:mode><GPI:mode>2_aarch64): Change output
statement and type.
(<optab>qihi2_aarch64): Likewise, and split into two.
(extendqihi2_aarch64): New.
(zero_extendqihi2_aarch64): New.
* config/aarch64/iterators.md (ldrxt): Remove.
* config/aarch64/aarch64.c (aarch64_rtx_costs): Change cost of
uxtb/uxth.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238821 138bc75d-0d04-0410-961f-82ee72b054a4

This patchset improves zero extend costs and code generation.

When zero extending a 32-bit register, we emit a "mov", but currently
report the cost of the "mov" incorrectly.

In terms of speed, we currently say the cost is that of an extend
operation. But the cost of a "mov" is the cost of 1 instruction, so fix
that.

In terms of size, we currently say that the "mov" takes 0 instructions.
Fix it by changing it to 1.

Bootstrapped and tested on aarch64-none-elf.

    gcc/
* config/aarch64/aarch64.c (aarch64_rtx_costs): Fix cost of zero extend.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238820 138bc75d-0d04-0410-961f-82ee72b054a4

This patch improves the readability of the prolog and epilog code by moving
some code into separate functions.  There is no difference in generated code.

gcc/
* config/aarch64/aarch64.c (aarch64_pushwb_pair_reg): Rename.
(aarch64_push_reg): New function to push 1 or 2 registers.
(aarch64_pop_reg): New function to pop 1 or 2 registers.
(aarch64_expand_prologue): Use aarch64_push_regs.
(aarch64_expand_epilogue): Use aarch64_pop_regs.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238818 138bc75d-0d04-0410-961f-82ee72b054a4

gcc/

2016-07-28  Yuri Rumyantsev  <ysrumyan@gmail.com>

PR tree-optimization/71734
* tree-ssa-loop-im.c (ref_indep_loop_p_1): Pass value of safelen
attribute instead of REF_LOOP and use it.
(ref_indep_loop_p_2): Use SAFELEN argument instead of REF_LOOP and
set it for Loops having non-zero safelen attribute.
(ref_indep_loop_p): Pass zero as initial value for safelen.

gcc/testsuite/

2016-07-28  Yuri Rumyantsev  <ysrumyan@gmail.com>

PR tree-optimization/71734
        * g++.dg/vect/pr70729-nest.cc: New test.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238817 138bc75d-0d04-0410-961f-82ee72b054a4

gcc/

PR middle-end/72657
PR target/72683
* tree-chkp.c (chkp_retbnd_call_by_val): Check for instrumentation
call using chkp_gimple_call_builtin_p.
(chkp_copy_bounds_for_assign): Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238816 138bc75d-0d04-0410-961f-82ee72b054a4

[PATCH] Revert Revert r238497 because of PR 71961.

This patch reverts the change for PR 71902 since it causes 178.gagel
miscompile in spec2000 as reported in PR 71961 which was observed in
x86_64, aarch64, powerpc64.

gcc/fortran/ChangeLog:

2016-07-28  Renlin Li  <renlin.li@arm.com>

Revert
2016-07-19  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/71902
* dependency.c (gfc_check_dependency): Use dep_ref.  Handle case
if identical is true and two array element references differ.
(gfc_dep_resovler):  Move most of the code to dep_ref.
(dep_ref):  New function.
* frontend-passes.c (realloc_string_callback):  Name temporary
variable "realloc_string".

gcc/testsuite/ChangeLog:

2016-07-28  Renlin Li  <renlin.li@arm.com>

Revert
2016-07-19  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/71902
* gfortran.dg/dependency_47.f90:  New test.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238815 138bc75d-0d04-0410-961f-82ee72b054a4

merge adjust_cost and adjust_cost_2 target hooks

gcc/ChangeLog:

2016-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* config/alpha/alpha.c (alpha_adjust_cost): Adjust.
* config/arm/arm-protos.h (struct tune_params): Likewise.
* config/arm/arm.c (xscale_sched_adjust_cost): Likewise.
(cortex_a9_sched_adjust_cost): Likewise.
(fa726te_sched_adjust_cost): Likewise.
(arm_adjust_cost): Likewise.
* config/bfin/bfin.c (bfin_adjust_cost): Likewise.
* config/c6x/c6x.c (c6x_adjust_cost): Likewise.
* config/epiphany/epiphany.c (epiphany_adjust_cost): Likewise.
* config/i386/i386.c (ix86_adjust_cost): Likewise.
* config/ia64/ia64.c: Likewise.
* config/m68k/m68k.c: Likewise.
* config/mep/mep.c (mep_adjust_cost): Likewise.
* config/microblaze/microblaze.c (microblaze_adjust_cost):
* Likewise.
* config/mips/mips.c (mips_adjust_cost): Likewise.
* config/mn10300/mn10300.c (mn10300_adjust_sched_cost):
* Likewise.
* config/pa/pa.c (pa_adjust_cost): Likewise.
* config/rs6000/rs6000.c (rs6000_adjust_cost): Likewise.
(rs6000_debug_adjust_cost): Likewise.
* config/sh/sh.c (sh_adjust_cost): Likewise.
* config/sparc/sparc.c (supersparc_adjust_cost): Likewise.
(hypersparc_adjust_cost): Likewise.
(sparc_adjust_cost): Likewise.
* config/spu/spu.c (spu_sched_adjust_cost): Likewise.
* config/tilegx/tilegx.c (tilegx_sched_adjust_cost): Likewise.
* config/tilepro/tilepro.c (tilepro_sched_adjust_cost):
* Likewise.
* config/visium/visium.c (visium_adjust_cost): Likewise.
* doc/tm.texi: Regenerate.
* haifa-sched.c (dep_cost_1): Adjust.
* target.def: Merge adjust_cost and adjust_cost_2.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238814 138bc75d-0d04-0410-961f-82ee72b054a4

haifa-sched.c: make twins a auto_vec<rtx_insn *>

gcc/ChangeLog:

2016-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* haifa-sched.c (add_to_speculative_block): Make twins a vector.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238813 138bc75d-0d04-0410-961f-82ee72b054a4

make pattern_regs a vec

gcc/ChangeLog:

2016-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* store-motion.c (struct st_expr): Make pattern_regs a vector.
(extract_mentioned_regs): Append to a vector instead of
returning a rtx_expr_list.
(st_expr_entry): Adjust.
(free_st_expr_entry): Likewise.
(store_ops_ok): Likewise.
(store_killed_in_insn): Likewise.
(find_moveable_store): Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238812 138bc75d-0d04-0410-961f-82ee72b054a4

Introduce no_profile_instrument_function attribute

PR gcov-profile/68025
* tree-profile.c (tree_profiling): Respect
no_profile_instrument_function attribute.
* doc/extend.texi: Document no_profile_instrument_function
attribute.
PR gcov-profile/68025
* c-common.c (handle_no_profile_instrument_function_attribute):
PR gcov-profile/68025
* gcc.dg/no_profile_instrument_function-attr-1.c: New test.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238811 138bc75d-0d04-0410-961f-82ee72b054a4

Add missing file.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238810 138bc75d-0d04-0410-961f-82ee72b054a4

Add mark_spam.py script

* mark_spam.py: New file.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238809 138bc75d-0d04-0410-961f-82ee72b054a4

Do not allow make_compound_operation for vector mode

* g++.dg/vect/pr70944.cc: New test.
PR rtl-optimization/70944
* combine.c (make_compound_operation):
Do not allow make_compound_operation for vector mode

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238808 138bc75d-0d04-0410-961f-82ee72b054a4

Daily bump.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238806 138bc75d-0d04-0410-961f-82ee72b054a4

gcc/testsuite/ChangeLog:

2016-07-28  Kugan Vivekanandarajah  <kuganv@linaro.org>

PR middle-end/71994
* gcc.dg/torture/pr71994.c: New test.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238803 138bc75d-0d04-0410-961f-82ee72b054a4

gcc/testsuite/ChangeLog:

2016-07-28  Kugan Vivekanandarajah  <kuganv@linaro.org>

PR middle-end/71994
* gcc.dg/torture/pr71994.c: New test.

gcc/ChangeLog:

2016-07-28  Kugan Vivekanandarajah  <kuganv@linaro.org>

PR middle-end/71994
* tree-ssa-reassoc.c (maybe_optimize_range_tests): Check tcc_comparison
 before calling get_ops.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238802 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        * defaults.h (LOG2_BITS_PER_UNIT): Move from here...
        * tree.h (LOG2_BITS_PER_UNIT): ...to here.
        (BITS_PER_UNIT_LOG): Remove.
        (int_bit_position): Use LOG2_BITS_PER_UNIT instead of BITS_PER_UNIT_LOG.
        * expr.c (expand_assignment): Likewise.
        * stor-layout.c (initialize_sizetypes): Likewise.

c-family:
2016-07-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        * c-common.c (check_user_alignment): Use LOG2_BITS_PER_UNIT instead of
        BITS_PER_UNIT_LOG.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238800 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-27  Michael Meissner  <meissner@linux.vnet.ibm.com>

* config/rs6000/vector.md (vec_extract<mode>): Change the calling
signature of rs6000_expand_vector_extract so that the element
number is a RTX instead of a constant integer.
* config/rs6000/rs6000-protos.h (rs6000_expand_vector_extract):
Likewise.
* config/rs6000/rs6000.c (rs6000_expand_vector_extract): Likewise.
(altivec_expand_vec_ext_builtin): Likewise.
* config/rs6000/altivec.md (reduc_plus_scal_<mode>): Likewise.
* config/rs6000/vsx.md (vsx_extract_<mode>): Fix spelling of the
MFVSRLD instruction.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238799 138bc75d-0d04-0410-961f-82ee72b054a4

* testsuite/20_util/forward/1_neg.cc: Move dg-error to right line.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238793 138bc75d-0d04-0410-961f-82ee72b054a4

Move make_location from tree.h/c to input.h/c

For some reason I added make_location and some related functions to
tree.h/c, rather than to input.h/c.  Move them there, so we can use them
without requiring tree, and add some selftest coverage.

gcc/ChangeLog:
* input.c (get_pure_location): Move here from tree.c.
(make_location): Likewise.  Add header comment.
(selftest::test_accessing_ordinary_linemaps): Verify
pure_location_p, make_location, get_location_from_adhoc_loc and
get_range_from_loc.
* input.h (get_pure_location): Move declaration here from tree.h.
(get_finish): Likewise for inline function.
(make_location): Likewise for declaration.
* tree.c (get_pure_location): Move to input.c.
(make_location): Likewise.
* tree.h (get_pure_location): Move declaration to tree.h.
(get_finish): Likewise for inline function.
(make_location): Likewise for declaration.

libcpp/ChangeLog:
* include/line-map.h (source_location): Fix line numbers in
comment.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238792 138bc75d-0d04-0410-961f-82ee72b054a4

Add missing PR marker to Changelog for 71216 fix

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238790 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>

PR middle-end/71078
* match.pd (x / abs(x) -> copysign(1.0, x)): New pattern.

testsuite/
* gcc.dg/tree-ssa/pr71078-1.c: New test-case.
* gcc.dg/tree-ssa/pr71078-2.c: Likewise.
* gcc.dg/tree-ssa/pr71078-3.c: Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238787 138bc75d-0d04-0410-961f-82ee72b054a4

Use static_assert for STATIC_ASSERT for C++11 onwards

C++11 has a
  static_assert (COND, MESSAGE)
which gives more readable error messages for STATIC_ASSERT than our
current implementation.

This patch makes us use it if __cplusplus >= 201103L

There's also a provisional static_assert (COND) in C++1z, but presumably
we should wait until that one is fully standardized before using it.

gcc/ChangeLog:
* system.h (STATIC_ASSERT): Use static_assert if building
with C++11 onwards.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238786 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/71747 - ICE with self-referential partial spec

* pt.c (get_partial_spec_bindings): Replace tparms and spec_args
parameters with spec_tmpl.  Call push_tinst_level.
(most_specialized_partial_spec): Adjust.
(more_specialized_partial_spec): Adjust.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238785 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-27  Richard Biener  <rguenther@suse.de>

* gcc.dg/vect/costmodel/x86_64/costmodel-pr68961.c: Remove.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238784 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-27  Richard Biener  <rguenther@suse.de>

PR tree-optimization/72517
* tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
Revert change to not compute read-read dependences.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238783 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-27  Richard Biener  <rguenther@suse.de>

* predict.c (set_even_probabilities): Make nedges unsigned.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238782 138bc75d-0d04-0410-961f-82ee72b054a4

predict.c: merge multi-edges

* gcc.dg/predict-13.c: New test.
* gcc.dg/predict-14.c: New test.
* predict.c (set_even_probabilities): Handle unlikely edges.
(combine_predictions_for_bb): Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238781 138bc75d-0d04-0410-961f-82ee72b054a4

Use __{U,}INTPTR_TYPE__ to avoid including stdint.h

gcc/testsuite/

* gcc.dg/torture/pr69352.c: Use  __INTPTR_TYPE__ instead of
including stdint.h.
* gcc.dg/torture/pr71866.c: Use __UINTPTR_TYPE__ isntead of
including stdint.h.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238780 138bc75d-0d04-0410-961f-82ee72b054a4

[gcc]
2016-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/71869
* config/rs6000/rs6000.c (rs6000_generate_compare): Rework
__float128 support when we don't have hardware support, so that
the IEEE built-in functions like isgreater, first call __unordkf3
to make sure neither operand is a NaN, and if both operands are
ordered, do the normal comparison.

[gcc/testsuite]
2016-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/71869
* gcc.target/powerpc/float128-cmp.c: New test to make sure that
IEEE built-in functions handle quiet and signalling NaNs
correctly.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238779 138bc75d-0d04-0410-961f-82ee72b054a4

Daily bump.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238778 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-22  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/71862
* class.c: Remove assert.  Iterate over component only if non-null.

2016-07-22  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/71862
* gfortran.dg/pr71862.f90: New test.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238774 138bc75d-0d04-0410-961f-82ee72b054a4

gcc/testsuite/ChangeLog:
* gcc.dg/atomic/pr71675.c: Replace the unsupported c11 target
selector with dg-options.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238766 138bc75d-0d04-0410-961f-82ee72b054a4

Missed changelog entries for commit 238252.  Sorry,
2016-07-12  Nathan Sidwell  <nathan@acm.org>

* config/arm/arm.c (arm_option_override): Set MASK_SINGLE_PIC_BASE
when -mno-pic-data-is-text-relative is in effect, by default.
* doc/invoke.texi (mpic-data-is-text-relative): Document new
behavior and clarify.

2016-07-12  Nathan Sidwell  <nathan@acm.org>

* gcc.target/arm/data-rel-1.c: New.
* gcc.target/arm/data-rel-2.c: New.
* gcc.target/arm/data-rel-3.c: New.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238763 138bc75d-0d04-0410-961f-82ee72b054a4

Minor changes in tree-vrp.c

gcc/ChangeLog:

* tree-vrp.c (dump_asserts_for): Print loc->expr instead of
name.
(extract_code_and_val_from_cond_with_ops): Verify that name is
either cond_op0 or cond_op1.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238762 138bc75d-0d04-0410-961f-82ee72b054a4

Teach VRP to register assertions along default switch labels (PR18046)

gcc/ChangeLog:

PR tree-optimization/18046
* genmodes.c (emit_mode_size_inline): Emit an assert that
verifies that mode is a valid array index.
(emit_mode_nuinits_inline): Likewise.
(emit_mode_inner_inline): Likewise.
(emit_mode_unit_size_inline): Likewise.
(emit_mode_unit_precision_inline): Likewise.
* tree-vrp.c: Include params.h.
(find_switch_asserts): Register edge assertions for the default
label which correspond to the anti-ranges of each case label.
* params.def (PARAM_MAX_VRP_SWITCH_ASSERTIONS): New.
* doc/invoke.texi: Document it.

gcc/testsuite/ChangeLog:

PR tree-optimization/18046
* gcc.dg/tree-ssa/ssa-dom-thread-6.c: Bump FSM count to 5.
* gcc.dg/tree-ssa/vrp103.c: New test.
* gcc.dg/tree-ssa/vrp104.c: New test.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238761 138bc75d-0d04-0410-961f-82ee72b054a4

Move ChangeLog entry to the right place

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238760 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* gcc.target/powerpc/pr63354.c: Require lp64 since
-mprofile-kernel is not legal with -m32.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238759 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* gimple-ssa-strength-reduction.c (slsr_process_phi): Remove dead
and unnecessary call to gimple_bb.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238758 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-26  Richard Biener  <rguenther@suse.de>

PR rtl-optimization/71984
* simplify-rtx.c (simplify_subreg): Use GET_MODE_SIZE and prepare
for VOIDmode.

* gcc.dg/torture/pr71984.c: New testcase.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238757 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-26  Richard Biener  <rguenther@suse.de>

PR middle-end/72517
* expmed.c (extract_bit_field_1): Constrain the vector mode
with element size matching the extraction mode size when
choosing a better vector mode to do the extraction from.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238756 138bc75d-0d04-0410-961f-82ee72b054a4

Add support to run auto-vectorization tests for multiple effective targets.

gcc/testsuite/

* g++.dg/vect/vect.exp: Add and set new global EFFECTIVE_TARGETS. Call
g++-dg-runtest via et-dg-runtest.
* gcc.dg/graphite/graphite.exp: Likewise, but for dg-runtest.
* gcc.dg/vect/vect.exp: Likewise.
* gfortran.dg/graphite/graphite.exp: Likewise, but for
gfortran-dg-runtest.
* gfortran.dg/vect/vect.exp: Likewise.
* lib/target-supports.exp (check_mpaired_single_hw_available): New.
(check_mips_loongson_hw_available): Likewise.
(check_effective_target_mpaired_single_runtime): Likewise.
(check_effective_target_mips_loongson_runtime): Likewise.
(add_options_for_mpaired_single): Likewise.
(check_effective_target_vect_int): Add global et_index.
Check and save the supported feature for a target selected by
the et_index target.  Break long lines where appropriate.  Call
et-is-effective-target for MIPS with an argument instead of
check_effective_target_* where appropriate.
(check_effective_target_vect_intfloat_cvt): Likewise.
(check_effective_target_vect_uintfloat_cvt): Likewise.
(check_effective_target_vect_floatint_cvt): Likewise.
(check_effective_target_vect_floatuint_cvt): Likewise.
(check_effective_target_vect_simd_clones): Likewise.
(check_effective_target_vect_shift): ewise.
(check_effective_target_whole_vector_shift): Likewise.
(check_effective_target_vect_bswap): Likewise.
(check_effective_target_vect_shift_char): Likewise.
(check_effective_target_vect_long): Likewise.
(check_effective_target_vect_float): Likewise.
(check_effective_target_vect_double): Likewise.
(check_effective_target_vect_long_long): Likewise.
(check_effective_target_vect_no_int_max): Likewise.
(check_effective_target_vect_no_int_add): Likewise.
(check_effective_target_vect_no_bitwise): Likewise.
(check_effective_target_vect_widen_shift): Likewise.
(check_effective_target_vect_no_align): Likewise.
(check_effective_target_vect_hw_misalign): Likewise.
(check_effective_target_vect_element_align): Likewise.
(check_effective_target_vect_condition): Likewise.
(check_effective_target_vect_cond_mixed): Likewise.
(check_effective_target_vect_char_mult): Likewise.
(check_effective_target_vect_short_mult): Likewise.
(check_effective_target_vect_int_mult): Likewise.
(check_effective_target_vect_extract_even_odd): Likewise.
(check_effective_target_vect_interleave): Likewise.
(check_effective_target_vect_stridedN): Likewise.
(check_effective_target_vect_multiple_sizes): Likewise.
(check_effective_target_vect64): Likewise.
(check_effective_target_vect_call_copysignf): Likewise.
(check_effective_target_vect_call_sqrtf): Likewise.
(check_effective_target_vect_call_btrunc): Likewise.
(check_effective_target_vect_call_btruncf): Likewise.
(check_effective_target_vect_call_ceil): Likewise.
(check_effective_target_vect_call_ceilf): Likewise.
(check_effective_target_vect_call_floor): Likewise.
(check_effective_target_vect_call_floorf): Likewise.
(check_effective_target_vect_call_lceil): Likewise.
(check_effective_target_vect_call_lfloor): Likewise.
(check_effective_target_vect_call_nearbyint): Likewise.
(check_effective_target_vect_call_nearbyintf): Likewise.
(check_effective_target_vect_call_round): Likewise.
(check_effective_target_vect_call_roundf): Likewise.
(check_effective_target_vect_perm): Likewise, but also append *_saved
to the existing global name to properly cache the result.
(check_effective_target_vect_perm_byte): Likewise.
(check_effective_target_vect_perm_short): Likewise.
(check_effective_target_vect_widen_sum_hi_to_si_pattern): Likewise.
(check_effective_target_vect_widen_sum_hi_to_si): Likewise.
(check_effective_target_vect_widen_sum_qi_to_hi): Likewise.
(check_effective_target_vect_widen_sum_qi_to_si): Likewise.
(check_effective_target_vect_widen_mult_qi_to_hi_pattern): Likewise.
(check_effective_target_vect_widen_mult_qi_to_hi): Likewise.
(check_effective_target_vect_widen_mult_hi_to_si_pattern): Likewise.
(check_effective_target_vect_widen_mult_si_to_di_pattern): Likewise.
(check_effective_target_vect_sdot_qi): Likewise.
(check_effective_target_vect_udot_qi): Likewise.
(check_effective_target_vect_sdot_hi): Likewise.
(check_effective_target_vect_udot_hi): Likewise.
(check_effective_target_vect_usad_char): Likewise.
(check_effective_target_vect_pack_trunc): Likewise.
(check_effective_target_vect_unpack): Likewise.
(check_effective_target_vect_aligned_arrays): Likewise.
(check_effective_target_vect_natural_alignment): Likewise.
(check_effective_target_vector_alignment_reachable): Likewise.
(check_effective_target_vector_alignment_reachable_for_64bit): Likewise.
(is-effective-target): Initialize et_index if undefined.
(et-dg-runtest): New.
(et-is-effective-target): Likewise.
(check_vect_support_and_set_flags): Add supported MIPS targets to
EFFECTIVE_TARGETS list.  Return the number of supported targets.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238755 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-26  Richard Biener  <rguenther@suse.de>
    Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>

PR middle-end/70920
* match.pd ((intptr)x eq/ne CST to x eq/ne (typeof x) CST): New
pattern.

testsuite/
* gcc.dg/pr70920-1.c: New test-case.
* gcc.dg/pr70902-2.c: Likewise.
* gcc.dg/pr70920-3.c: Likewise.
* gcc.dg/pr70920-4.c: Likewise
* gcc.dg/tree-ssa/ssa-dom-branch-1.c: Change scan-tree-dump-times to
2 instead of 3.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238754 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-26  Richard Biener  <rguenther@suse.de>

* update_version_svn: Ignore the GCC 4.9 branch.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238753 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-26  Richard Biener  <rguenther@suse.de>

* crontab: Remove entry for the GCC 4.9 branch.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238752 138bc75d-0d04-0410-961f-82ee72b054a4

add [cd]tors to scc_info

gcc/ChangeLog:

2016-07-26  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* tree-ssa-structalias.c (struct scc_info): Change types of
members to auto_sbitmap and auto_vec.
(scc_info::scc_info): New constructor.
(scc_info::~scc_info): New destructor.
(init_scc_info): Remove.
(free_scc_info): Remove.
(find_indirect_cycles): Adjust.
(perform_var_substitution): Likewise.
(free_var_substitution_info): Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238751 138bc75d-0d04-0410-961f-82ee72b054a4

add a constructor to elim_graph

gcc/ChangeLog:

2016-07-26  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* tree-outof-ssa.c (struct elim_graph): Change type of members
to auto_vec and auto_sbitmap.
(elim_graph::elim_graph): New constructor.
(delete_elim_graph): Remove.
(expand_phi_nodes): Adjust.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238750 138bc75d-0d04-0410-961f-82ee72b054a4

remove elim_graph typedef

gcc/ChangeLog:

2016-07-26  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* tree-outof-ssa.c (struct elim_graph): Remove typedef.
(new_elim_graph): Adjust.
(clear_elim_graph): Likewise.
(delete_elim_graph): Likewise.
(elim_graph_size): Likewise.
(elim_graph_add_node): Likewise.
(elim_graph_add_edge): Likewise.
(elim_graph_remove_succ_edge): Likewise.
(eliminate_name): Likewise.
(eliminate_build): Likewise.
(elim_forward): Likewise.
(elim_unvisited_predecessor): Likewise.
(elim_backward): Likewise.
(elim_create): Likewise.
(eliminate_phi): Likewise.
(expand_phi_nodes): Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238749 138bc75d-0d04-0410-961f-82ee72b054a4

use auto_sbitmap in various places

gcc/ChangeLog:

2016-07-26  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* bt-load.c (compute_out): Use auto_sbitmap class.
(link_btr_uses): Likewise.
* cfganal.c (mark_dfs_back_edges): Likewise.
(post_order_compute): Likewise.
(inverted_post_order_compute): Likewise.
(pre_and_rev_post_order_compute_fn): Likewise.
(single_pred_before_succ_order): Likewise.
* cfgexpand.c (pass_expand::execute): Likewise.
* cfgloop.c (verify_loop_structure): Likewise.
* cfgloopmanip.c (fix_bb_placements): Likewise.
(remove_path): Likewise.
(update_dominators_in_loop): Likewise.
* cfgrtl.c (break_superblocks): Likewise.
* ddg.c (check_sccs): Likewise.
(create_ddg_all_sccs): Likewise.
* df-core.c (df_worklist_dataflow): Likewise.
* dse.c (dse_step3): Likewise.
* except.c (eh_region_outermost): Likewise.
* function.c (thread_prologue_and_epilogue_insns): Likewise.
* gcse.c (prune_expressions): Likewise.
(prune_insertions_deletions): Likewise.
* gimple-ssa-backprop.c (backprop::~backprop): Likewise.
* graph.c (draw_cfg_nodes_no_loops): Likewise.
* ira-lives.c (remove_some_program_points_and_update_live_ranges): Likewise.
* lcm.c (compute_earliest): Likewise.
(compute_farthest): Likewise.
* loop-unroll.c (unroll_loop_constant_iterations): Likewise.
(unroll_loop_runtime_iterations): Likewise.
(unroll_loop_stupid): Likewise.
* lower-subreg.c (decompose_multiword_subregs): Likewise.
* lra-lives.c: Likewise.
* lra.c (lra): Likewise.
* modulo-sched.c (schedule_reg_moves): Likewise.
(optimize_sc): Likewise.
(get_sched_window): Likewise.
(sms_schedule_by_order): Likewise.
(check_nodes_order): Likewise.
(order_nodes_of_sccs): Likewise.
(order_nodes_in_scc): Likewise.
* recog.c (split_all_insns): Likewise.
* regcprop.c (pass_cprop_hardreg::execute): Likewise.
* reload1.c (reload): Likewise.
* sched-rgn.c (haifa_find_rgns): Likewise.
(split_edges): Likewise.
(compute_trg_info): Likewise.
* sel-sched.c (init_seqno): Likewise.
* store-motion.c (remove_reachable_equiv_notes): Likewise.
* tree-into-ssa.c (update_ssa): Likewise.
* tree-ssa-live.c (live_worklist): Likewise.
* tree-ssa-loop-im.c (fill_always_executed_in): Likewise.
* tree-ssa-loop-ivcanon.c (try_unroll_loop_completely):
* Likewise.
(try_peel_loop): Likewise.
* tree-ssa-loop-manip.c (tree_transform_and_unroll_loop):
* Likewise.
* tree-ssa-pre.c (compute_antic): Likewise.
* tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
* tree-stdarg.c (reachable_at_most_once): Likewise.
* tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Likewise.
* var-tracking.c (vt_find_locations): Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238748 138bc75d-0d04-0410-961f-82ee72b054a4

add auto_sbitmap class

gcc/ChangeLog:

2016-07-26  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>

* sbitmap.h (auto_sbitmap): New class.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238747 138bc75d-0d04-0410-961f-82ee72b054a4

[RS6000] push_secondary_reload ICE

PR target/72103
* config/rs6000/rs6000.c (rs6000_secondary_reload): Initialize
sri->t_icode.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238744 138bc75d-0d04-0410-961f-82ee72b054a4

libgo: bump library version number for 1.7

    Reviewed-on: https://go-review.googlesource.com/25211

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238743 138bc75d-0d04-0410-961f-82ee72b054a4

os/user: fix Solaris declaration.

    Patch from Rainer Orth.

    Reviewed-on: https://go-review.googlesource.com/25210

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238742 138bc75d-0d04-0410-961f-82ee72b054a4

Daily bump.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238741 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/65970 - revert loop location change

* cp-gimplify.c (genericize_cp_loop): Revert location change.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238737 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/71837 - pack expansion in init-capture

* lambda.c (add_capture): Leave a pack expansion in a TREE_LIST.
(build_lambda_object): Call build_x_compound_expr_from_list.
* pt.c (tsubst) [DECLTYPE_TYPE]: Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238733 138bc75d-0d04-0410-961f-82ee72b054a4

Fix selftest::temp_source_file ctor

gcc/ChangeLog:
* input.c (selftest::temp_source_file::temp_source_file): Fix
missing "%s" in fprintf.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238732 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/71833 - member template with two parameter packs

PR c++/54440
* pt.c (coerce_template_parameter_pack): Fix logic for
pack index.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238731 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/65970 - constexpr infinite loop

gcc/c-family/
* c.opt (fconstexpr-loop-limit): New.
gcc/cp/
* constexpr.c (cxx_eval_loop_expr): Count iterations.
* cp-gimplify.c (genericize_cp_loop): Use start_locus even for
infinite loops.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238730 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/71972 - constexpr array self-modification

* constexpr.c (cxx_eval_array_reference): Handle looking for the
value of an element we're currently modifying.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238729 138bc75d-0d04-0410-961f-82ee72b054a4

* g++.dg/init/elide5.C: Don't use unsigned long for size_t.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238728 138bc75d-0d04-0410-961f-82ee72b054a4

PR middle-end/71732
* cselib.c (cselib_process_insn): Invalidate argument slots for
const/pure calls.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238727 138bc75d-0d04-0410-961f-82ee72b054a4

testsuite: add two missing label_values annotations

2016-07-25  Alexander Monakov  <amonakov@ispras.ru>

* gcc.c-torture/execute/pr71494.c: Require label_values.
* gcc.dg/pr16973.c: Ditto.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238726 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][10/10] ARMv8.2-A FP16 lane scalar intrinsics

gcc/
* config/aarch64/arm_neon.h (vfmah_lane_f16, vfmah_laneq_f16,
vfmsh_lane_f16, vfmsh_laneq_f16, vmulh_lane_f16, vmulh_laneq_f16,
vmulxh_lane_f16, vmulxh_laneq_f16): New.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238725 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][9/10] ARMv8.2-A FP16 three operands scalar intrinsics

gcc/
* config/aarch64/aarch64-simd-builtins.def: Register new builtins.
* config/aarch64/aarch64.md (fma, fnma): Support HF.
* config/aarch64/arm_fp16.h (vfmah_f16, vfmsh_f16): New.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238724 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][8/10] ARMv8.2-A FP16 two operands scalar intrinsics

gcc/
* config/aarch64/aarch64-simd-builtins.def: Register new builtins.
* config/aarch64/aarch64.md (<FCVT_F2FIXED:fcvt_fixed_insn>hf<mode>3):
New.
(<FCVT_FIXED2F:fcvt_fixed_insn><mode>hf3): Likewise.
(add<mode>3): Likewise.
(sub<mode>3): Likewise.
(mul<mode>3): Likewise.
(div<mode>3): Likewise.
(*div<mode>3): Likewise.
(<fmaxmin><mode>3): Extend to HF.
* config/aarch64/aarch64-simd.md (aarch64_rsqrts<mode>): Likewise.
(fabd<mode>3): Likewise.
(<FCVT_F2FIXED:fcvt_fixed_insn><VHSDF_HSDF:mode>3): Likewise.
(<FCVT_FIXED2F:fcvt_fixed_insn><VHSDI_HSDI:mode>3): Likewise.
(aarch64_fmulx<mode>): Likewise.
(aarch64_fac<optab><mode>): Likewise.
(aarch64_frecps<mode>): Likewise.
(<FCVT_F2FIXED:fcvt_fixed_insn>hfhi3): New.
(<FCVT_FIXED2F:fcvt_fixed_insn>hihf3): Likewise.
* config/aarch64/iterators.md (VHSDF_SDF): Delete.
(VSDQ_HSDI): Support HI.
(fcvt_target, FCVT_TARGET): Likewise.
* config/aarch64/arm_fp16.h (vaddh_f16, vsubh_f16, vabdh_f16,
vcageh_f16, vcagth_f16, vcaleh_f16, vcalth_f16, vceqh_f16, vcgeh_f16,
vcgth_f16, vcleh_f16, vclth_f16, vcvth_n_f16_s16, vcvth_n_f16_s32,
vcvth_n_f16_s64, vcvth_n_f16_u16, vcvth_n_f16_u32, vcvth_n_f16_u64,
vcvth_n_s16_f16, vcvth_n_s32_f16, vcvth_n_s64_f16, vcvth_n_u16_f16,
vcvth_n_u32_f16, vcvth_n_u64_f16, vdivh_f16, vmaxh_f16, vmaxnmh_f16,
vminh_f16, vminnmh_f16, vmulh_f16, vmulxh_f16, vrecpsh_f16,
vrsqrtsh_f16): New.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238723 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][7/10] ARMv8.2-A FP16 one operand scalar intrinsics

gcc/
* config.gcc (aarch64*-*-*): Install arm_fp16.h.
* config/aarch64/aarch64-builtins.c (hi_UP): New.
* config/aarch64/aarch64-simd-builtins.def: Register new builtins.
* config/aarch64/aarch64-simd.md (aarch64_frsqrte<mode>): Extend to HF
mode.
(aarch64_frecp<FRECP:frecp_suffix><mode>): Likewise.
(aarch64_cm<optab><mode>): Likewise.
* config/aarch64/aarch64.md (<frint_pattern><mode>2): Likewise.
(l<fcvt_pattern><su_optab><GPF:mode><GPI:mode>2): Likewise.
(fix_trunc<GPF:mode><GPI:mode>2): Likewise.
(sqrt<mode>2): Likewise.
(abs<mode>2): Likewise.
(<optab><mode>hf2): New pattern for HF mode.
(<optab>hihf2): Likewise.
* config/aarch64/arm_neon.h: Include arm_fp16.h.
* config/aarch64/iterators.md (GPF_F16, GPI_F16, VHSDF_HSDF): New.
(w1, w2, v, s, q, Vmtype, V_cmp_result, fcvt_iesize, FCVT_IESIZE):
Support HF mode.
* config/aarch64/arm_fp16.h: New file.
(vabsh_f16, vceqzh_f16, vcgezh_f16, vcgtzh_f16, vclezh_f16, vcltzh_f16,
vcvth_f16_s16, vcvth_f16_s32, vcvth_f16_s64, vcvth_f16_u16,
vcvth_f16_u32, vcvth_f16_u64, vcvth_s16_f16, vcvth_s32_f16,
vcvth_s64_f16, vcvth_u16_f16, vcvth_u32_f16, vcvth_u64_f16,
vcvtah_s16_f16, vcvtah_s32_f16, vcvtah_s64_f16, vcvtah_u16_f16,
vcvtah_u32_f16, vcvtah_u64_f16, vcvtmh_s16_f16, vcvtmh_s32_f16,
vcvtmh_s64_f16, vcvtmh_u16_f16, vcvtmh_u32_f16, vcvtmh_u64_f16,
vcvtnh_s16_f16, vcvtnh_s32_f16, vcvtnh_s64_f16, vcvtnh_u16_f16,
vcvtnh_u32_f16, vcvtnh_u64_f16, vcvtph_s16_f16, vcvtph_s32_f16,
vcvtph_s64_f16, vcvtph_u16_f16, vcvtph_u32_f16, vcvtph_u64_f16,
vnegh_f16, vrecpeh_f16, vrecpxh_f16, vrndh_f16, vrndah_f16, vrndih_f16,
vrndmh_f16, vrndnh_f16, vrndph_f16, vrndxh_f16, vrsqrteh_f16,
vsqrth_f16): New.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238722 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][6/14] ARMv8.2-A FP16 reduction vector intrinsics

gcc/
* config/aarch64/aarch64-simd-builtins.def (reduc_smax_scal_,
reduc_smin_scal_): Use VDQIF_F16.
(reduc_smax_nan_scal_, reduc_smin_nan_scal_): Use VHSDF.
* config/aarch64/aarch64-simd.md (reduc_<maxmin_uns>_scal_<mode>):
Use VHSDF.
(aarch64_reduc_<maxmin_uns>_internal<mode>): Likewise.
* config/aarch64/iterators.md (VDQIF_F16): New.
(vp): Support HF modes.
* config/aarch64/arm_neon.h (vmaxv_f16, vmaxvq_f16, vminv_f16,
vminvq_f16, vmaxnmv_f16, vmaxnmvq_f16, vminnmv_f16, vminnmvq_f16): New.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238721 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][5/10] ARMv8.2-A FP16 lane vector intrinsics

gcc/
* config/aarch64/aarch64-simd.md (*aarch64_mulx_elt_to_64v2df): Rename to
"*aarch64_mulx_elt_from_dup<mode>".
(*aarch64_mul3_elt<mode>): Update schedule type.
(*aarch64_mul3_elt_from_dup<mode>): Likewise.
(*aarch64_fma4_elt_from_dup<mode>): Likewise.
(*aarch64_fnma4_elt_from_dup<mode>): Likewise.
* config/aarch64/iterators.md (VMUL): Supprt half precision float modes.
(f, fp): Support HF modes.
* config/aarch64/arm_neon.h (vfma_lane_f16, vfmaq_lane_f16,
vfma_laneq_f16, vfmaq_laneq_f16, vfma_n_f16, vfmaq_n_f16, vfms_lane_f16,
        vfmsq_lane_f16, vfms_laneq_f16, vfmsq_laneq_f16, vfms_n_f16,
vfmsq_n_f16, vmul_lane_f16, vmulq_lane_f16, vmul_laneq_f16,
vmulq_laneq_f16, vmul_n_f16, vmulq_n_f16, vmulx_lane_f16,
vmulxq_lane_f16, vmulx_laneq_f16, vmulxq_laneq_f16): New.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238719 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][4/10] ARMv8.2-A FP16 three operands vector intrinsics

gcc/
* config/aarch64/aarch64-simd-builtins.def: Register new builtins.
* config/aarch64/aarch64-simd.md (fma<mode>4, fnma<mode>4): Extend to HF
modes.
* config/aarch64/arm_neon.h (vfma_f16, vfmaq_f16, vfms_f16,
vfmsq_f16): New.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238718 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][3/10] ARMv8.2-A FP16 two operands vector intrinsics

gcc/
* config/aarch64/aarch64-simd-builtins.def: Register new builtins.
* config/aarch64/aarch64-simd.md
(aarch64_rsqrts<mode>): Extend to HF modes.
(fabd<mode>3): Likewise.
(<FCVT_F2FIXED:fcvt_fixed_insn><VHSDF_SDF:mode>3): Likewise.
(<FCVT_FIXED2F:fcvt_fixed_insn><VHSDI_SDI:mode>3): Likewise.
(aarch64_<maxmin_uns>p<mode>): Likewise.
(<su><maxmin><mode>3): Likewise.
(<maxmin_uns><mode>3): Likewise.
(<fmaxmin><mode>3): Likewise.
(aarch64_faddp<mode>): Likewise.
(aarch64_fmulx<mode>): Likewise.
(aarch64_frecps<mode>): Likewise.
(*aarch64_fac<optab><mode>): Rename to aarch64_fac<optab><mode>.
(add<mode>3): Extend to HF modes.
(sub<mode>3): Likewise.
(mul<mode>3): Likewise.
(div<mode>3): Likewise.
(*div<mode>3): Likewise.
* config/aarch64/aarch64.c (aarch64_emit_approx_div): Return false for
HF, V4HF and V8HF.
* config/aarch64/iterators.md (VDQ_HSDI, VSDQ_HSDI): New mode iterator.
* config/aarch64/arm_neon.h (vadd_f16): New.
(vaddq_f16, vabd_f16, vabdq_f16, vcage_f16, vcageq_f16, vcagt_f16,
vcagtq_f16, vcale_f16, vcaleq_f16, vcalt_f16, vcaltq_f16, vceq_f16,
vceqq_f16, vcge_f16, vcgeq_f16, vcgt_f16, vcgtq_f16, vcle_f16,
vcleq_f16, vclt_f16, vcltq_f16, vcvt_n_f16_s16, vcvtq_n_f16_s16,
vcvt_n_f16_u16, vcvtq_n_f16_u16, vcvt_n_s16_f16, vcvtq_n_s16_f16,
vcvt_n_u16_f16, vcvtq_n_u16_f16, vdiv_f16, vdivq_f16, vdup_lane_f16,
vdup_laneq_f16, vdupq_lane_f16, vdupq_laneq_f16, vdups_lane_f16,
vdups_laneq_f16, vmax_f16, vmaxq_f16, vmaxnm_f16, vmaxnmq_f16, vmin_f16,
vminq_f16, vminnm_f16, vminnmq_f16, vmul_f16, vmulq_f16, vmulx_f16,
vmulxq_f16, vpadd_f16, vpaddq_f16, vpmax_f16, vpmaxq_f16, vpmaxnm_f16,
vpmaxnmq_f16, vpmin_f16, vpminq_f16, vpminnm_f16, vpminnmq_f16,
vrecps_f16, vrecpsq_f16, vrsqrts_f16, vrsqrtsq_f16, vsub_f16,
vsubq_f16): Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238717 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][2/10] ARMv8.2-A FP16 one operand vector intrinsics

gcc/
* config/aarch64/aarch64-builtins.c (TYPES_BINOP_USS): New.
* config/aarch64/aarch64-simd-builtins.def: Register new builtins.
* config/aarch64/aarch64-simd.md (aarch64_rsqrte<mode>): Extend to HF modes.
(neg<mode>2): Likewise.
(abs<mode>2): Likewise.
(<frint_pattern><mode>2): Likewise.
(l<fcvt_pattern><su_optab><VDQF:mode><fcvt_target>2): Likewise.
(<optab><VDQF:mode><fcvt_target>2): Likewise.
(<fix_trunc_optab><VDQF:mode><fcvt_target>2): Likewise.
(ftrunc<VDQF:mode>2): Likewise.
(<optab><fcvt_target><VDQF:mode>2): Likewise.
(sqrt<mode>2): Likewise.
(*sqrt<mode>2): Likewise.
(aarch64_frecpe<mode>): Likewise.
(aarch64_cm<optab><mode>): Likewise.
* config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Return
false for V4HF and V8HF.
* config/aarch64/iterators.md (VHSDF, VHSDF_DF, VHSDF_SDF): New.
(VDQF_COND, fcvt_target, FCVT_TARGET, hcon): Extend mode attribute to HF modes.
(stype): New.
* config/aarch64/arm_neon.h (vdup_n_f16): New.
(vdupq_n_f16): Likewise.
(vld1_dup_f16): Use vdup_n_f16.
(vld1q_dup_f16): Use vdupq_n_f16.
(vabs_f16): New.
(vabsq_f16, vceqz_f16, vceqzq_f16, vcgez_f16, vcgezq_f16, vcgtz_f16,
vcgtzq_f16, vclez_f16, vclezq_f16, vcltz_f16, vcltzq_f16, vcvt_f16_s16,
vcvtq_f16_s16, vcvt_f16_u16, vcvtq_f16_u16, vcvt_s16_f16, vcvtq_s16_f16,
vcvt_u16_f16, vcvtq_u16_f16, vcvta_s16_f16, vcvtaq_s16_f16,
vcvta_u16_f16, vcvtaq_u16_f16, vcvtm_s16_f16, vcvtmq_s16_f16,
vcvtm_u16_f16, vcvtmq_u16_f16, vcvtn_s16_f16, vcvtnq_s16_f16,
vcvtn_u16_f16, vcvtnq_u16_f16, vcvtp_s16_f16, vcvtpq_s16_f16,
vcvtp_u16_f16, vcvtpq_u16_f16, vneg_f16, vnegq_f16, vrecpe_f16,
vrecpeq_f16, vrnd_f16, vrndq_f16, vrnda_f16, vrndaq_f16, vrndi_f16,
vrndiq_f16, vrndm_f16, vrndmq_f16, vrndn_f16, vrndnq_f16, vrndp_f16,
vrndpq_f16, vrndx_f16, vrndxq_f16, vrsqrte_f16, vrsqrteq_f16, vsqrt_f16,
vsqrtq_f16): Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238716 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][1/10] ARMv8.2-A FP16 data processing intrinsics

gcc/
* config/aarch64/aarch64-simd.md
(aarch64_<PERMUTE:perm_insn><PERMUTE:perm_hilo><mode>): Use VALL_F16.
(aarch64_ext<mode>): Likewise.
(aarch64_rev<REVERSE:rev_op><mode>): Likewise.
* config/aarch64/aarch64.c (aarch64_evpc_trn): Support V4HFmode and
V8HFmode.
(aarch64_evpc_uzp): Likewise.
(aarch64_evpc_zip): Likewise.
(aarch64_evpc_ext): Likewise.
(aarch64_evpc_rev): Likewise.
* config/aarch64/arm_neon.h (__aarch64_vdup_lane_f16): New.
(__aarch64_vdup_laneq_f16): New..
(__aarch64_vdupq_lane_f16): New.
(__aarch64_vdupq_laneq_f16): New.
(vbsl_f16): New.
(vbslq_f16): New.
(vdup_n_f16): New.
(vdupq_n_f16): New.
(vdup_lane_f16): New.
(vdup_laneq_f16): New.
(vdupq_lane_f16): New.
(vdupq_laneq_f16): New.
(vduph_lane_f16): New.
(vduph_laneq_f16): New.
(vext_f16): New.
(vextq_f16): New.
(vmov_n_f16): New.
(vmovq_n_f16): New.
(vrev64_f16): New.
(vrev64q_f16): New.
(vtrn1_f16): New.
(vtrn1q_f16): New.
(vtrn2_f16): New.
(vtrn2q_f16): New.
(vtrn_f16): New.
(vtrnq_f16): New.
(__INTERLEAVE_LIST): Support float16x4_t, float16x8_t.
(vuzp1_f16): New.
(vuzp1q_f16): New.
(vuzp2_f16): New.
(vuzp2q_f16): New.
(vzip1_f16): New.
(vzip2q_f16): New.
(vmov_n_f16): Reimplement using vdup_n_f16.
(vmovq_n_f16): Reimplement using vdupq_n_f16..

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238715 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][3/3] Migrate aarch64_expand_prologue/epilogue to aarch64_add_constant

gcc/
* config/aarch64/aarch64.c (aarch64_add_constant): New parameter
"frame_related_p".  Generate CFA annotation when it's necessary.
(aarch64_expand_prologue): Use aarch64_add_constant.
(aarch64_expand_epilogue): Likewise.
(aarch64_output_mi_thunk): Pass "false" when calling
aarch64_add_constant.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238714 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][2/3] Optimize aarch64_add_constant to generate better addition sequences

gcc/
* config/aarch64/aarch64.c (aarch64_add_constant): Optimize instruction
sequences.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238713 138bc75d-0d04-0410-961f-82ee72b054a4

[AArch64][1/3] Migrate aarch64_add_constant to new interface & kill aarch64_build_constant

gcc/
* config/aarch64/aarch64.c (aarch64_add_constant): New parameter "mode".
Use aarch64_internal_mov_immediate instead of aarch64_build_constant.
(aarch64_output_mi_thunk): Pass Pmode when calling aarch64_add_constant.
(aarch64_build_constant): Delete.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238712 138bc75d-0d04-0410-961f-82ee72b054a4

Fix missing qualification in <ext/rope>

2016-07-25  Georeth Chow  <georeth2010@gmail.com>

* include/ext/ropeimpl.h (rope<>::_S_dump(_RopeRep*, int)): Qualify
_S_concat enumerator.
* testsuite/ext/rope/6.cc: New test.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238711 138bc75d-0d04-0410-961f-82ee72b054a4

revert: nvptx: do not implicitly enable -ftoplevel-reorder

Revert
2016-07-20  Alexander Monakov  <amonakov@ispras.ru>

* config/nvptx/nvptx.c (nvptx_option_override): Do not set
flag_toplevel_reorder.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238710 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-25  Richard Biener  <rguenther@suse.de>

* cgraph.c (cgraph_node::verify_node): Compare against builtin
by using DECL_BUILT_IN_CLASS and DECL_FUNCTION_CODE.
* tree-chkp.c (chkp_gimple_call_builtin_p): Likewise.
* tree-streamer.h (streamer_handle_as_builtin_p): Remove.
(streamer_get_builtin_tree): Likewise.
(streamer_write_builtin): Likewise.
* lto-streamer.h (LTO_builtin_decl): Remove.
* lto-streamer-in.c (lto_read_tree_1): Remove assert.
(lto_input_scc): Remove LTO_builtin_decl handling.
(lto_input_tree_1): Liekwise.
* lto-streamer-out.c (lto_output_tree_1): Remove special
handling of builtins.
(DFS::DFS): Likewise.
* tree-streamer-in.c (streamer_get_builtin_tree): Remove.
* tree-streamer-out.c (pack_ts_function_decl_value_fields): Remove
assert.
(streamer_write_builtin): Remove.

lto/
* lto.c (compare_tree_sccs_1): Remove streamer_handle_as_builtin_p uses.
(unify_scc): Likewise.
(lto_read_decls): Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238709 138bc75d-0d04-0410-961f-82ee72b054a4

Fix tests for targets with sizeof(int) != 32.

gcc/testsuite/

* gcc.dg/torture/pr69352.c (foo): Cast to intptr_t instead of long.
* gcc.dg/torture/pr69771.c: Require int32plus.
* gcc.dg/torture/pr71866.c (inb): Add cast to intptr_t.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238708 138bc75d-0d04-0410-961f-82ee72b054a4

Don't call get_working_sets w/ LTO and -fauto-profile (PR

* lto-cgraph.c (input_symtab): Don't call get_working_sets
if flag_auto_profile is set to true.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238707 138bc75d-0d04-0410-961f-82ee72b054a4

Handle loops with loop->latch == NULL (PR gcov-profile/71868)

PR gcov-profile/71868
* cfgloopanal.c (expected_loop_iterations_unbounded): When we
have a function with multiple latches, count them all.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238706 138bc75d-0d04-0410-961f-82ee72b054a4

Fix memory leak introduced in r238336

* tree-ssa-loop-niter.c (loop_only_exit_p): Release body array.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238705 138bc75d-0d04-0410-961f-82ee72b054a4

Call get_ops just for SSA_NAMEs (PR tree-optimization/71987)

PR tree-optimization/71987
* tree-ssa-reassoc.c (maybe_optimize_range_tests): Call get_ops
just for SSA_NAMEs. Fix GNU coding style.
* gcc.dg/torture/pr71987.c: New test.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238704 138bc75d-0d04-0410-961f-82ee72b054a4

Use std::mt19937, std::thread and std::atomic to simplify tests

* testsuite/20_util/shared_ptr/thread/default_weaktoshared.cc: Use
std::mt19937, std::thread and std::atomic to simplify test.
* testsuite/20_util/shared_ptr/thread/mutex_weaktoshared.cc: Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238703 138bc75d-0d04-0410-961f-82ee72b054a4

Adapt the numbering scheme (PR gcov-profile/64874)

PR gcov-profile/64874
* gcov-io.h: Update command about file format.
* gcov-iov.c (main): Adapt the numbering scheme.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238702 138bc75d-0d04-0410-961f-82ee72b054a4

Revert 2015-11-09 sanitizer/obstack configury

The 2015-11-23 sanitizer merge from upstream lost the changes from
f6528435 to sanitizer_common/sanitizer_common_interceptors.inc, which
made use of _OBSTACK_SIZE_T.  So the configury changes to define
_OBSTACK_SIZE_T don't do anything.  This wasn't such a bad thing
anyway..  The configure test wrongly adds -I${srcdir}/../include,
effectively resulting in a test of libiberty/obstack rather than libc
obstack support, and it's the latter that asan and tsan need to work
with.

So, remove the useless configure test.  Upstream santizer project has
been made aware of the problem if glibc obstack support is ever
updated.  Bootsrapped etc. x86_64-linux and committed as obvious.

Revert 2015-11-09  Alan Modra  <amodra@gmail.com>
* configure.ac: Don't substitute OBSTACK_DEFS.
* asan/Makefile.am: Remove OBSTACK_DEFS from DEFS.
* tsan/Makefile.am: Likewise.
* configure: Regenerate.
* Makefile.in: Regenerate.
* asan/Makefile.in: Regenerate.
* interception/Makefile.in: Regenerate.
* libbacktrace/Makefile.in: Regenerate.
* lsan/Makefile.in: Regenerate.
* sanitizer_common/Makefile.in: Regenerate.
* tsan/Makefile.in: Regenerate.
* ubsan/Makefile.in: Regenerate.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238701 138bc75d-0d04-0410-961f-82ee72b054a4

Daily bump.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238700 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/71515 - typename in partial specialization

* pt.c (resolve_typename_type): Try to avoid calling
currently_open_class.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238696 138bc75d-0d04-0410-961f-82ee72b054a4

gcc/ChangeLog:

2016-07-24  Kugan Vivekanandarajah  <kuganv@linaro.org>

PR middle-end/66726
* tree-ssa-reassoc.c (optimize_vec_cond_expr): Handle tcc_compare stmt
whose result is used in PHI.
(final_range_test_p): Likewise.
(maybe_optimize_range_tests): Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238695 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/66617 - virtual base list-initialization

* call.c (add_list_candidates): Handle VTT parm.
(build_new_method_call_1): Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238689 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/55922 - list-value-initialization of base

PR c++/63151
* init.c (expand_aggr_init_1): Handle list-initialization from {}.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238688 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/70709 - zero-length array member

* class.c (walk_subobject_offsets): Handle 0-length array.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238687 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/70778 - member template template parameter

* pt.c (tsubst): Also substitute into the template of a
BOUND_TEMPLATE_TEMPLATE_PARM.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238686 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/71738 - nested template friend

* pt.c (lookup_template_class_1): Handle getting template from tsubst.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238685 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/71350 - error recursion with initializer-list

* decl.c (reshape_init_r): Check complain for missing braces warning.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238684 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/71576 - bitfield and rvalue reference

* call.c (convert_like_real): Use lvalue_kind.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238683 138bc75d-0d04-0410-961f-82ee72b054a4

PR c++/71748 - call to base destructor in template.

PR c++/52746
* pt.c (tsubst_baselink): Call
adjust_result_of_qualified_name_lookup for unqualified
destructors.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238681 138bc75d-0d04-0410-961f-82ee72b054a4

Daily bump.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238680 138bc75d-0d04-0410-961f-82ee72b054a4

2016-07-22  Michael Meissner  <meissner@linux.vnet.ibm.com>

* config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
Reformat two multi-line strings.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@238675 138bc75d-0d04-0410-961f-82ee72b054a4