PR tree-optimization/69207
* tree-vect-slp.c (vect_get_constant_vectors): For
VECTOR_BOOLEAN_TYPE_P, assert op has integral type instead of
fold_convertible_p to vector_type's element type, and always
use VCE for non-VECTOR_BOOLEAN_TYPE_P.
From-SVN: r232233
2016-01-11 Richard Biener <rguenther@suse.de>
PR tree-optimization/69173
* tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Only
fixup the cycle if all stmts are in a pattern.
* gcc.dg/torture/pr69173.c: New testcase.
From-SVN: r232230
PR middle-end/68999
* alias.c (base_alias_check): Move check for addresses with
alignment ANDs before the call for compare_base_decls.
(memrefs_conflict_p): Return -1 for different decls
that went through alignment adjustments.
From-SVN: r232229
PR rtl-optimization/68796
* config/aarch64/aarch64.md (*and<mode>_compare0): New pattern.
* config/aarch64/aarch64.c (aarch64_select_cc_mode): Handle HImode
and QImode comparisons against zero with CC_NZmode.
* config/aarch64/iterators.md (short_mask): New mode_attr.
* gcc.target/aarch64/tst_5.c: New test.
* gcc.target/aarch64/tst_6.c: Likewise.
From-SVN: r232228
When link_ptx runs, a CUDA device is already bound to current thread, so the
driver library knows the target architecture. There isn't any benefit from
forcing a specific target here; on the contrary, hardcoding sm_30 breaks
offloading on later (Maxwell, sm_5x) devices.
* plugin/plugin-nvptx.c (link_ptx): Do not set CU_JIT_TARGET.
From-SVN: r232227
PR rtl-optimization/68841
* ifcvt.c (struct noce_if_info): Add orig_x field.
(bbs_ok_for_cmove_arith): Add to_rename parameter.
Don't record conflicts on to_rename if it's present.
Allow memory destinations in sets.
(noce_try_cmove_arith): Call bbs_ok_for_cmove_arith even on simple
blocks, passing orig_x to the checks.
(noce_process_if_block): Set if_info->orig_x appropriately.
* gcc.dg/pr68841.c: New test.
* gcc.c-torture/execute/pr68841.c: New test.
From-SVN: r232223
gcc/
2016-01-11 Yuri Rumyantsev <ysrumyan@gmail.com>
PR rtl-optimization/68920
* config/i386/i386.c (ix86_option_override_internal): Restrict number
of conditional moves for RTL if-conversion to 1 for
TARGET_ONE_IF_CONV_INSN.
* config/i386/i386.h (TARGET_ONE_IF_CONV_INSN): New macros.
* config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): New macros.
* params.def (PARAM_MAX_RTL_IF_CONVERSION_INSNS) : Introduce new
parameter to restirct number of conditional moves for
RTL if-conversion.
* doc/invoke.texi (max-rtl-if-conversion-insns): Document it.
* ifcvt.c (bb_ok_for_noce_convert_multiple_sets): Limit number of
conditionl moves.
gcc/testsuite/
2016-01-11 Yuri Rumyantsev <ysrumyan@gmail.com>
PR rtl-optimization/68920
* gcc.dg/ifcvt-4.c: Add "--param max-rtl-if-conversion-insns=3" option
for ix86 targets.
* gcc.dg/ifcvt-5.c: New test.
From-SVN: r232220
The problem arises because we used to drop overwritten MEMs from loc
lists of VALUEs, but not of other onepart variables, and it just so
happens that, by doing so, block 6 in the testcase has no D#5 in its
output in the first pass, because the MEM holding its (previous) value
was correctly dropped from value 88:88, but gains it in the second
pass because D#5 has the MEM location incoming directly in its loc
list, rather than indirectly in a VALUE.
This incorrect binding enables other blocks to believe they have a
tentative binding for D#5 in some cycles, but others, still operating
on the early conclusion, believe there isn't, and they oscillate from
that.
Since we check for escaping MEMs in clobbers, we won't lose anything
relevant by dropping call-clobbered or overwritten MEMs in all onepart
variables, and this ensures the loc intersection operation in onepart
vars won't let a MEM through that wasn't present in earlier
iterations.
for gcc/ChangeLog
PR bootstrap/69123
* var-tracking.c (drop_overlapping_mem_locs): Operate on all
onepart vars. Fix typo in comment. Fix reversed condition in
unshare test.
(dataflow_set_remove_mem_locs): Operate on all onepart vars.
for gcc/testsuite/ChangeLog
PR bootstrap/69123
* g++.dg/pr69123.C: New.
From-SVN: r232218
for gcc/ChangeLog
PR bootstrap/69123
* var-tracking.c (dump_onepart_variable_differences): New.
(dataflow_set_different): If a detailed dump is requested,
delay early returns and dump differences between onepart
variables present before and after, and added variables.
From-SVN: r232217
gcc/
PR target/69010
* expr.c (expand_expr_real_1): For boolean vector constants
with a scalar mode use const_scalar_mask_from_tree.
(const_scalar_mask_from_tree): New.
* optabs.c (expand_vec_cond_mask_expr): Use mask mode
assigned to a mask type to handle constants.
gcc/testsuite/
PR target/69010
* gcc.target/i386/pr69010.c: New test.
From-SVN: r232216
2016-01-11 Martin Jambor <mjambor@suse.cz>
PR ipa/69044
* ipa-cp.c (estimate_local_effects): Do not clone for removal of
useless parameters if we cannot change function signature.
testsuite/
* gcc.target/i386/chkp-pr69044.c: New test.
From-SVN: r232215
2016-01-11 Tom de Vries <tom@codesourcery.com>
PR tree-optimization/69109
* tree-parloops.c (try_transform_to_exit_first_loop_alt): Don't allow
latch with phi.
* gcc.dg/autopar/pr69109-2.c: New test.
* gcc.dg/autopar/pr69109.c: New test.
From-SVN: r232212
2016-01-11 Tom de Vries <tom@codesourcery.com>
PR tree-optimization/69108
* tree-parloops.c (gather_scalar_reductions): Handle case that outer phi
res is not used in a phi.
* gcc.dg/autopar/pr69108.c: New test.
From-SVN: r232211
2016-01-11 Tom de Vries <tom@codesourcery.com>
PR tree-optimization/69058
* tree-parloops.c (pass_parallelize_loops::execute): Return 0 if libgomp
not supported.
From-SVN: r232208
2016-01-11 Andrew Burgess <andrew.burgess@embecosm.com>
* config/arc/arc.opt (mdiv-rem): Add period to the end.
(mcode-density): Likewise.
From-SVN: r232207
2016-01-10 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/69154
* frontend-passes.c (in_where): New variable.
(inline_matmul_assign): Don't try this if we are within
a WHERE statement.
(gfc_code_walker): Keep track of in_where.
2016-01-10 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/69154
* gfortran.dg/inline_matmul_12.f90: New test.
From-SVN: r232201
2016-01-10 Paul Thomas <pault@gcc.gnu.org>
PR fortran/67779
* trans_array.c (gfc_conv_scalarized_array_ref): Add missing
se->use_offset from condition for calculation of 'base'.
2016-01-10 Paul Thomas <pault@gcc.gnu.org>
PR fortran/67779
* gfortran.dg/actual_array_offset_1: New test.
From-SVN: r232200
2016-01-10 Tom de Vries <tom@codesourcery.com>
PR tree-optimization/69062
* tree-parloops.c (loop_has_phi_with_address_arg): New function.
(parallelize_loops): Don't paralelize loop that has phi with address
arg.
* gcc.dg/autopar/pr69062.c: New test.
From-SVN: r232199
2016-01-10 Tom de Vries <tom@codesourcery.com>
PR tree-optimization/69039
* tree-parloops.c (try_create_reduction_list): Only allow single exit
phi for reduction.
* gcc.dg/autopar/pr69039.c: New test.
From-SVN: r232196
PR middle-end/68743
* match.pd: Require target has function_c99_misc before doing
truncl(extend(x)) and trunc(extend(x)) -> extend(truncf(x)), etc.
From-SVN: r232191
PR middle-end/50865
PR tree-optimization/69097
* fold-const.h (expr_not_equal_to): New prototype.
* fold-const.c: Include stringpool.h and tree-ssanames.h.
(expr_not_equal_to): New function.
* match.pd (X % -Y is the same as X % Y): Don't optimize
unless X is known not to be equal to minimum or Y is known
not to be equal to -1.
* tree-vrp.c (simplify_div_or_mod_using_ranges): Add GSI argument.
fold TRUNC_MOD_EXPR if the second argument is not a power of two.
(simplify_stmt_using_ranges): Adjust caller.
(vrp_finalize): Call set_value_range on SSA_NAMEs before calling
substitute_and_fold.
* gcc.c-torture/execute/pr50865.c: New test.
* gcc.c-torture/execute/pr69097-1.c: New test.
* gcc.c-torture/execute/pr69097-2.c: New test.
* gcc.dg/pr69097-1.c: New test.
* gcc.dg/pr69097-2.c: New test.
From-SVN: r232188
PR tree-optimization/69167
* gimple-fold.c (replace_stmt_with_simplification): Also punt if
new SSA_NAME_OCCURS_IN_ABNORMAL_PHI SSA_NAMEs appear in operands of
ops[0] comparison.
* gimple-match-head.c (maybe_push_res_to_seq): Likewise.
* gcc.dg/pr69167.c: New test.
From-SVN: r232178
gcc/:
* tree-vect-slp.c (vect_analyze_slp_instance): Cancel permuted SLP
instances that can be handled via vect_load_lanes.
gcc/testsuite/:
* lib/target-supports.exp (check_effective_target_vect_load_lanes): New.
* gcc.dg/vect/slp-perm-1.c: Look for vect_load_lanes instead of SLP
on platforms supporting it.
* gcc.dg/vect/slp-perm-2.c: Likewise.
* gcc.dg/vect/slp-perm-3.c: Likewise.
* gcc.dg/vect/slp-perm-5.c: Likewise.
* gcc.dg/vect/slp-perm-7.c: Likewise.
* gcc.dg/vect/slp-perm-8.c: Likewise.
* gcc.dg/vect/slp-perm-6.c: Look for vect_load_lanes in addition to SLP
on platforms supporting it.
Co-Authored-By: Richard Biener <rguenther@suse.de>
From-SVN: r232174
* symtab.c (symtab_node::equal_address_to): Return -1 instead of 2
if we can't determine address equivalence.
* alias.c (compare_base_decl): Update for changed return value of
symtab_node::equal_address_to.
From-SVN: r232172
PR tree-optimization/69162
* gimplify.c (gimplify_va_arg_expr): Encode original type of
valist argument in another argument.
(gimplify_modify_expr): Adjust for the above change. Cleanup.
* tree-stdarg.c (expand_ifn_va_arg_1): Use new 3rd argument
to determine the va_list type, build a MEM_REF instead of
build_fold_indirect_ref.
* gcc.dg/pr69162.c: New test.
From-SVN: r232156
