alahay01 [Fri, 3 Jun 2016 13:00:06 +0000 (13:00 +0000)]
2016-06-03 Alan Hayward <alan.hayward@arm.com>
[2/3] Vectorize inductions that are live after the loop
gcc/
* tree-vect-loop.c (vect_analyze_loop_operations): Allow live stmts.
(vectorizable_reduction): Check for new relevant state.
(vectorizable_live_operation): vectorize live stmts using
BIT_FIELD_REF. Remove special case for gimple assigns stmts.
* tree-vect-stmts.c (is_simple_and_all_uses_invariant): New function.
(vect_stmt_relevant_p): Check for stmts which are only used live.
(process_use): Use of a stmt does not inherit it's live value.
(vect_mark_stmts_to_be_vectorized): Simplify relevance inheritance.
(vect_analyze_stmt): Check for new relevant state.
* tree-vectorizer.h (vect_relevant): New entry for a stmt which is used
outside the loop, but not inside it.
testsuite/
* gcc.dg/tree-ssa/pr64183.c: Ensure test does not vectorize.
* testsuite/gcc.dg/vect/no-scevccp-vect-iv-2.c: Remove xfail.
* gcc.dg/vect/vect-live-1.c: New test.
* gcc.dg/vect/vect-live-2.c: New test.
* gcc.dg/vect/vect-live-3.c: New test.
* gcc.dg/vect/vect-live-4.c: New test.
* gcc.dg/vect/vect-live-5.c: New test.
* gcc.dg/vect/vect-live-slp-1.c: New test.
* gcc.dg/vect/vect-live-slp-2.c: New test.
* gcc.dg/vect/vect-live-slp-3.c: New test.
alahay01 [Fri, 3 Jun 2016 12:48:21 +0000 (12:48 +0000)]
2016-06-03 Alan Hayward <alan.hayward@arm.com>
[1/3] Split vect_get_vec_def_for_operand into two
gcc/
* tree-vectorizer.h (vect_get_vec_def_for_operand_1): New
* tree-vect-stmts.c (vect_get_vec_def_for_operand_1): New
(vect_get_vec_def_for_operand): Split out code.
segher [Fri, 3 Jun 2016 12:33:07 +0000 (12:33 +0000)]
rs6000: Remove the ancient mfcr peepholes
These peepholes replace two mfcr;mask sequences by one mfcr;mask;mask
sequence. On modern cpus, the original mfcr's were actually mfocrf,
but the new insn is an actual heavy-weight mfcr. This is very bad
for performance.
The comment says there is a three cycle delay between two consecutive
mfcr insns. This may have been true on rios, and it's true on 604,
but on 603, 750, 7400 it is just a single cycle (on 7450 it is two).
This is also a define_peephole, and we should get rid of those.
So this patch just removes the peepholes; the benefit is marginal at
best, and it so very hurts in other cases.
* config/rs6000/rs6000.md (define_peepholes for two mfcr's): Delete.
jakub [Fri, 3 Jun 2016 08:03:11 +0000 (08:03 +0000)]
PR middle-end/71387
* cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): If redirecting
to noreturn e->callee->decl that has void return type and void
arguments, adjust gimple_call_fntype and remove lhs even if it had
previously addressable type.
law [Fri, 3 Jun 2016 05:20:16 +0000 (05:20 +0000)]
PR tree-optimization/71328
* tree-ssa-threadupdate.c (duplicate_thread_path): Fix off-by-one
error when checking for a jump back onto the copied path. */
PR tree-optimization/71328
* gcc.c-torture/compile/pr71328.c: New test.
jakub [Thu, 2 Jun 2016 16:36:04 +0000 (16:36 +0000)]
PR c++/71372
* cp-gimplify.c (cp_fold): For INDIRECT_REF, if the folded expression
is INDIRECT_REF or MEM_REF, copy over TREE_READONLY, TREE_SIDE_EFFECTS
and TREE_THIS_VOLATILE flags. For ARRAY_REF and ARRAY_RANGE_REF, copy
over TREE_READONLY, TREE_SIDE_EFFECTS and TREE_THIS_VOLATILE flags
to the newly built tree.
hjl [Thu, 2 Jun 2016 13:46:20 +0000 (13:46 +0000)]
Update TARGET_FUNCTION_INCOMING_ARG documentation
On x86, interrupt handlers are only called by processors which push
interrupt data onto stack at the address where the normal return address
is. Since interrupt handlers must access interrupt data via pointers so
that they can update interrupt data, the pointer argument is passed as
"argument pointer - word".
TARGET_FUNCTION_INCOMING_ARG defines how callee sees its argument.
Normally it returns REG, NULL, or CONST_INT. This patch adds arbitrary
address computation based on hard register, which can be forced into a
register, to the list.
When copying an incoming argument onto stack, assign_parm_setup_stack
has:
if (argument in memory)
copy argument in memory to stack
else
move argument to stack
Since an arbitrary address computation may be passed as an argument, we
change it to:
if (argument in memory)
copy argument in memory to stack
else
{
if (argument isn't in register)
force argument into a register
move argument to stack
}
* function.c (assign_parm_setup_stack): Force source into a
register if needed.
* target.def (function_incoming_arg): Update documentation to
allow arbitrary address computation based on hard register.
* doc/tm.texi: Regenerated.
marxin [Thu, 2 Jun 2016 12:51:57 +0000 (12:51 +0000)]
Fix first match heuristics
* predict.c (combine_predictions_for_bb): Fix first match in
cases where a first predictor contains more than one occurence
in list of predictors. Take the best value in such case.
* gcc.dg/predict-9.c: New test.
ktkachov [Thu, 2 Jun 2016 12:26:42 +0000 (12:26 +0000)]
[rtlanal] Fix rtl-optimization/71295
PR rtl-optimization/71295
* rtlanal.c (subreg_get_info): If taking a subreg at the requested
offset would go over the size of the inner mode reject it.
nathan [Thu, 2 Jun 2016 12:22:01 +0000 (12:22 +0000)]
On behalf of Aaron Conole <aconole@redhat.com>
* libgcov-driver-system.c (__gcov_error_file): New.
(get_gcov_error_file): New.
(gcov_error): Use and set __gcov_error_file.
(gcov_error_exit): New.
* libgcov-driver.c (gcov_exit): Call gcov_error_exit.
jakub [Thu, 2 Jun 2016 09:43:52 +0000 (09:43 +0000)]
* config/i386/sse.md (*vec_concatv2di): Add x86_avx512dq v=Yv,rm
alternative. Change x=xm,C alternative to v=vm,C, x=x,x alternative
to v=Yv,Yv and x=x,m to v=v,m. Use maybe_evex prefix attribute
instead of vex for the last two above mentioned alternatives.
* gcc.target/i386/avx512dq-concatv2di-1.c: New test.
* gcc.target/i386/avx512vl-concatv2di-1.c: New test.
* gcc.target/i386/sse2-init-v2di-2.c: Adjust expected vec_concatv2di
alternative number.
meissner [Wed, 1 Jun 2016 22:24:29 +0000 (22:24 +0000)]
Fix ChangeLog entry:
2016-05-24 Michael Meissner <meissner@linux.vnet.ibm.com>
* config/rs6000/altivec.md (VNEG iterator): New iterator for
VNEGW/VNEGD instructions.
(p9_neg<mode>2): New insns for ISA 3.0 VNEGW/VNEGD.
(neg<mode>2): Add expander for V2DImode added in ISA 2.07, and
support for ISA 3.0 VNEGW/VNEGD instructions.
[gcc]
2016-05-31 Michael Meissner <meissner@linux.vnet.ibm.com>
* config/rs6000/vsx.md (vsx_splat_<mode>, V2DI/V2DF): Simplify
alternatives, eliminating preferred register class. Add support
for the MTVSRDD instruction in ISA 3.0.
(vsx_splat_v4si_internal): Use splat_input_operand instead of
reg_or_indexed_operand.
(vsx_splat_v4sf_internal): Likewise.
[gcc/testsuite]
2016-05-31 Michael Meissner <meissner@linux.vnet.ibm.com>
jvdelisle [Wed, 1 Jun 2016 17:06:50 +0000 (17:06 +0000)]
2016-06-01 Jerry DeLisle <jvdelisle@gcc.gnu.org>
PR fortran/52393
* io.c (match_io): For READ, try to match a default character
expression. If found, set the dt format expression to this,
otherwise go back and try control list.
PR fortran/52393
* gfortran.dg/fmt_read_3.f90: New test.
doko [Wed, 1 Jun 2016 16:42:41 +0000 (16:42 +0000)]
gcc/c-family/ChangeLog:
2016-05-13 Eduard Sanou <dhole@openmailbox.org>
* c-common.c (get_source_date_epoch): Rename to
cb_get_source_date_epoch.
* c-common.c (cb_get_source_date_epoch): Use a single generic erorr
message when the parsing fails. Use error_at instead of fatal_error.
* c-common.h (get_source_date_epoch): Rename to
cb_get_source_date_epoch.
* c-common.h (cb_get_source_date_epoch): Prototype.
* c-common.h (MAX_SOURCE_DATE_EPOCH): Define.
* c-common.h (c_omp_region_type): Remove trailing comma.
* c-lex.c (init_c_lex): Set cb->get_source_date_epoch callback.
* c-lex.c (c_lex_with_flags): Remove initialization of
pfile->source_date_epoch.
gcc/ChangeLog:
2016-05-13 Eduard Sanou <dhole@openmailbox.org>
* doc/cppenv.texi: Note that the `%s` in `date` is a non-standard
extension.
* gcc.c (driver_handle_option): Call set_source_date_epoch_envvar.
* gcc.c (set_source_date_epoch_envvar): New function, sets
the SOURCE_DATE_EPOCH environment variable to the current time.
gcc/testsuite/ChangeLog:
2016-05-13 Eduard Sanou <dhole@openmailbox.org>
* gcc.dg/cpp/source_date_epoch-1.c: New file, test the proper
behaviour of the macros __DATE__ and __TIME__ when SOURCE_DATE_EPOCH
env var is set.
* gcc.dg/cpp/source_date_epoch-2.c: New file, test the error output
when parsing the SOURCE_DATE_EPOCH env var, and make sure it is only
shown once.
* lib/gcc-dg.exp (dg-set-compiler-env-var): New function, set env vars
during compilation.
* lib/gcc-dg.exp (restore-compiler-env-var): New function, restore env
vars set by dg-set-compiler-env-var.
libcpp/ChangeLog:
2016-05-13 Eduard Sanou <dhole@openmailbox.org>
* include/cpplib.h (cpp_callbacks): Add get_source_date_epoch
callback.
* include/cpplib.h (cpp_init_source_date_epoch): Remove prototype.
* init.c (cpp_init_source_date_epoch): Remove function.
* init.c (cpp_create_reader): Initialize pfile->source_date_epoch.
* internal.h (cpp_reader): Extend comment about source_date_epoch.
* macro.c (_cpp_builtin_macro_text): Use get_source_date_epoch
callback only once, read pfile->source_date_epoch on future passes.
Check that get_source_date_epoch callback is not NULL.
pault [Wed, 1 Jun 2016 14:30:00 +0000 (14:30 +0000)]
2016-06-01 Paul Thomas <pault@gcc.gnu.org>
PR fortran/71156
* decl.c (copy_prefix): Add checks that the module procedure
declaration prefixes are compliant with the interface. Invert
order of existing elemental and pure checks.
* resolve.c (resolve_fl_procedure): Invert order of elemental
and pure errors.
2016-06-01 Paul Thomas <pault@gcc.gnu.org>
PR fortran/71156
* gfortran.dg/submodule_14.f08: Add missing recursive prefix
to the module procedure declaration.
* gfortran.dg/submodule_16.f08: New test.
rguenth [Wed, 1 Jun 2016 14:05:22 +0000 (14:05 +0000)]
2016-06-01 Richard Biener <rguenther@suse.de>
PR tree-optimization/71366
* tree-ssa-loop-ivcanon.c (edges_to_remove): New global.
(unloop_loops): Move removing edges here ...
(try_unroll_loop_completely): ... from here.
(try_peel_loop): ... and here.
(tree_unroll_loops_completely_1): Track parent loops via
bitmap of header BBs.
(tree_unroll_loops_completely): Adjust for that.
* gcc.dg/torture/pr71366-1.c: New testcase.
* gcc.dg/torture/pr71366-2.c: Likewise.
kelvin [Wed, 1 Jun 2016 14:01:10 +0000 (14:01 +0000)]
gcc/ChangeLog:
2016-06-01 Kelvin Nilsen <kelvin@gcc.gnu.org>
* config/rs6000/altivec.h (vec_slv): New macro.
(vec_srv): New macro.
* config/rs6000/altivec.md (UNSPEC_VSLV): New value.
(UNSPEC_VSRV): New value.
(vslv): New insn.
(vsrv): New insn.
* config/rs6000/rs6000-builtin.def (vslv): New builtin definition.
(vsrv): New builtin definition.
* config/rs6000/rs6000-c.c (P9V_BUILTIN_VSLV): Macro expansion to
define argument types for new builtin.
(P9V_BUILTIN_VSRV): Macro expansion to define argument types for
new builtin.
* doc/extend.texi: Document the new vec_vslv and vec_srv built-in
functions.
gcc/testsuite/ChangeLog:
2016-06-01 Kelvin Nilsen <kelvin@gcc.gnu.org>
* gcc.target/powerpc/vslv-0.c: New test.
* gcc.target/powerpc/vslv-1.c: New test.
* gcc.target/powerpc/vsrv-0.c: New test.
* gcc.target/powerpc/vsrv-1.c: New test.
rguenth [Wed, 1 Jun 2016 13:08:24 +0000 (13:08 +0000)]
2016-06-01 Richard Biener <rguenther@suse.de>
PR tree-optimization/71261
* tree-vect-patterns.c (check_bool_pattern): Gather a hash-set
of stmts successfully put in the bool pattern. Remove
single-use restriction.
(adjust_bool_pattern_cast): Add cast at the use site via the
pattern def sequence.
(adjust_bool_pattern): Remove recursion, maintain a hash-map
of patterned defs. Use the pattern def seqence instead of
multiple independent patterns.
(sort_after_uid): New qsort compare function.
(adjust_bool_stmts): New function to process stmts in the bool
pattern in IL order.
(vect_recog_bool_pattern): Adjust.
* tree-if-conv.c (ifcvt_split_def_stmt): Remove.
(ifcvt_walk_pattern_tree): Likewise.
(stmt_is_root_of_bool_pattern): Likewise.
(ifcvt_repair_bool_pattern): Likewise.
(tree_if_conversion): Do not call ifcvt_repair_bool_pattern.
jakub [Wed, 1 Jun 2016 12:21:38 +0000 (12:21 +0000)]
* parse.c (case_decl): Move ST_OMP_* to ...
(case_omp_decl): ... here, new macro.
(verify_st_order): For case_omp_decl, complain about
p->state >= ORDER_EXEC, but don't change p->state otherwise.
* gfortran.dg/gomp/order-1.f90: New test.
* gfortran.dg/gomp/order-2.f90: New test.
ktkachov [Wed, 1 Jun 2016 10:44:07 +0000 (10:44 +0000)]
[ARM] Use proper output modifier for DImode register in store exclusive patterns
* config/arm/sync.md (arm_store_exclusive<mode>):
Use 'H' output modifier on operands[2] rather than creating a new
entry in out-of-bounds memory of the operands array.
(arm_store_release_exclusivedi): Likewise.
rguenth [Wed, 1 Jun 2016 09:00:57 +0000 (09:00 +0000)]
2016-06-01 Richard Biener <rguenther@suse.de>
* genmatch.c (comparison_code_p): New predicate.
(swap_tree_comparison): New function.
(commutate): Add for_vec parameter to append new for entries.
Support commutating relational operators by swapping it alongside
operands.
(lower_commutative): Adjust.
(dt_simplify::gen): Do not pass artificial operators to gen
functions.
(decision_tree::gen): Do not add artificial operators as parameters.
(parser::parse_expr): Verify operator commutativity when :c is
applied. Allow :C to override this.
* match.pd: Adjust patterns to use :C instead of :c where required.
ppalka [Wed, 1 Jun 2016 02:36:27 +0000 (02:36 +0000)]
Fix PR tree-optimization/71077
gcc/ChangeLog:
PR tree-optimization/71077
* tree-ssa-threadedge.c (simplify_control_stmt_condition_1): In
the combining step, use boolean_false_node and boolean_true_node
as the designated false/true return values.
gcc/testsuite/ChangeLog:
PR tree-optimization/71077
* gcc.dg/tree-ssa/pr71077.c: New test.
hubicka [Tue, 31 May 2016 23:33:08 +0000 (23:33 +0000)]
* g++.d/predict-lop-exit-1.C: Update template for new predictor name.
* g++.d/predict-lop-exit-2.C: Update template for new predictor name.
* g++.d/predict-lop-exit-2.C: Update template for new predictor name.
* predict.def (PRED_LOOP_EXTRA_EXIT): Define.
* predict.c (predict_iv_comparison): Also check PRED_LOOP_EXTRA_EXIT.
(predict_extra_loop_exits): Use PRED_LOOP_EXTRA_EXIT instead of
PRED_LOOP_EXIT.
hubicka [Tue, 31 May 2016 23:23:40 +0000 (23:23 +0000)]
* doc/invoke.texi (-frename-registers): Drop -fpeel-loops from list
of flags impliying the register renaming.
* toplev.c (process_options): Do not imply flag_rename_registers with
loop peeling.
olegendo [Tue, 31 May 2016 15:06:25 +0000 (15:06 +0000)]
Fix wrong-code issues of RX atomic operations.
gcc/
* config/rx/rx.md (FETCHOP_NO_MINUS): New code iterator.
(atomic_<fetchop_name>_fetchsi): Extract minus operator into ...
(atomic_sub_fetchsi): ... this new pattern.
(mvtc): Add CC_REG clobber.
sayle [Tue, 31 May 2016 11:30:56 +0000 (11:30 +0000)]
2016-05-31 Roger Sayle <roger@nextmovesoftware.com>
gcc/java:
* builtins.c (java_builtins): Use popcount* and bswap* builtins to
implement bitCount() and reverseBytes() methods in java.lang.Integer
and friends.
(initialize_builtins): Annotate math builtins with ECF_LEAF. Call
define_builtin for the new popcount* and bswap* builtins.
libjava:
* testsuite/libjava.lang/BuiltinBitCount.java: New test case.
* testsuite/libjava.lang/BuiltinReverseBytes.java: Likewise.
hubicka [Tue, 31 May 2016 10:42:59 +0000 (10:42 +0000)]
* loop-init.c (gate): Do not enale RTL loop unroller with -fpeel-loops.
It no longer does that.
* toplev.c (process_options): Do not enable flag_web with -fpeel-loops.
msebor [Mon, 30 May 2016 22:56:43 +0000 (22:56 +0000)]
PR c++/71306 - bogus -Wplacement-new with an array element
gcc/cp/ChangeLog:
2016-05-27 Martin Sebor <msebor@redhat.com>
PR c++/71306
* init.c (warn_placement_new_too_small): Handle placement new arguments
that are elements of arrays more carefully. Remove a pointless loop.
gcc/testsuite/ChangeLog:
2016-05-27 Martin Sebor <msebor@redhat.com>
PR c++/71306
* g++.dg/warn/Wplacement-new-size-3.C: New test.
jakub [Mon, 30 May 2016 21:36:24 +0000 (21:36 +0000)]
PR c++/71349
* c-parser.c (c_parser_omp_for): Don't disallow nowait clause
when combined with target construct.
* parser.c (cp_parser_omp_for): Don't disallow nowait clause
when combined with target construct.
(cp_parser_omp_parallel): Pass cclauses == NULL as last argument
to cp_parser_omp_all_clauses.
* c-omp.c (c_omp_split_clauses): Put OMP_CLAUSE_DEPEND to
C_OMP_CLAUSE_SPLIT_TARGET. Put OMP_CLAUSE_NOWAIT to
C_OMP_CLAUSE_SPLIT_TARGET if combined with target construct,
instead of C_OMP_CLAUSE_SPLIT_FOR.
paolo [Mon, 30 May 2016 19:18:13 +0000 (19:18 +0000)]
/cp
2016-05-30 Paolo Carlini <paolo.carlini@oracle.com>
PR c++/71238
* lex.c (unqualified_name_lookup_error): Take a location too.
(unqualified_fn_lookup_error): Take a cp_expr.
* cp-tree.h (unqualified_name_lookup_error,
unqualified_fn_lookup_error): Adjust declarations.
* semantics.c (perform_koenig_lookup): Adjust
unqualified_fn_lookup_error call, pass the location of
the identifier too as part of a cp_expr.
/testsuite
2016-05-30 Paolo Carlini <paolo.carlini@oracle.com>
ak [Mon, 30 May 2016 18:13:12 +0000 (18:13 +0000)]
Don't cause ICEs when auto profile file is not found with checking
Currently, on a checking enabled compiler when -fauto-profile does
not find the profile feedback file it errors out with assertation
failures. Add proper errors for this case.
gcc/:
2016-05-30 Andi Kleen <ak@linux.intel.com>
* auto-profile.c (read_profile): Replace asserts with errors
when file does not exist.
* gcov-io.c (gcov_read_words): Dito.
marxin [Mon, 30 May 2016 16:04:50 +0000 (16:04 +0000)]
Add profiling support for IVOPTS
* tree-ssa-loop-ivopts.c (get_computation_cost_at): Scale
computed costs by frequency of BB they belong to.
(get_scaled_computation_cost_at): New function.
projects / thirdparty / gcc.git / log
