PR tree-optimization/71625
* c-common.c (braced_list_to_string): New function.
* c-common.h (braced_list_to_string): Declare it.
gcc/cp/ChangeLog:
PR tree-optimization/71625
* decl.c (check_initializer): Call braced_list_to_string.
(eval_check_narrowing): New function.
* gcc/cp/typeck2.c (digest_init_r): Accept strings literals
as initilizers for all narrow character types.
gcc/testsuite/ChangeLog:
PR tree-optimization/71625
* g++.dg/init/string2.C: New test.
* g++.dg/init/string3.C: New test.
* g++.dg/init/string4.C: New test.
* gcc.dg/init-string-3.c: New test.
* gcc.dg/strlenopt-55.c: New test.
* gcc.dg/strlenopt-56.c: New test.
mpolacek [Mon, 13 Aug 2018 15:46:42 +0000 (15:46 +0000)]
P0806R2 - Deprecate implicit capture of this via [=]
* lambda.c (add_default_capture): Formatting fixes. Warn about
deprecated implicit capture of this via [=].
* g++.dg/cpp2a/lambda-this1.C: New test.
* g++.dg/cpp2a/lambda-this2.C: New test.
* g++.dg/cpp2a/lambda-this3.C: New test.
vries [Mon, 13 Aug 2018 12:04:24 +0000 (12:04 +0000)]
[nvptx] Use CUDA driver API to select default runtime launch geometry
The CUDA driver API starting version 6.5 offers a set of runtime functions to
calculate several occupancy-related measures, as a replacement for the occupancy
calculator spreadsheet.
This patch adds a heuristic for default runtime launch geometry, based on the
new runtime function cuOccupancyMaxPotentialBlockSize.
Build on x86_64 with nvptx accelerator and ran libgomp testsuite.
2018-08-13 Cesar Philippidis <cesar@codesourcery.com>
Tom de Vries <tdevries@suse.de>
PR target/85590
* plugin/cuda/cuda.h (CUoccupancyB2DSize): New typedef.
(cuOccupancyMaxPotentialBlockSize): Declare.
* plugin/cuda-lib.def (cuOccupancyMaxPotentialBlockSize): New
CUDA_ONE_CALL_MAYBE_NULL.
* plugin/plugin-nvptx.c (CUDA_VERSION < 6050): Define
CUoccupancyB2DSize and declare
cuOccupancyMaxPotentialBlockSize.
(nvptx_exec): Use cuOccupancyMaxPotentialBlockSize to set the
default num_gangs and num_workers when the driver supports it.
pault [Sun, 12 Aug 2018 17:19:09 +0000 (17:19 +0000)]
2018-08-12 Paul Thomas <pault@gcc.gnu.org>
PR fortran/66679
* trans-intrinsic.c (gfc_conv_intrinsic_transfer): Class array
elements are returned as references to the data element. Get
the class expression by stripping back the references. Use this
for the element size.
2018-08-12 Paul Thomas <pault@gcc.gnu.org>
PR fortran/66679
* gfortran.dg/transfer_class_3.f90: New test.
pault [Sun, 12 Aug 2018 10:55:13 +0000 (10:55 +0000)]
2018-08-12 Paul Thomas <pault@gcc.gnu.org>
PR fortran/86906
* resolve.c (resolve_fl_variable_derived): Check if the derived
type is use associated before checking for the host association
error.
2018-08-12 Paul Thomas <pault@gcc.gnu.org>
PR fortran/86906
* gfortran.dg/use_rename_9.f90: New test.
danglin [Sat, 11 Aug 2018 21:37:55 +0000 (21:37 +0000)]
gcc
* config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
Update comment for atomic instructions.
(atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
Remove.
(atomic_loaddi): Revise fence expansion to only emit fence prior to
load for __ATOMIC_SEQ_CST model.
(atomic_loaddi_1): Remove float register target.
(atomic_storedi): Handle CONST_INT values.
(atomic_storedi_1): Remove float register source. Add special case
for zero value.
(memory_barrier): New expander and insn.
libgcc
* config/pa/linux-atomic.c: Update comment.
(FETCH_AND_OP_2, OP_AND_FETCH_2, FETCH_AND_OP_WORD, OP_AND_FETCH_WORD,
COMPARE_AND_SWAP_2, __sync_val_compare_and_swap_4,
SYNC_LOCK_TEST_AND_SET_2, __sync_lock_test_and_set_4): Use
__ATOMIC_RELAXED for atomic loads.
(SYNC_LOCK_RELEASE_1): New define. Use __sync_synchronize() and
unordered store to release lock.
(__sync_lock_release_8): Likewise.
(SYNC_LOCK_RELEASE_2): Remove define.
jakub [Sat, 11 Aug 2018 11:29:12 +0000 (11:29 +0000)]
PR tree-optimization/86835
* tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
new_stmt after def_gsi, make sure to insert new_square_stmt after
that stmt, not 2 stmts before it.
janus [Fri, 10 Aug 2018 14:08:53 +0000 (14:08 +0000)]
2018-08-10 Janus Weil <janus@gcc.gnu.org>
PR fortran/57160
* invoke.texi (frontend-optimize): Mention short-circuiting.
* options.c (gfc_post_options): Disable -ffrontend-optimize with -Og.
* resolve.c (resolve_operator): Warn about short-circuiting only with
-ffrontend-optimize.
* trans-expr.c (gfc_conv_expr_op): Use short-circuiting operators only
with -ffrontend-optimize. Without that flag, make sure that both
operands are evaluated.
2018-08-10 Janus Weil <janus@gcc.gnu.org>
PR fortran/57160
* gfortran.dg/actual_pointer_function_1.f90: Fix invalid test case.
* gfortran.dg/inline_matmul_23.f90: Add option "-ffrontend-optimize".
* gfortran.dg/short_circuiting_2.f90: New test case.
* gfortran.dg/short_circuiting_3.f90: New test case.
PR target/83610
* builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
function type.
* builtins.c (expand_builtin_expect_with_probability):
New function.
(expand_builtin_expect_with_probability): New function.
(build_builtin_expect_predicate): Add new argumnet probability
for BUILT_IN_EXPECT_WITH_PROBABILITY.
(fold_builtin_expect):
(fold_builtin_2):
(fold_builtin_3):
* builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
* builtins.h (fold_builtin_expect): Set new argument.
* doc/extend.texi: Document __builtin_expect_with_probability.
* doc/invoke.texi: Likewise.
* gimple-fold.c (gimple_fold_call): Pass new argument.
* ipa-fnsummary.c (find_foldable_builtin_expect): Handle
also BUILT_IN_EXPECT_WITH_PROBABILITY.
* predict.c (get_predictor_value): New function.
(expr_expected_value): Add new argument probability. Assume
that predictor and probability are always non-null.
(expr_expected_value_1): Likewise. For __builtin_expect and
__builtin_expect_with_probability set probability. Handle
combination in binary expressions.
(tree_predict_by_opcode): Simplify code by simply calling
get_predictor_value.
(pass_strip_predict_hints::execute): Add handling of
BUILT_IN_EXPECT_WITH_PROBABILITY.
* predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
new predictor.
* tree.h (DECL_BUILT_IN_P): New function.
2018-08-10 Martin Liska <mliska@suse.cz>
PR target/83610
* gcc.dg/predict-17.c: New test.
* gcc.dg/predict-18.c: New test.
* gcc.dg/predict-19.c: New test.
marxin [Fri, 10 Aug 2018 09:31:51 +0000 (09:31 +0000)]
Strip only selected predictors after early tree passes (PR tree-optimization/85799).
2018-08-10 Martin Liska <mliska@suse.cz>
PR tree-optimization/85799
* passes.def: Add argument for pass_strip_predict_hints.
* predict.c (class pass_strip_predict_hints): Add new argument
early_p.
(strip_predictor_early): New function.
(pass_strip_predict_hints::execute): Call the function to
strip predictors.
(strip_predict_hints): New function.
* predict.def: Fix comment.
2018-08-10 Martin Liska <mliska@suse.cz>
PR tree-optimization/85799
* gcc.dg/pr85799.c: New test.
thopre01 [Fri, 10 Aug 2018 08:48:34 +0000 (08:48 +0000)]
Clarify source of tm.texi to copy for GFDL grant
When tm.texi.in is updated in the source tree, the following message
gets displayed:
Verify that you have permission to grant a GFDL license for all
new text in tm.texi, then copy it to <gcc src dir>/gcc/doc/tm.texi.
Having been myself and some colleagues confused several time by that
message as to what tm.texi to copy, I think it would be clearer to
indicate the absolute path for the source as well. This patch achieves
that.
2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
gcc/
* Makefile.in: Clarify which tm.texi to copy over to assert the
right to grant a GFDL license for all.
redi [Thu, 9 Aug 2018 23:25:53 +0000 (23:25 +0000)]
Define aliases for containers using polymorphic_allocator
These aliases are placed in the top-level header, e.g. <vector> not
<bits/stl_vector.h>. This ensures that they refer to whichever of
std::vector or __debug::vector or __profile::vector is in use when the
header is included.
* include/std/deque (std::pmr::deque): Declare alias.
* include/std/forward_list (std::pmr::forward_list): Likewise.
* include/std/list (std::pmr::list): Likewise.
* include/std/map (std::pmr::map, std::pmr::multimap): Likewise.
* include/std/regex (std::pmr::match_results, std::pmr::cmatch)
(std::pmr::smatch, std::pmr::wcmatch, std::pmr::wsmatch): Likewise.
* include/std/set (std::pmr::set, std::pmr::multiset): Likewise.
* include/std/string (std::pmr::basic_string, std::pmr::string)
(std::pmr::u16string, std::pmr::u32string, std::pmr::wstring):
Likewise.
* include/std/unordered_map (std::pmr::unordered_map)
(std::pmr::unordered_multimap): Likewise.
* include/std/unordered_set (std::pmr::unordered_set)
(std::pmr::unordered_multiset): Likewise.
* include/std/vector (std::pmr::vector): Likewise.
* testsuite/21_strings/basic_string/types/pmr_typedefs.cc: New test.
* testsuite/23_containers/deque/types/pmr_typedefs.cc: New test.
* testsuite/23_containers/forward_list/pmr_typedefs.cc: New test.
* testsuite/23_containers/list/pmr_typedefs.cc: New test.
* testsuite/23_containers/map/pmr_typedefs.cc: New test.
* testsuite/23_containers/multimap/pmr_typedefs.cc: New test.
* testsuite/23_containers/multiset/pmr_typedefs.cc: New test.
* testsuite/23_containers/set/pmr_typedefs.cc: New test.
* testsuite/23_containers/unordered_map/pmr_typedefs.cc: New test.
* testsuite/23_containers/unordered_multimap/pmr_typedefs.cc: New
test.
* testsuite/23_containers/unordered_multiset/pmr_typedefs.cc: New
test.
* testsuite/23_containers/unordered_set/pmr_typedefs.cc: New test.
* testsuite/23_containers/vector/pmr_typedefs.cc: New test.
* testsuite/28_regex/match_results/pmr_typedefs.cc: New test.
dmalcolm [Thu, 9 Aug 2018 15:32:13 +0000 (15:32 +0000)]
diagnostics: add line numbers to source (PR other/84889)
This patch adds a left margin to the lines of source (and annotations)
printed by diagnostic_show_locus, so that e.g. rather than:
test.c: In function 'test':
test.c:12:15: error: 'struct foo' has no member named 'm_bar'; did you mean 'bar'?
return ptr->m_bar;
^~~~~
bar
we print:
test.c: In function 'test':
test.c:12:15: error: 'struct foo' has no member named 'm_bar'; did you mean 'bar'?
12 | return ptr->m_bar;
| ^~~~~
| bar
Similarly, for a multiline case (in C++ this time), this:
bad-binary-ops.C: In function 'int test_2()':
bad-binary-ops.C:26:4: error: no match for 'operator+' (operand types are 's' and 't')
return (some_function ()
~~~~~~~~~~~~~~~~
+ some_other_function ());
^~~~~~~~~~~~~~~~~~~~~~~~
becomes:
bad-binary-ops.C: In function 'int test_2()':
bad-binary-ops.C:26:4: error: no match for 'operator+' (operand types are 's' and 't')
25 | return (some_function ()
| ~~~~~~~~~~~~~~~~
26 | + some_other_function ());
| ^~~~~~~~~~~~~~~~~~~~~~~~
I believe this slightly improves the readability of the output, in that it:
- distinguishes between the user's source code vs the annotation lines
that we're adding (the underlinings and fix-it hints here)
- shows the line numbers in another place (potentially helpful for
multiline diagnostics, where the user can see the line numbers directly,
rather than have to figure them out relative to the caret: in the 2nd
example, note how the diagnostic is reported at line 26, but the first
line printed is actually line 25)
I'm not sure that this is the precise format we want to go with [1], but
I think it's an improvement over the status quo, and we're in stage 1
of gcc 9, so there's plenty of time to shake out issues.
I've turned it on by default; it can be disabled via
-fno-diagnostics-show-line-numbers (it's also turned off in the testsuite, to
avoid breaking numerous existing test cases).
[1] Some possible variants:
- maybe just "LL|" rather than "LL | "
- maybe ':' rather than '|'
- maybe we should have some leading indentation, to better split up
the diagnostics visually via the left-hand column
- etc
gcc/ChangeLog:
PR other/84889
* common.opt (fdiagnostics-show-line-numbers): New option.
* diagnostic-show-locus.c (class layout): Add fields
"m_show_line_numbers_p" and "m_linenum_width";
(num_digits): New function.
(test_num_digits): New function.
(layout::layout): Initialize new fields. Update m_x_offset
logic to handle any left margin.
(layout::print_source_line): Print line number when requested.
(layout::start_annotation_line): New member function.
(layout::print_annotation_line): Call it.
(layout::print_leading_fixits): Likewise.
(layout::print_trailing_fixits): Likewise. Update calls to
move_to_column for new parameter.
(layout::get_x_bound_for_row): Add "add_left_margin" param and use
it to potentially call start_annotation_line.
(layout::show_ruler): Call start_annotation_line.
(selftest::test_line_numbers_multiline_range): New selftest.
(selftest::diagnostic_show_locus_c_tests): Call test_num_digits
and selftest::test_line_numbers_multiline_range.
* diagnostic.c (diagnostic_initialize): Initialize
show_line_numbers_p.
* diagnostic.h (struct diagnostic_context): Add field
"show_line_numbers_p".
* doc/invoke.texi (Diagnostic Message Formatting Options): Add
-fno-diagnostics-show-line-numbers.
* dwarf2out.c (gen_producer_string): Add
OPT_fdiagnostics_show_line_numbers to the ignored options.
* lto-wrapper.c (merge_and_complain): Likewise to the "pick
one setting" options.
(append_compiler_options): Likewise to the dropped options.
(append_diag_options): Likewise to the passed-on options.
* opts.c (common_handle_option): Handle the new option.
* toplev.c (general_init): Set up global_dc->show_line_numbers_p.
gcc/testsuite/ChangeLog:
PR other/84889
* gcc.dg/plugin/diagnostic-test-show-locus-bw-line-numbers.c: New
test.
* gcc.dg/plugin/diagnostic-test-show-locus-color-line-numbers.c:
New test.
* gcc.dg/plugin/plugin.exp (plugin_test_list): Add the new tests.
* lib/prune.exp: Add -fno-diagnostics-show-line-numbers to
TEST_ALWAYS_FLAGS.
kelvin [Thu, 9 Aug 2018 15:01:55 +0000 (15:01 +0000)]
gcc/ChangeLog:
2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
* doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
third argument of type "const signed char" to descriptions of
__builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
__builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
__builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
__builtin_bcdsub_ov functions.
rsandifo [Thu, 9 Aug 2018 14:45:02 +0000 (14:45 +0000)]
Restore flow_bb_inside_loop_p tests (PR 86858)
The series to remove vinfo_for_stmt also removed tests of
flow_bb_inside_loop_p if the call was simply testing whether the
statement was in the vectorisation region. I'd tried to keep calls
that were testing whether the statement was in a particular loop
(inner or outer), but messed up in vect_is_simple_reduction and
removed calls that were actually needed. This patch restores them.
I double-checked the other removed calls and I think these are
the only ones affected.
2018-08-08 Richard Sandiford <richard.sandiford@arm.com>
rsandifo [Thu, 9 Aug 2018 14:37:24 +0000 (14:37 +0000)]
Fix invalid assumption in vect_transform_stmt (PR 86871)
The handling of outer-loop uses of inner-loop definitions assumed
that anything that wasn't a PHI would be a gassign. It's also
possible for it to be a gcall.
2018-08-08 Richard Sandiford <richard.sandiford@arm.com>
gcc/
PR tree-optimization/86871
* tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
instead of gimple_assign_lhs.
gcc/testsuite/
PR tree-optimization/86871
* gcc.dg/vect/pr86871.c: New test.
Some of the carryin insn patterns are missing a register constraint.
That means that the register allocator can pick practically anything
to hold that value, including memory locations, or registers of the
wrong class.
marxin [Thu, 9 Aug 2018 08:08:44 +0000 (08:08 +0000)]
Fix typos in params.def.
2018-08-09 Martin Liska <mliska@suse.cz>
* params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
at the end of a line, make first letter capital and end up
a sentence with a dot.
(PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
(PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
(PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
(PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
(PARAM_MAX_ISL_OPERATIONS): Likewise.
(PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
(PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
(PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
(PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
(PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
(PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
(PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
(PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
(PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
(PARAM_TREE_REASSOC_WIDTH): Likewise.
(PARAM_HSA_GEN_DEBUG_STORES): Likewise.
(PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
(PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
krebbel [Thu, 9 Aug 2018 07:06:23 +0000 (07:06 +0000)]
S/390: Fix PR84332 ICE with stack clash protection
Our implementation of the stack probe requires the probe interval to
be used as displacement in an address operand. The maximum probe
interval currently is 64k. This would exceed short displacements.
Trim that value down to 4k if that happens. This might result in too
many probes being generated only on the oldest supported machine level
z900.
gcc/ChangeLog:
2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
PR target/84332
* config/s390/s390.c (s390_option_override_internal): Reduce the
stack-clash-protection-probe-interval param if it would be too big
for z900.
gcc/testsuite/ChangeLog:
2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
PR target/84332
* gcc.target/s390/pr84332.c: New testcase.
schwab [Wed, 8 Aug 2018 20:01:26 +0000 (20:01 +0000)]
m68k: handle more cases of TLS symbols with offset
PR target/46179
* config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
* config/m68k/m68k.c (handle_move_double): Don't call
m68k_final_prescan_insn.
(m68k_adjust_decorated_operand): Renamed from
m68k_final_prescan_insn, remove first and third operand and
simplify.
(print_operand): Call it.
(print_operand_address): Call it.
PR target/46179
* gcc.target/m68k/tls-dimode.c: New file.
redi [Wed, 8 Aug 2018 15:16:43 +0000 (15:16 +0000)]
Prevent internal aligned_alloc clashing with libc version
If configure fails to detect aligned_alloc we will try to define our
own in new_opa.cc but that could clash with the libcversion in
<stdlib.h>. Use a namespace to keep them distinct.
* libsupc++/new_opa.cc (aligned_alloc): Declare inside namespace to
avoid clashing with an ::aligned_alloc function that was not detected
by configure.
vries [Wed, 8 Aug 2018 14:26:37 +0000 (14:26 +0000)]
[libgomp, nvptx] Fall back to cuLinkAddData/cuLinkCreate if _v2 not found
Cuda driver api functions cuLinkAddData and cuLinkCreate are available starting
version 5.5. In version 6.5, they are remapped onto _v2 versions.
The dlopen interface of the libgomp nvptx plugin uses the _v2 versions, so it
won't work with a cuda driver with driver api version lower than 6.5.
This patch fixes the problem by testing for the presence of the _v2 versions,
and falling back to the original versions in case of absence of the _v2
versions.
Build on x86_64 with nvptx accelerator and reg-tested libgomp, both with and
without --without-cuda-driver.
2018-08-08 Tom de Vries <tdevries@suse.de>
* plugin/cuda-lib.def (cuLinkAddData_v2, cuLinkCreate_v2): Declare using
CUDA_ONE_CALL_MAYBE_NULL.
* plugin/plugin-nvptx.c (cuLinkAddData, cuLinkCreate): Undef and declare.
(cuLinkAddData_v2, cuLinkCreate_v2): Declare.
(link_ptx): Fall back to cuLinkAddData/cuLinkCreate if the _v2 versions
are not found.
vries [Wed, 8 Aug 2018 14:26:28 +0000 (14:26 +0000)]
[libgomp, nvptx] Allow cuGetErrorString to be NULL
Cuda driver api function cuGetErrorString is available in version 6.0 and
higher.
Currently, when the driver that is used does not contain this function, the
libgomp nvptx plugin will not build (PLUGIN_NVPTX_DYNAMIC == 0) or run
(PLUGIN_NVPTX_DYNAMIC == 1).
This patch fixes this problem by testing for the presence of the function, and
handling absence.
Build on x86_64 with nvptx accelerator and reg-tested libgomp, both with and
without --without-cuda-driver.
2018-08-08 Tom de Vries <tdevries@suse.de>
* plugin/cuda-lib.def (cuGetErrorString): Use CUDA_ONE_CALL_MAYBE_NULL.
* plugin/plugin-nvptx.c (cuda_error): Handle if cuGetErrorString is not
present.
vries [Wed, 8 Aug 2018 14:26:19 +0000 (14:26 +0000)]
[libgomp, nvptx] Remove hard-coded const in nvptx_open_device
CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_PER_MULTIPROCESSOR is defined in cuda driver
api version 6.0 and higher.
Currently nvptx_open_device uses a hard-coded constant instead.
This patch fixes that by:
- defining CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_PER_MULTIPROCESSOR to the hardcoded
constant at toplevel, if not present in cuda.h, and
- using CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_PER_MULTIPROCESSOR in nvptx_open_device
Build on x86_64 with nvptx accelerator and reg-tested libgomp.
2018-08-08 Tom de Vries <tdevries@suse.de>
* plugin/plugin-nvptx.c
(CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_PER_MULTIPROCESSOR): Define.
(nvptx_open_device): Use
CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_PER_MULTIPROCESSOR.
vries [Wed, 8 Aug 2018 14:26:10 +0000 (14:26 +0000)]
[libgomp, nvptx] Note that cuGetErrorString is in CUDA_VERSION >= 6000
Cuda driver api function cuGetErrorString is available in version 6.0 and
higher.
This patch:
- removes a comment saying the declaration is not available in cuda.h 6.0
- fixes the presence test to use CUDA_VERSION < 6000
- moves the declaration to toplevel
Build on x86_64 with nvptx accelerator and reg-tested libgomp.
2018-08-08 Tom de Vries <tdevries@suse.de>
* plugin/plugin-nvptx.c (cuda_error): Move declaration of cuGetErrorString ...
(cuGetErrorString): ... here. Guard with CUDA_VERSION < 6000.
redi [Wed, 8 Aug 2018 13:09:19 +0000 (13:09 +0000)]
PR libstdc++/86597 directory_entry observers should clear error_code
PR libstdc++/86597
* include/bits/fs_dir.h (directory_entry::_M_file_type(error_code&)):
Clear error_code when cached type is used.
* testsuite/27_io/filesystem/directory_entry/86597.cc: New test.
krebbel [Wed, 8 Aug 2018 12:38:51 +0000 (12:38 +0000)]
S/390: Fix PR85295
gcc/ChangeLog:
2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
PR target/85295
* config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
definitions.
* config/s390/s390.md ("movti"): Add more alternatives for
constant to GPR copies.
gcc/testsuite/ChangeLog:
2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
PR target/85295
* gcc.target/s390/TI-constants-lra.c: New testcase.
* gcc.target/s390/TI-constants-nolra.c: New testcase.
krebbel [Wed, 8 Aug 2018 11:25:31 +0000 (11:25 +0000)]
S/390: Remove TARGET_CPU_ZARCH
TARGET_CPU_ZARCH allowed to distinguish between g5/g6 and newer
machines. Since the former are now gone, we can assume that
TARGET_CPU_ZARCH is always true. As a side-effect, branch splitting
is now completely gone. Some parts of literal pool splitting are also
gone, but it's still there: we need to support it because floating
point and vector instructions still cannot use relative addressing.
jakub [Wed, 8 Aug 2018 09:00:51 +0000 (09:00 +0000)]
P0595R1 - is_constant_evaluated
cp/
* cp-tree.h (enum cp_built_in_function): New.
(maybe_constant_init): Add pretend_const_required argument.
* typeck2.c (store_init_value): Pass true as new argument to
maybe_constant_init.
* constexpr.c (constexpr_fn_retval): Check also DECL_BUILT_IN_CLASS
for BUILT_IN_UNREACHABLE.
(struct constexpr_ctx): Add pretend_const_required field.
(cxx_eval_builtin_function_call): Use DECL_IS_BUILTIN_CONSTANT_P
macro. Handle CP_BUILT_IN_IS_CONSTANT_EVALUATED. Check also
DECL_BUILT_IN_CLASS for BUILT_IN_UNREACHABLE.
(cxx_eval_outermost_constant_expr): Add pretend_const_required
argument, initialize pretend_const_required field in ctx. If the
result is TREE_CONSTANT and non_constant_p, retry with
pretend_const_required false if it was true.
(is_sub_constant_expr): Initialize pretend_const_required_field in
ctx.
(cxx_constant_value): Pass true as pretend_const_required to
cxx_eval_outermost_constant_expr.
(maybe_constant_value): Pass false as pretend_const_required to
cxx_eval_outermost_constant_expr.
(fold_non_dependent_expr): Likewise.
(maybe_constant_init_1): Add pretend_const_required argument, pass it
down to cxx_eval_outermost_constant_expr. Pass !allow_non_constant
instead of false as strict to cxx_eval_outermost_constant_expr.
(maybe_constant_init): Add pretend_const_required argument, pass it
down to maybe_constant_init_1.
(cxx_constant_init): Pass true as pretend_const_required to
maybe_constant_init_1.
* cp-gimplify.c (cp_gimplify_expr): Handle CALL_EXPRs to
CP_BUILT_IN_IS_CONSTANT_EVALUATED.
(cp_fold): Don't fold CP_BUILT_IN_IS_CONSTANT_EVALUATED calls.
* decl.c: Include langhooks.h.
(cxx_init_decl_processing): Register __builtin_is_constant_evaluated
built-in.
* tree.c (builtin_valid_in_constant_expr_p): Return true for
CP_BUILT_IN_IS_CONSTANT_EVALUATED.
* pt.c (declare_integer_pack): Initialize DECL_FUNCTION_CODE.
testsuite/
* g++.dg/cpp2a/is-constant-evaluated1.C: New test.
jakub [Wed, 8 Aug 2018 08:31:40 +0000 (08:31 +0000)]
PR c++/86738
* constexpr.c (cxx_eval_binary_expression): For arithmetics involving
NULL pointer set *non_constant_p to true.
(cxx_eval_component_reference): For dereferencing of a NULL pointer,
set *non_constant_p to true and return t.
siddhesh [Wed, 8 Aug 2018 07:54:14 +0000 (07:54 +0000)]
[AArch64, Falkor] Switch to using Falkor-specific vector costs.
The adjusted vector costs give Falkor a reasonable boost in performance for FP
benchmarks (both CPU2017 and CPU2006) and doesn't change INT benchmarks that
much. There are some regressions that will be investigated as follow on work.
Adjust Falkor's register_sextend cost from 4 to 3. This fixes a testsuite
failure in gcc.target/aarch64/extend.c:ldr_sxtw where GCC was generating
a sbfiz instruction rather than a load with sign extension.
No performance changes.
gcc/ChangeLog:
2018-08-08 Luis Machado <luis.machado@linaro.org>
* config/aarch64/aarch64.c (qdf24xx_addrcost_table)
<register_sextend>: Set to 3.
nathan [Tue, 7 Aug 2018 21:28:51 +0000 (21:28 +0000)]
[PATCH] Line map table allocation
https://gcc.gnu.org/ml/gcc-patches/2018-08/msg00434.html
* line-map.c: (linemap_init): Set default allocator here.
(new_linemap): Rather than here. Refactor allocation logic.
PR libstdc++/86874
* include/std/variant (_Copy_ctor_base::_M_destructive_move): Define
here instead of in _Move_assign_base.
(_Copy_ctor_base<true, _Types...>::_M_destructive_move): Define.
(_Copy_assign_base::operator=): Use _M_destructive_move when changing
the contained value to another alternative.
(_Move_assign_base::operator=): Likewise.
(_Move_assign_base::_M_destructive_move): Remove.
* testsuite/20_util/variant/86874.cc: New test.
ian [Tue, 7 Aug 2018 17:28:22 +0000 (17:28 +0000)]
libgo: uncomment trace.Stop() call in testing package
Fix up the testing package to insure that execution traces
work properly (e.g. "-test.trace=<XXX>" command line option). The
call to stop tracing and emit the output file was stubbed out.
rsandifo [Tue, 7 Aug 2018 17:22:19 +0000 (17:22 +0000)]
[AArch64] Fix -mlow-precision-div (PR 86838)
The "@" handling broke -mlow-precision-div, because the scalar forms of
the instruction were provided by a pattern that also provided FRECPX
(and so were parameterised on an unspec code as well as a mode),
while the SIMD versions had a dedicated FRECPE pattern. This patch
moves the scalar FRECPE handling to the SIMD pattern too (as for FRECPS)
and uses a separate pattern for FRECPX.
The convention in aarch64-simd-builtins.def seemed to be to add
comments only if the mapping wasn't obvious (i.e. not just sticking
"aarch64_" on the beginning and "<mode>" on the end), so the patch
deletes the reference to the combined pattern instead of rewording it.
There didn't seem to be any coverage of -mlow-precision-div in the
testsuite, so the patch adds some tests for it.
2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
gcc/
PR target/86838
* config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
* config/aarch64/aarch64-simd.md
(aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
(@aarch64_frecpe<mode>): ...here and the move FRECPX to...
(aarch64_frecpx<mode>): ...this new pattern.
* config/aarch64/aarch64-simd-builtins.def: Remove comment
about aarch64_frecp<FRECP:frecp_suffix><mode>.
gcc/testsuite/
PR target/86838
* gcc.target/aarch64/frecpe_1.c: New test.
* gcc.target/aarch64/frecpe_2.c: Likewise.
paolo [Tue, 7 Aug 2018 16:40:18 +0000 (16:40 +0000)]
/cp
2018-08-07 Paolo Carlini <paolo.carlini@oracle.com>
PR c++/59480, DR 136
* decl.c (check_no_redeclaration_friend_default_args): New.
(duplicate_decls): Use the latter; also check that a friend
declaration specifying default arguments is a definition.
/testsuite
2018-08-07 Paolo Carlini <paolo.carlini@oracle.com>
redi [Tue, 7 Aug 2018 16:10:29 +0000 (16:10 +0000)]
PR libstdc++/86861 Meet precondition for Solaris memalign
Solaris memalign requires alignment to be at least sizeof(int), so
increase it as needed.
Also move the check for valid alignments from the fallback
implementation of aligned_alloc into operator new, as it's required for
all of aligned_alloc, memalign, posix_memalign and __aligned_malloc.
This adds a branch to check for undefined behaviour which we could just
ignore, so the check could just be removed. It should certainly be
removed if PR 86878 is implemented to issue a warning about calls with
invalid alignments.
PR libstdc++/86861
* libsupc++/new_opa.cc [_GLIBCXX_HAVE_MEMALIGN] (aligned_alloc):
Replace macro with inline function.
[__sun]: Increase alignment to meet memalign precondition.
[!HAVE__ALIGNED_MALLOC && !HAVE_POSIX_MEMALIGN && !HAVE_MEMALIGN]
(aligned_alloc): Move check for valid alignment to operator new.
Remove redundant check for non-zero size, it's enforced by the caller.
(operator new): Move check for valid alignment here. Use
__builtin_expect on check for zero size.
marxin [Tue, 7 Aug 2018 11:59:13 +0000 (11:59 +0000)]
Add malloc predictor (PR middle-end/83023).
2018-08-07 Martin Liska <mliska@suse.cz>
PR middle-end/83023
* predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
* predict.def (PRED_MALLOC_NONNULL): New predictor.
* doc/extend.texi: Document that malloc attribute adds
hit to compiler.
2018-08-07 Martin Liska <mliska@suse.cz>
PR middle-end/83023
* gcc.dg/predict-16.c: New test.
* g++.dg/predict-1.C: New test.
vries [Mon, 6 Aug 2018 22:13:56 +0000 (22:13 +0000)]
[libgomp, nvptx] Handle CUDA_ONE_CALL_MAYBE_NULL
This patch adds handling of functions that may not be present in the cuda
driver.
Such a function can be declared using CUDA_ONE_CALL_MAYBE_NULL in cuda-lib.def,
it can be called with the usual convenience macros, but before calling its
presence needs to be tested using new macro CUDA_CALL_EXISTS.
When using the dlopen interface (PLUGIN_NVPTX_DYNAMIC == 1), we allow
non-present functions by allowing dlsym to return NULL. Otherwise
(PLUGIN_NVPTX_DYNAMIC == 0) we declare the non-present function to be weak.
Build and reg-tested libgomp on x86_64 with nvidia accelerator, with and without
--disable-cuda-driver, in combination with a trigger patch that adds a
non-existing function foo to cuda-lib.def:
...
CUDA_ONE_CALL_MAYBE_NULL (foo)
...
and declares it in plugin-nvptx.c:
...
CUresult foo (void);
...
and then uses it in nvptx_init after the init_cuda_lib call:
...
if (CUDA_CALL_EXISTS (foo))
CUDA_CALL (foo);
...
Also build and reg-tested on x86_64 with nvidia accelerator, with and without
--disable-cuda-driver, in combination with a trigger patch that replaces all
CUDA_ONE_CALLs in cuda-lib.def with CUDA_ONE_CALL_MAYBE_NULL, and guards two
CUDA_CALLs with CUDA_CALL_EXISTS, one for a regular fn, and one for a fn that is
a define in cuda/cuda.h.
2018-08-07 Tom de Vries <tdevries@suse.de>
* plugin/plugin-nvptx.c (DO_PRAGMA): Define.
(struct cuda_lib_s): Add def/undef of CUDA_ONE_CALL_MAYBE_NULL.
(init_cuda_lib): Add new param to CUDA_ONE_CALL_1. Add arg to
corresponding call in CUDA_ONE_CALL. Add def/undef of
CUDA_ONE_CALL_MAYBE_NULL.
(CUDA_CALL_EXISTS): Define.
vries [Mon, 6 Aug 2018 22:13:46 +0000 (22:13 +0000)]
[libgomp, nvptx] Minimize lifetime of CUDA_ONE_CALL defines
This patch makes sure that the lifetimes of the CUDA_ONE_CALL macro (which is
defined twice in plugin-nvptx.c) are minimized, to make it obvious that the
definitions are used only in the lib-cuda.def include.
Build on x86_64 with nvptx accelerator and reg-tested libgomp.
2018-08-07 Tom de Vries <tdevries@suse.de>
* plugin/plugin-nvptx.c (struct cuda_lib_s, init_cuda_lib): Put
CUDA_ONE_CALL defines right before the cuda-lib.def include, and the
corresponding undefs right after.
law [Mon, 6 Aug 2018 20:13:20 +0000 (20:13 +0000)]
* tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
the vr_values instance to cprop_into_stmt.
(cprop_into_stmt): Pass vr_values instance down to cprop_operand.
(cprop_operand): Also query EVRP to determine if OP is a constant.
dmalcolm [Mon, 6 Aug 2018 16:25:27 +0000 (16:25 +0000)]
Fix memory leak in selftest::test_expansion_to_rtl
"make selftest-valgrind" shows:
187 bytes in 1 blocks are definitely lost in loss record 567 of 669
at 0x4A081D4: calloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
by 0x1F08260: xcalloc (xmalloc.c:162)
by 0xB24F32: init_emit() (emit-rtl.c:5843)
by 0xC10080: prepare_function_start() (function.c:4803)
by 0xC10254: init_function_start(tree_node*) (function.c:4877)
by 0x1CDF92A: selftest::test_expansion_to_rtl() (function-tests.c:595)
by 0x1CE007C: selftest::function_tests_c_tests() (function-tests.c:676)
by 0x1E010E7: selftest::run_tests() (selftest-run-tests.c:98)
by 0x1062D1E: toplev::run_self_tests() (toplev.c:2225)
by 0x1062F40: toplev::main(int, char**) (toplev.c:2303)
by 0x1E5B90A: main (main.c:39)
This patch fixes this leak (and makes the output of
"make selftest-valgrind" clean) by calling free_after_compilation at the
end of the selftest in question.
alahay01 [Mon, 6 Aug 2018 15:33:00 +0000 (15:33 +0000)]
Enable clobber high for tls descs on Aarch64
gcc/
* config/aarch64/aarch64.md: Add clobber highs to tls_desc.
gcc/testsuite/
* gcc.target/aarch64/sve/tls_preserve_1.c: New test.
* gcc.target/aarch64/sve/tls_preserve_2.c: New test.
* gcc.target/aarch64/sve/tls_preserve_3.c: New test.
projects / thirdparty / gcc.git / log
