iii [Wed, 24 Oct 2018 12:04:53 +0000 (12:04 +0000)]
S/390: Fix ICE in s390_check_qrst_address ()
In r265371 (S/390: Make "b" constraint match literal pool references)
the CONSTANT_POOL_ADDRESS_P () check was moved from
s390_loadrelative_operand_p () to s390_check_qrst_address (). However,
in the original code it was guarded by SYMBOL_REF_P (), which was not
added to the new code.
gcc/ChangeLog:
2018-10-24 Ilya Leoshkevich <iii@linux.ibm.com>
* config/s390/s390.c (s390_check_qrst_address): Add the missing
SYMBOL_REF_P () check.
rguenth [Wed, 24 Oct 2018 11:46:58 +0000 (11:46 +0000)]
2018-10-24 Richard Biener <rguenther@suse.de>
PR tree-optimization/87105
* tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
dump classification.
(vect_analyze_data_ref_accesses): Handle duplicate loads and
stores by splitting the affected group after the fact.
* tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
fail the SLP build because of size constraints.
* gcc.dg/vect/bb-slp-39.c: New testcase.
* gfortran.dg/vect/pr83232.f90: Un-XFAIL.
rguenth [Wed, 24 Oct 2018 09:42:19 +0000 (09:42 +0000)]
2018-10-24 Richard Biener <rguenther@suse.de>
PR tree-optimization/84013
* tree-ssa-structalias.c (struct msdi_data): New struct for
marshalling data to walk_stmt_load_store_ops.
(maybe_set_dependence_info): Refactor as callback for
walk_stmt_load_store_ops.
(compute_dependence_clique): Set restrict info on all stmt kinds.
Using a delegating constructor to implement these constructors means
that they instantiate the destructor, which requires the element_type to
be complete. In C++11 and C++14 they were specified to be delegating,
but that was changed as part of LWG 2801 so in C++17 they don't require
a complete type (as was intended all along).
PR libstdc++/87704
* include/bits/unique_ptr.h (unique_ptr::unique_ptr(nullptr_t)): Do
not delegate to default constructor.
(unique_ptr<T[], D>::unique_ptr(nullptr_t)): Likewise.
* testsuite/20_util/unique_ptr/cons/incomplete.cc: New test.
rguenth [Tue, 23 Oct 2018 11:34:56 +0000 (11:34 +0000)]
2018-10-23 Richard Biener <rguenther@suse.de>
PR tree-optimization/87105
PR tree-optimization/87608
* passes.def (pass_all_early_optimizations): Add early phi-opt
after dce.
* tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
addition to debug stmts.
(tree_ssa_phiopt_worker): Add early_p argument, do only min/max
and abs replacement early.
* tree-cfg.c (gimple_empty_block_p): Likewise.
* g++.dg/tree-ssa/phiopt-1.C: New testcase.
g++.dg/vect/slp-pr87105.cc: Likewise.
* g++.dg/tree-ssa/pr21463.C: Scan phiopt2 because this testcase
relies on phiprop run before.
* g++.dg/tree-ssa/pr30738.C: Likewise.
* g++.dg/tree-ssa/pr57380.C: Likewise.
* gcc.dg/tree-ssa/pr84859.c: Likewise.
* gcc.dg/tree-ssa/pr45397.c: Scan phiopt2 because phiopt1 is
confused by copies in the IL left by EVRP.
* gcc.dg/tree-ssa/phi-opt-5.c: Likewise, this time confused
by predictors.
* gcc.dg/tree-ssa/phi-opt-12.c: Scan phiopt2.
* gcc.dg/pr24574.c: Likewise.
* g++.dg/tree-ssa/pr86544.C: Scan phiopt4.
rearnsha [Tue, 23 Oct 2018 10:19:15 +0000 (10:19 +0000)]
[arm] Update default CPUs during configure
There are a couple of places in config.gcc where the default CPU is
still arm6, but that was removed as a supported CPU earlier this year.
This patch fixes those entries.
The default CPU for configurations that do not explicitly set a
default is now arm7tdmi (so assumes thumb is available). Given that
StrongArm is on the deprecated list, this is a better default than we
had previously.
For NetBSD the default is StrongArm; this is the only remaining port
that uses the old ABI and really still carries support for non-thumb
based targets.
PR target/86383
* config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
specified to configure.
(arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
pault [Tue, 23 Oct 2018 08:27:14 +0000 (08:27 +0000)]
2018-10-23 Paul Thomas <pault@gcc.gnu.org>
PR fortran/85603
* frontend-passes.c (get_len_call): New function to generate a
call to intrinsic LEN.
(create_var): Use this to make length expressions for variable
rhs string lengths.
Clean up some white space issues.
2018-10-23 Paul Thomas <pault@gcc.gnu.org>
PR fortran/85603
* gfortran.dg/deferred_character_23.f90 : Check reallocation is
occurring as it should and a regression caused by version 1 of
this patch.
ian [Tue, 23 Oct 2018 02:46:41 +0000 (02:46 +0000)]
compiler: export indexed type data, read unexported types lazily
Introduce a new "types" command to the export data to record the
number of types and the size of their export data. It is immediately
followed by new "type" commands that can be indexed. Parse all the
exported types immediately so that we register them, but parse other
type data only as needed.
segher [Mon, 22 Oct 2018 20:23:39 +0000 (20:23 +0000)]
combine: Do not combine moves from hard registers
On most targets every function starts with moves from the parameter
passing (hard) registers into pseudos. Similarly, after every call
there is a move from the return register into a pseudo. These moves
usually combine with later instructions (leaving pretty much the same
instruction, just with a hard reg instead of a pseudo).
This isn't a good idea. Register allocation can get rid of unnecessary
moves just fine, and moving the parameter passing registers into many
later instructions tends to prevent good register allocation. This
patch disallows combining moves from a hard (non-fixed) register.
This also avoid the problem mentioned in PR87600 #c3 (combining hard
registers into inline assembler is problematic).
Because the register move can often be combined with other instructions
*itself*, for example for setting some condition code, this patch adds
extra copies via new pseudos after every copy-from-hard-reg.
On some targets this reduces average code size. On others it increases
it a bit, 0.1% or 0.2% or so. (I tested this on all *-linux targets).
PR rtl-optimization/87600
* combine.c: Add include of expr.h.
(cant_combine_insn_p): Do not combine moves from any hard non-fixed
register to a pseudo.
(make_more_copies): New function, add a copy to a new pseudo after
the moves from hard registers into pseudos.
(rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
later. Call make_more_copies.
GCC will currently increment "reject" once, for operand 0, and then decrement
it once for each of the other operands, ending with reject == -2 and an
assertion failure. If there's a conflict then it might try to decrement reject
yet again.
Incidentally, what these patterns are trying to achieve is an allocation in
which operand 0 may match one of the other operands, but may not partially
overlap any of them. Ideally there'd be a better way to do this.
In any case, it will affect any pattern in which multiple operands may (or
must) match an early-clobber operand.
The patch only allows a reject-- when one has not already occurred, for that
operand.
2018-10-22 Andrew Stubbs <ams@codesourcery.com>
gcc/
* lra-constraints.c (process_alt_operands): New local array,
matching_early_clobber. Check matching_early_clobber before
decrementing reject, and set matching_early_clobber after.
segher [Mon, 22 Oct 2018 14:03:22 +0000 (14:03 +0000)]
rs6000: Handle print_operand_address for unexpected RTL (PR87598)
As the PR shows, the user can force this to be called on at least some
RTL that is not a valid address. Most targets treat this as if the
user knows best; let's do the same.
PR target/87598
* config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
call output_addr_const and hope for the best.
rguenth [Mon, 22 Oct 2018 13:57:47 +0000 (13:57 +0000)]
2018-10-22 Richard Biener <rguenther@suse.de>
* gimple-ssa-evrp-analyze.c
(evrp_range_analyzer::record_ranges_from_incoming_edge): Be
smarter about what ranges to use.
* tree-vrp.c (add_assert_info): Dump here.
(register_edge_assert_for_2): Instead of here at multiple but
not all places.
rguenth [Mon, 22 Oct 2018 13:54:23 +0000 (13:54 +0000)]
2018-10-22 Steven Bosscher <steven@gcc.gnu.org>
Richard Biener <rguenther@suse.de>
* bitmap.h: Update data structure documentation, including a
description of bitmap views as either linked-lists or splay trees.
(struct bitmap_element_def): Update comments for splay tree bitmaps.
(struct bitmap_head_def): Likewise.
(bitmap_list_view, bitmap_tree_view): New prototypes.
(bitmap_initialize_stat): Initialize a bitmap_head's indx and
tree_form fields.
(bmp_iter_set_init): Assert the iterated bitmaps are in list form.
(bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
* bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
released bitmap element here.
(bitmap_element_free): Remove.
(bitmap_elt_clear_from): Work on splay tree bitmaps.
(bitmap_list_link_element): Renamed from bitmap_element_link. Move
this function similar ones such that linked-list bitmap implementation
functions are grouped.
(bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
and moved for grouping.
(bitmap_list_insert_element_after): Renamed from
bitmap_elt_insert_after, and moved for grouping.
(bitmap_list_find_element): New function spliced from bitmap_find_bit.
(bitmap_tree_link_left, bitmap_tree_link_right,
bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
bitmap_tree_link_element, bitmap_tree_unlink_element,
bitmap_tree_find_element): New functions for splay-tree bitmap
implementation.
(bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
Renamed and moved, see above entries.
(bitmap_tree_listify_from): New function to convert part of a splay
tree bitmap to a linked-list bitmap.
(bitmap_list_view): Convert a splay tree bitmap to linked-list form.
(bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
(bitmap_find_bit): Remove.
(bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
Handle splay tree bitmaps.
(bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
bitmap_intersect_compl_p, bitmap_ior_and_compl,
bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
bitmap_hash): Reject trying to act on splay tree bitmaps. Make
corresponding changes to use linked-list specific bitmap_element
manipulation functions as applicable for efficiency.
(bitmap_tree_to_vec): New function.
(debug_bitmap_elt_file): New function split out from ...
(debug_bitmap_file): ... here. Handle splay tree bitmaps.
(bitmap_print): Likewise.
PR tree-optimization/63155
* tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
SSA edge worklists.
* tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
in tree-view.
+/* Unaligned version of the same type. */
+typedef float __m128_u __attribute__ ((__vector_size__ (16), __may_alias__,
+ __aligned__ (1)));
+
/* Internal data types for implementing the intrinsics. */
typedef float __v4sf __attribute__ ((__vector_size__ (16)));
ebotcazou [Mon, 22 Oct 2018 11:03:17 +0000 (11:03 +0000)]
* gcc-interface/utils.c (unchecked_convert): Use local variables for
the biased and reverse SSO attributes of both types.
Further extend the processing of integral types in the presence of
reverse SSO to all scalar types.
jamborm [Mon, 22 Oct 2018 08:27:50 +0000 (08:27 +0000)]
Add a fun parameter to three stmt_could_throw... functions
This long patch only does one simple thing, adds an explicit function
parameter to predicates stmt_could_throw_p, stmt_can_throw_external
and stmt_can_throw_internal.
My motivation was ability to use stmt_can_throw_external in IPA
analysis phase without the need to push cfun. As I have discovered,
we were already doing that in cgraph.c, which this patch avoids as
well. In the process, I had to add a struct function parameter to
stmt_could_throw_p and decided to also change the interface of
stmt_can_throw_internal just for the sake of some minimal consistency.
In the process I have discovered that calling method
cgraph_node::create_version_clone_with_body (used by ipa-split,
ipa-sra, OMP simd and multiple_target) leads to calls of
stmt_can_throw_external with NULL cfun. I have worked around this by
making stmt_can_throw_external and stmt_could_throw_p gracefully
accept NULL and just be pessimistic in that case. The problem with
fixing this in a better way is that struct function for the clone is
created after cloning edges where we attempt to push the yet not
existing cfun, and moving it before would require a bit of surgery in
tree-inline.c. A slightly hackish but simpler fix might be to
explicitely pass the "old" function to symbol_table::create_edge
because it should be just as good at that moment. In any event, that
is a topic for another patch.
I believe that currently we incorrectly use cfun in
maybe_clean_eh_stmt_fn and maybe_duplicate_eh_stmt_fn, both in
tree-eh.c, and so I have fixed these cases too. The bulk of other
changes is just mechanical adding of cfun to all users.
Bootstrapped and tested on x86_64-linux (also with extra NULLing and
restoring cfun to double check it is not used in a place I missed), OK
for trunk?
Thanks,
Martin
2018-10-22 Martin Jambor <mjambor@suse.cz>
* tree-eh.h (stmt_could_throw_p): Add function parameter.
(stmt_can_throw_external): Likewise.
(stmt_can_throw_internal): Likewise.
* tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
(lower_eh_constructs_2): Likewise.
(stmt_could_throw_p): Add fun parameter, use it instead of cfun.
(stmt_can_throw_external): Likewise.
(stmt_can_throw_internal): Likewise.
(maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
(maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
(maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
(maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
(pass_lower_eh_dispatch::execute): Pass cfun to
stmt_can_throw_external.
(cleanup_empty_eh): Likewise.
(verify_eh_edges): Pass cfun to stmt_could_throw_p.
* cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
stmt_can_throw_external instead of pushing it to cfun.
(symbol_table::create_edge): Likewise.
* gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
stmt_can_throw_internal.
* gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
to stmt_could_throw_p.
* gimple-ssa-store-merging.c (handled_load): Pass cfun to
stmt_can_throw_internal.
(pass_store_merging::execute): Likewise.
* gimple-ssa-strength-reduction.c
(find_candidates_dom_walker::before_dom_children): Pass cfun to
stmt_could_throw_p.
* gimplify-me.c (gimple_regimplify_operands): Pass cfun to
stmt_can_throw_internal.
* ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
to stmt_can_throw_external.
(check_stmt): Pass cfun to stmt_could_throw_p.
(check_stmt): Pass cfun to stmt_can_throw_external.
(pass_nothrow::execute): Likewise.
* trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
* tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
stmt_can_throw_internal.
(verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
(stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
(gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
* tree-complex.c (expand_complex_libcall): Pass cfun to
stmt_could_throw_p and to stmt_can_throw_internal.
(expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
* tree-inline.c (copy_edges_for_bb): Likewise.
(maybe_move_debug_stmts_to_successors): Likewise.
* tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
stmt_could_throw_p.
* tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
* tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
* tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
stmt_can_throw_internal.
* tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
* tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
stmt_could_throw_p.
(mark_aliased_reaching_defs_necessary_1): Pass cfun to
stmt_can_throw_internal.
* tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
* tree-ssa-loop-im.c (movement_possibility): Pass cfun to
stmt_could_throw_p.
* tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
(add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
* tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
(convert_mult_to_fma_1): Likewise.
(convert_to_divmod): Likewise.
* tree-ssa-phiprop.c (propagate_with_phi): Likewise.
* tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
* tree-ssa-propagate.c
(substitute_and_fold_dom_walker::before_dom_children): Likewise.
* tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
(maybe_optimize_range_tests): Likewise.
(linearize_expr_tree): Likewise.
(reassociate_bb): Likewise.
* tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
* tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
* tree-ssa-strlen.c (adjust_last_stmt): Likewise.
(handle_char_store): Likewise.
* tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
stmt_can_throw_internal.
* tree-vect-patterns.c (check_bool_pattern): Pass cfun to
stmt_could_throw_p.
* tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
(vectorizable_call): Pass cfun to stmt_can_throw_internal.
(vectorizable_simd_clone_call): Likewise.
* value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
(gimple_stringop_fixed_value): Likewise.
iii [Mon, 22 Oct 2018 08:21:03 +0000 (08:21 +0000)]
S/390: Make "b" constraint match literal pool references
Improves the code generation by getting rid of redundant LAs, as seen
in the following example:
- la %r1,0(%r13)
- lg %r4,0(%r1)
+ lg %r4,0(%r13)
Also allows to proceed with the merge of movdi_64 and movdi_larl.
Currently LRA decides to spill literal pool references back to the
literal pool, because it preliminarily chooses alternatives with
CT_MEMORY constraints without calling
satisfies_memory_constraint_p (). Later on it notices that the
constraint is wrong and fixes it by spilling. The constraint in this
case is "b", and the operand is a literal pool reference. There is
no reason to reject them. The current behavior was introduced,
apparently unintentionally, by
https://gcc.gnu.org/ml/gcc-patches/2010-09/msg00812.html
The patch affects a little bit more than mentioned in the subject,
because it changes s390_loadrelative_operand_p (), which is called not
only for checking the "b" constraint. However, the only caller for
which it should really not accept literal pool references is
s390_check_qrst_address (), so it was changed to explicitly do so.
gcc/ChangeLog:
2018-10-22 Ilya Leoshkevich <iii@linux.ibm.com>
* config/s390/s390.c (s390_loadrelative_operand_p): Accept
literal pool references.
(s390_check_qrst_address): Adapt to the new behavior of
s390_loadrelative_operand_p ().
hjl [Sun, 21 Oct 2018 20:28:56 +0000 (20:28 +0000)]
i386: Enable AVX512 memory broadcast for FNMSUB
Many AVX512 vector operations can broadcast from a scalar memory source.
This patch enables memory broadcast for FNMSUB operations. In order to
support AVX512 memory broadcast for FNMSUB, FNMSUB builtin functions are
also added, instead of passing the negated value to FMA builtin functions.
gcc/
PR target/72782
* config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
__builtin_ia32_vfnmsubpd512_mask.
(_mm512_mask_fnmsub_round_pd): Likewise.
(_mm512_fnmsub_pd): Likewise.
(_mm512_mask_fnmsub_pd): Likewise.
(_mm512_maskz_fnmsub_round_pd): Use
__builtin_ia32_vfnmsubpd512_maskz.
(_mm512_maskz_fnmsub_pd): Likewise.
(_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
(_mm512_mask_fnmsub_round_ps): Likewise.
(_mm512_fnmsub_ps): Likewise.
(_mm512_mask_fnmsub_ps): Likewise.
(_mm512_maskz_fnmsub_round_ps): Use
__builtin_ia32_vfnmsubps512_maskz.
(_mm512_maskz_fnmsub_ps): Likewise.
* config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
__builtin_ia32_vfnmsubpd256_mask.
(_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
(_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
(_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
(_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
(_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
(_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
(_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
(_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
* config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
__builtin_ia32_vfnmsubpd.
(_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
(_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
(_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
(_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
(_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
* config/i386/i386-builtin.def: Add
__builtin_ia32_vfnmsubpd256_mask,
__builtin_ia32_vfnmsubpd256_maskz,
__builtin_ia32_vfnmsubpd128_mask,
__builtin_ia32_vfnmsubpd128_maskz,
__builtin_ia32_vfnmsubps256_mask,
__builtin_ia32_vfnmsubps256_maskz,
__builtin_ia32_vfnmsubps128_mask,
__builtin_ia32_vfnmsubps128_maskz,
__builtin_ia32_vfnmsubpd512_mask,
__builtin_ia32_vfnmsubpd512_maskz,
__builtin_ia32_vfnmsubps512_mask,
__builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
__builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
__builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
__builtin_ia32_vfnmsubpd256.
* config/i386/sse.md (fma4i_fnmsub_<mode>): New.
(<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
(*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
Likewise.
(*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
Likewise.
(*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
Likewise.
(fmai_vmfnmsub_<mode><round_name>): Likewise.
hjl [Sun, 21 Oct 2018 20:27:09 +0000 (20:27 +0000)]
i386: Enable AVX512 memory broadcast for FNMADD
Many AVX512 vector operations can broadcast from a scalar memory source.
This patch enables memory broadcast for FNMADD operations. In order to
support AVX512 memory broadcast for FNMADD, FNMADD builtin functions are
also added, instead of passing the negated value to FMA builtin functions.
gcc/
PR target/72782
* config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
__builtin_ia32_vfnmaddpd512_mask.
(_mm512_mask_fnmadd_round_pd): Likewise.
(_mm512_fnmadd_pd): Likewise.
(_mm512_mask_fnmadd_pd): Likewise.
(_mm512_maskz_fnmadd_round_pd): Use
__builtin_ia32_vfnmaddpd512_maskz.
(_mm512_maskz_fnmadd_pd): Likewise.
(_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
(_mm512_mask_fnmadd_round_ps): Likewise.
(_mm512_fnmadd_ps): Likewise.
(_mm512_mask_fnmadd_ps): Likewise.
(_mm512_maskz_fnmadd_round_ps): Use
__builtin_ia32_vfnmaddps512_maskz.
(_mm512_maskz_fnmadd_ps): Likewise.
* config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
__builtin_ia32_vfnmaddpd256_mask.
(_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
(_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
(_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
(_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
(_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
(_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
(_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
(_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
* config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
__builtin_ia32_vfnmaddpd.
(_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
(_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
(_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
(_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
(_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
* config/i386/i386-builtin.def: Add
__builtin_ia32_vfnmaddpd256_mask,
__builtin_ia32_vfnmaddpd256_maskz,
__builtin_ia32_vfnmaddpd128_mask,
__builtin_ia32_vfnmaddpd128_maskz,
__builtin_ia32_vfnmaddps256_mask,
__builtin_ia32_vfnmaddps256_maskz,
__builtin_ia32_vfnmaddps128_mask,
__builtin_ia32_vfnmaddps128_maskz,
__builtin_ia32_vfnmaddpd512_mask,
__builtin_ia32_vfnmaddpd512_maskz,
__builtin_ia32_vfnmaddps512_mask,
__builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
__builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
__builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
__builtin_ia32_vfnmaddpd256.
* config/i386/sse.md (fma4i_fnmadd_<mode>): New.
(<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
(*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
Likewise.
(*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
Likewise.
(*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
Likewise.
(fmai_vmfnmadd_<mode><round_name>): Likewise.
hjl [Sun, 21 Oct 2018 20:24:50 +0000 (20:24 +0000)]
Enable AVX512 memory broadcast for FMSUB
Many AVX512 vector operations can broadcast from a scalar memory source.
This patch enables memory broadcast for FMSUB operations. In order to
support AVX512 memory broadcast for FMSUB, FMSUB builtin functions are
also added, instead of passing the negated value to FMA builtin functions.
gcc/
PR target/72782
* config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
__builtin_ia32_vfmsubpd512_mask.
(_mm512_mask_fmsub_round_pd): Likewise.
(_mm512_fmsub_pd): Likewise.
(_mm512_mask_fmsub_pd): Likewise.
(_mm512_maskz_fmsub_round_pd): Use
__builtin_ia32_vfmsubpd512_maskz.
(_mm512_maskz_fmsub_pd): Likewise.
(_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
(_mm512_mask_fmsub_round_ps): Likewise.
(_mm512_fmsub_ps): Likewise.
(_mm512_mask_fmsub_ps): Likewise.
(_mm512_maskz_fmsub_round_ps): Use
__builtin_ia32_vfmsubps512_maskz.
(_mm512_maskz_fmsub_ps): Likewise.
* config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
__builtin_ia32_vfmsubpd256_mask.
(_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
(_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
(_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
(_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
(_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
(_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
(_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
(_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
* config/i386/fmaintrin.h (_mm_fmsub_pd): Use
__builtin_ia32_vfmsubpd.
(_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
(_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
(_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
(_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
(_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
* config/i386/i386-builtin.def: Add
__builtin_ia32_vfmsubpd256_mask,
__builtin_ia32_vfmsubpd256_maskz,
__builtin_ia32_vfmsubpd128_mask,
__builtin_ia32_vfmsubpd128_maskz,
__builtin_ia32_vfmsubps256_mask,
__builtin_ia32_vfmsubps256_maskz,
__builtin_ia32_vfmsubps128_mask,
__builtin_ia32_vfmsubps128_maskz,
__builtin_ia32_vfmsubpd512_mask,
__builtin_ia32_vfmsubpd512_maskz,
__builtin_ia32_vfmsubps512_mask,
__builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
__builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
__builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
__builtin_ia32_vfmsubpd256.
* config/i386/sse.md (fma4i_fmsub_<mode>): New.
(<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
(*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
Likewise.
(*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
Likewise.
(*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
Likewise.
(fmai_vmfmsub_<mode><round_name>): Likewise.
fdumont [Sat, 20 Oct 2018 20:00:45 +0000 (20:00 +0000)]
2018-10-20 François Dumont <fdumont@gcc.gnu.org>
* testsuite/util/testsuite_containers.h
(forward_members_unordered<>::forward_members_unordered
(const value_type&)): Add local_iterator pre and post increment checks.
* config/abi/pre/gnu.ver: Add GLIBCXX_3.4.26 new symbol.
mpolacek [Sat, 20 Oct 2018 17:21:19 +0000 (17:21 +0000)]
* g++.dg/*.C: Use target c++17 instead of explicit dg-options.
* lib/g++-dg.exp: Don't test C++11 by default. Add C++17 to
the list of default stds to test.
jakub [Fri, 19 Oct 2018 22:52:06 +0000 (22:52 +0000)]
PR middle-end/85488
PR middle-end/87649
* omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
depend closely nested inside of loop with ordered clause with
a parameter.
* c-c++-common/gomp/doacross-2.c: New test.
* c-c++-common/gomp/sink-3.c: Expect another error during error
recovery.
redi [Fri, 19 Oct 2018 21:50:03 +0000 (21:50 +0000)]
Fix testsuite failures due to extra errors in strict dialects
When __STRICT_ANSI__ is defined the incorrect allocators used in these
tests also trigger and additional static assertion. Prune those extra
errors so that the tests don't fail when built with strict dialects.
redi [Fri, 19 Oct 2018 21:49:49 +0000 (21:49 +0000)]
Conditionally disable tests of non-standard extensions
These tests include uses of the extension to allow allocators with the
wrong value_type in containers. Skip those parts of the tests when
__STRICT_ANIS__ is defined.
redi [Fri, 19 Oct 2018 21:49:40 +0000 (21:49 +0000)]
Fix tests that use allocators with incorrect value types
As a GNU extension we allow containers to be instantiated with
allocators that use a different value type from the container, and
automatically rebind the allocator to the correct type. This extension
is disabled in strict modes (when __STRICT_ANSI__ is defined, i.e.
-std=c++NN dialects). These testcases unintentionally rely on the
extension and so fail for strict modes.
Tests which intentionally make use of the extension will still fail in
strict dialects, but will be addressed in a later change.
redi [Fri, 19 Oct 2018 21:49:19 +0000 (21:49 +0000)]
Remove duplicate tests
These tests originally existed to check the containers in C++11 mode,
when the default was C++98 mode. Now that the default is C++14 (and we
run most tests for all modes) it serves no purpose to have two copies of
the tests when neither is explicitly using -std=gnu++98 anyway.
* testsuite/23_containers/list/requirements/explicit_instantiation/
5_c++0x.cc: Remove redundant test that is functionally identical to
the 5.cc test.
* testsuite/23_containers/map/requirements/explicit_instantiation/
5_c++0x.cc: Likewise.
* testsuite/23_containers/multimap/requirements/explicit_instantiation/
5_c++0x.cc: Likewise.
* testsuite/23_containers/multiset/requirements/explicit_instantiation/
5_c++0x.cc: Likewise.
* testsuite/23_containers/set/requirements/explicit_instantiation/
5_c++0x.cc: Likewise.
dmalcolm [Fri, 19 Oct 2018 19:50:02 +0000 (19:50 +0000)]
gccint.texi: add user experience guidelines
gcc/ChangeLog:
* Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
* doc/gccint.texi: Include ux.texi and use it in top-level menu.
* doc/ux.texi: New file.
ian [Fri, 19 Oct 2018 19:43:47 +0000 (19:43 +0000)]
compiler: don't export any functions with special names
This keeps init functions from appearing in the export data. Checking
for special names in general means that we don't need to check
specifically for nested functions or thunks, which have special names.
wschmidt [Fri, 19 Oct 2018 18:28:11 +0000 (18:28 +0000)]
[gcc]
2018-10-19 Bill Schmidt <wschmidt@linux.ibm.com>
PR tree-optimization/87473
* gimple-ssa-strength-reduction.c (record_phi_increments_1): For
phi arguments identical to the base expression of the phi
candidate, record a phi-adjust increment of zero minus the index
expression of the hidden basis.
(phi_incr_cost_1): For phi arguments identical to the base
expression of the phi candidate, the difference to compare against
the increment is zero minus the index expression of the hidden
basis, and there is no potential savings from replacing the (phi)
statement.
(ncd_with_phi): For phi arguments identical to the base expression
of the phi candidate, the difference to compare against the
increment is zero minus the index expression of the hidden basis.
(all_phi_incrs_profitable_1): For phi arguments identical to the
base expression of the phi candidate, the increment to be checked
for profitability is zero minus the index expression of the hidden
basis.
[gcc/testsuite]
2018-10-19 Bill Schmidt <wschmidt@linux.ibm.com>
PR tree-optimization/87473
* gcc.c-torture/compile/pr87473.c: New file.
segher [Fri, 19 Oct 2018 15:40:57 +0000 (15:40 +0000)]
rs6000: Put CR0 first in REG_ALLOC_ORDER
IRA and LRA prefer to use CR7 (which is first in REG_ALLOC_ORDER) over
CR0, although the latter often is cheaper ("x" vs. "y" constraints).
We should figure out why this is and fix it; but until that is done,
this patch makes CR0 the first allocated register: it improves the
current code, and it is required for later patches to be effective.
(It changes two testcases to no longer look at what CR field is
allocated).
* config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
be the first CR field allocated.
gcc/testsuite/
* gcc.target/powerpc/safe-indirect-jump-2.c: Do not check assigned CR
field number.
* gcc.target/powerpc/safe-indirect-jump-3.c: Ditto.
redi [Fri, 19 Oct 2018 13:46:24 +0000 (13:46 +0000)]
Fix compilation error with _GLIBCXX_PARALLEL
* include/bits/regex_executor.tcc (_Backref_matcher::_M_apply): Use
_GLIBCXX_STD_A to refer to normal mode algorithms.
* testsuite/28_regex/headers/regex/parallel_mode.cc: New test.
* testsuite/28_regex/headers/regex/std_c++0x_neg.cc: Remove empty
whitespace.
iii [Fri, 19 Oct 2018 08:33:52 +0000 (08:33 +0000)]
lra: fix spill_hard_reg_in_range clobber check
FROM..TO range might contain NOTE_INSN_DELETED insns, for which the
corresponding entries in lra_insn_recog_data[] are NULLs. Example from
the problematic code from PR87596:
(note 148 154 68 7 NOTE_INSN_DELETED)
lra_insn_recog_data[] is used directly only when the insn in question
is taken from insn_bitmap, which is not the case here. In other
situations lra_get_insn_recog_data () guarded by INSN_P () or other
stricter predicate are used. So we need to do this here as well.
A tiny detail worth noting: I put the INSN_P () check before the
insn_bitmap check, because I believe that insn_bitmap can contain only
real insns anyway.
gcc/ChangeLog:
2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
PR rtl-optimization/87596
* lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
lra_get_insn_recog_data () instead of lra_insn_recog_data[]
for instructions in FROM..TO range.
gcc/testsuite/ChangeLog:
2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
PR rtl-optimization/87596
* gcc.target/i386/pr87596.c: New test.
ebotcazou [Fri, 19 Oct 2018 07:17:20 +0000 (07:17 +0000)]
* cfgexpand.c (expand_one_var): Use specific wording in error message
for non-local frame variables.
* stor-layout.c (layout_decl): Do not issue a warning for them.
ian [Thu, 18 Oct 2018 23:26:20 +0000 (23:26 +0000)]
compiler: add COMPARE_ALIASES flag for type compare and hash
Normally aliases compare as identical to the underlying type. Add a
COMPARE_ALIASES flag to let them compare (and hash) differently. This
will be used by later patches in this series.
ian [Thu, 18 Oct 2018 23:22:01 +0000 (23:22 +0000)]
compiler: list indirect imports separately in export data
Previously when export data referred to a type that was not defined in
a directly imported package, we would write the package name as
additional information in the type's export data. That approach
required all type information to be read in order. This patch changes
the compiler to find all references to indirectly imported packages,
and write them out as an indirectimport line in the import data. This
will permit us to read exported type data out of order.
The type traversal used to find indirect imports is a little more
complicated than necessary in preparation for later patches in this
series.
redi [Thu, 18 Oct 2018 20:04:55 +0000 (20:04 +0000)]
Fix tests that fail when built with different options
* testsuite/20_util/duration/cons/2.cc: Add -ffloat-store to fix
failure when compiled without optimisation.
* testsuite/ext/profile/mutex_extensions_neg.cc: Prune additional
errors caused by C++17 std::pmr alias templates.
redi [Thu, 18 Oct 2018 19:57:25 +0000 (19:57 +0000)]
PR libstdc++/87642 handle multibyte thousands separators from libc
If a locale's THOUSANDS_SEP or MON_THOUSANDS_SEP string is not a
single character we either need to narrow it to a single char or
ignore it (and therefore disable digit grouping for that facet).
PR libstdc++/87642
* config/locale/gnu/monetary_members.cc
(moneypunct<char, true>::_M_initialize_moneypunct): Use
__narrow_multibyte_chars to convert multibyte thousands separators
to a single char.
* config/locale/gnu/numeric_members.cc
(numpunct<char>::_M_initialize_numpunct): Likewise.
(__narrow_multibyte_chars): New function.
projects / thirdparty / gcc.git / log
