gcc/
2018-07-06 Kito Cheng <kito.cheng@gmail.com>
* config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
(riscv_expand_epilogue): Add assertion to check interrupt mode.
(riscv_set_current_function): Extract getting interrupt type to new
function.
(riscv_get_interrupt_type): New function.
(riscv_merge_decl_attributes): New function, checking interrupt type is
same.
(TARGET_MERGE_DECL_ATTRIBUTES): Define.
gcc/testsuite/
2018-07-06 Kito Cheng <kito.cheng@gmail.com>
* gcc.target/riscv/interrupt-conflict-mode.c: New.
From-SVN: r262607
* c-attribs.c (c_common_attribute_table): Add
"omp declare target implicit" attribute.
* c-decl.c (c_decl_attributes): Don't diagnose vars without mappable
type here, instead add "omp declare target implicit" attribute.
(finish_decl): Diagnose vars without mappable type here.
* decl2.c (cplus_decl_attributes): Don't diagnose vars without mappable
type here, instead add "omp declare target implicit" attribute. Add
that attribute instead of "omp declare target" also when
processing_template_decl.
* decl.c (cp_finish_decl): Diagnose vars without mappable type here,
and before calling cp_omp_mappable_type call complete_type.
* c-c++-common/gomp/declare-target-3.c: New test.
* g++.dg/gomp/declare-target-2.C: New test.
From-SVN: r262605
This patch uses IFN_COND_* to vectorise conditionally-executed,
potentially-trapping arithmetic, such as most floating-point
ops with -ftrapping-math. E.g.:
if (cond) { ... x = a + b; ... }
becomes:
...
x = .COND_ADD (cond, a, b, else_value);
...
When this transformation is done on its own, the value of x for
!cond isn't important, so else_value is simply the target's
preferred_else_value (i.e. the value it can handle the most
efficiently).
However, the patch also looks for the equivalent of:
y = cond ? x : c;
in which the "then" value is the result of the conditionally-executed
operation and the "else" value "c" is some value that is available at x.
In that case we can instead use:
x = .COND_ADD (cond, a, b, c);
and replace uses of y with uses of x.
The patch also looks for:
y = !cond ? c : x;
which can be transformed in the same way. This involved adding a new
utility function inverse_conditions_p, which was already open-coded
in a more limited way in match.pd.
2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
gcc/
* fold-const.h (inverse_conditions_p): Declare.
* fold-const.c (inverse_conditions_p): New function.
* match.pd: Use inverse_conditions_p. Add folds of view_converts
that test the inverse condition of a conditional internal function.
* internal-fn.h (vectorized_internal_fn_supported_p): Declare.
* internal-fn.c (internal_fn_mask_index): Handle conditional
internal functions.
(vectorized_internal_fn_supported_p): New function.
* tree-if-conv.c: Include internal-fn.h and fold-const.h.
(any_pred_load_store): Replace with...
(need_to_predicate): ...this new variable.
(redundant_ssa_names): New variable.
(ifcvt_can_use_mask_load_store): Move initial checks to...
(ifcvt_can_predicate): ...this new function. Handle tree codes
for which a conditional internal function exists.
(if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
instead of ifcvt_can_use_mask_load_store. Update after variable
name change.
(predicate_load_or_store): New function, split out from
predicate_mem_writes.
(check_redundant_cond_expr): New function.
(value_available_p): Likewise.
(predicate_rhs_code): Likewise.
(predicate_mem_writes): Rename to...
(predicate_statements): ...this. Use predicate_load_or_store
and predicate_rhs_code.
(combine_blocks, tree_if_conversion): Update after above name changes.
(ifcvt_local_dce): Handle redundant_ssa_names.
* tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
general conditional functions.
* tree-vect-stmts.c (vectorizable_call): Likewise.
gcc/testsuite/
* gcc.dg/vect/vect-cond-arith-4.c: New test.
* gcc.dg/vect/vect-cond-arith-5.c: Likewise.
* gcc.target/aarch64/sve/cond_arith_1.c: Likewise.
* gcc.target/aarch64/sve/cond_arith_1_run.c: Likewise.
* gcc.target/aarch64/sve/cond_arith_2.c: Likewise.
* gcc.target/aarch64/sve/cond_arith_2_run.c: Likewise.
* gcc.target/aarch64/sve/cond_arith_3.c: Likewise.
* gcc.target/aarch64/sve/cond_arith_3_run.c: Likewise.
From-SVN: r262589
This patch adds support for fusing a conditional add or subtract
with a multiplication, so that we can use fused multiply-add and
multiply-subtract operations for fully-masked reductions. E.g.
for SVE we vectorise:
double res = 0.0;
for (int i = 0; i < n; ++i)
res += x[i] * y[i];
using a fully-masked loop in which the loop body has the form:
res_1 = PHI<0(preheader), res_2(latch)>;
avec = .MASK_LOAD (loop_mask, a)
bvec = .MASK_LOAD (loop_mask, b)
prod = avec * bvec;
res_2 = .COND_ADD (loop_mask, res_1, prod, res_1);
where the last statement does the equivalent of:
res_2 = loop_mask ? res_1 + prod : res_1;
(operating elementwise). The point of the patch is to convert the last
two statements into:
res_s = .COND_FMA (loop_mask, avec, bvec, res_1, res_1);
which is equivalent to:
res_2 = loop_mask ? fma (avec, bvec, res_1) : res_1;
(again operating elementwise).
2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
Alan Hayward <alan.hayward@arm.com>
David Sherwood <david.sherwood@arm.com>
gcc/
* internal-fn.h (can_interpret_as_conditional_op_p): Declare.
* internal-fn.c (can_interpret_as_conditional_op_p): New function.
* tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
plus and minus and convert them into IFN_COND_FMA-based sequences.
(convert_mult_to_fma): Handle conditional plus and minus.
gcc/testsuite/
* gcc.dg/vect/vect-fma-2.c: New test.
* gcc.target/aarch64/sve/reduc_4.c: Likewise.
* gcc.target/aarch64/sve/reduc_6.c: Likewise.
* gcc.target/aarch64/sve/reduc_7.c: Likewise.
Co-Authored-By: Alan Hayward <alan.hayward@arm.com>
Co-Authored-By: David Sherwood <david.sherwood@arm.com>
From-SVN: r262588
This patch adds match.pd support for applying normal folds to their
IFN_COND_* forms. E.g. the rule:
(plus @0 (negate @1)) -> (minus @0 @1)
also allows the fold:
(IFN_COND_ADD @0 @1 (negate @2) @3) -> (IFN_COND_SUB @0 @1 @2 @3)
Actually doing this by direct matches in gimple-match.c would
probably lead to combinatorial explosion, so instead, the patch
makes gimple_match_op carry a condition under which the operation
happens ("cond"), and the value to use when the condition is false
("else_value"). Thus in the example above we'd do the following
(a) convert:
cond:NULL_TREE (IFN_COND_ADD @0 @1 @4 @3) else_value:NULL_TREE
to:
cond:@0 (plus @1 @4) else_value:@3
(b) apply gimple_resimplify to (plus @1 @4)
(c) reintroduce cond and else_value when constructing the result.
Nested operations inherit the condition of the outer operation
(so that we don't introduce extra faults) but have a null else_value.
If we try to build such an operation, the target gets to choose what
else_value it can handle efficiently: obvious choices include one of
the operands or a zero constant. (The alternative would be to have some
representation for an undefined value, but that seems a bit invasive,
and isn't likely to be useful here.)
I've made the condition a mandatory part of the gimple_match_op
constructor so that it doesn't accidentally get dropped.
2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
gcc/
* target.def (preferred_else_value): New target hook.
* doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
* doc/tm.texi: Regenerate.
* targhooks.h (default_preferred_else_value): Declare.
* targhooks.c (default_preferred_else_value): New function.
* internal-fn.h (conditional_internal_fn_code): Declare.
* internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
(get_conditional_internal_fn): Use it.
(conditional_internal_fn_code): New function.
* gimple-match.h (gimple_match_cond): New struct.
(gimple_match_op): Add a cond member function.
(gimple_match_op::gimple_match_op): Update all forms to take a
gimple_match_cond.
* genmatch.c (expr::gen_transform): Use the same condition as res_op
for the suboperation, but don't specify a particular else_value.
* tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
(visit_nary_op, visit_reference_op_load): Pass
gimple_match_cond::UNCOND to the gimple_match_op constructor.
* gimple-match-head.c: Include tree-eh.h
(convert_conditional_op): New function.
(maybe_resimplify_conditional_op): Likewise.
(gimple_resimplify1): Call maybe_resimplify_conditional_op.
(gimple_resimplify2): Likewise.
(gimple_resimplify3): Likewise.
(gimple_resimplify4): Likewise.
(maybe_push_res_to_seq): Return null for conditional operations.
(try_conditional_simplification): New function.
(gimple_simplify): Call it. Pass conditions to the gimple_match_op
constructor.
* match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
IFN_COND_* call.
* config/aarch64/aarch64.c (aarch64_preferred_else_value): New
function.
(TARGET_PREFERRED_ELSE_VALUE): Redefine.
gcc/testsuite/
* gcc.dg/vect/vect-cond-arith-2.c: New test.
* gcc.target/aarch64/sve/loop_add_6.c: Likewise.
From-SVN: r262586
* lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
DECL_FCONTEXT
(hash_tree): Do not hash DECL_FCONTEXT
* tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
Do not stream DECL_FCONTEXT.
* tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
* tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
From-SVN: r262585
2018-07-12 Richard Biener <rguenther@suse.de>
PR debug/86462
* dwarf2out.c (gen_block_die): Only output blocks when they have
at least one !DECL_IGNORED_P variable.
From-SVN: r262583
2018-07-12 Richard Biener <rguenther@suse.de>
* tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
left-over from last patch.
From-SVN: r262578
PR tree-optimization/86492
* gimple-ssa-store-merging.c
(imm_store_chain_info::coalesce_immediate_stores): Call
check_no_overlap even for the merge_overlapping case. Formatting fix.
* gcc.c-torture/execute/pr86492.c: New test.
From-SVN: r262576
2018-07-12 Richard Biener <rguenther@suse.de>
PR c/86453
* c-attribs.c (handle_packed_attribute): Do not build a variant
type with TYPE_PACKED, instead ignore the attribute if we may
not apply to the original type.
* g++.dg/warn/pr86453.C: New testcase.
From-SVN: r262575
2018-07-12 Richard Biener <rguenther@suse.de>
PR middle-end/86479
* fold-const.c (fold_binary_op_with_conditional_arg): Do not
move possibly trapping operations into the conditional.
* gcc.dg/graphite/pr86479.c: New testcase.
From-SVN: r262574
A type alias and its underlying type are identical. Build a
single backend type for them. Previously we build two backend
types, which sometimes confuse the backend's type system.
Also don't include type aliases into the list of named type
declarations, since they are not named types.
Reviewed-on: https://go-review.googlesource.com/123362
From-SVN: r262572
* config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
Use __mmask64 type instead of __mmask8 for __M argument.
* config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
_mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
__U argument.
(_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
__mmask16 for __M argument.
(_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
_mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
to __mmask16 instead of __mmask8.
* config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
_mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
_mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
_mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
instead of __mmask16 for __U argument.
* config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
__mmask16 instead of __mmask8 for __U argument.
(_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
__U argument.
(_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
__mmask16.
(_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
argument.
(_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
__U argument.
(_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
__mmask16.
(_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
of __mmask16.
(_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
__U argument.
(_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
__U argument.
(_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
__U argument.
(_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
__U argument.
(_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
_mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
return type as well as __M argument type and all casts from __mmask8
to __mmask32.
(_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
_mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
return type as well as __M argument type and all casts from __mmask8
to __mmask16.
(_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
_mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
return type as well as __M argument type and all casts from __mmask8
to __mmask32.
(_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
_mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
return type as well as __M argument type and all casts from __mmask8
to __mmask16.
* config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
_mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
__mmask16.
* gcc.target/i386/avx512bw-vpcmpb-2.c (CMP): Use SIZE macro instead
of hardcoding size. Cast (rel) to MASK_TYPE.
* gcc.target/i386/avx512bw-vpcmpub-2.c (CMP): Likewise.
* gcc.target/i386/avx512f-vinserti32x4-3.c: New test.
* gcc.target/i386/avx512f-vinsertf32x4-3.c: New test.
* gcc.target/i386/avx512vl-vpcmpnequb-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpgeub-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpleb-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpgeb-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpltb-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpltub-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpleub-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpneqb-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpnequw-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpgeuw-2.c: New test.
* gcc.target/i386/avx512vl-vpcmplew-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpgew-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpltw-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpltuw-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpleuw-2.c: New test.
* gcc.target/i386/avx512vl-vpcmpneqw-2.c: New test.
2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
* config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
_mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
for __U argument.
* gcc.target/i386/avx512bw-vpcmpb-2.c (SIZE): Define to
(AVX512F_LEN / 8) instead of (AVX512F_LEN / 16).
* gcc.target/i386/avx512bw-vpcmpub-2.c (SIZE): Likewise.
From-SVN: r262566
* lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
stream DECL_ORIGINAL_TYPE.
(DFS::DFS_write_tree_body): Drop hack handling local external decls.
(hash_tree): Do not walk DECL_ORIGINAL_TYPE.
* tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
Do not walk original type.
* tree-streamer-out.c (streamer_write_chain): Drop hack handling
external decls.
(write_ts_decl_non_common_tree_pointers): Do not stream
DECL_ORIGINAL_TYPE
* tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
(find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
From-SVN: r262560
As part of the transition from the original support for named FPUs to
general FPU properties I defined an entry in the CPU definitions in
arm-cpus.in to use a named FPU. However, that has now outlived its
usefulness and increasingly we are likely to find that newer cores do
not fit the legacy FPU names very well. Furthermore it is now possible
to encode all the FPU capatilities directly in the ISA definitions, often
as simply as using +fp or +simd.
So this patch removes the fpu field from the "define cpu" entries and
instead encodes the same information in the isa field. This also alows us
to remove a bit of now-dead code from parsecpu.awk.
* config/arm/arm-cpus.in: Move information from fpu field of each
cpu definition to the isa field.
* config/arm/parsecpu.awk (fpu): Delete match rule.
(gen_comm_data): Don't add bits from the CPU's FPU entry.
From-SVN: r262556
The spec says that when an index expression appears on the left hand
side of an assignment, the operands should be evaluated. The
gofrontend code was assuming that that only referred to the index
operand. But discussion of https://golang.org/issue/23188 has
clarified that this means both the slice/map/string operand and the
index operand. Adjust the gofrontend code accordingly.
Fixesgolang/go#23188
Reviewed-on: https://go-review.googlesource.com/123155
From-SVN: r262554
2018-07-11 Richard Biener <rguenther@suse.de>
PR debug/86457
* dwarf2out.c (init_sections_and_labels): Use
output_asm_line_debug_info consistently.
(dwarf2out_early_finish): Likewise.
(dwarf2out_finish): Remove DW_AT_stmt_list from early generated
type units.
From-SVN: r262551
2018-07-10 Richard Biener <rguenther@suse.de>
* tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
Rework father_bb setting in a way to avoid propagating constants
multiple times on a loop body.
From-SVN: r262550
There was a typo in the output_macinfo_op gcc_assert.
The function is called dwarf_FORM, not dwarf_form.
Add the provided testcase from the bug to test -gsplit-dwarf -g3.
gcc/ChangeLog:
PR debug/86459
* dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
gcc/testsuite/ChangeLog:
PR debug/86459
* gcc.dg/pr86459.c: New test.
From-SVN: r262545
