| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
OpenACC 2.6 specifies that the array descriptor (when present) must be
copied to the target before attaching pointers in Fortran. This patch
reverses the stripping of GOMP_MAP_TO_PSET and GOMP_MAP_POINTER that
was introduced by the "OpenACC reference count overhaul" patch.
2020-07-10 Julian Brown <julian@codesourcery.com>
Thomas Schwinge <thomas@codesourcery.com>
gcc/
* gimplify.c (gimplify_scan_omp_clauses): Do not strip
GOMP_MAP_TO_PSET/GOMP_MAP_POINTER for OpenACC enter/exit data
directives (see also PR92929).
gcc/testsuite/
* gfortran.dg/goacc/finalize-1.f: Update expected dump output.
libgomp/
* testsuite/libgomp.oacc-fortran/dynamic-pointer-1.f90: New test.
Co-Authored-By: Thomas Schwinge <thomas@codesourcery.com>
|
|
The GIMPLE store merging pass doesn't merge STRING_CSTs in the general
case, although they are accepted by native_encode_expr; the reason is
that the pass only works with integral modes, i.e. with chunks whose
size is a power of two.
There are two possible ways of extending it to handle STRING_CSTs:
1) lift the condition of integral modes and treat STRING_CSTs as
other _CST nodes but with arbitrary size; 2) implement a specific
and separate handling for STRING_CSTs.
The attached patch implements 2) for the following reasons: on the
one hand, even in Ada where character strings are first-class citizens,
cases where merging STRING_CSTs with other *_CST nodes would be possible
are quite rare in practice; on the other hand, string concatenations
happen more naturally and frequently thanks to the "&" operator, giving
rise to merging opportunities.
gcc/ChangeLog:
* gimple-fold.c (gimple_fold_builtin_memory_op): Fold calls that
were initially created for the assignment of a variable-sized
object and whose source is now a string constant.
* gimple-ssa-store-merging.c (struct merged_store_group): Document
STRING_CST for rhs_code field.
Add string_concatenation boolean field.
(merged_store_group::merged_store_group): Initialize it as well as
bit_insertion here.
(merged_store_group::do_merge): Set it upon seeing a STRING_CST.
Also set bit_insertion here upon seeing a BIT_INSERT_EXPR.
(merged_store_group::apply_stores): Clear it for small regions.
Do not create a power-of-2-sized buffer if it is still true.
And do not set bit_insertion here again.
(encode_tree_to_bitpos): Deal with BLKmode for the expression.
(merged_store_group::can_be_merged_into): Deal with STRING_CST.
(imm_store_chain_info::coalesce_immediate_stores): Set bit_insertion
to true after changing MEM_REF stores into BIT_INSERT_EXPR stores.
(count_multiple_uses): Return 0 for STRING_CST.
(split_group): Do not split the group for a string concatenation.
(imm_store_chain_info::output_merged_store): Constify and rename
some local variables. Build an array type as destination type
for a string concatenation, as well as a zero mask, and call
build_string to build the source.
(lhs_valid_for_store_merging_p): Return true for VIEW_CONVERT_EXPR.
(pass_store_merging::process_store): Accept STRING_CST on the RHS.
* gimple.h (gimple_call_alloca_for_var_p): New accessor function.
* gimplify.c (gimplify_modify_expr_to_memcpy): Set alloca_for_var.
* tree.h (CALL_ALLOCA_FOR_VAR_P): Document it for BUILT_IN_MEMCPY.
gcc/testsuite/ChangeLog:
* gnat.dg/opt87.adb: New test.
* gnat.dg/opt87_pkg.ads: New helper.
* gnat.dg/opt87_pkg.adb: Likewise.
|
|
This is not a thorough review of the code, just a few cases I noticed while
scanning 'gcc/gimplify.c' for "enclosing".
gcc/
* gimplify.c (omp_notice_threadprivate_variable)
(omp_default_clause, omp_notice_variable): 'inform' after 'error'
diagnostic. Adjust all users.
gcc/testsuite/
* c-c++-common/gomp/default-1.c: Update.
* c-c++-common/gomp/defaultmap-3.c: Likewise.
* c-c++-common/gomp/order-4.c: Likewise.
* g++.dg/gomp/parallel-2.C: Likewise.
* g++.dg/gomp/predetermined-1.C: Likewise.
* g++.dg/gomp/sharing-1.C: Likewise.
* gcc.dg/gomp/appendix-a/a.24.1.c: Likewise.
* gcc.dg/gomp/parallel-2.c: Likewise.
* gcc.dg/gomp/pr44085.c: Likewise.
* gcc.dg/gomp/sharing-1.c: Likewise.
* gcc.dg/gomp/vla-1.c: Likewise.
* gfortran.dg/gomp/appendix-a/a.24.1.f90: Likewise.
* gfortran.dg/gomp/crayptr3.f90: Likewise.
* gfortran.dg/gomp/pr33439.f90: Likewise.
* gfortran.dg/gomp/pr44036-1.f90: Likewise.
* gfortran.dg/gomp/pr44085.f90: Likewise.
* gfortran.dg/gomp/pr44536.f90: Likewise.
* gfortran.dg/gomp/pr94672.f90: Likewise.
* gfortran.dg/gomp/sharing-1.f90: Likewise.
* gfortran.dg/gomp/sharing-2.f90: Likewise.
* gfortran.dg/gomp/sharing-3.f90: Likewise.
|
|
gcc/ChangeLog:
* Makefile.in: Add new file.
* expr.c (expand_expr_real_2): Add gcc_unreachable as we should
not meet this condition.
(do_store_flag): Likewise.
* gimplify.c (gimplify_expr): Gimplify first argument of
VEC_COND_EXPR to be a SSA name.
* internal-fn.c (vec_cond_mask_direct): New.
(vec_cond_direct): Likewise.
(vec_condu_direct): Likewise.
(vec_condeq_direct): Likewise.
(expand_vect_cond_optab_fn): New.
(expand_vec_cond_optab_fn): Likewise.
(expand_vec_condu_optab_fn): Likewise.
(expand_vec_condeq_optab_fn): Likewise.
(expand_vect_cond_mask_optab_fn): Likewise.
(expand_vec_cond_mask_optab_fn): Likewise.
(direct_vec_cond_mask_optab_supported_p): Likewise.
(direct_vec_cond_optab_supported_p): Likewise.
(direct_vec_condu_optab_supported_p): Likewise.
(direct_vec_condeq_optab_supported_p): Likewise.
* internal-fn.def (VCOND): New OPTAB.
(VCONDU): Likewise.
(VCONDEQ): Likewise.
(VCOND_MASK): Likewise.
* optabs.c (get_rtx_code): Make it global.
(expand_vec_cond_mask_expr): Removed.
(expand_vec_cond_expr): Removed.
* optabs.h (expand_vec_cond_expr): Likewise.
(vector_compare_rtx): Make it global.
* passes.def: Add new pass_gimple_isel pass.
* tree-cfg.c (verify_gimple_assign_ternary): Add check
for VEC_COND_EXPR about first argument.
* tree-pass.h (make_pass_gimple_isel): New.
* tree-ssa-forwprop.c (pass_forwprop::execute): Prevent
propagation of the first argument of a VEC_COND_EXPR.
* tree-ssa-reassoc.c (ovce_extract_ops): Support SSA_NAME as
first argument of a VEC_COND_EXPR.
(optimize_vec_cond_expr): Likewise.
* tree-vect-generic.c (expand_vector_divmod): Make SSA_NAME
for a first argument of created VEC_COND_EXPR.
(expand_vector_condition): Fix coding style.
* tree-vect-stmts.c (vectorizable_condition): Gimplify
first argument.
* gimple-isel.cc: New file.
gcc/testsuite/ChangeLog:
* g++.dg/vect/vec-cond-expr-eh.C: New test.
|
|
OpenMP 5.0 adds support for non-rectangular loop collapses, e.g.
triangular and more complex.
This patch deals just with the diagnostics so that they aren't rejected
immediately as before. As the spec generally requires as before that the
iteration variable initializer and bound in the comparison as invariant
vs. the outermost loop, and just add some exceptional forms that can violate
that, we need to avoid folding the expressions until we can detect them and
in order to avoid folding it later on, I chose to use a TREE_VEC in those
expressions to hold the var_outer * expr1 + expr2 triplet, the patch adds
pretty-printing of that, gimplification etc. and just sorry_at during
omp expansion for now.
The next step will be to implement the different cases of that one by one.
2020-06-16 Jakub Jelinek <jakub@redhat.com>
gcc/
* tree.h (OMP_FOR_NON_RECTANGULAR): Define.
* gimplify.c (gimplify_omp_for): Diagnose schedule, ordered
or dist_schedule clause on non-rectangular loops. Handle
gimplification of non-rectangular lb/b expressions. When changing
iteration variable, adjust also non-rectangular lb/b expressions
referencing that.
* omp-general.h (struct omp_for_data_loop): Add m1, m2 and outer
members.
(struct omp_for_data): Add non_rect member.
* omp-general.c (omp_extract_for_data): Handle non-rectangular
loops. Fill in non_rect, m1, m2 and outer.
* omp-low.c (lower_omp_for): Handle non-rectangular lb/b expressions.
* omp-expand.c (expand_omp_for): Emit sorry_at for unsupported
non-rectangular loop cases and assert for cases that can't be
non-rectangular.
* tree-pretty-print.c (dump_mem_ref): Formatting fix.
(dump_omp_loop_non_rect_expr): New function.
(dump_generic_node): Handle non-rectangular OpenMP loops.
* tree-pretty-print.h (dump_omp_loop_non_rect_expr): Declare.
* gimple-pretty-print.c (dump_gimple_omp_for): Handle non-rectangular
OpenMP loops.
gcc/c-family/
* c-common.h (c_omp_check_loop_iv_exprs): Add an int argument.
* c-omp.c (struct c_omp_check_loop_iv_data): Add maybe_nonrect and
idx members.
(c_omp_is_loop_iterator): New function.
(c_omp_check_loop_iv_r): Use it. Add support for silent scanning
if outer loop iterator is present. Perform duplicate checking through
hash_set in the function rather than expecting caller to do that.
Pass NULL instead of d->ppset to walk_tree_1.
(c_omp_check_nonrect_loop_iv): New function.
(c_omp_check_loop_iv): Use it. Fill in new members, allow
non-rectangular loop forms, diagnose multiple associated loops with
the same iterator. Pass NULL instead of &pset to walk_tree_1.
(c_omp_check_loop_iv_exprs): Likewise.
gcc/c/
* c-parser.c (c_parser_expr_no_commas): Save, clear and restore
c_in_omp_for.
(c_parser_omp_for_loop): Set c_in_omp_for around some calls to avoid
premature c_fully_fold. Defer explicit c_fully_fold calls to after
c_finish_omp_for.
* c-tree.h (c_in_omp_for): Declare.
* c-typeck.c (c_in_omp_for): Define.
(build_modify_expr): Avoid c_fully_fold if c_in_omp_for.
(digest_init): Likewise.
(build_binary_op): Likewise.
gcc/cp/
* semantics.c (handle_omp_for_class_iterator): Adjust
c_omp_check_loop_iv_exprs caller.
(finish_omp_for): Likewise. Don't call fold_build_cleanup_point_expr
before calling c_finish_omp_for and c_omp_check_loop_iv, move it
after those calls.
* pt.c (tsubst_omp_for_iterator): Handle non-rectangular loops.
gcc/testsuite/
* c-c++-common/gomp/loop-6.c: New test.
* gcc.dg/gomp/loop-1.c: Don't expect diagnostics on valid
non-rectangular loops.
* gcc.dg/gomp/loop-2.c: New test.
* g++.dg/gomp/loop-1.C: Don't expect diagnostics on valid
non-rectangular loops.
* g++.dg/gomp/loop-2.C: Likewise.
* g++.dg/gomp/loop-5.C: New test.
* g++.dg/gomp/loop-6.C: New test.
|
|
gimplify_modify_expr_rhs has an optimization whereby the assignment to
an aggregate variable from a read-only object with a DECL_INITIAL is
optimized into the direct assignment of the DECL_INITIAL, provided that
no temporary is created in the process.
The optimization is blocked if the read-only object is volatile, which
is OK as per the semantics of volatile, but also if the target variable
is volatile, on the grounds that the modified assignment might end up
being done on a per field basis, which is also OK. But this latter
restriction is enforced a priori and there are cases where the modified
assignment would be OK, for example if there is only one field or the
DECL_INITIAL is equivalent to the empty CONSTRUCTOR, i.e. all zeros.
So, in the latter case, the patch changes gimplify_modify_expr_rhs to ask
gimplify_init_constructor whether the assignment would be done as a block,
which is easy because gimplify_init_constructor knows that it must create
a temporary if the LHS is volatile and this would not be the case, so it's
just a matter of completing the NOTIFY_TEMP_CREATION mechanism.
gcc/ChangeLog
* gimplify.c (gimplify_init_constructor) <AGGREGATE_TYPE>: Declare
new ENSURE_SINGLE_ACCESS constant and move variables down. If it is
true and all elements are zero, then always clear. Return GS_ERROR
if a temporary would be created for it and NOTIFY_TEMP_CREATION set.
(gimplify_modify_expr_rhs) <VAR_DECL>: If the target is volatile but
the type is aggregate non-addressable, ask gimplify_init_constructor
whether it can generate a single access to the target.
gcc/testsuite/ChangeLog
* gnat.dg/aggr30.ads, gnat.dg/aggr30.adb: New test.
|
|
Adds param tsan-instrument-func-entry-exit, which controls if
__tsan_func_{entry,exit} calls should be emitted or not. The default
behaviour is to emit the calls.
This may be required by alternative race detection runtimes. One such
runtime is the Kernel Concurrency Sanitizer (KCSAN):
https://github.com/google/ktsan/wiki/KCSAN
After this change, GCC should satisfy all requirements for KCSAN:
https://lore.kernel.org/lkml/20200515150338.190344-7-elver@google.com/
gcc/ChangeLog:
* gimplify.c (gimplify_function_tree): Optimize and do not emit
IFN_TSAN_FUNC_EXIT in a finally block if we do not need it.
* params.opt: Add --param=tsan-instrument-func-entry-exit=.
* tsan.c (instrument_memory_accesses): Make
fentry_exit_instrument bool depend on new param.
gcc/testsuite/ChangeLog:
* c-c++-common/tsan/func_entry_exit.c: New test.
* c-c++-common/tsan/func_entry_exit_disabled.c: New test.
|
|
These are not itself necessary for OpenACC 'exit data' directives, and are
skipped over (now) in libgomp. We might as well not emit them to start with,
in line with the equivalent OpenMP directive. We keep the no-op handling in
libgomp for the reason of backward compatibility.
gcc/
* gimplify.c (gimplify_adjust_omp_clauses): Remove
'GOMP_MAP_STRUCT' mapping from OpenACC 'exit data' directives.
gcc/testsuite/
* c-c++-common/goacc/struct-enter-exit-data-1.c: New file.
libgomp/
* oacc-mem.c (goacc_exit_data_internal) <GOMP_MAP_STRUCT>: Explain
special handling.
Co-Authored-By: Julian Brown <julian@codesourcery.com>
|
|
gcc/c-family/ChangeLog:
* c-common.h (c_omp_predetermined_mapping): Declare.
* c-omp.c (c_omp_predetermined_mapping): New.
gcc/c/ChangeLog:
* c-objc-common.h (LANG_HOOKS_OMP_PREDETERMINED_MAPPING): Redefine.
gcc/cp/ChangeLog:
* cp-gimplify.c (cxx_omp_predetermined_mapping): New.
* cp-objcp-common.h (LANG_HOOKS_OMP_PREDETERMINED_MAPPING): Redfine.
* cp-tree.h (cxx_omp_predetermined_mapping): Declare.
gcc/fortran/ChangeLog:
* f95-lang.c (LANG_HOOKS_OMP_PREDETERMINED_MAPPING): Redefine.
* trans-openmp.c (gfc_omp_predetermined_mapping): New.
* trans.h (gfc_omp_predetermined_mapping): Declare.
gcc/ChangeLog:
* gimplify.c (omp_notice_variable): Use new hook.
* langhooks-def.h (lhd_omp_predetermined_mapping): Declare.
(LANG_HOOKS_OMP_PREDETERMINED_MAPPING): Define
(LANG_HOOKS_DECLS): Add it.
* langhooks.c (lhd_omp_predetermined_sharing): Remove bogus unused attr.
(lhd_omp_predetermined_mapping): New.
* langhooks.h (struct lang_hooks_for_decls): Add new hook.
gcc/testsuite/ChangeLog
2020-06-03 Thomas Schwinge <thomas@codesourcery.com>
Tobias Burnus <tobias@codesourcery.com>
PR middle-end/94874
* c-c++-common/gomp/pr94874.c: New.
|
|
iterators is instantiated [PR95197]
During gimplification omp_finish_clause langhook is called in several places
to add the language specific info to the clause like what default/copy ctors,
dtors and assignment operators should be used.
Unfortunately, if it refers to some not yet instantiated method, during
gimplification it is too late and the methods will not be instantiated
anymore. For other cases, the genericizer has code to detect those and
instantiate whatever is needed, this change adds the same for
distribute parallel for class iterators where we under the hood need
a copy constructor for the iterator to implement it.
2020-05-26 Jakub Jelinek <jakub@redhat.com>
PR c++/95197
* gimplify.c (find_combined_omp_for): Move to omp-general.c.
* omp-general.h (find_combined_omp_for): Declare.
* omp-general.c: Include tree-iterator.h.
(find_combined_omp_for): New function, moved from gimplify.c.
* cp-gimplify.c: Include omp-general.h.
(cp_genericize_r) <case OMP_DISTRIBUTE>: For class iteration
variables in composite distribute parallel for, instantiate copy
ctor of their types.
|
|
gcc/
PR middle-end/94635
* gimplify.c (gimplify_scan_omp_clauses): For MAP_TO_PSET with
OMP_TARGET_EXIT_DATA, use 'release:' unless the associated
item is 'delete:'.
gcc/testsuite
PR middle-end/94635
* gfortran.dg/gomp/target-exit-data.f90: New.
|
|
This extends DECL_GIMPLE_REG_P to all types so we can clear
TREE_ADDRESSABLE even for integers with partial defs, not just
complex and vector variables. To make that transition easier
the patch inverts DECL_GIMPLE_REG_P to DECL_NOT_GIMPLE_REG_P
since that makes the default the current state for all other
types besides complex and vectors.
For the testcase in PR94703 we're able to expand the partial
def'ed local integer to a register then, producing a single
movl rather than going through the stack.
On i?86 this execute FAILs gcc.dg/torture/pr71522.c because
we now expand a round-trip through a long double automatic var
to a register fld/fst which normalizes the value. For that
during RTL expansion we're looking for problematic punnings
of decls and avoid pseudos for those - I chose integer or
BLKmode accesses on decls with modes where precision doesn't
match bitsize which covers the XFmode case.
2020-05-07 Richard Biener <rguenther@suse.de>
PR middle-end/94703
* tree-core.h (tree_decl_common::gimple_reg_flag): Rename ...
(tree_decl_common::not_gimple_reg_flag): ... to this.
* tree.h (DECL_GIMPLE_REG_P): Rename ...
(DECL_NOT_GIMPLE_REG_P): ... to this.
* gimple-expr.c (copy_var_decl): Copy DECL_NOT_GIMPLE_REG_P.
(create_tmp_reg): Simplify.
(create_tmp_reg_fn): Likewise.
(is_gimple_reg): Check DECL_NOT_GIMPLE_REG_P for all regs.
* gimplify.c (create_tmp_from_val): Simplify.
(gimplify_bind_expr): Likewise.
(gimplify_compound_literal_expr): Likewise.
(gimplify_function_tree): Likewise.
(prepare_gimple_addressable): Set DECL_NOT_GIMPLE_REG_P.
* asan.c (create_odr_indicator): Do not clear DECL_GIMPLE_REG_P.
(asan_add_global): Copy it.
* cgraphunit.c (cgraph_node::expand_thunk): Force args
to be GIMPLE regs.
* function.c (gimplify_parameters): Copy
DECL_NOT_GIMPLE_REG_P.
* ipa-param-manipulation.c
(ipa_param_body_adjustments::common_initialization): Simplify.
(ipa_param_body_adjustments::reset_debug_stmts): Copy
DECL_NOT_GIMPLE_REG_P.
* omp-low.c (lower_omp_for_scan): Do not set DECL_GIMPLE_REG_P.
* sanopt.c (sanitize_rewrite_addressable_params): Likewise.
* tree-cfg.c (make_blocks_1): Simplify.
(verify_address): Do not verify DECL_GIMPLE_REG_P setting.
* tree-eh.c (lower_eh_constructs_2): Simplify.
* tree-inline.c (declare_return_variable): Adjust and
generalize.
(copy_decl_to_var): Copy DECL_NOT_GIMPLE_REG_P.
(copy_result_decl_to_var): Likewise.
* tree-into-ssa.c (pass_build_ssa::execute): Adjust comment.
* tree-nested.c (create_tmp_var_for): Simplify.
* tree-parloops.c (separate_decls_in_region_name): Copy
DECL_NOT_GIMPLE_REG_P.
* tree-sra.c (create_access_replacement): Adjust and
generalize partial def support.
* tree-ssa-forwprop.c (pass_forwprop::execute): Set
DECL_NOT_GIMPLE_REG_P on decls we introduce partial defs on.
* tree-ssa.c (maybe_optimize_var): Handle clearing of
TREE_ADDRESSABLE and setting/clearing DECL_NOT_GIMPLE_REG_P
independently.
* lto-streamer-out.c (hash_tree): Hash DECL_NOT_GIMPLE_REG_P.
* tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
DECL_NOT_GIMPLE_REG_P.
* tree-streamer-in.c (unpack_ts_decl_common_value_fields): Likewise.
* cfgexpand.c (avoid_type_punning_on_regs): New.
(discover_nonconstant_array_refs): Call
avoid_type_punning_on_regs to avoid unsupported mode punning.
lto/
* lto-common.c (compare_tree_sccs_1): Compare
DECL_NOT_GIMPLE_REG_P.
c/
* gimple-parser.c (c_parser_parse_ssa_name): Do not set
DECL_GIMPLE_REG_P.
cp/
* optimize.c (update_cloned_parm): Copy DECL_NOT_GIMPLE_REG_P.
* gcc.dg/tree-ssa/pr94703.c: New testcase.
|
|
The CONSTRUCTOR_NO_CLEARING flag was invented to avoid generating a memset
for CONSTRUCTORS that lack elements, but it turns out that the gimplifier
can generate a memcpy for them instead, which is worse performance-wise,
so this prevents it from doing that for them.
* gimplify.c (gimplify_init_constructor): Do not put the constructor
into static memory if it is not complete.
|
|
PR middle-end/94635
* gimplify.c (gimplify_scan_omp_clauses): Turn MAP_TO_PSET to
MAP_DELETE.
PR middle-end/94635
* testsuite/libgomp.fortran/target-enter-data-2.F90: New.
|
|
When gimplifying an address operand we may expose an indirect
ref via DECL_VALUE_EXPR for example. This is dealt with in the
code already but it fails to consider that INDIRECT_REFs get
gimplified to MEM_REFs.
Fixed which makes the ICE observed on x86_64-netbsd go away.
2020-04-07 Richard Biener <rguenther@suse.de>
PR middle-end/94479
* gimplify.c (gimplify_addr_expr): Also consider generated
MEM_REFs.
* gcc.dg/torture/pr94479.c: New testcase.
|
|
The following testcase FAILs, because gimplify_body adds a GIMPLE_NOP only
when there are no statements in the function and with -g there is a
DEBUG_BEGIN_STMT, so it doesn't add it and due to -fno-tree-dce that never
gets removed afterwards. Similarly, if the body seq after gimplification
contains some DEBUG_BEGIN_STMTs plus a single gbind, then we could behave
differently between -g0 and -g, by using that gbind as the body in the -g0
case and not in the -g case.
This patch fixes that by ignoring DEBUG_BEGIN_STMTs (other debug stmts can't
appear at this point yet thankfully) during decisions and if we pick the
single gbind and there are DEBUG_BEGIN_STMTs next to it, it moves them into
the gbind.
While debugging this, I found also a bug in the gimple_seq_last_nondebug_stmt
function, for a seq that has a single non-DEBUG_BEGIN_STMT statement
followed by one or more DEBUG_BEGIN_STMTs it would return NULL rather than
the first statement.
2020-03-26 Jakub Jelinek <jakub@redhat.com>
PR debug/94281
* gimple.h (gimple_seq_first_nondebug_stmt): New function.
(gimple_seq_last_nondebug_stmt): Don't return NULL if seq contains
a single non-debug stmt followed by one or more debug stmts.
* gimplify.c (gimplify_body): Use gimple_seq_first_nondebug_stmt
instead of gimple_seq_first_stmt, use gimple_seq_first_nondebug_stmt
and gimple_seq_last_nondebug_stmt instead of gimple_seq_first and
gimple_seq_last to check if outer_stmt gbind could be reused and
if yes and it is surrounded by any debug stmts, move them into the
gbind body.
* g++.dg/debug/pr94281.C: New test.
|
|
The following testcase is rejected, while it was accepted in 3.4 and earlier
(before tree-ssa merge).
The problem is that we decide to promote the const variable to TREE_STATIC,
but TREE_STATIC DECL_REGISTER VAR_DECLs may only be the global register vars
and so assemble_variable/make_decl_rtl diagnoses it.
Either we do what the following patch does, where we could consider
register as a hint the user doesn't want such optimization, because if
something is forced static, it is not "register" anymore and register static
is not valid in C either, or we could clear DECL_REGISTER instead, but would
still need to punt at least on DECL_HARD_REGISTER cases.
2020-02-27 Jakub Jelinek <jakub@redhat.com>
PR c/93949
* gimplify.c (gimplify_init_constructor): Don't promote readonly
DECL_REGISTER variables to TREE_STATIC.
* gcc.c-torture/compile/pr93949.c: New test.
|
|
DECL_IN_CONSTANT_POOL are shared and thus don't really get emitted in the
BLOCK where they are used, so for OpenMP target regions that have initializers
gimplified into copying from them we actually map them at runtime from host to
offload devices. This patch instead marks them as "omp declare target", so
that they are on the target device from the beginning and don't need to be
copied there.
2020-02-09 Jakub Jelinek <jakub@redhat.com>
* gimplify.c (gimplify_adjust_omp_clauses_1): Promote
DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
copying them around between host and target.
* testsuite/libgomp.c/target-38.c: New test.
|
|
[PR93515]
If we call omp_add_variable, following omp_notice_variable will already find it
on that construct and not go through outer constructs, the following patch fixes that.
Note, this still doesn't follow OpenMP 5.0 semantics on target combined with other
constructs with reduction/lastprivate/linear clauses, will handle that for GCC11.
2020-02-06 Jakub Jelinek <jakub@redhat.com>
PR libgomp/93515
* gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
shared clause, call omp_notice_variable on outer context if any.
|
|
If a TARGET_EXPR has poly-int size, the gimplifier would treat it
like a VLA and use gimplify_vla_decl. gimplify_vla_decl in turn
would use an alloca and expect all references to be gimplified
via the DECL_VALUE_EXPR. This caused confusion later in
gimplify_var_or_parm_decl_1 when we (correctly) had direct rather
than indirect references.
For completeness, the patch also fixes similar tests in the RETURN_EXPR
handling and OpenMP depend clauses.
2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
than testing directly for INTEGER_CST.
(gimplify_target_expr, gimplify_omp_depend): Likewise.
gcc/testsuite/
* g++.target/aarch64/sve/acle/general-c++/gimplify_1.C: New test.
|
|
This is a pretty rare situation since the C++11 change to make all
destructors default to noexcept, but it is still possible to define throwing
destructors, and if a destructor for a local variable throws during the
return, we've already constructed the return value, so now we need to
destroy it. I handled this somewhat like the new-expression cleanup; as in
that case, this cleanup can't properly nest with the cleanups for local
variables, so I introduce a cleanup region around the whole function and a
flag variable to indicate whether the return value actually needs to be
destroyed.
Setting the flag requires giving a COMPOUND_EXPR as the operand of a
RETURN_EXPR, so I adjust gimplify_return_expr to handle that.
This doesn't currently work with deduced return type because we don't know
the type when we're deciding whether to introduce the cleanup region.
gcc/
* gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
gcc/cp/
* cp-tree.h (current_retval_sentinel): New macro.
* decl.c (start_preparsed_function): Set up cleanup for retval.
* typeck.c (check_return_expr): Set current_retval_sentinel.
|
|
gcc/
* tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
* tree-core.h: Document it.
* gimplify.c (gimplify_omp_workshare): Set it.
* omp-low.c (lower_omp_target): Use it.
* tree-pretty-print.c (dump_omp_clause): Print it.
gcc/testsuite/
* c-c++-common/goacc/host_data-1.c: Extend.
* gfortran.dg/goacc/host_data-tree.f95: Likewise.
gcc/
* omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
libgomp/
* target.c (gomp_map_vars_internal)
<GOMP_MAP_USE_DEVICE_PTR_IF_PRESENT>: Clean up/elaborate code
paths.
From-SVN: r280149
|
|
My patch for 93033 wasn't sufficient to handle all the possible sharing
introduced by split_nonconstant_init, and it occurred to me that it would
make sense to use the same unsharing technique as unshare_body, namely
copy_if_shared.
PR c++/93033
gcc/
* gimplify.c (copy_if_shared): No longer static.
* gimplify.h: Declare it.
gcc/cp/
* cp-gimplify.c (cp_gimplify_init_expr, cp_gimplify_expr): Use
copy_if_shared after cp_genericize_tree.
* typeck2.c (split_nonconstant_init): Don't unshare here.
From-SVN: r280126
|
|
2020-01-09 Richard Biener <rguenther@suse.de>
PR middle-end/93054
* gimplify.c (gimplify_expr): Deal with NOP definitions.
* gcc.dg/pr93054.c: New testcase.
From-SVN: r280039
|
|
From-SVN: r279813
|
|
gcc/
* gimplify.c (gimplify_omp_var_data): Add GOVD_MAP_HAS_ATTACHMENTS.
(insert_struct_comp_map): Support derived-type member mappings
for arrays with descriptors which use GOMP_MAP_TO_PSET. Support
GOMP_MAP_ATTACH_DETACH.
(gimplify_scan_omp_clauses): Tidy up OACC_ENTER_DATA/OACC_EXIT_DATA
mappings. Handle attach/detach clauses and component references.
(gimplify_adjust_omp_clauses_1): Skip adjustments for explicit
attach/detach clauses.
(gimplify_omp_target_update): Handle struct mappings and finalize for
detach operations.
* omp-low.c (lower_omp_target): Support GOMP_MAP_ATTACH,
GOMP_MAP_DETACH, GOMP_MAP_FORCE_DETACH.
* tree-pretty-print.c (dump_omp_clause): Likewise, plus
GOMP_MAP_ATTACH_DETACH.
include/
* gomp-constants.h (gomp_map_kind): Add GOMP_MAP_ATTACH_DETACH.
From-SVN: r279626
|
|
gcc/
* gimplify.c (insert_struct_comp_map, extract_base_bit_offset): New.
(gimplify_scan_omp_clauses): Outline duplicated code into calls to
above two functions.
From-SVN: r279623
|
|
No functional changes.
gcc/
* gimplify.c (gimplify_omp_target_update): Elaborate 'exit data'
'finalize' handling.
gcc/testsuite/
* c-c++-common/goacc/finalize-1.c: Extend.
* gfortran.dg/goacc/finalize-1.f: Likewise.
libgomp/
* oacc-mem.c (GOACC_enter_exit_data): Simplify 'exit data'
'finalize' handling.
From-SVN: r279531
|
|
Even EXACT_DIV_EXPR doesn't distribute across addition for wrapping
types, so in general we can't fold EXACT_DIV_EXPRs of POLY_INT_CSTs
at compile time. This was causing an ICE when trying to gimplify the
element size field in an ARRAY_REF.
If the result of that EXACT_DIV_EXPR is an invariant, we don't bother
recording it in the ARRAY_REF and simply read the element size from the
element type. This avoids the overhead of doing:
/* ??? tree_ssa_useless_type_conversion will eliminate casts to
sizetype from another type of the same width and signedness. */
if (TREE_TYPE (aligned_size) != sizetype)
aligned_size = fold_convert_loc (loc, sizetype, aligned_size);
return size_binop_loc (loc, MULT_EXPR, aligned_size,
size_int (TYPE_ALIGN_UNIT (elmt_type)));
each time array_ref_element_size is called.
So rather than read array_ref_element_size, do some arithmetic on it,
and only then check whether the result is an invariant, we might as
well check whether the element size is an invariant to start with.
We're then directly testing whether array_ref_element_size gives
a reusable value.
For consistency, the patch makes the same change for the offset field
in a COMPONENT_REF, although I don't think that can trigger yet.
2019-12-03 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* gimplify.c (gimplify_compound_lval): Don't gimplify and install
an array element size if array_element_size is already an invariant.
Similarly don't gimplify and install a field offset if
component_ref_field_offset is already an invariant.
gcc/testsuite/
* gcc.target/aarch64/sve/acle/general-c/struct_1.c: New test.
From-SVN: r278942
|
|
Various bad uses of the [[fallthrough]] attribute are constraint
violations in C2x, so need pedwarns rather than warnings.
This patch duly turns the relevant warnings into pedwarns. The
relevant code is not specific to C, and does not know which form the
attribute was given in ([[fallthrough]] or [[gnu::fallthrough]] or
__attribute__((fallthrough))), but as I understand it these usages are
also erroneous for C++ and it seems reasonable to give a pedwarn here
even when a form other than [[fallthrough]] is being used.
The precise meaning of the standard wording about "The next statement
that would be executed" seems a but unclear in some corner cases; the
tests added keep to cases where it is clear whether or not the next
statement executed is of the required form.
Bootstrapped with no regressions for x86_64-pc-linux-gnu.
gcc:
* gimplify.c (expand_FALLTHROUGH_r, expand_FALLTHROUGH): Use
pedwarn instead of warning_at for fallthrough not preceding a case
or default label.
gcc/c-family:
* c-attribs.c (handle_fallthrough_attribute): Use pedwarn instead
of warning.
gcc/testsuite:
* gcc.dg/c2x-attr-fallthrough-6.c: New test. Split out from
c2x-attr-fallthrough-3.c.
* gcc.dg/c2x-attr-fallthrough-1.c: Add more tests.
* gcc.dg/c2x-attr-fallthrough-2.c: Update expected diagnostics.
* gcc.dg/c2x-attr-fallthrough-3.c: Split inside-switch part of
test out to c2x-attr-fallthrough-6.c.
From-SVN: r278599
|
|
after gimplification.
* gimplify.c (gimplify_call_expr): Don't call
omp_resolve_declare_variant after gimplification.
* omp-general.c (omp_context_selector_matches): For isa that might
match in some other function, defer if in declare simd function.
(omp_context_compute_score): Don't look for " score" in construct
trait set. Set *score to -1 if it can't ever match.
(omp_resolve_declare_variant): If any variants need to be deferred,
don't punt immediately, but compute scores of all variants and if
ther eis a score winner that doesn't need to be deferred, return that.
* c-c++-common/gomp/declare-variant-13.c: New test.
From-SVN: r278280
|
|
The `serial' construct (cf. section 2.5.3 of the OpenACC 2.6 standard)
is equivalent to a `parallel' construct with clauses `num_gangs(1)
num_workers(1) vector_length(1)' implied.
These clauses are therefore not supported with the `serial'
construct. All the remaining clauses accepted with `parallel' are also
accepted with `serial'.
The `serial' construct is implemented like `parallel', except for
hardcoding dimensions rather than taking them from the relevant
clauses, in `expand_omp_target'.
Separate codes are used to denote the `serial' construct throughout the
middle end, even though the mapping of `serial' to an equivalent
`parallel' construct could have been done in the individual language
frontends. In particular, this allows to distinguish between compute
constructs in warnings, error messages, dumps etc.
2019-11-12 Maciej W. Rozycki <macro@codesourcery.com>
Tobias Burnus <tobias@codesourcery.com>
Frederik Harwath <frederik@codesourcery.com>
Thomas Schwinge <thomas@codesourcery.com>
gcc/
* gimple.h (gf_mask): Add GF_OMP_TARGET_KIND_OACC_SERIAL
enumeration constant.
(is_gimple_omp_oacc): Handle GF_OMP_TARGET_KIND_OACC_SERIAL.
(is_gimple_omp_offloaded): Likewise.
* gimplify.c (omp_region_type): Add ORT_ACC_SERIAL enumeration
constant. Adjust the value of ORT_NONE accordingly.
(is_gimple_stmt): Handle OACC_SERIAL.
(oacc_default_clause): Handle ORT_ACC_SERIAL.
(gomp_needs_data_present): Likewise.
(gimplify_adjust_omp_clauses): Likewise.
(gimplify_omp_workshare): Handle OACC_SERIAL.
(gimplify_expr): Likewise.
* omp-expand.c (expand_omp_target):
Handle GF_OMP_TARGET_KIND_OACC_SERIAL.
(build_omp_regions_1, omp_make_gimple_edges): Likewise.
* omp-low.c (is_oacc_parallel): Rename function to...
(is_oacc_parallel_or_serial): ... this.
Handle GF_OMP_TARGET_KIND_OACC_SERIAL.
(scan_sharing_clauses): Adjust accordingly.
(scan_omp_for): Likewise.
(lower_oacc_head_mark): Likewise.
(convert_from_firstprivate_int): Likewise.
(lower_omp_target): Likewise.
(check_omp_nesting_restrictions): Handle
GF_OMP_TARGET_KIND_OACC_SERIAL.
(lower_oacc_reductions): Likewise.
(lower_omp_target): Likewise.
* tree.def (OACC_SERIAL): New tree code.
* tree-pretty-print.c (dump_generic_node): Handle OACC_SERIAL.
* doc/generic.texi (OpenACC): Document OACC_SERIAL.
gcc/c-family/
* c-pragma.h (pragma_kind): Add PRAGMA_OACC_SERIAL enumeration
constant.
* c-pragma.c (oacc_pragmas): Add "serial" entry.
gcc/c/
* c-parser.c (OACC_SERIAL_CLAUSE_MASK): New macro.
(c_parser_oacc_kernels_parallel): Rename function to...
(c_parser_oacc_compute): ... this. Handle PRAGMA_OACC_SERIAL.
(c_parser_omp_construct): Update accordingly.
gcc/cp/
* constexpr.c (potential_constant_expression_1): Handle
OACC_SERIAL.
* parser.c (OACC_SERIAL_CLAUSE_MASK): New macro.
(cp_parser_oacc_kernels_parallel): Rename function to...
(cp_parser_oacc_compute): ... this. Handle PRAGMA_OACC_SERIAL.
(cp_parser_omp_construct): Update accordingly.
(cp_parser_pragma): Handle PRAGMA_OACC_SERIAL. Fix alphabetic
order.
* pt.c (tsubst_expr): Handle OACC_SERIAL.
gcc/fortran/
* gfortran.h (gfc_statement): Add ST_OACC_SERIAL_LOOP,
ST_OACC_END_SERIAL_LOOP, ST_OACC_SERIAL and ST_OACC_END_SERIAL
enumeration constants.
(gfc_exec_op): Add EXEC_OACC_SERIAL_LOOP and EXEC_OACC_SERIAL
enumeration constants.
* match.h (gfc_match_oacc_serial): New prototype.
(gfc_match_oacc_serial_loop): Likewise.
* dump-parse-tree.c (show_omp_node, show_code_node): Handle
EXEC_OACC_SERIAL_LOOP and EXEC_OACC_SERIAL.
* match.c (match_exit_cycle): Handle EXEC_OACC_SERIAL_LOOP.
* openmp.c (OACC_SERIAL_CLAUSES): New macro.
(gfc_match_oacc_serial_loop): New function.
(gfc_match_oacc_serial): Likewise.
(oacc_is_loop): Handle EXEC_OACC_SERIAL_LOOP.
(resolve_omp_clauses): Handle EXEC_OACC_SERIAL.
(oacc_code_to_statement): Handle EXEC_OACC_SERIAL and
EXEC_OACC_SERIAL_LOOP.
(gfc_resolve_oacc_directive): Likewise.
* parse.c (decode_oacc_directive) <'s'>: Add case for "serial"
and "serial loop".
(next_statement): Handle ST_OACC_SERIAL_LOOP and ST_OACC_SERIAL.
(gfc_ascii_statement): Likewise. Handle ST_OACC_END_SERIAL_LOOP
and ST_OACC_END_SERIAL.
(parse_oacc_structured_block): Handle ST_OACC_SERIAL.
(parse_oacc_loop): Handle ST_OACC_SERIAL_LOOP and
ST_OACC_END_SERIAL_LOOP.
(parse_executable): Handle ST_OACC_SERIAL_LOOP and
ST_OACC_SERIAL.
(is_oacc): Handle EXEC_OACC_SERIAL_LOOP and EXEC_OACC_SERIAL.
* resolve.c (gfc_resolve_blocks, gfc_resolve_code): Likewise.
* st.c (gfc_free_statement): Likewise.
* trans-openmp.c (gfc_trans_oacc_construct): Handle
EXEC_OACC_SERIAL.
(gfc_trans_oacc_combined_directive): Handle
EXEC_OACC_SERIAL_LOOP.
(gfc_trans_oacc_directive): Handle EXEC_OACC_SERIAL_LOOP and
EXEC_OACC_SERIAL.
* trans.c (trans_code): Likewise.
gcc/testsuite/
* c-c++-common/goacc/parallel-dims.c: New test.
* gfortran.dg/goacc/parallel-dims.f90: New test.
libgomp/
* testsuite/libgomp.oacc-c-c++-common/parallel-dims.c: New test.
* testsuite/libgomp.oacc-fortran/parallel-dims-aux.c: New test.
* testsuite/libgomp.oacc-fortran/parallel-dims.f89: New test.
* testsuite/libgomp.oacc-fortran/parallel-dims-2.f90: New test.
Reviewed-by: Thomas Schwinge <thomas@codesourcery.com>
Co-Authored-By: Frederik Harwath <frederik@codesourcery.com>
Co-Authored-By: Thomas Schwinge <thomas@codesourcery.com>
Co-Authored-By: Tobias Burnus <tobias@codesourcery.com>
From-SVN: r278082
|
|
'gcc/gimplify.c:gimplify_scan_omp_clauses'
... to fix a long-time typo/copy'n'past-o.
gcc/
* gimplify.c (gimplify_scan_omp_clauses): Assert 'offset2' instead
of 'offset'.
From-SVN: r278038
|
|
This was first submitted many years ago
https://gcc.gnu.org/ml/gcc-patches/2010-10/msg02468.html
The command line option -fcallgraph-info is added and makes the
compiler generate another output file (xxx.ci) for each compilation
unit (or LTO partitoin), which is a valid VCG file (you can launch
your favorite VCG viewer on it unmodified) and contains the "final"
callgraph of the unit. "final" is a bit of a misnomer as this is
actually the callgraph at RTL expansion time, but since most
high-level optimizations are done at the Tree level and RTL doesn't
usually fiddle with calls, it's final in almost all cases. Moreover,
the nodes can be decorated with additional info: -fcallgraph-info=su
adds stack usage info and -fcallgraph-info=da dynamic allocation info.
for gcc/ChangeLog
From Eric Botcazou <ebotcazou@adacore.com>, Alexandre Oliva <oliva@adacore.com>
* common.opt (-fcallgraph-info[=]): New option.
* doc/invoke.texi (Developer options): Document it.
* opts.c (common_handle_option): Handle it.
* builtins.c (expand_builtin_alloca): Record allocation if
-fcallgraph-info=da.
* calls.c (expand_call): If -fcallgraph-info, record the call.
(emit_library_call_value_1): Likewise.
* flag-types.h (enum callgraph_info_type): New type.
* explow.c: Include stringpool.h.
(set_stack_check_libfunc): Set SET_SYMBOL_REF_DECL on the symbol.
* function.c (allocate_stack_usage_info): New.
(allocate_struct_function): Call it for -fcallgraph-info.
(prepare_function_start): Call it otherwise.
(record_final_call, record_dynamic_alloc): New.
* function.h (struct callinfo_callee): New.
(CALLEE_FROM_CGRAPH_P): New.
(struct callinfo_dalloc): New.
(struct stack_usage): Add callees and dallocs.
(record_final_call, record_dynamic_alloc): Declare.
* gimplify.c (gimplify_decl_expr): Record dynamically-allocated
object if -fcallgraph-info=da.
* optabs-libfuncs.c (build_libfunc_function): Keep SYMBOL_REF_DECL.
* print-tree.h (print_decl_identifier): Declare.
(PRINT_DECL_ORIGIN, PRINT_DECL_NAME, PRINT_DECL_UNIQUE_NAME): New.
* print-tree.c: Include print-tree.h.
(print_decl_identifier): New function.
* toplev.c: Include print-tree.h.
(callgraph_info_file): New global variable.
(callgraph_info_external_printed): Likewise.
(output_stack_usage): Rename to...
(output_stack_usage_1): ... this. Make it static, add cf
parameter. If -fcallgraph-info=su, print stack usage to cf.
If -fstack-usage, use print_decl_identifier for
pretty-printing.
(INDIRECT_CALL_NAME): New.
(dump_final_node_vcg_start): New.
(dump_final_callee_vcg, dump_final_node_vcg): New.
(output_stack_usage): New.
(lang_dependent_init): Open and start file if
-fcallgraph-info. Allocated callgraph_info_external_printed.
(finalize): If callgraph_info_file is not null, finish it,
close it, and release callgraph_info_external_printed.
for gcc/ada/ChangeLog
* gcc-interface/misc.c (callgraph_info_file): Delete.
Co-Authored-By: Alexandre Oliva <oliva@adacore.com>
From-SVN: r277876
|
|
PR inline-asm/92352
* gimplify.c (gimplify_asm_expr): Reject VLA in output or input
operands with non-memory constraints.
* c-c++-common/pr92352.c: New test.
From-SVN: r277873
|
|
a new SCORES argument.
* gimplify.h (omp_construct_selector_matches): Change return
type to int, add a new SCORES argument.
* gimplify.c (omp_construct_selector_matches): Likewise. If
SCORES is non-NULL, compute scores of each construct.
* omp-general.h (omp_get_context_selector): Declare.
* omp-general.c (omp_maybe_offloaded, omp_context_selector_matches):
Adjust omp_construct_selector_matches callers.
(omp_get_context_selector): New function, moved from c-family/c-omp.c.
(omp_context_compute_score): New function.
(omp_resolve_declare_variant): Compute scores and decide based on
that.
c-family/
* c-common.h (c_omp_get_context_selector): Remove.
* c-omp.c (c_omp_get_context_selector): Moved to omp-general.c
and renamed to omp_get_context_selector.
c/
* c-parser.c (c_finish_omp_declare_variant): Use
omp_get_context_selector instead of c_omp_get_context_selector.
cp/
* decl.c (omp_declare_variant_finalize_one): Use
omp_get_context_selector instead of c_omp_get_context_selector.
testsuite/
* c-c++-common/gomp/declare-variant-12.c: New test.
From-SVN: r277742
|
|
gcc/
* gimplify.c (gimplify_scan_omp_clauses): Remove FE-generated
GOMP_MAP_TO_PSET and GOMP_MAP_POINTER mapping for 'target update'
and 'target exit data'.
libgomp/
* testsuite/libgomp.fortran/target9.f90: New.
From-SVN: r277631
|
|
2019-10-25 Cesar Philippidis <cesar@codesourcery.com>
Tobias Burnus <tobias@codesourcery.com>
gcc/fortran/
* openmp.c (gfc_match_omp_map_clause): Add and pass allow_commons
argument.
(gfc_match_omp_clauses): Update calls to permit common blocks for
OpenACC's copy/copyin/copyout, create/delete, host,
pcopy/pcopy_in/pcopy_out, present_or_copy, present_or_copy_in,
present_or_copy_out, present_or_create and self.
gcc/
* gimplify.c (oacc_default_clause): Privatize fortran common blocks.
(omp_notice_variable): Defer the expansion of DECL_VALUE_EXPR for
common block decls.
gcc/testsuite/
* gfortran.dg/goacc/common-block-1.f90: New test.
* gfortran.dg/goacc/common-block-2.f90: New test.
* gfortran.dg/goacc/common-block-3.f90: New test.
libgomp/
* testsuite/libgomp.oacc-fortran/common-block-1.f90: New test.
* testsuite/libgomp.oacc-fortran/common-block-2.f90: New test.
* testsuite/libgomp.oacc-fortran/common-block-3.f90: New test.
Reviewed-by: Thomas Schwinge <thomas@codesourcery.com>
Co-Authored-By: Tobias Burnus <tobias@codesourcery.com>
From-SVN: r277451
|
|
* gimplify.h (omp_construct_selector_matches): Declare.
* gimplify.c (struct gimplify_omp_ctx): Add code member.
(gimplify_call_expr): Call omp_resolve_declare_variant and remap
called function if needed for flag_openmp.
(gimplify_scan_omp_clauses): Set ctx->code.
(omp_construct_selector_matches): New function.
* omp-general.h (omp_constructor_traits_to_codes,
omp_context_selector_matches, omp_resolve_declare_variant): Declare.
* omp-general.c (omp_constructor_traits_to_codes,
omp_context_selector_matches, omp_resolve_declare_variant): New
functions.
c-family/
* c-common.h (c_omp_context_selector_matches): Remove.
* c-omp.c (c_omp_context_selector_matches): Remove.
* c-attribs.c (c_common_attribute_table): Add
"omp declare target {host,nohost,block}" attributes.
c/
* c-parser.c (c_finish_omp_declare_variant): Use
omp_context_selector_matches instead of
c_omp_context_selector_matches.
* c-decl.c (c_decl_attributes): Add "omp declare target block"
attribute in between declare target and end declare target
pragmas.
cp/
* decl2.c (cplus_decl_attributes): Add "omp declare target block"
attribute in between declare target and end declare target
pragmas.
testsuite/
* c-c++-common/gomp/declare-variant-8.c: New test.
From-SVN: r277427
|
|
* cp-gimplify.c (cp_gimplify_expr): Use get_initialized_tmp_var.
The comment for get_formal_tmp_var says that it shouldn't be used for
expressions whose value might change between initialization and use, and in
this case we're creating a temporary precisely because the value might
change, so we should use get_initialized_tmp_var instead.
I also noticed that many callers of get_initialized_tmp_var pass NULL for
post_p, so it seems appropriate to make it a default argument.
gcc/
* gimplify.h (get_initialized_tmp_var): Add default argument to
post_p.
* gimplify.c (gimplify_self_mod_expr, gimplify_omp_atomic): Remove
NULL post_p argument.
* targhooks (std_gimplify_va_arg_expr): Likewise.
From-SVN: r277128
|
|
Right now gimplifier does not allow VEC_COND_EXPR's condition to trap
and introduces a temporary if this could happen, for example, generating
_5 = _4 > { 2.0e+0, 2.0e+0, 2.0e+0, 2.0e+0 };
_6 = VEC_COND_EXPR <_5, { -1, -1, -1, -1 }, { 0, 0, 0, 0 }>;
from GENERIC
VEC_COND_EXPR < (*b > { 2.0e+0, 2.0e+0, 2.0e+0, 2.0e+0 }) ,
{ -1, -1, -1, -1 } ,
{ 0, 0, 0, 0 } >
This is not necessary and makes the resulting GIMPLE harder to analyze.
Change the gimplifier so as to allow COND_EXPR and VEC_COND_EXPR
conditions to trap.
This patch takes special care to avoid introducing trapping comparisons
in GIMPLE_COND. They are not allowed, because they would require 3
outgoing edges (then, else and EH), which is awkward to say the least.
Therefore, computations of such conditions should live in their own basic
blocks.
gcc/ChangeLog:
2019-10-07 Ilya Leoshkevich <iii@linux.ibm.com>
PR target/77918
* gimple-expr.c (gimple_cond_get_ops_from_tree): Assert that the
caller passes a non-trapping condition.
(is_gimple_condexpr): Allow trapping conditions.
(is_gimple_condexpr_1): New helper function.
(is_gimple_condexpr_for_cond): New function, acts like old
is_gimple_condexpr.
* gimple-expr.h (is_gimple_condexpr_for_cond): New function.
* gimple.c (gimple_could_trap_p_1): Handle COND_EXPR and
VEC_COND_EXPR. Fix an issue with statements like i = (fp < 1.).
* gimplify.c (gimplify_cond_expr): Use
is_gimple_condexpr_for_cond.
(gimplify_expr): Allow is_gimple_condexpr_for_cond.
* tree-eh.c (operation_could_trap_p): Assert on COND_EXPR and
VEC_COND_EXPR.
(tree_could_trap_p): Handle COND_EXPR and VEC_COND_EXPR.
* tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Use
is_gimple_condexpr_for_cond, remove pointless tmp check
(forward_propagate_into_cond): Remove pointless tmp check.
From-SVN: r276659
|
|
PR middle-end/91920
* gimplify.c (omp_default_clause): Predetermine DECL_IN_CONSTANT_POOL
variables as shared.
* c-c++-common/gomp/pr91920.c: New test.
From-SVN: r276211
|
|
Source-level SVE vectors should be gimplified in the same way
as normal fixed-length vectors rather than as VLAs.
This is tested by later SVE patches.
2019-09-18 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* gimplify.c (gimplify_decl_expr): Use poly_int_tree_p instead
of checking specifically for INTEGER_CST.
From-SVN: r275870
|
|
helper variables on target data even if...
* gimplify.c (omp_add_variable): Use GOVD_PRIVATE | GOVD_EXPLICIT
for VLA helper variables on target data even if not GOVD_FIRSTPRIVATE.
(gimplify_scan_omp_clauses): For OMP_CLAUSE_USE_DEVICE_* use just
GOVD_EXPLICIT flags.
(gimplify_omp_workshare): For OMP_TARGET_DATA move all
OMP_CLAUSE_USE_DEVICE_* clauses to the end of clauses chain.
* omp-low.c (scan_sharing_clauses): For OMP_CLAUSE_USE_DEVICE_*
call install_var_field with mask 11 instead of 3.
(lower_omp_target): For OMP_CLAUSE_USE_DEVICE_* use pass
(splay_tree_key) &DECL_UID (var) to build_sender_ref instead of var.
gcc/c/
* c-typeck.c (c_finish_omp_clauses): For C_ORT_OMP
OMP_CLAUSE_USE_DEVICE_* clauses use oacc_reduction_head bitmap
instead of generic_head to track duplicates.
gcc/cp/
* semantics.c (finish_omp_clauses): For C_ORT_OMP
OMP_CLAUSE_USE_DEVICE_* clauses use oacc_reduction_head bitmap
instead of generic_head to track duplicates.
libgomp/
* target.c (gomp_map_vars_internal): For GOMP_MAP_USE_DEVICE_PTR
perform the lookup in the first loop only if !not_found_cnt, otherwise
perform lookups for it in the second loop guarded with
if (not_found_cnt || has_firstprivate).
* testsuite/libgomp.c/target-37.c: New test.
* testsuite/libgomp.c++/target-22.C: New test.
From-SVN: r274206
|
|
description.
* tree-core.h (enum omp_clause_code): Adjust OMP_CLAUSE_USE_DEVICE_PTR
OpenMP description. Add OMP_CLAUSE_USE_DEVICE_ADDR clause.
* tree.c (omp_clause_num_ops, omp_clause_code_name): Add entries
for OMP_CLAUSE_USE_DEVICE_ADDR clause.
(walk_tree_1): Handle OMP_CLAUSE_USE_DEVICE_ADDR.
* tree-pretty-print.c (dump_omp_clause): Likewise.
* tree-nested.c (convert_nonlocal_omp_clauses,
convert_local_omp_clauses): Likewise.
* gimplify.c (gimplify_scan_omp_clauses, gimplify_adjust_omp_clauses):
Likewise.
* omp-low.c (scan_sharing_clauses, lower_omp_target): Likewise.
Treat OMP_CLAUSE_USE_DEVICE_ADDR like OMP_CLAUSE_USE_DEVICE_PTR
clause with array or reference to array types, no matter what type
except for reference it has.
gcc/c-family/
* c-pragma.h (enum pragma_omp_clause): Add
PRAGMA_OMP_CLAUSE_USE_DEVICE_ADDR. Set PRAGMA_OACC_CLAUSE_USE_DEVICE
equal to PRAGMA_OMP_CLAUSE_USE_DEVICE_PTR instead of being a separate
enumeration value.
gcc/c/
* c-parser.c (c_parser_omp_clause_name): Parse use_device_addr clause.
(c_parser_omp_clause_use_device_addr): New function.
(c_parser_omp_all_clauses): Handle PRAGMA_OMP_CLAUSE_USE_DEVICE_ADDR.
(OMP_TARGET_DATA_CLAUSE_MASK): Add PRAGMA_OMP_CLAUSE_USE_DEVICE_ADDR.
(c_parser_omp_target_data): Handle PRAGMA_OMP_CLAUSE_USE_DEVICE_ADDR
like PRAGMA_OMP_CLAUSE_USE_DEVICE_PTR, adjust diagnostics about no
map or use_device_* clauses.
* c-typeck.c (c_finish_omp_clauses): For OMP_CLAUSE_USE_DEVICE_PTR
in OpenMP, require pointer type rather than pointer or array type.
Handle OMP_CLAUSE_USE_DEVICE_ADDR.
gcc/cp/
* parser.c (cp_parser_omp_clause_name): Parse use_device_addr clause.
(cp_parser_omp_all_clauses): Handle PRAGMA_OMP_CLAUSE_USE_DEVICE_ADDR.
(OMP_TARGET_DATA_CLAUSE_MASK): Add PRAGMA_OMP_CLAUSE_USE_DEVICE_ADDR.
(cp_parser_omp_target_data): Handle PRAGMA_OMP_CLAUSE_USE_DEVICE_ADDR
like PRAGMA_OMP_CLAUSE_USE_DEVICE_PTR, adjust diagnostics about no
map or use_device_* clauses.
* semantics.c (finish_omp_clauses): For OMP_CLAUSE_USE_DEVICE_PTR
in OpenMP, require pointer or reference to pointer type rather than
pointer or array or reference to pointer or array type. Handle
OMP_CLAUSE_USE_DEVICE_ADDR.
* pt.c (tsubst_omp_clauses): Handle OMP_CLAUSE_USE_DEVICE_ADDR.
gcc/testsuite/
* c-c++-common/gomp/target-data-1.c (foo): Use use_device_addr clause
instead of use_device_ptr clause where required by OpenMP 5.0, add
further tests for both use_device_ptr and use_device_addr clauses.
libgomp/
* testsuite/libgomp.c/target-18.c (struct S): New type.
(foo): Use use_device_addr clause instead of use_device_ptr clause
where required by OpenMP 5.0, add further tests for both use_device_ptr
and use_device_addr clauses.
* testsuite/libgomp.c++/target-9.C (struct S): New type.
(foo): Use use_device_addr clause instead of use_device_ptr clause
where required by OpenMP 5.0, add further tests for both use_device_ptr
and use_device_addr clauses. Add t and u arguments.
(main): Adjust caller.
From-SVN: r274159
|
|
* tree.h (OMP_CLAUSE_LASTPRIVATE_TASKLOOP_IV): Rename to ...
(OMP_CLAUSE_LASTPRIVATE_LOOP_IV): ... this. Adjust comment.
* gimplify.c (gimple_add_tmp_var): In SIMD contexts, turn addressable
new vars into GOVD_PRIVATE rather than GOVD_LOCAL.
(gimplify_omp_for): Don't do C++ random access iterator clause
adjustments on combined constructs from OMP_LOOP. For OMP_LOOP,
don't predetermine the artificial iterator in case of C++ random
access iterators as lastprivate, but private. For OMP_LOOP, force
bind expr around simd body and force for_pre_body before the
construct. Use OMP_CLAUSE_LASTPRIVATE_LOOP_IV instead of
OMP_CLAUSE_LASTPRIVATE_TASKLOOP_IV.
(gimplify_omp_loop): Add firstprivate clauses on OMP_PARALLEL for
diff var of C++ random access iterators. Handle
OMP_CLAUSE_FIRSTPRIVATE. For OMP_CLAUSE_LASTPRIVATE_LOOP_IV, if
not outermost also add OMP_CLAUSE_FIRSTPRIVATE, and in both cases
clear OMP_CLAUSE_LASTPRIVATE_LOOP_IV on the lastprivate clause
on the OMP_FOR and OMP_DISTRIBUTE constructs if any.
* omp-low.c (lower_rec_input_clauses): For
OMP_CLAUSE_LASTPRIVATE_LOOP_IV on simd copy construct the private
variables instead of default constructing them.
(lower_lastprivate_clauses): Use OMP_CLAUSE_LASTPRIVATE_LOOP_IV
instead of OMP_CLAUSE_LASTPRIVATE_TASKLOOP_IV and move the
is_taskloop_ctx check from the assert to the guarding condition.
gcc/cp/
* parser.c (cp_parser_omp_for_loop): For OMP_LOOP, ignore parallel
clauses and predetermine iterator as lastprivate.
* semantics.c (handle_omp_for_class_iterator): Use
OMP_CLAUSE_LASTPRIVATE_LOOP_IV instead of
OMP_CLAUSE_LASTPRIVATE_TASKLOOP_IV, set it for lastprivate also
on OMP_LOOP construct. If a clause is missing for class iterator
on OMP_LOOP, add firstprivate clause, and if there is private
clause, turn it into firstprivate too.
(finish_omp_for): Formatting fix. For OMP_LOOP, adjust
OMP_CLAUSE_LASTPRIVATE_LOOP_IV clause CP_CLAUSE_INFO, so that it
uses copy ctor instead of default ctor.
* cp-gimplify.c (cp_gimplify_expr): Handle OMP_LOOP like
OMP_DISTRIBUTE etc.
(cp_fold_r): Likewise.
(cp_genericize_r): Likewise.
(cxx_omp_finish_clause): Also finish lastprivate clause with
OMP_CLAUSE_LASTPRIVATE_LOOP_IV flag.
* pt.c (tsubst_omp_clauses): Handle OMP_CLAUSE_BIND.
(tsubst_omp_for_iterator): For OMP_LOOP, ignore parallel
clauses and predetermine iterator as lastprivate.
* constexpr.c (potential_constant_expression_1): Handle OMP_LOOP
like OMP_DISTRIBUTE etc.
libgomp/
* testsuite/libgomp.c++/loop-13.C: New test.
* testsuite/libgomp.c++/loop-14.C: New test.
* testsuite/libgomp.c++/loop-15.C: New test.
From-SVN: r274138
|
|
distribute parallel for private (iterator))
PR middle-end/91301
* gimplify.c (gimplify_omp_for): If for class iterator on
distribute parallel for there is no data sharing clause
on inner_for_stmt, look for private clause on combined
parallel too and if found, move it to inner_for_stmt.
* testsuite/libgomp.c++/for-27.C: New test.
From-SVN: r273922
|
|
information in the printable name.
* gimplify.c (flag_instrument_functions_exclude_p): Include
namespace/class information in the printable name.
* opts.c (add_comma_separated_to_vector): Add NUL terminator
to tokens entered into the vector.
From-SVN: r273766
|
|
* tree.def (OMP_LOOP): New tree code.
* tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_BIND.
(enum omp_clause_bind_kind): New enum.
(struct tree_omp_clause): Add subcode.bind_kind.
* tree.h (OMP_LOOP_CHECK): Rename to ...
(OMP_LOOPING_CHECK): ... this.
(OMP_FOR_BODY, OMP_FOR_CLAUSES, OMP_FOR_INIT, OMP_FOR_COND,
OMP_FOR_INCR, OMP_FOR_PRE_BODY, OMP_FOR_ORIG_DECLS): Use
OMP_LOOPING_CHECK instead of OMP_LOOP_CHECK.
(OMP_CLAUSE_BIND_KIND): Define.
* tree.c (omp_clause_num_ops, omp_clause_code_name): Add
bind clause entries.
(walk_tree_1): Handle OMP_CLAUSE_BIND.
* tree-pretty-print.c (dump_omp_clause): Likewise.
(dump_generic_node): Handle OMP_LOOP.
* gimplify.c (enum omp_region_type): Add ORT_IMPLICIT_TARGET.
(in_omp_construct): New variable.
(is_gimple_stmt): Handle OMP_LOOP.
(gimplify_scan_omp_clauses): For lastprivate don't set
check_non_private if code == OMP_LOOP. For reduction clause
on OMP_LOOP combined with parallel or teams propagate as shared
on the combined construct. Handle OMP_CLAUSE_BIND.
(gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_BIND.
(gimplify_omp_for): Pass OMP_LOOP instead of OMP_{FOR,DISTRIBUTE}
for constructs from a loop construct to gimplify_scan_omp_clauses.
Don't predetermine iterator linear on OMP_SIMD from loop construct.
(replace_reduction_placeholders, gimplify_omp_loop): New functions.
(gimplify_omp_workshare): Use ORT_IMPLICIT_TARGET instead of trying
to match the implicit ORT_TARGET construct around whole body.
Temporarily clear in_omp_construct when processing body.
(gimplify_expr): Handle OMP_LOOP. For OMP_MASTER, OMP_TASKGROUP
etc. temporarily set in_omp_construct when processing body.
(gimplify_body): Create ORT_IMPLICIT_TARGET instead of ORT_TARGET.
* omp-low.c (struct omp_context): Add loop_p.
(build_outer_var_ref): Treat ctx->loop_p similarly to simd construct
in that the original var might be private.
(scan_sharing_clauses): Handle OMP_CLAUSE_BIND.
(check_omp_nesting_restrictions): Adjust nesting restrictions for
addition of loop construct.
(scan_omp_1_stmt): Allow setjmp inside of loop construct.
gcc/c-family/
* c-pragma.h (enum pragma_kind): Add PRAGMA_OMP_LOOP.
(enum pragma_omp_clause): Add PRAGMA_OMP_CLAUSE_BIND.
* c-pragma.c (omp_pragmas_simd): Add PRAGMA_OMP_LOOP entry.
* c-common.h (enum c_omp_clause_split): Add C_OMP_CLAUSE_SPLIT_LOOP.
* c-omp.c (c_omp_split_clauses): Add support for 4 new combined
constructs with the loop construct.
gcc/c/
* c-parser.c (c_parser_omp_clause_name): Handle bind clause.
(c_parser_omp_clause_bind): New function.
(c_parser_omp_all_clauses): Handle PRAGMA_OMP_CLAUSE_BIND.
(OMP_LOOP_CLAUSE_MASK): Define.
(c_parser_omp_loop): New function.
(c_parser_omp_parallel, c_parser_omp_teams): Handle parsing of
loop combined with parallel or teams.
(c_parser_omp_construct): Handle PRAGMA_OMP_LOOP.
* c-typeck.c (c_finish_omp_clauses): Handle OMP_CLAUSE_BIND.
gcc/cp/
* cp-tree.h (OMP_FOR_GIMPLIFYING_P): Use OMP_LOOPING_CHECK
instead of OMP_LOOP_CHECK.
* parser.c (cp_parser_omp_clause_name): Handle bind clause.
(cp_parser_omp_clause_bind): New function.
(cp_parser_omp_all_clauses): Handle PRAGMA_OMP_CLAUSE_BIND.
(OMP_LOOP_CLAUSE_MASK): Define.
(cp_parser_omp_loop): New function.
(cp_parser_omp_parallel, cp_parser_omp_teams): Handle parsing of
loop combined with parallel or teams.
(cp_parser_omp_construct): Handle PRAGMA_OMP_LOOP.
(cp_parser_pragma): Likewise.
* pt.c (tsubst_expr): Handle OMP_LOOP.
* semantics.c (finish_omp_clauses): Handle OMP_CLAUSE_BIND.
gcc/testsuite/
* c-c++-common/gomp/cancel-1.c: Adjust expected diagnostic wording.
* c-c++-common/gomp/clauses-1.c (foo, baz, bar): Add order(concurrent)
clause where allowed. Add combined constructs with loop with all
possible clauses.
(qux): New function.
* c-c++-common/gomp/loop-1.c: New test.
* c-c++-common/gomp/loop-2.c: New test.
* c-c++-common/gomp/loop-3.c: New test.
* c-c++-common/gomp/loop-4.c: New test.
* c-c++-common/gomp/loop-5.c: New test.
* c-c++-common/gomp/order-3.c: Adjust expected diagnostic wording.
* c-c++-common/gomp/simd-setjmp-1.c: New test.
* c-c++-common/gomp/teams-2.c: Adjust expected diagnostic wording.
libgomp/
* testsuite/libgomp.c-c++-common/loop-1.c: New test.
From-SVN: r273621
|
|
* gimplify.c (struct gimplify_omp_ctx): Add order_concurrent member.
(omp_notice_threadprivate_variable): Diagnose threadprivate variable
uses inside of order(concurrent) constructs.
(gimplify_scan_omp_clauses): Set ctx->order_concurrent if
OMP_CLAUSE_ORDER is seen.
* omp-low.c (struct omp_context): Add order_concurrent member.
(scan_sharing_clauses): Set ctx->order_concurrent if
OMP_CLAUSE_ORDER is seen.
(check_omp_nesting_restrictions): Diagnose ordered or atomic inside
of simd order(concurrent). Diagnose constructs not allowed inside of
for order(concurrent).
(setjmp_or_longjmp_p): Add a context and TREE_PUBLIC check to avoid
complaining about static double setjmp (double); or class static
methods or non-global namespace setjmps.
(omp_runtime_api_call): New function.
(scan_omp_1_stmt): Diagnose OpenMP runtime API calls inside of
order(concurrent) loops.
* c-c++-common/gomp/order-3.c: New test.
* c-c++-common/gomp/order-4.c: New test.
From-SVN: r273464
|
