openmp: Add OMPT variant of GOMP_has_masked_thread_num
Checks
| Context |
Check |
Description |
| linaro-tcwg-bot/tcwg_gcc_build--master-arm |
success
|
Build passed
|
Commit Message
Add GOMP_has_masked_thread_num_with_end, which is semantically identically but
meant for -fopenmp-ompt.
gcc/ChangeLog:
* omp-builtins.def (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END):
New builtin.
* omp-low.cc (lower_omp_master): Use it when -fopenmp-ompt.
libgomp/ChangeLog:
* libgomp.map: Add GOMP_has_masked_thread_num_with_end.
* libgomp.texi: Document it.
* libgomp_g.h (GOMP_has_masked_thread_num_with_end): Declare.
* parallel.c (GOMP_has_masked_thread_num_with_end): New function.
gcc/testsuite/ChangeLog:
* c-c++-common/gomp/masked-1.c: Check new builtin is not emitted.
* c-c++-common/gomp/masked-3.c: Check new builtin is emitted with
-fopenmp-ompt.
---
gcc/omp-builtins.def | 3 +++
gcc/omp-low.cc | 6 +++++-
gcc/testsuite/c-c++-common/gomp/masked-1.c | 1 +
gcc/testsuite/c-c++-common/gomp/masked-3.c | 4 +++-
libgomp/libgomp.map | 1 +
libgomp/libgomp.texi | 7 ++++---
libgomp/libgomp_g.h | 1 +
libgomp/parallel.c | 8 ++++++++
8 files changed, 26 insertions(+), 5 deletions(-)
Comments
Paul-Antoine Arras wrote:
> Add GOMP_has_masked_thread_num_with_end, which is semantically identically but
> meant for -fopenmp-ompt.
Which solves the issue I mentioned in my other email
- sorry for the belated review due to vaccation - namely:
For OMPT, libgomp needs to know whether to pass ompt_scope_beginend
or ompt_scope_begin followed by a later ompt_scope_end to the
registered tool.
> gcc/ChangeLog:
> * omp-builtins.def (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END):
> New builtin.
> * omp-low.cc (lower_omp_master): Use it when -fopenmp-ompt.
>
> libgomp/ChangeLog:
> * libgomp.map: Add GOMP_has_masked_thread_num_with_end.
> * libgomp.texi: Document it.
> * libgomp_g.h (GOMP_has_masked_thread_num_with_end): Declare.
> * parallel.c (GOMP_has_masked_thread_num_with_end): New function.
>
> gcc/testsuite/ChangeLog:
> * c-c++-common/gomp/masked-1.c: Check new builtin is not emitted.
> * c-c++-common/gomp/masked-3.c: Check new builtin is emitted with
> -fopenmp-ompt.
...
> @@ -9098,7 +9098,11 @@ lower_omp_master (gimple_stmt_iterator *gsi_p, omp_context *ctx)
> gsi_replace (gsi_p, bind, true);
> - bfn_decl = builtin_decl_explicit (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
> + if (flag_openmp_ompt)
> + bfn_decl
> + = builtin_decl_explicit (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END);
> + else
> + bfn_decl = builtin_decl_explicit (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
Elsewhere, you use such version:
bfn_decl = builtin_decl_explicit (
flag_openmp_ompt ? BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM
: BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END);
which seems to be tad more readable - however, not that it really
matters.
> --- a/libgomp/libgomp.texi
> +++ b/libgomp/libgomp.texi
>
> -@code{GOMP_has_masked_thread_num} encapsulates the following check, so that OMPT
> -can distinguish the compiler-generated code from user-level thread-number
> -queries:
> +@code{GOMP_has_masked_thread_num} (or its variant
> +@code{GOMP_has_masked_thread_num_with_end} enabled by @option{-fopenmp-ompt})
> +encapsulates the following check, so that OMPT can distinguish the
> +compiler-generated code from user-level thread-number queries:
I think it is more readable and clearer (+ more details) to keep the
current wording and add a sentence, i.e.
@code{GOMP_has_masked_thread_num} encapsulates the following check, so that OMPT
can distinguish the compiler-generated code from user-level thread-number
queries; with @option{-fopenmp-ompt}, a paired call to
@code{GOMP_has_masked_thread_num_with_end} and @code{GOMP_masked_end} is
generated instead.
Otherwise, LGTM.
Thanks,
Tobias
See the updated patch in attachment and a few comments below.
On 16/07/2026 16:45, Tobias Burnus wrote:
> Paul-Antoine Arras wrote:
>> Add GOMP_has_masked_thread_num_with_end, which is semantically
>> identically but
>> meant for -fopenmp-ompt.
>
> Which solves the issue I mentioned in my other email
> - sorry for the belated review due to vaccation - namely:
>
> For OMPT, libgomp needs to know whether to pass ompt_scope_beginend
> or ompt_scope_begin followed by a later ompt_scope_end to the
> registered tool.
>
>
>> gcc/ChangeLog:
>> * omp-builtins.def (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END):
>> New builtin.
>> * omp-low.cc (lower_omp_master): Use it when -fopenmp-ompt.
>>
>> libgomp/ChangeLog:
>> * libgomp.map: Add GOMP_has_masked_thread_num_with_end.
>> * libgomp.texi: Document it.
>> * libgomp_g.h (GOMP_has_masked_thread_num_with_end): Declare.
>> * parallel.c (GOMP_has_masked_thread_num_with_end): New function.
>>
>> gcc/testsuite/ChangeLog:
>> * c-c++-common/gomp/masked-1.c: Check new builtin is not emitted.
>> * c-c++-common/gomp/masked-3.c: Check new builtin is emitted with
>> -fopenmp-ompt.
> ...
>> @@ -9098,7 +9098,11 @@ lower_omp_master (gimple_stmt_iterator *gsi_p,
>> omp_context *ctx)
>> gsi_replace (gsi_p, bind, true);
>> - bfn_decl = builtin_decl_explicit
>> (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
>> + if (flag_openmp_ompt)
>> + bfn_decl
>> + = builtin_decl_explicit
>> (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END);
>> + else
>> + bfn_decl = builtin_decl_explicit
>> (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
>
> Elsewhere, you use such version:
>
> bfn_decl = builtin_decl_explicit (
> flag_openmp_ompt ? BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM
> :
> BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END);
>
> which seems to be tad more readable - however, not that it really
> matters.
Agreed! Refactored in the updated patch.
>> --- a/libgomp/libgomp.texi
>> +++ b/libgomp/libgomp.texi
>> -@code{GOMP_has_masked_thread_num} encapsulates the following check,
>> so that OMPT
>> -can distinguish the compiler-generated code from user-level thread-
>> number
>> -queries:
>> +@code{GOMP_has_masked_thread_num} (or its variant
>> +@code{GOMP_has_masked_thread_num_with_end} enabled by @option{-
>> fopenmp-ompt})
>> +encapsulates the following check, so that OMPT can distinguish the
>> +compiler-generated code from user-level thread-number queries:
>
> I think it is more readable and clearer (+ more details) to keep the
> current wording and add a sentence, i.e.
>
> @code{GOMP_has_masked_thread_num} encapsulates the following check, so
> that OMPT
> can distinguish the compiler-generated code from user-level thread-number
> queries; with @option{-fopenmp-ompt}, a paired call to
> @code{GOMP_has_masked_thread_num_with_end} and @code{GOMP_masked_end} is
> generated instead.
Rephrased as suggested.
Thanks,
Hi!
On 2026-07-16T19:34:51+0200, Paul-Antoine Arras <parras@baylibre.com> wrote:
> On 16/07/2026 16:45, Tobias Burnus wrote:
>> Paul-Antoine Arras wrote:
>>> Add GOMP_has_masked_thread_num_with_end, which is semantically
>>> identically but
>>> meant for -fopenmp-ompt.
>>
>> Which solves the issue I mentioned in my other email
>> - sorry for the belated review due to vaccation - namely: [...]
We've managed to keep things under control, and moving forward.
>>> @@ -9098,7 +9098,11 @@ lower_omp_master (gimple_stmt_iterator *gsi_p,
>>> omp_context *ctx)
>>> gsi_replace (gsi_p, bind, true);
>>> - bfn_decl = builtin_decl_explicit
>>> (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
>>> + if (flag_openmp_ompt)
>>> + bfn_decl
>>> + = builtin_decl_explicit
>>> (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END);
>>> + else
>>> + bfn_decl = builtin_decl_explicit
>>> (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
>>
>> Elsewhere, you use such version:
>>
>> bfn_decl = builtin_decl_explicit (
>> flag_openmp_ompt ? BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM
>> :
>> BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END);
>>
>> which seems to be tad more readable - however, not that it really
>> matters.
>
> Agreed! Refactored in the updated patch.
> --- a/gcc/omp-low.cc
> +++ b/gcc/omp-low.cc
> @@ -9098,7 +9098,9 @@ lower_omp_master (gimple_stmt_iterator *gsi_p, omp_context *ctx)
> - bfn_decl = builtin_decl_explicit (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
> + bfn_decl = builtin_decl_explicit (
> + flag_openmp_ompt ? BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM
> + : BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END);
Please test your patches; v1 vs. v2:
PASS: c-c++-common/gomp/masked-1.c (test for excess errors)
PASS: c-c++-common/gomp/masked-1.c scan-tree-dump omplower "GOMP_has_masked_thread_num"
[-PASS:-]{+FAIL:+} c-c++-common/gomp/masked-1.c scan-tree-dump-not omplower "GOMP_has_masked_thread_num_with_end"
PASS: c-c++-common/gomp/masked-1.c scan-tree-dump-not omplower "GOMP_masked_end"
PASS: c-c++-common/gomp/masked-1.c scan-tree-dump-not omplower "omp_get_thread_num"
PASS: c-c++-common/gomp/masked-3.c (test for excess errors)
[-PASS:-]{+FAIL:+} c-c++-common/gomp/masked-3.c scan-tree-dump-times omplower "GOMP_has_masked_thread_num_with_end" 1
PASS: c-c++-common/gomp/masked-3.c scan-tree-dump-times omplower "GOMP_masked_end" 1
PASS: c-c++-common/gomp/masked-3.c scan-tree-dump-not omplower "omp_get_thread_num"
Similar for C++.
Sorry if I sound grumpy, but I am; such things are a waste of time.
Grüße
Thomas
On 16/07/2026 23:28, Thomas Schwinge wrote:
>> --- a/gcc/omp-low.cc
>> +++ b/gcc/omp-low.cc
>> @@ -9098,7 +9098,9 @@ lower_omp_master (gimple_stmt_iterator *gsi_p, omp_context *ctx)
>> - bfn_decl = builtin_decl_explicit (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
>> + bfn_decl = builtin_decl_explicit (
>> + flag_openmp_ompt ? BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM
>> + : BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END);
> Please test your patches; v1 vs. v2:
>
> PASS: c-c++-common/gomp/masked-1.c (test for excess errors)
> PASS: c-c++-common/gomp/masked-1.c scan-tree-dump omplower "GOMP_has_masked_thread_num"
> [-PASS:-]{+FAIL:+} c-c++-common/gomp/masked-1.c scan-tree-dump-not omplower "GOMP_has_masked_thread_num_with_end"
> PASS: c-c++-common/gomp/masked-1.c scan-tree-dump-not omplower "GOMP_masked_end"
> PASS: c-c++-common/gomp/masked-1.c scan-tree-dump-not omplower "omp_get_thread_num"
>
> PASS: c-c++-common/gomp/masked-3.c (test for excess errors)
> [-PASS:-]{+FAIL:+} c-c++-common/gomp/masked-3.c scan-tree-dump-times omplower "GOMP_has_masked_thread_num_with_end" 1
> PASS: c-c++-common/gomp/masked-3.c scan-tree-dump-times omplower "GOMP_masked_end" 1
> PASS: c-c++-common/gomp/masked-3.c scan-tree-dump-not omplower "omp_get_thread_num"
>
> Similar for C++.
>
>
> Sorry if I sound grumpy, but I am; such things are a waste of time.
Fixed in v3.
Hi!
On 2026-07-17T11:07:55+0200, Paul-Antoine Arras <parras@baylibre.com> wrote:
> Subject: [PATCH v3] openmp: Add OMPT variant of GOMP_has_masked_thread_num
Thanks, looks good to me! Tobias, any further comments?
Grüße
Thomas
> Add GOMP_has_masked_thread_num_with_end, which is semantically identically but
> meant for -fopenmp-ompt.
>
> gcc/ChangeLog:
>
> * omp-builtins.def (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END):
> New builtin.
> * omp-low.cc (lower_omp_master): Use it when -fopenmp-ompt.
>
> libgomp/ChangeLog:
>
> * libgomp.map: Add GOMP_has_masked_thread_num_with_end.
> * libgomp.texi: Document it.
> * libgomp_g.h (GOMP_has_masked_thread_num_with_end): Declare.
> * parallel.c (GOMP_has_masked_thread_num_with_end): New function.
>
> gcc/testsuite/ChangeLog:
>
> * c-c++-common/gomp/masked-1.c: Check new builtin is not emitted.
> * c-c++-common/gomp/masked-3.c: Check new builtin is emitted with
> -fopenmp-ompt.
> ---
> gcc/omp-builtins.def | 3 +++
> gcc/omp-low.cc | 4 +++-
> gcc/testsuite/c-c++-common/gomp/masked-1.c | 1 +
> gcc/testsuite/c-c++-common/gomp/masked-3.c | 4 +++-
> libgomp/libgomp.map | 1 +
> libgomp/libgomp.texi | 4 +++-
> libgomp/libgomp_g.h | 1 +
> libgomp/parallel.c | 8 ++++++++
> 8 files changed, 23 insertions(+), 3 deletions(-)
>
> diff --git a/gcc/omp-builtins.def b/gcc/omp-builtins.def
> index 121bdcb61a3..29b3b72c2cd 100644
> --- a/gcc/omp-builtins.def
> +++ b/gcc/omp-builtins.def
> @@ -508,6 +508,9 @@ DEF_GOMP_BUILTIN (BUILT_IN_GOMP_ERROR, "GOMP_error",
> DEF_GOMP_BUILTIN (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM,
> "GOMP_has_masked_thread_num", BT_FN_BOOL_INT,
> ATTR_CONST_NOTHROW_LEAF_LIST)
> +DEF_GOMP_BUILTIN (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END,
> + "GOMP_has_masked_thread_num_with_end", BT_FN_BOOL_INT,
> + ATTR_CONST_NOTHROW_LEAF_LIST)
> DEF_GOMP_BUILTIN (BUILT_IN_GOMP_MASKED_END, "GOMP_masked_end", BT_FN_VOID,
> ATTR_NOTHROW_LEAF_LIST)
> /* TODO review these function attributes once OMPT implementation done */
> diff --git a/gcc/omp-low.cc b/gcc/omp-low.cc
> index 50da8581a8d..ea140224fd9 100644
> --- a/gcc/omp-low.cc
> +++ b/gcc/omp-low.cc
> @@ -9098,7 +9098,9 @@ lower_omp_master (gimple_stmt_iterator *gsi_p, omp_context *ctx)
> gsi_replace (gsi_p, bind, true);
> gimple_bind_add_stmt (bind, stmt);
>
> - bfn_decl = builtin_decl_explicit (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
> + bfn_decl = builtin_decl_explicit (
> + flag_openmp_ompt ? BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END
> + : BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
> x = build_call_expr_loc (loc, bfn_decl, 1, filter);
> x = build3 (COND_EXPR, void_type_node, x, NULL, build_and_jump (&lab));
> tseq = NULL;
> diff --git a/gcc/testsuite/c-c++-common/gomp/masked-1.c b/gcc/testsuite/c-c++-common/gomp/masked-1.c
> index 28d2735d96e..92d005dbb9b 100644
> --- a/gcc/testsuite/c-c++-common/gomp/masked-1.c
> +++ b/gcc/testsuite/c-c++-common/gomp/masked-1.c
> @@ -26,5 +26,6 @@ foo (int x, int *a)
> }
>
> /* { dg-final { scan-tree-dump "GOMP_has_masked_thread_num" "omplower" } } */
> +/* { dg-final { scan-tree-dump-not "GOMP_has_masked_thread_num_with_end" "omplower" } } */
> /* { dg-final { scan-tree-dump-not "GOMP_masked_end" "omplower" } } */
> /* { dg-final { scan-tree-dump-not "omp_get_thread_num" "omplower" } } */
> diff --git a/gcc/testsuite/c-c++-common/gomp/masked-3.c b/gcc/testsuite/c-c++-common/gomp/masked-3.c
> index adea59fdb69..cb2209284c3 100644
> --- a/gcc/testsuite/c-c++-common/gomp/masked-3.c
> +++ b/gcc/testsuite/c-c++-common/gomp/masked-3.c
> @@ -1,6 +1,8 @@
> /* { dg-do compile } */
> /* { dg-additional-options "-fopenmp-ompt -fdump-tree-omplower" } */
>
> +/* Check that OMPT variants of libgomp calls are emitted. */
> +
> void bar (void);
>
> void
> @@ -10,6 +12,6 @@ foo (void)
> bar ();
> }
>
> -/* { dg-final { scan-tree-dump-times "GOMP_has_masked_thread_num" 1 "omplower" } } */
> +/* { dg-final { scan-tree-dump-times "GOMP_has_masked_thread_num_with_end" 1 "omplower" } } */
> /* { dg-final { scan-tree-dump-times "GOMP_masked_end" 1 "omplower" } } */
> /* { dg-final { scan-tree-dump-not "omp_get_thread_num" "omplower" } } */
> diff --git a/libgomp/libgomp.map b/libgomp/libgomp.map
> index 3f465cc1491..81bad4d2329 100644
> --- a/libgomp/libgomp.map
> +++ b/libgomp/libgomp.map
> @@ -485,6 +485,7 @@ GOMP_6.0.1 {
> GOMP_6.0.2 {
> global:
> GOMP_has_masked_thread_num;
> + GOMP_has_masked_thread_num_with_end;
> GOMP_masked_end;
> GOMP_loop_static_worksharing;
> GOMP_loop_static_worksharing_start;
> diff --git a/libgomp/libgomp.texi b/libgomp/libgomp.texi
> index 5d29da0984d..b6314b75aba 100644
> --- a/libgomp/libgomp.texi
> +++ b/libgomp/libgomp.texi
> @@ -8029,7 +8029,9 @@ if (GOMP_has_masked_thread_num(thread_num))
>
> @code{GOMP_has_masked_thread_num} encapsulates the following check, so that OMPT
> can distinguish the compiler-generated code from user-level thread-number
> -queries:
> +queries; with @option{-fopenmp-ompt}, a paired call to
> +@code{GOMP_has_masked_thread_num_with_end} and @code{GOMP_masked_end} is
> +generated instead.
>
> @smallexample
> omp_get_thread_num() == @var{thread_num}
> diff --git a/libgomp/libgomp_g.h b/libgomp/libgomp_g.h
> index 1ea8123056e..be3e349a417 100644
> --- a/libgomp/libgomp_g.h
> +++ b/libgomp/libgomp_g.h
> @@ -307,6 +307,7 @@ extern unsigned GOMP_parallel_reductions (void (*) (void *), void *, unsigned,
> extern bool GOMP_cancel (int, bool);
> extern bool GOMP_cancellation_point (int);
> extern bool GOMP_has_masked_thread_num (int);
> +extern bool GOMP_has_masked_thread_num_with_end (int);
> extern void GOMP_masked_end (void);
>
> /* task.c */
> diff --git a/libgomp/parallel.c b/libgomp/parallel.c
> index bb522b39ecb..286798ff902 100644
> --- a/libgomp/parallel.c
> +++ b/libgomp/parallel.c
> @@ -280,6 +280,14 @@ GOMP_has_masked_thread_num (int tid)
> return tid == gomp_thread ()->ts.team_id;
> }
>
> +/* OMPT variant enabled by -fopenmp-ompt. */
> +
> +bool
> +GOMP_has_masked_thread_num_with_end (int tid)
> +{
> + return tid == gomp_thread ()->ts.team_id;
> +}
> +
> /* Stub for OMPT callback enabled by -fopenmp-ompt. */
>
> void
> --
> 2.53.0
Thomas Schwinge wrote:
> On 2026-07-17T11:07:55+0200, Paul-Antoine Arras <parras@baylibre.com> wrote:
>> Subject: [PATCH v3] openmp: Add OMPT variant of GOMP_has_masked_thread_num
> Thanks, looks good to me! Tobias, any further comments?
No LGTM – and sorry for mixing up the order in '? :' in my previous
review comment.
Tobias
@@ -508,6 +508,9 @@ DEF_GOMP_BUILTIN (BUILT_IN_GOMP_ERROR, "GOMP_error",
DEF_GOMP_BUILTIN (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM,
"GOMP_has_masked_thread_num", BT_FN_BOOL_INT,
ATTR_CONST_NOTHROW_LEAF_LIST)
+DEF_GOMP_BUILTIN (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END,
+ "GOMP_has_masked_thread_num_with_end", BT_FN_BOOL_INT,
+ ATTR_CONST_NOTHROW_LEAF_LIST)
DEF_GOMP_BUILTIN (BUILT_IN_GOMP_MASKED_END, "GOMP_masked_end", BT_FN_VOID,
ATTR_NOTHROW_LEAF_LIST)
/* TODO review these function attributes once OMPT implementation done */
@@ -9098,7 +9098,11 @@ lower_omp_master (gimple_stmt_iterator *gsi_p, omp_context *ctx)
gsi_replace (gsi_p, bind, true);
gimple_bind_add_stmt (bind, stmt);
- bfn_decl = builtin_decl_explicit (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
+ if (flag_openmp_ompt)
+ bfn_decl
+ = builtin_decl_explicit (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM_WITH_END);
+ else
+ bfn_decl = builtin_decl_explicit (BUILT_IN_GOMP_HAS_MASKED_THREAD_NUM);
x = build_call_expr_loc (loc, bfn_decl, 1, filter);
x = build3 (COND_EXPR, void_type_node, x, NULL, build_and_jump (&lab));
tseq = NULL;
@@ -26,5 +26,6 @@ foo (int x, int *a)
}
/* { dg-final { scan-tree-dump "GOMP_has_masked_thread_num" "omplower" } } */
+/* { dg-final { scan-tree-dump-not "GOMP_has_masked_thread_num_with_end" "omplower" } } */
/* { dg-final { scan-tree-dump-not "GOMP_masked_end" "omplower" } } */
/* { dg-final { scan-tree-dump-not "omp_get_thread_num" "omplower" } } */
@@ -1,6 +1,8 @@
/* { dg-do compile } */
/* { dg-additional-options "-fopenmp-ompt -fdump-tree-omplower" } */
+/* Check that OMPT variants of libgomp calls are emitted. */
+
void bar (void);
void
@@ -10,6 +12,6 @@ foo (void)
bar ();
}
-/* { dg-final { scan-tree-dump-times "GOMP_has_masked_thread_num" 1 "omplower" } } */
+/* { dg-final { scan-tree-dump-times "GOMP_has_masked_thread_num_with_end" 1 "omplower" } } */
/* { dg-final { scan-tree-dump-times "GOMP_masked_end" 1 "omplower" } } */
/* { dg-final { scan-tree-dump-not "omp_get_thread_num" "omplower" } } */
@@ -485,6 +485,7 @@ GOMP_6.0.1 {
GOMP_6.0.2 {
global:
GOMP_has_masked_thread_num;
+ GOMP_has_masked_thread_num_with_end;
GOMP_masked_end;
GOMP_loop_static_worksharing;
GOMP_loop_static_worksharing_start;
@@ -8027,9 +8027,10 @@ if (GOMP_has_masked_thread_num(thread_num))
block
@end smallexample
-@code{GOMP_has_masked_thread_num} encapsulates the following check, so that OMPT
-can distinguish the compiler-generated code from user-level thread-number
-queries:
+@code{GOMP_has_masked_thread_num} (or its variant
+@code{GOMP_has_masked_thread_num_with_end} enabled by @option{-fopenmp-ompt})
+encapsulates the following check, so that OMPT can distinguish the
+compiler-generated code from user-level thread-number queries:
@smallexample
omp_get_thread_num() == @var{thread_num}
@@ -307,6 +307,7 @@ extern unsigned GOMP_parallel_reductions (void (*) (void *), void *, unsigned,
extern bool GOMP_cancel (int, bool);
extern bool GOMP_cancellation_point (int);
extern bool GOMP_has_masked_thread_num (int);
+extern bool GOMP_has_masked_thread_num_with_end (int);
extern void GOMP_masked_end (void);
/* task.c */
@@ -280,6 +280,14 @@ GOMP_has_masked_thread_num (int tid)
return tid == gomp_thread ()->ts.team_id;
}
+/* OMPT variant enabled by -fopenmp-ompt. */
+
+bool
+GOMP_has_masked_thread_num_with_end (int tid)
+{
+ return tid == gomp_thread ()->ts.team_id;
+}
+
/* Stub for OMPT callback enabled by -fopenmp-ompt. */
void