* [Patch] Fortran/OpenMP: Support (parallel) master taskloop (simd) [PR99928]
@ 2021-06-01 9:53 Tobias Burnus
2021-06-01 10:02 ` Jakub Jelinek
0 siblings, 1 reply; 3+ messages in thread
From: Tobias Burnus @ 2021-06-01 9:53 UTC (permalink / raw)
To: gcc-patches, Jakub Jelinek, fortran
[-- Attachment #1: Type: text/plain, Size: 988 bytes --]
(PR 99928 is mainly about handling clauses in combined constructs.
but it sees some related commits like this one. The PR is not
fully fixed for C/C++ and for Fortran a bit more is missing.)
This patch adds '(parallel) master taskloop (simd)' support to gfortran.
For full parsing support for the Fortran version of the C/C++ testcases pr99928-*.c,
parsing support for 'loop' (+ combined constructs) and reduction with an array (slice)
is missing. – And, of course, doing FE fixes for handling the clauses,
similar to what has been/is being done for C/C++.
(No attempt has been made in this patch to fix long-standing
combined-clause handling issues.)
As bug fix, 'taskloop simd reduction(...)' is now supported and
'taskloop simd reduction(task,...' is now diagnosed by the FE.
Tobias
-----------------
Mentor Graphics (Deutschland) GmbH, Arnulfstrasse 201, 80634 München Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Frank Thürauf
[-- Attachment #2: master-taskloop.diff --]
[-- Type: text/x-patch, Size: 48046 bytes --]
Fortran/OpenMP: Support (parallel) master taskloop (simd) [PR99928]
PR middle-end/99928
gcc/fortran/ChangeLog:
* dump-parse-tree.c (show_omp_node, show_code_node): Handle
(parallel) master taskloop (simd).
* frontend-passes.c (gfc_code_walker): Set in_omp_workshare
to false for parallel master taskloop (simd).
* gfortran.h (enum gfc_statement):
Add ST_OMP_(END_)(PARALLEL_)MASTER_TASKLOOP(_SIMD).
(enum gfc_exec_op): EXEC_OMP_(PARALLEL_)MASTER_TASKLOOP(_SIMD).
* match.h (gfc_match_omp_master_taskloop,
gfc_match_omp_master_taskloop_simd,
gfc_match_omp_parallel_master_taskloop,
gfc_match_omp_parallel_master_taskloop_simd): New prototype.
* openmp.c (gfc_match_omp_parallel_master_taskloop,
gfc_match_omp_parallel_master_taskloop_simd,
gfc_match_omp_master_taskloop,
gfc_match_omp_master_taskloop_simd): New.
(gfc_match_omp_taskloop_simd): Permit 'reduction' clause.
(resolve_omp_clauses): Handle new combined directives; remove
inscan-reduction check to reduce multiple errors; add
task-reduction error for 'taskloop simd'.
(gfc_resolve_omp_parallel_blocks,
resolve_omp_do, omp_code_to_statement,
gfc_resolve_omp_directive): Handle new combined constructs.
* parse.c (decode_omp_directive, next_statement,
gfc_ascii_statement, parse_omp_do, parse_omp_structured_block,
parse_executable): Likewise.
* resolve.c (gfc_resolve_blocks, gfc_resolve_code): Likewise.
* st.c (gfc_free_statement): Likewise.
* trans.c (trans_code): Likewise.
* trans-openmp.c (gfc_split_omp_clauses,
gfc_trans_omp_directive): Likewise.
(gfc_trans_omp_parallel_master): Move after gfc_trans_omp_master_taskloop;
handle parallel master taskloop (simd) as well.
(gfc_trans_omp_taskloop): Take gfc_exec_op as arg.
(gfc_trans_omp_master_taskloop): New.
gcc/testsuite/ChangeLog:
* gfortran.dg/gomp/reduction5.f90: Remove dg-error the issue is
now diagnosed with error output less.
* gfortran.dg/gomp/scan-1.f90: Likewise.
* gfortran.dg/gomp/pr99928-3.f90: New test.
* gfortran.dg/gomp/taskloop-1.f90: New test.
gcc/fortran/dump-parse-tree.c | 12 +++
gcc/fortran/frontend-passes.c | 2 +
gcc/fortran/gfortran.h | 10 +-
gcc/fortran/match.h | 4 +
gcc/fortran/openmp.c | 85 ++++++++++++++--
gcc/fortran/parse.c | 73 +++++++++++++-
gcc/fortran/resolve.c | 10 ++
gcc/fortran/st.c | 4 +
gcc/fortran/trans-openmp.c | 112 ++++++++++++++++-----
gcc/fortran/trans.c | 4 +
gcc/testsuite/gfortran.dg/gomp/pr99928-3.f90 | 139 ++++++++++++++++++++++++++
gcc/testsuite/gfortran.dg/gomp/reduction5.f90 | 4 +-
gcc/testsuite/gfortran.dg/gomp/scan-1.f90 | 4 +-
gcc/testsuite/gfortran.dg/gomp/taskloop-1.f90 | 126 +++++++++++++++++++++++
14 files changed, 550 insertions(+), 39 deletions(-)
gcc/fortran/dump-parse-tree.c | 12 +++
gcc/fortran/frontend-passes.c | 2 +
gcc/fortran/gfortran.h | 10 +-
gcc/fortran/match.h | 4 +
gcc/fortran/openmp.c | 85 ++++++++++++++--
gcc/fortran/parse.c | 73 +++++++++++++-
gcc/fortran/resolve.c | 10 ++
gcc/fortran/st.c | 4 +
gcc/fortran/trans-openmp.c | 112 ++++++++++++++++-----
gcc/fortran/trans.c | 4 +
gcc/testsuite/gfortran.dg/gomp/pr99928-3.f90 | 139 ++++++++++++++++++++++++++
gcc/testsuite/gfortran.dg/gomp/reduction5.f90 | 4 +-
gcc/testsuite/gfortran.dg/gomp/scan-1.f90 | 4 +-
gcc/testsuite/gfortran.dg/gomp/taskloop-1.f90 | 126 +++++++++++++++++++++++
14 files changed, 550 insertions(+), 39 deletions(-)
diff --git a/gcc/fortran/dump-parse-tree.c b/gcc/fortran/dump-parse-tree.c
index 93ff572cbd2..0e7fe1cc3f3 100644
--- a/gcc/fortran/dump-parse-tree.c
+++ b/gcc/fortran/dump-parse-tree.c
@@ -1898,12 +1898,18 @@ show_omp_node (int level, gfc_code *c)
case EXEC_OMP_DO_SIMD: name = "DO SIMD"; break;
case EXEC_OMP_FLUSH: name = "FLUSH"; break;
case EXEC_OMP_MASTER: name = "MASTER"; break;
+ case EXEC_OMP_MASTER_TASKLOOP: name = "MASTER TASKLOOP"; break;
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD: name = "MASTER TASKLOOP SIMD"; break;
case EXEC_OMP_ORDERED: name = "ORDERED"; break;
case EXEC_OMP_DEPOBJ: name = "DEPOBJ"; break;
case EXEC_OMP_PARALLEL: name = "PARALLEL"; break;
case EXEC_OMP_PARALLEL_DO: name = "PARALLEL DO"; break;
case EXEC_OMP_PARALLEL_DO_SIMD: name = "PARALLEL DO SIMD"; break;
case EXEC_OMP_PARALLEL_MASTER: name = "PARALLEL MASTER"; break;
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ name = "PARALLEL MASTER TASKLOOP"; break;
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
+ name = "PARALLEL MASTER TASKLOOP SIMD"; break;
case EXEC_OMP_PARALLEL_SECTIONS: name = "PARALLEL SECTIONS"; break;
case EXEC_OMP_PARALLEL_WORKSHARE: name = "PARALLEL WORKSHARE"; break;
case EXEC_OMP_SCAN: name = "SCAN"; break;
@@ -1976,6 +1982,8 @@ show_omp_node (int level, gfc_code *c)
case EXEC_OMP_PARALLEL_DO:
case EXEC_OMP_PARALLEL_DO_SIMD:
case EXEC_OMP_PARALLEL_MASTER:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_PARALLEL_SECTIONS:
case EXEC_OMP_PARALLEL_WORKSHARE:
case EXEC_OMP_SCAN:
@@ -3184,11 +3192,15 @@ show_code_node (int level, gfc_code *c)
case EXEC_OMP_DO_SIMD:
case EXEC_OMP_FLUSH:
case EXEC_OMP_MASTER:
+ case EXEC_OMP_MASTER_TASKLOOP:
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_ORDERED:
case EXEC_OMP_PARALLEL:
case EXEC_OMP_PARALLEL_DO:
case EXEC_OMP_PARALLEL_DO_SIMD:
case EXEC_OMP_PARALLEL_MASTER:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_PARALLEL_SECTIONS:
case EXEC_OMP_PARALLEL_WORKSHARE:
case EXEC_OMP_SCAN:
diff --git a/gcc/fortran/frontend-passes.c b/gcc/fortran/frontend-passes.c
index ffe2db4881d..e3b1d1544aa 100644
--- a/gcc/fortran/frontend-passes.c
+++ b/gcc/fortran/frontend-passes.c
@@ -5543,6 +5543,8 @@ gfc_code_walker (gfc_code **c, walk_code_fn_t codefn, walk_expr_fn_t exprfn,
case EXEC_OMP_PARALLEL_DO:
case EXEC_OMP_PARALLEL_DO_SIMD:
case EXEC_OMP_PARALLEL_MASTER:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_PARALLEL_SECTIONS:
in_omp_workshare = false;
diff --git a/gcc/fortran/gfortran.h b/gcc/fortran/gfortran.h
index 55fba04642a..2020ab4a503 100644
--- a/gcc/fortran/gfortran.h
+++ b/gcc/fortran/gfortran.h
@@ -267,7 +267,11 @@ enum gfc_statement
ST_GET_FCN_CHARACTERISTICS, ST_LOCK, ST_UNLOCK, ST_EVENT_POST,
ST_EVENT_WAIT, ST_FAIL_IMAGE, ST_FORM_TEAM, ST_CHANGE_TEAM,
ST_END_TEAM, ST_SYNC_TEAM, ST_OMP_PARALLEL_MASTER,
- ST_OMP_END_PARALLEL_MASTER, ST_NONE
+ ST_OMP_END_PARALLEL_MASTER, ST_OMP_PARALLEL_MASTER_TASKLOOP,
+ ST_OMP_END_PARALLEL_MASTER_TASKLOOP, ST_OMP_PARALLEL_MASTER_TASKLOOP_SIMD,
+ ST_OMP_END_PARALLEL_MASTER_TASKLOOP_SIMD, ST_OMP_MASTER_TASKLOOP,
+ ST_OMP_END_MASTER_TASKLOOP, ST_OMP_MASTER_TASKLOOP_SIMD,
+ ST_OMP_END_MASTER_TASKLOOP_SIMD, ST_NONE
};
/* Types of interfaces that we can have. Assignment interfaces are
@@ -2711,7 +2715,9 @@ enum gfc_exec_op
EXEC_OMP_TARGET_PARALLEL, EXEC_OMP_TARGET_PARALLEL_DO,
EXEC_OMP_TARGET_PARALLEL_DO_SIMD, EXEC_OMP_TARGET_SIMD,
EXEC_OMP_TASKLOOP, EXEC_OMP_TASKLOOP_SIMD, EXEC_OMP_SCAN, EXEC_OMP_DEPOBJ,
- EXEC_OMP_PARALLEL_MASTER
+ EXEC_OMP_PARALLEL_MASTER, EXEC_OMP_PARALLEL_MASTER_TASKLOOP,
+ EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD, EXEC_OMP_MASTER_TASKLOOP,
+ EXEC_OMP_MASTER_TASKLOOP_SIMD
};
typedef struct gfc_code
diff --git a/gcc/fortran/match.h b/gcc/fortran/match.h
index 09c572324bc..bcedf8ef806 100644
--- a/gcc/fortran/match.h
+++ b/gcc/fortran/match.h
@@ -169,12 +169,16 @@ match gfc_match_omp_do (void);
match gfc_match_omp_do_simd (void);
match gfc_match_omp_flush (void);
match gfc_match_omp_master (void);
+match gfc_match_omp_master_taskloop (void);
+match gfc_match_omp_master_taskloop_simd (void);
match gfc_match_omp_ordered (void);
match gfc_match_omp_ordered_depend (void);
match gfc_match_omp_parallel (void);
match gfc_match_omp_parallel_do (void);
match gfc_match_omp_parallel_do_simd (void);
match gfc_match_omp_parallel_master (void);
+match gfc_match_omp_parallel_master_taskloop (void);
+match gfc_match_omp_parallel_master_taskloop_simd (void);
match gfc_match_omp_parallel_sections (void);
match gfc_match_omp_parallel_workshare (void);
match gfc_match_omp_requires (void);
diff --git a/gcc/fortran/openmp.c b/gcc/fortran/openmp.c
index 4ed6a0d64c8..9dba165f990 100644
--- a/gcc/fortran/openmp.c
+++ b/gcc/fortran/openmp.c
@@ -3995,6 +3995,22 @@ gfc_match_omp_parallel_master (void)
return match_omp (EXEC_OMP_PARALLEL_MASTER, OMP_PARALLEL_CLAUSES);
}
+match
+gfc_match_omp_parallel_master_taskloop (void)
+{
+ return match_omp (EXEC_OMP_PARALLEL_MASTER_TASKLOOP,
+ (OMP_PARALLEL_CLAUSES | OMP_TASKLOOP_CLAUSES)
+ & ~(omp_mask (OMP_CLAUSE_IN_REDUCTION)));
+}
+
+match
+gfc_match_omp_parallel_master_taskloop_simd (void)
+{
+ return match_omp (EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD,
+ (OMP_PARALLEL_CLAUSES | OMP_TASKLOOP_CLAUSES
+ | OMP_SIMD_CLAUSES)
+ & ~(omp_mask (OMP_CLAUSE_IN_REDUCTION)));
+}
match
gfc_match_omp_parallel_sections (void)
@@ -4429,8 +4445,7 @@ match
gfc_match_omp_taskloop_simd (void)
{
return match_omp (EXEC_OMP_TASKLOOP_SIMD,
- (OMP_TASKLOOP_CLAUSES | OMP_SIMD_CLAUSES)
- & ~(omp_mask (OMP_CLAUSE_REDUCTION)));
+ OMP_TASKLOOP_CLAUSES | OMP_SIMD_CLAUSES);
}
@@ -4533,6 +4548,18 @@ gfc_match_omp_master (void)
return MATCH_YES;
}
+match
+gfc_match_omp_master_taskloop (void)
+{
+ return match_omp (EXEC_OMP_MASTER_TASKLOOP, OMP_TASKLOOP_CLAUSES);
+}
+
+match
+gfc_match_omp_master_taskloop_simd (void)
+{
+ return match_omp (EXEC_OMP_MASTER_TASKLOOP_SIMD,
+ OMP_TASKLOOP_CLAUSES | OMP_SIMD_CLAUSES);
+}
match
gfc_match_omp_ordered (void)
@@ -5073,6 +5100,16 @@ resolve_omp_clauses (gfc_code *code, gfc_omp_clauses *omp_clauses,
ok = ifc == OMP_IF_PARALLEL || ifc == OMP_IF_SIMD;
break;
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ ok = ifc == OMP_IF_PARALLEL || ifc == OMP_IF_TASKLOOP;
+ break;
+
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
+ ok = (ifc == OMP_IF_PARALLEL
+ || ifc == OMP_IF_TASKLOOP
+ || ifc == OMP_IF_SIMD);
+ break;
+
case EXEC_OMP_SIMD:
case EXEC_OMP_DO_SIMD:
case EXEC_OMP_DISTRIBUTE_SIMD:
@@ -5085,10 +5122,12 @@ resolve_omp_clauses (gfc_code *code, gfc_omp_clauses *omp_clauses,
break;
case EXEC_OMP_TASKLOOP:
+ case EXEC_OMP_MASTER_TASKLOOP:
ok = ifc == OMP_IF_TASKLOOP;
break;
case EXEC_OMP_TASKLOOP_SIMD:
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
ok = ifc == OMP_IF_TASKLOOP || ifc == OMP_IF_SIMD;
break;
@@ -5848,11 +5887,16 @@ resolve_omp_clauses (gfc_code *code, gfc_omp_clauses *omp_clauses,
n->sym->name, name, &n->where);
switch (list)
{
- case OMP_LIST_REDUCTION_INSCAN:
case OMP_LIST_REDUCTION_TASK:
- if (code && (code->op == EXEC_OMP_TASKLOOP
- || code->op == EXEC_OMP_TEAMS
- || code->op == EXEC_OMP_TEAMS_DISTRIBUTE))
+ if (code
+ && (code->op == EXEC_OMP_TASKLOOP
+ || code->op == EXEC_OMP_TASKLOOP_SIMD
+ || code->op == EXEC_OMP_MASTER_TASKLOOP
+ || code->op == EXEC_OMP_MASTER_TASKLOOP_SIMD
+ || code->op == EXEC_OMP_PARALLEL_MASTER_TASKLOOP
+ || code->op == EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD
+ || code->op == EXEC_OMP_TEAMS
+ || code->op == EXEC_OMP_TEAMS_DISTRIBUTE))
{
gfc_error ("Only DEFAULT permitted as reduction-"
"modifier in REDUCTION clause at %L",
@@ -5863,6 +5907,7 @@ resolve_omp_clauses (gfc_code *code, gfc_omp_clauses *omp_clauses,
case OMP_LIST_REDUCTION:
case OMP_LIST_IN_REDUCTION:
case OMP_LIST_TASK_REDUCTION:
+ case OMP_LIST_REDUCTION_INSCAN:
switch (n->u.reduction_op)
{
case OMP_REDUCTION_PLUS:
@@ -6766,6 +6811,10 @@ gfc_resolve_omp_parallel_blocks (gfc_code *code, gfc_namespace *ns)
case EXEC_OMP_DISTRIBUTE_PARALLEL_DO_SIMD:
case EXEC_OMP_PARALLEL_DO:
case EXEC_OMP_PARALLEL_DO_SIMD:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
+ case EXEC_OMP_MASTER_TASKLOOP:
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_TARGET_PARALLEL_DO:
case EXEC_OMP_TARGET_PARALLEL_DO_SIMD:
case EXEC_OMP_TARGET_TEAMS_DISTRIBUTE:
@@ -6909,6 +6958,18 @@ resolve_omp_do (gfc_code *code)
name = "!$OMP PARALLEL DO SIMD";
is_simd = true;
break;
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ name = "!$OMP PARALLEL MASTER TASKLOOP";
+ break;
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
+ name = "!$OMP PARALLEL MASTER TASKLOOP SIMD";
+ is_simd = true;
+ break;
+ case EXEC_OMP_MASTER_TASKLOOP: name = "!$OMP MASTER TASKLOOP"; break;
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
+ name = "!$OMP MASTER TASKLOOP SIMD";
+ is_simd = true;
+ break;
case EXEC_OMP_SIMD: name = "!$OMP SIMD"; is_simd = true; break;
case EXEC_OMP_TARGET_PARALLEL_DO: name = "!$OMP TARGET PARALLEL DO"; break;
case EXEC_OMP_TARGET_PARALLEL_DO_SIMD:
@@ -7063,6 +7124,10 @@ omp_code_to_statement (gfc_code *code)
return ST_OMP_PARALLEL;
case EXEC_OMP_PARALLEL_MASTER:
return ST_OMP_PARALLEL_MASTER;
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ return ST_OMP_PARALLEL_MASTER_TASKLOOP;
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
+ return ST_OMP_PARALLEL_MASTER_TASKLOOP_SIMD;
case EXEC_OMP_PARALLEL_SECTIONS:
return ST_OMP_PARALLEL_SECTIONS;
case EXEC_OMP_SECTIONS:
@@ -7073,6 +7138,10 @@ omp_code_to_statement (gfc_code *code)
return ST_OMP_CRITICAL;
case EXEC_OMP_MASTER:
return ST_OMP_MASTER;
+ case EXEC_OMP_MASTER_TASKLOOP:
+ return ST_OMP_MASTER_TASKLOOP;
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
+ return ST_OMP_MASTER_TASKLOOP_SIMD;
case EXEC_OMP_SINGLE:
return ST_OMP_SINGLE;
case EXEC_OMP_TASK:
@@ -7561,6 +7630,10 @@ gfc_resolve_omp_directive (gfc_code *code, gfc_namespace *ns)
case EXEC_OMP_DO_SIMD:
case EXEC_OMP_PARALLEL_DO:
case EXEC_OMP_PARALLEL_DO_SIMD:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
+ case EXEC_OMP_MASTER_TASKLOOP:
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_SIMD:
case EXEC_OMP_TARGET_PARALLEL_DO:
case EXEC_OMP_TARGET_PARALLEL_DO_SIMD:
diff --git a/gcc/fortran/parse.c b/gcc/fortran/parse.c
index 6efb3fd5e7b..c44e23ccb62 100644
--- a/gcc/fortran/parse.c
+++ b/gcc/fortran/parse.c
@@ -920,11 +920,19 @@ decode_omp_directive (void)
matchs ("end do simd", gfc_match_omp_end_nowait, ST_OMP_END_DO_SIMD);
matcho ("end do", gfc_match_omp_end_nowait, ST_OMP_END_DO);
matchs ("end simd", gfc_match_omp_eos_error, ST_OMP_END_SIMD);
+ matcho ("end master taskloop simd", gfc_match_omp_eos_error,
+ ST_OMP_END_MASTER_TASKLOOP_SIMD);
+ matcho ("end master taskloop", gfc_match_omp_eos_error,
+ ST_OMP_END_MASTER_TASKLOOP);
matcho ("end master", gfc_match_omp_eos_error, ST_OMP_END_MASTER);
matchs ("end ordered", gfc_match_omp_eos_error, ST_OMP_END_ORDERED);
matchs ("end parallel do simd", gfc_match_omp_eos_error,
ST_OMP_END_PARALLEL_DO_SIMD);
matcho ("end parallel do", gfc_match_omp_eos_error, ST_OMP_END_PARALLEL_DO);
+ matcho ("end parallel master taskloop simd", gfc_match_omp_eos_error,
+ ST_OMP_END_PARALLEL_MASTER_TASKLOOP_SIMD);
+ matcho ("end parallel master taskloop", gfc_match_omp_eos_error,
+ ST_OMP_END_PARALLEL_MASTER_TASKLOOP);
matcho ("end parallel master", gfc_match_omp_eos_error,
ST_OMP_END_PARALLEL_MASTER);
matcho ("end parallel sections", gfc_match_omp_eos_error,
@@ -974,6 +982,10 @@ decode_omp_directive (void)
matcho ("flush", gfc_match_omp_flush, ST_OMP_FLUSH);
break;
case 'm':
+ matcho ("master taskloop simd", gfc_match_omp_master_taskloop_simd,
+ ST_OMP_MASTER_TASKLOOP_SIMD);
+ matcho ("master taskloop", gfc_match_omp_master_taskloop,
+ ST_OMP_MASTER_TASKLOOP);
matcho ("master", gfc_match_omp_master, ST_OMP_MASTER);
break;
case 'o':
@@ -992,6 +1004,12 @@ decode_omp_directive (void)
matchs ("parallel do simd", gfc_match_omp_parallel_do_simd,
ST_OMP_PARALLEL_DO_SIMD);
matcho ("parallel do", gfc_match_omp_parallel_do, ST_OMP_PARALLEL_DO);
+ matcho ("parallel master taskloop simd",
+ gfc_match_omp_parallel_master_taskloop_simd,
+ ST_OMP_PARALLEL_MASTER_TASKLOOP_SIMD);
+ matcho ("parallel master taskloop",
+ gfc_match_omp_parallel_master_taskloop,
+ ST_OMP_PARALLEL_MASTER_TASKLOOP);
matcho ("parallel master", gfc_match_omp_parallel_master,
ST_OMP_PARALLEL_MASTER);
matcho ("parallel sections", gfc_match_omp_parallel_sections,
@@ -1610,8 +1628,11 @@ next_statement (void)
case ST_IF_BLOCK: case ST_BLOCK: case ST_ASSOCIATE: \
case ST_WHERE_BLOCK: case ST_SELECT_CASE: case ST_SELECT_TYPE: \
case ST_SELECT_RANK: case ST_OMP_PARALLEL: case ST_OMP_PARALLEL_MASTER: \
+ case ST_OMP_PARALLEL_MASTER_TASKLOOP: \
+ case ST_OMP_PARALLEL_MASTER_TASKLOOP_SIMD: \
case ST_OMP_PARALLEL_SECTIONS: case ST_OMP_SECTIONS: case ST_OMP_ORDERED: \
- case ST_OMP_CRITICAL: case ST_OMP_MASTER: case ST_OMP_SINGLE: \
+ case ST_OMP_CRITICAL: case ST_OMP_MASTER: case ST_OMP_MASTER_TASKLOOP: \
+ case ST_OMP_MASTER_TASKLOOP_SIMD: case ST_OMP_SINGLE: \
case ST_OMP_DO: case ST_OMP_PARALLEL_DO: case ST_OMP_ATOMIC: \
case ST_OMP_WORKSHARE: case ST_OMP_PARALLEL_WORKSHARE: \
case ST_OMP_TASK: case ST_OMP_TASKGROUP: case ST_OMP_SIMD: \
@@ -2341,6 +2362,12 @@ gfc_ascii_statement (gfc_statement st)
case ST_OMP_END_MASTER:
p = "!$OMP END MASTER";
break;
+ case ST_OMP_END_MASTER_TASKLOOP:
+ p = "!$OMP END MASTER TASKLOOP";
+ break;
+ case ST_OMP_END_MASTER_TASKLOOP_SIMD:
+ p = "!$OMP END MASTER TASKLOOP SIMD";
+ break;
case ST_OMP_END_ORDERED:
p = "!$OMP END ORDERED";
break;
@@ -2356,6 +2383,12 @@ gfc_ascii_statement (gfc_statement st)
case ST_OMP_END_PARALLEL_MASTER:
p = "!$OMP END PARALLEL MASTER";
break;
+ case ST_OMP_END_PARALLEL_MASTER_TASKLOOP:
+ p = "!$OMP END PARALLEL MASTER TASKLOOP";
+ break;
+ case ST_OMP_END_PARALLEL_MASTER_TASKLOOP_SIMD:
+ p = "!$OMP END PARALLEL MASTER TASKLOOP SIMD";
+ break;
case ST_OMP_END_PARALLEL_SECTIONS:
p = "!$OMP END PARALLEL SECTIONS";
break;
@@ -2437,6 +2470,12 @@ gfc_ascii_statement (gfc_statement st)
case ST_OMP_MASTER:
p = "!$OMP MASTER";
break;
+ case ST_OMP_MASTER_TASKLOOP:
+ p = "!$OMP MASTER TASKLOOP";
+ break;
+ case ST_OMP_MASTER_TASKLOOP_SIMD:
+ p = "!$OMP MASTER TASKLOOP SIMD";
+ break;
case ST_OMP_ORDERED:
case ST_OMP_ORDERED_DEPEND:
p = "!$OMP ORDERED";
@@ -2453,6 +2492,12 @@ gfc_ascii_statement (gfc_statement st)
case ST_OMP_PARALLEL_MASTER:
p = "!$OMP PARALLEL MASTER";
break;
+ case ST_OMP_PARALLEL_MASTER_TASKLOOP:
+ p = "!$OMP PARALLEL MASTER TASKLOOP";
+ break;
+ case ST_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
+ p = "!$OMP PARALLEL MASTER TASKLOOP SIMD";
+ break;
case ST_OMP_PARALLEL_SECTIONS:
p = "!$OMP PARALLEL SECTIONS";
break;
@@ -5025,6 +5070,16 @@ parse_omp_do (gfc_statement omp_st)
break;
case ST_OMP_TASKLOOP: omp_end_st = ST_OMP_END_TASKLOOP; break;
case ST_OMP_TASKLOOP_SIMD: omp_end_st = ST_OMP_END_TASKLOOP_SIMD; break;
+ case ST_OMP_MASTER_TASKLOOP: omp_end_st = ST_OMP_END_MASTER_TASKLOOP; break;
+ case ST_OMP_MASTER_TASKLOOP_SIMD:
+ omp_end_st = ST_OMP_END_MASTER_TASKLOOP_SIMD;
+ break;
+ case ST_OMP_PARALLEL_MASTER_TASKLOOP:
+ omp_end_st = ST_OMP_END_PARALLEL_MASTER_TASKLOOP;
+ break;
+ case ST_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
+ omp_end_st = ST_OMP_END_PARALLEL_MASTER_TASKLOOP_SIMD;
+ break;
case ST_OMP_TEAMS_DISTRIBUTE:
omp_end_st = ST_OMP_END_TEAMS_DISTRIBUTE;
break;
@@ -5268,6 +5323,12 @@ parse_omp_structured_block (gfc_statement omp_st, bool workshare_stmts_only)
case ST_OMP_PARALLEL_MASTER:
omp_end_st = ST_OMP_END_PARALLEL_MASTER;
break;
+ case ST_OMP_PARALLEL_MASTER_TASKLOOP:
+ omp_end_st = ST_OMP_END_PARALLEL_MASTER_TASKLOOP;
+ break;
+ case ST_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
+ omp_end_st = ST_OMP_END_PARALLEL_MASTER_TASKLOOP_SIMD;
+ break;
case ST_OMP_PARALLEL_SECTIONS:
omp_end_st = ST_OMP_END_PARALLEL_SECTIONS;
break;
@@ -5283,6 +5344,12 @@ parse_omp_structured_block (gfc_statement omp_st, bool workshare_stmts_only)
case ST_OMP_MASTER:
omp_end_st = ST_OMP_END_MASTER;
break;
+ case ST_OMP_MASTER_TASKLOOP:
+ omp_end_st = ST_OMP_END_MASTER_TASKLOOP;
+ break;
+ case ST_OMP_MASTER_TASKLOOP_SIMD:
+ omp_end_st = ST_OMP_END_MASTER_TASKLOOP_SIMD;
+ break;
case ST_OMP_SINGLE:
omp_end_st = ST_OMP_END_SINGLE;
break;
@@ -5624,6 +5691,10 @@ parse_executable (gfc_statement st)
case ST_OMP_DO_SIMD:
case ST_OMP_PARALLEL_DO:
case ST_OMP_PARALLEL_DO_SIMD:
+ case ST_OMP_PARALLEL_MASTER_TASKLOOP:
+ case ST_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
+ case ST_OMP_MASTER_TASKLOOP:
+ case ST_OMP_MASTER_TASKLOOP_SIMD:
case ST_OMP_SIMD:
case ST_OMP_TARGET_PARALLEL_DO:
case ST_OMP_TARGET_PARALLEL_DO_SIMD:
diff --git a/gcc/fortran/resolve.c b/gcc/fortran/resolve.c
index 747516fbc1d..fed6dce3d98 100644
--- a/gcc/fortran/resolve.c
+++ b/gcc/fortran/resolve.c
@@ -10798,11 +10798,15 @@ gfc_resolve_blocks (gfc_code *b, gfc_namespace *ns)
case EXEC_OMP_DO:
case EXEC_OMP_DO_SIMD:
case EXEC_OMP_MASTER:
+ case EXEC_OMP_MASTER_TASKLOOP:
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_ORDERED:
case EXEC_OMP_PARALLEL:
case EXEC_OMP_PARALLEL_DO:
case EXEC_OMP_PARALLEL_DO_SIMD:
case EXEC_OMP_PARALLEL_MASTER:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_PARALLEL_SECTIONS:
case EXEC_OMP_PARALLEL_WORKSHARE:
case EXEC_OMP_SECTIONS:
@@ -11765,6 +11769,8 @@ gfc_resolve_code (gfc_code *code, gfc_namespace *ns)
case EXEC_OMP_PARALLEL_DO:
case EXEC_OMP_PARALLEL_DO_SIMD:
case EXEC_OMP_PARALLEL_MASTER:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_PARALLEL_SECTIONS:
case EXEC_OMP_TARGET_PARALLEL:
case EXEC_OMP_TARGET_PARALLEL_DO:
@@ -12214,6 +12220,8 @@ start:
case EXEC_OMP_DO:
case EXEC_OMP_DO_SIMD:
case EXEC_OMP_MASTER:
+ case EXEC_OMP_MASTER_TASKLOOP:
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_ORDERED:
case EXEC_OMP_SCAN:
case EXEC_OMP_SECTIONS:
@@ -12252,6 +12260,8 @@ start:
case EXEC_OMP_PARALLEL_DO:
case EXEC_OMP_PARALLEL_DO_SIMD:
case EXEC_OMP_PARALLEL_MASTER:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_PARALLEL_SECTIONS:
case EXEC_OMP_PARALLEL_WORKSHARE:
omp_workshare_save = omp_workshare_flag;
diff --git a/gcc/fortran/st.c b/gcc/fortran/st.c
index 7d0e2c14731..9f6fe49db02 100644
--- a/gcc/fortran/st.c
+++ b/gcc/fortran/st.c
@@ -226,11 +226,15 @@ gfc_free_statement (gfc_code *p)
case EXEC_OMP_DO:
case EXEC_OMP_DO_SIMD:
case EXEC_OMP_END_SINGLE:
+ case EXEC_OMP_MASTER_TASKLOOP:
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_ORDERED:
case EXEC_OMP_PARALLEL:
case EXEC_OMP_PARALLEL_DO:
case EXEC_OMP_PARALLEL_DO_SIMD:
case EXEC_OMP_PARALLEL_MASTER:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_PARALLEL_SECTIONS:
case EXEC_OMP_PARALLEL_WORKSHARE:
case EXEC_OMP_SCAN:
diff --git a/gcc/fortran/trans-openmp.c b/gcc/fortran/trans-openmp.c
index 7ea7aa37efd..2917d3d058b 100644
--- a/gcc/fortran/trans-openmp.c
+++ b/gcc/fortran/trans-openmp.c
@@ -5380,6 +5380,14 @@ gfc_split_omp_clauses (gfc_code *code,
mask = GFC_OMP_MASK_PARALLEL | GFC_OMP_MASK_DO | GFC_OMP_MASK_SIMD;
innermost = GFC_OMP_SPLIT_SIMD;
break;
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ mask = GFC_OMP_MASK_PARALLEL | GFC_OMP_MASK_TASKLOOP | GFC_OMP_MASK_SIMD;
+ innermost = GFC_OMP_SPLIT_TASKLOOP;
+ break;
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
+ mask = GFC_OMP_MASK_PARALLEL | GFC_OMP_MASK_TASKLOOP | GFC_OMP_MASK_SIMD;
+ innermost = GFC_OMP_SPLIT_SIMD;
+ break;
case EXEC_OMP_SIMD:
innermost = GFC_OMP_SPLIT_SIMD;
break;
@@ -5427,9 +5435,11 @@ gfc_split_omp_clauses (gfc_code *code,
| GFC_OMP_MASK_DISTRIBUTE | GFC_OMP_MASK_SIMD;
innermost = GFC_OMP_SPLIT_SIMD;
break;
+ case EXEC_OMP_MASTER_TASKLOOP:
case EXEC_OMP_TASKLOOP:
innermost = GFC_OMP_SPLIT_TASKLOOP;
break;
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_TASKLOOP_SIMD:
mask = GFC_OMP_MASK_TASKLOOP | GFC_OMP_MASK_SIMD;
innermost = GFC_OMP_SPLIT_SIMD;
@@ -5820,28 +5830,6 @@ gfc_trans_omp_parallel_do_simd (gfc_code *code, stmtblock_t *pblock,
return gfc_finish_block (&block);
}
-static tree
-gfc_trans_omp_parallel_master (gfc_code *code)
-{
- stmtblock_t block;
- tree stmt, omp_clauses;
-
- gfc_start_block (&block);
- omp_clauses = gfc_trans_omp_clauses (&block, code->ext.omp_clauses,
- code->loc);
- pushlevel ();
- stmt = gfc_trans_omp_master (code);
- if (TREE_CODE (stmt) != BIND_EXPR)
- stmt = build3_v (BIND_EXPR, NULL, stmt, poplevel (1, 0));
- else
- poplevel (0, 0);
- stmt = build2_loc (gfc_get_location (&code->loc), OMP_PARALLEL,
- void_type_node, stmt, omp_clauses);
- OMP_PARALLEL_COMBINED (stmt) = 1;
- gfc_add_expr_to_block (&block, stmt);
- return gfc_finish_block (&block);
-}
-
static tree
gfc_trans_omp_parallel_sections (gfc_code *code)
{
@@ -6217,7 +6205,7 @@ gfc_trans_omp_target (gfc_code *code)
}
static tree
-gfc_trans_omp_taskloop (gfc_code *code)
+gfc_trans_omp_taskloop (gfc_code *code, gfc_exec_op op)
{
stmtblock_t block;
gfc_omp_clauses clausesa[GFC_OMP_SPLIT_NUM];
@@ -6229,7 +6217,7 @@ gfc_trans_omp_taskloop (gfc_code *code)
omp_clauses
= gfc_trans_omp_clauses (&block, &clausesa[GFC_OMP_SPLIT_TASKLOOP],
code->loc);
- switch (code->op)
+ switch (op)
{
case EXEC_OMP_TASKLOOP:
/* This is handled in gfc_trans_omp_do. */
@@ -6258,6 +6246,75 @@ gfc_trans_omp_taskloop (gfc_code *code)
return gfc_finish_block (&block);
}
+static tree
+gfc_trans_omp_master_taskloop (gfc_code *code, gfc_exec_op op)
+{
+ stmtblock_t block;
+ tree stmt;
+
+ gfc_start_block (&block);
+ pushlevel ();
+ if (op == EXEC_OMP_MASTER_TASKLOOP_SIMD)
+ stmt = gfc_trans_omp_taskloop (code, EXEC_OMP_TASKLOOP_SIMD);
+ else
+ {
+ gfc_omp_clauses clausesa[GFC_OMP_SPLIT_NUM];
+ gcc_assert (op == EXEC_OMP_MASTER_TASKLOOP);
+ if (op != code->op)
+ gfc_split_omp_clauses (code, clausesa);
+ stmt = gfc_trans_omp_do (code, EXEC_OMP_TASKLOOP, NULL,
+ op != code->op
+ ? &clausesa[GFC_OMP_SPLIT_TASKLOOP]
+ : code->ext.omp_clauses, NULL);
+ }
+ if (TREE_CODE (stmt) != BIND_EXPR)
+ stmt = build3_v (BIND_EXPR, NULL, stmt, poplevel (1, 0));
+ else
+ poplevel (0, 0);
+ stmt = build1_v (OMP_MASTER, stmt);
+ gfc_add_expr_to_block (&block, stmt);
+ return gfc_finish_block (&block);
+}
+
+static tree
+gfc_trans_omp_parallel_master (gfc_code *code)
+{
+ stmtblock_t block;
+ tree stmt, omp_clauses;
+ gfc_omp_clauses clausesa[GFC_OMP_SPLIT_NUM];
+
+ if (code->op != EXEC_OMP_PARALLEL_MASTER)
+ gfc_split_omp_clauses (code, clausesa);
+
+ gfc_start_block (&block);
+ omp_clauses = gfc_trans_omp_clauses (&block,
+ code->op == EXEC_OMP_PARALLEL_MASTER
+ ? code->ext.omp_clauses
+ : &clausesa[GFC_OMP_SPLIT_PARALLEL],
+ code->loc);
+ pushlevel ();
+ if (code->op == EXEC_OMP_PARALLEL_MASTER)
+ stmt = gfc_trans_omp_master (code);
+ else
+ {
+ gcc_assert (code->op == EXEC_OMP_PARALLEL_MASTER_TASKLOOP
+ || code->op == EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD);
+ gfc_exec_op op = (code->op == EXEC_OMP_PARALLEL_MASTER_TASKLOOP
+ ? EXEC_OMP_MASTER_TASKLOOP
+ : EXEC_OMP_MASTER_TASKLOOP_SIMD);
+ stmt = gfc_trans_omp_master_taskloop (code, op);
+ }
+ if (TREE_CODE (stmt) != BIND_EXPR)
+ stmt = build3_v (BIND_EXPR, NULL, stmt, poplevel (1, 0));
+ else
+ poplevel (0, 0);
+ stmt = build2_loc (gfc_get_location (&code->loc), OMP_PARALLEL,
+ void_type_node, stmt, omp_clauses);
+ OMP_PARALLEL_COMBINED (stmt) = 1;
+ gfc_add_expr_to_block (&block, stmt);
+ return gfc_finish_block (&block);
+}
+
static tree
gfc_trans_omp_target_data (gfc_code *code)
{
@@ -6568,6 +6625,9 @@ gfc_trans_omp_directive (gfc_code *code)
return gfc_trans_omp_flush (code);
case EXEC_OMP_MASTER:
return gfc_trans_omp_master (code);
+ case EXEC_OMP_MASTER_TASKLOOP:
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
+ return gfc_trans_omp_master_taskloop (code, code->op);
case EXEC_OMP_ORDERED:
return gfc_trans_omp_ordered (code);
case EXEC_OMP_PARALLEL:
@@ -6577,6 +6637,8 @@ gfc_trans_omp_directive (gfc_code *code)
case EXEC_OMP_PARALLEL_DO_SIMD:
return gfc_trans_omp_parallel_do_simd (code, NULL, NULL);
case EXEC_OMP_PARALLEL_MASTER:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
return gfc_trans_omp_parallel_master (code);
case EXEC_OMP_PARALLEL_SECTIONS:
return gfc_trans_omp_parallel_sections (code);
@@ -6610,7 +6672,7 @@ gfc_trans_omp_directive (gfc_code *code)
case EXEC_OMP_TASKGROUP:
return gfc_trans_omp_taskgroup (code);
case EXEC_OMP_TASKLOOP_SIMD:
- return gfc_trans_omp_taskloop (code);
+ return gfc_trans_omp_taskloop (code, code->op);
case EXEC_OMP_TASKWAIT:
return gfc_trans_omp_taskwait (code);
case EXEC_OMP_TASKYIELD:
diff --git a/gcc/fortran/trans.c b/gcc/fortran/trans.c
index 9f296bd8c03..cbbfcd9d1bc 100644
--- a/gcc/fortran/trans.c
+++ b/gcc/fortran/trans.c
@@ -2170,11 +2170,15 @@ trans_code (gfc_code * code, tree cond)
case EXEC_OMP_DO_SIMD:
case EXEC_OMP_FLUSH:
case EXEC_OMP_MASTER:
+ case EXEC_OMP_MASTER_TASKLOOP:
+ case EXEC_OMP_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_ORDERED:
case EXEC_OMP_PARALLEL:
case EXEC_OMP_PARALLEL_DO:
case EXEC_OMP_PARALLEL_DO_SIMD:
case EXEC_OMP_PARALLEL_MASTER:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP:
+ case EXEC_OMP_PARALLEL_MASTER_TASKLOOP_SIMD:
case EXEC_OMP_PARALLEL_SECTIONS:
case EXEC_OMP_PARALLEL_WORKSHARE:
case EXEC_OMP_SECTIONS:
diff --git a/gcc/testsuite/gfortran.dg/gomp/pr99928-3.f90 b/gcc/testsuite/gfortran.dg/gomp/pr99928-3.f90
new file mode 100644
index 00000000000..ce43dfb8ccd
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/gomp/pr99928-3.f90
@@ -0,0 +1,139 @@
+! PR middle-end/99928
+! { dg-do compile }
+! { dg-options "-fopenmp -fdump-tree-gimple" }
+
+module m
+ implicit none
+ integer :: l00, l01, l02, l03, l04, l07, l08, l09
+ integer :: l10, l11
+
+contains
+
+subroutine bar ()
+ integer :: l05, l06
+ integer :: i
+ l05 = 0; l06 = 0
+ ! { dg-final { scan-tree-dump "omp for\[^\n\r]*firstprivate\\(l00\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp for\[^\n\r]*lastprivate\\(l00\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp simd\[^\n\r]*firstprivate\\(l00\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp simd\[^\n\r]*lastprivate\\(l00\\)" "gimple" } }
+ !$omp do simd firstprivate (l00) lastprivate (l00)
+ do i = 1, 64
+ l00 = i
+ end do
+ ! { dg-final { scan-tree-dump-not "omp master\[^\n\r]*firstprivate\\(l01\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp master\[^\n\r]*lastprivate\\(l01\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp taskloop\[^\n\r]*firstprivate\\(l01\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp taskloop\[^\n\r]*lastprivate\\(l01\\)" "gimple" } }
+ !$omp master taskloop firstprivate (l01) lastprivate (l01)
+ do i = 1, 64
+ l01 = i
+ end do
+ ! { dg-final { scan-tree-dump-not "omp master\[^\n\r]*firstprivate\\(l02\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp master\[^\n\r]*lastprivate\\(l02\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp taskloop\[^\n\r]*firstprivate\\(l02\\)" "gimple" { xfail *-*-* } } }
+ ! { dg-final { scan-tree-dump "omp taskloop\[^\n\r]*lastprivate\\(l02\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp simd\[^\n\r]*firstprivate\\(l02\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp simd\[^\n\r]*lastprivate\\(l02\\)" "gimple" } }
+ !$omp master taskloop simd firstprivate (l02) lastprivate (l02)
+ do i = 1, 64
+ l02 = i
+ end do
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*firstprivate\\(l03\\)" "gimple" } } ! FIXME: This should be on for instead.
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*lastprivate\\(l03\\)" "gimple" } } ! FIXME: This should be on for instead.
+ ! { dg-final { scan-tree-dump-not "omp for\[^\n\r]*firstprivate\\(l03\\)" "gimple" } } ! FIXME.
+ ! { dg-final { scan-tree-dump-not "omp for\[^\n\r]*lastprivate\\(l03\\)" "gimple" } } ! FIXME.
+ !$omp parallel do firstprivate (l03) lastprivate (l03)
+ do i = 1, 64
+ l03 = i
+ end do
+ !$omp end parallel do
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*firstprivate\\(l04\\)" "gimple" } } ! FIXME: This should be on for instead.
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*lastprivate\\(l04\\)" "gimple" } } ! FIXME: This should be on for instead.
+ ! { dg-final { scan-tree-dump-not "omp for\[^\n\r]*firstprivate\\(l04\\)" "gimple" } } ! FIXME.
+ ! { dg-final { scan-tree-dump-not "omp for\[^\n\r]*lastprivate\\(l04\\)" "gimple" } } ! FIXME.
+ ! { dg-final { scan-tree-dump-not "omp simd\[^\n\r]*firstprivate\\(l04\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp simd\[^\n\r]*lastprivate\\(l04\\)" "gimple" } }
+ !$omp parallel do simd firstprivate (l04) lastprivate (l04)
+ do i = 1, 64
+ l04 = i
+ end do
+ !$omp end parallel do simd
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*shared\\(l05\\)" "gimple" { xfail *-*-* } } }
+ ! { dg-final { scan-tree-dump-not "omp master\[^\n\r]*firstprivate\\(l05\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp master\[^\n\r]*lastprivate\\(l05\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp taskloop\[^\n\r]*firstprivate\\(l05\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp taskloop\[^\n\r]*lastprivate\\(l05\\)" "gimple" { xfail *-*-* } } }
+ !$omp parallel master taskloop firstprivate (l05) lastprivate (l05)
+ do i = 1, 64
+ l05 = i
+ end do
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*shared\\(l06\\)" "gimple" { xfail *-*-* } } }
+ ! { dg-final { scan-tree-dump-not "omp master\[^\n\r]*firstprivate\\(l06\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp master\[^\n\r]*lastprivate\\(l06\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp taskloop\[^\n\r]*firstprivate\\(l06\\)" "gimple" { xfail *-*-* } } }
+ ! { dg-final { scan-tree-dump "omp taskloop\[^\n\r]*lastprivate\\(l06\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp simd\[^\n\r]*firstprivate\\(l06\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp simd\[^\n\r]*lastprivate\\(l06\\)" "gimple" } }
+ !$omp parallel master taskloop simd firstprivate (l06) lastprivate (l06)
+ do i = 1, 64
+ l06 = i
+ end do
+ !$omp end parallel master taskloop simd
+ ! FIXME: OpenMP 5.0/5.1 broken here, conceptually it should be shared on parallel and
+ ! firstprivate+lastprivate on sections, in GCC implementation we put firstprivate+lastprivate
+ ! on parallel for historic reasons, but OpenMP 5.0/5.1 mistakenly say firstprivate
+ ! should be on parallel and lastprivate on sections.
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*firstprivate\\(l07\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*lastprivate\\(l07\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp sections\[^\n\r]*firstprivate\\(l07\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp sections\[^\n\r]*lastprivate\\(l07\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp section \[^\n\r]*firstprivate\\(l07\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp section \[^\n\r]*lastprivate\\(l07\\)" "gimple" } }
+ !$omp parallel sections firstprivate (l07) lastprivate (l07)
+ l07 = 1
+ !$omp section
+ l07 = 2
+ !$omp end parallel sections
+ ! { dg-final { scan-tree-dump "omp target\[^\n\r]*map\\(tofrom:l08" "gimple" { xfail *-*-* } } }
+ ! { dg-final { scan-tree-dump-not "omp target\[^\n\r]*firstprivate\\(l08\\)" "gimple" { xfail *-*-* } } }
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*firstprivate\\(l08\\)" "gimple" } } ! FIXME: This should be on for instead.
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*lastprivate\\(l08\\)" "gimple" } } ! FIXME: This should be on for instead.
+ ! { dg-final { scan-tree-dump-not "omp for\[^\n\r]*firstprivate\\(l08\\)" "gimple" } } ! FIXME.
+ ! { dg-final { scan-tree-dump-not "omp for\[^\n\r]*lastprivate\\(l08\\)" "gimple" } } ! FIXME.
+ !$omp target parallel do firstprivate (l08) lastprivate (l08)
+ do i = 1, 64
+ l08 = i
+ end do
+ !$omp end target parallel do
+ ! { dg-final { scan-tree-dump "omp target\[^\n\r]*map\\(tofrom:l09" "gimple" { xfail *-*-* } } }
+ ! { dg-final { scan-tree-dump-not "omp target\[^\n\r]*firstprivate\\(l09\\)" "gimple" { xfail *-*-* } } }
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*firstprivate\\(l09\\)" "gimple" } } ! FIXME: This should be on for instead.
+ ! { dg-final { scan-tree-dump "omp parallel\[^\n\r]*lastprivate\\(l09\\)" "gimple" } } ! FIXME: This should be on for instead.
+ ! { dg-final { scan-tree-dump-not "omp for\[^\n\r]*firstprivate\\(l09\\)" "gimple" } } ! FIXME.
+ ! { dg-final { scan-tree-dump-not "omp for\[^\n\r]*lastprivate\\(l09\\)" "gimple" } } ! FIXME.
+ ! { dg-final { scan-tree-dump-not "omp simd\[^\n\r]*firstprivate\\(l09\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp simd\[^\n\r]*lastprivate\\(l09\\)" "gimple" } }
+ !$omp target parallel do simd firstprivate (l09) lastprivate (l09)
+ do i = 1, 64
+ l09 = i
+ end do
+ ! { dg-final { scan-tree-dump "omp target\[^\n\r]*map\\(tofrom:l10" "gimple" { xfail *-*-* } } }
+ ! { dg-final { scan-tree-dump-not "omp target\[^\n\r]*firstprivate\\(l10\\)" "gimple" { xfail *-*-* } } }
+ ! { dg-final { scan-tree-dump-not "omp simd\[^\n\r]*firstprivate\\(l10\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp simd\[^\n\r]*lastprivate\\(l10\\)" "gimple" } }
+ !$omp target simd firstprivate (l10) lastprivate (l10)
+ do i = 1, 64
+ l10 = i
+ end do
+ ! { dg-final { scan-tree-dump "omp taskloop\[^\n\r]*firstprivate\\(l11\\)" "gimple" { xfail *-*-* } } }
+ ! { dg-final { scan-tree-dump "omp taskloop\[^\n\r]*lastprivate\\(l11\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump-not "omp simd\[^\n\r]*firstprivate\\(l11\\)" "gimple" } }
+ ! { dg-final { scan-tree-dump "omp simd\[^\n\r]*lastprivate\\(l11\\)" "gimple" } }
+ !$omp taskloop simd firstprivate (l11) lastprivate (l11)
+ do i = 1, 64
+ l11 = i
+ end do
+ !$omp end taskloop simd
+end
+end module m
diff --git a/gcc/testsuite/gfortran.dg/gomp/reduction5.f90 b/gcc/testsuite/gfortran.dg/gomp/reduction5.f90
index 032703d3b81..44f89d84c78 100644
--- a/gcc/testsuite/gfortran.dg/gomp/reduction5.f90
+++ b/gcc/testsuite/gfortran.dg/gomp/reduction5.f90
@@ -22,8 +22,7 @@ end do
!$omp taskloop reduction(inscan,+:a) in_reduction(+:b) ! { dg-error "'inscan' REDUCTION clause on construct other than DO, SIMD, DO SIMD, PARALLEL DO, PARALLEL DO SIMD" }
! { dg-error "34: With INSCAN at .1., expected loop body with ..OMP SCAN between two structured-block-sequences" "" { target *-*-* } .-1 }
- ! { dg-error "Only DEFAULT permitted as reduction-modifier in REDUCTION clause" "" { target *-*-* } .-2 }
- ! { dg-error "'inscan' and non-'inscan' 'reduction' clauses on the same construct" "" { target *-*-* } .-3 }
+ ! { dg-error "'inscan' and non-'inscan' 'reduction' clauses on the same construct" "" { target *-*-* } .-2 }
do i=1,10
a = a + 1
end do
@@ -34,7 +33,6 @@ do i=1,10
end do
!$omp teams reduction(inscan,+:b) ! { dg-error "'inscan' REDUCTION clause on construct other than DO, SIMD, DO SIMD, PARALLEL DO, PARALLEL DO SIMD" }
- ! { dg-error "Only DEFAULT permitted as reduction-modifier in REDUCTION clause" "" { target *-*-* } .-1 }
a = a + 1
!$omp end teams
diff --git a/gcc/testsuite/gfortran.dg/gomp/scan-1.f90 b/gcc/testsuite/gfortran.dg/gomp/scan-1.f90
index 8c879fd98b9..61d89259c48 100644
--- a/gcc/testsuite/gfortran.dg/gomp/scan-1.f90
+++ b/gcc/testsuite/gfortran.dg/gomp/scan-1.f90
@@ -100,7 +100,7 @@ subroutine f3 (c, d)
use m
implicit none
integer i, c(64), d(64)
- !$omp teams reduction (inscan, +: a) ! { dg-error "Only DEFAULT permitted as reduction-modifier in REDUCTION clause at" }
+ !$omp teams reduction (inscan, +: a)
! { dg-error "'inscan' REDUCTION clause on construct other than DO, SIMD, DO SIMD, PARALLEL DO, PARALLEL DO SIMD" "" { target *-*-* } .-1 }
! ...
!$omp end teams
@@ -135,7 +135,7 @@ subroutine f4 (c, d)
use m
implicit none
integer i, c(64), d(64)
- !$omp taskloop reduction (inscan, +: a) ! { dg-error "Only DEFAULT permitted as reduction-modifier in REDUCTION clause" }
+ !$omp taskloop reduction (inscan, +: a)
! { dg-error "'inscan' REDUCTION clause on construct other than DO, SIMD, DO SIMD, PARALLEL DO, PARALLEL DO SIMD" "" { target *-*-* } .-1 }
do i = 1, 64
d(i) = a
diff --git a/gcc/testsuite/gfortran.dg/gomp/taskloop-1.f90 b/gcc/testsuite/gfortran.dg/gomp/taskloop-1.f90
new file mode 100644
index 00000000000..7060a7a5c96
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/gomp/taskloop-1.f90
@@ -0,0 +1,126 @@
+module m
+ implicit none
+ integer :: t
+ !$omp threadprivate (t)
+ integer :: f, l, ll, r, r2
+ !$omp declare target to(f, l, ll, r, r2)
+end module m
+
+subroutine foo(fi, p, pp, g, s, nta, nth, ntm, i1, i2, i3, q)
+ use m
+ implicit none
+ integer, value :: p, pp, g, s, nta, nth, ntm
+ logical, value :: fi, i1, i2, i3
+ integer, pointer :: q(:)
+ integer :: i
+
+ !$omp taskgroup task_reduction(+:r2) !allocate (r2)
+ !$omp taskloop simd &
+ !$omp& private (p) firstprivate (f) lastprivate (l) shared (s) default(shared) grainsize (g) collapse(1) untied if(taskloop: i1) &
+ !$omp& if(simd: i2) final(fi) mergeable priority (pp) &
+ !$omp& safelen(8) simdlen(4) linear(ll: 1) aligned(q: 32) reduction(default, +:r) in_reduction(+:r2) nontemporal(ntm) &
+ !$omp& order(concurrent) !allocate (f)
+ do i = 1, 64
+ ll = ll + 1
+ end do
+ !$omp end taskgroup
+
+ !$omp taskgroup task_reduction(+:r) !allocate (r)
+ !$omp taskloop simd &
+ !$omp& private (p) firstprivate (f) lastprivate (l) shared (s) default(shared) grainsize (g) &
+ !$omp& collapse(1) untied if(i1) final(fi) mergeable nogroup priority (pp) &
+ !$omp& safelen(8) simdlen(4) linear(ll: 1) aligned(q: 32) in_reduction(+:r) nontemporal(ntm) &
+ !$omp& order(concurrent) !allocate (f)
+ do i = 1, 64
+ ll = ll + 1
+ end do
+ !$omp taskwait
+
+ !$omp taskloop simd &
+ !$omp& private (p) firstprivate (f) lastprivate (l) shared (s) default(shared) num_tasks (nta) &
+ !$omp& collapse(1) if(taskloop: i1) final(fi) priority (pp) &
+ !$omp& safelen(8) simdlen(4) linear(ll: 1) aligned(q: 32) reduction(+:r) if (simd: i3) nontemporal(ntm) &
+ !$omp& order(concurrent) !allocate (f)
+ do i = 1, 64
+ ll = ll + 1
+ end do
+ !$omp end taskgroup
+
+ !$omp taskgroup task_reduction (+:r2) !allocate (r2)
+ !$omp master taskloop &
+ !$omp& private (p) firstprivate (f) lastprivate (l) shared (s) default(shared) grainsize (g) &
+ !$omp& collapse(1) untied if(taskloop: i1) final(fi) mergeable priority (pp) &
+ !$omp& reduction(default, +:r) in_reduction(+:r2) !allocate (f)
+ do i = 1, 64
+ ll = ll + 1
+ end do
+ !$omp end taskgroup
+
+ !$omp taskgroup task_reduction (+:r2) !allocate (r2)
+ !$omp master taskloop simd &
+ !$omp& private (p) firstprivate (f) lastprivate (l) shared (s) default(shared) grainsize (g) &
+ !$omp& collapse(1) untied if(taskloop: i1) if(simd: i2) final(fi) mergeable priority (pp) &
+ !$omp& safelen(8) simdlen(4) linear(ll: 1) aligned(q: 32) reduction(default, +:r) in_reduction(+:r2) nontemporal(ntm) &
+ !$omp& order(concurrent) !allocate (f)
+ do i = 1, 64
+ ll = ll + 1
+ end do
+ !$omp end taskgroup
+
+ !$omp parallel master taskloop &
+ !$omp& private (p) firstprivate (f) lastprivate (l) shared (s) default(shared) grainsize (g) &
+ !$omp& collapse(1) untied if(taskloop: i1) final(fi) mergeable priority (pp) &
+ !$omp& reduction(default, +:r) if (parallel: i2) num_threads (nth) proc_bind(spread) copyin(t) !allocate (f)
+ do i = 1, 64
+ ll = ll + 1
+ end do
+
+ !$omp parallel master taskloop simd &
+ !$omp& private (p) firstprivate (f) lastprivate (l) shared (s) default(shared) grainsize (g) collapse(1) &
+ !$omp& untied if(taskloop: i1) if(simd: i2) final(fi) mergeable priority (pp) &
+ !$omp& safelen(8) simdlen(4) linear(ll: 1) aligned(q: 32) reduction(default, +:r) nontemporal(ntm) &
+ !$omp& if (parallel: i2) num_threads (nth) proc_bind(spread) copyin(t) &
+ !$omp& order(concurrent) !allocate (f)
+ do i = 1, 64
+ ll = ll + 1
+ end do
+
+ !$omp taskgroup task_reduction (+:r2) !allocate (r2)
+ !$omp master taskloop &
+ !$omp& private (p) firstprivate (f) lastprivate (l) shared (s) default(shared) num_tasks (nta) &
+ !$omp& collapse(1) untied if(i1) final(fi) mergeable priority (pp) &
+ !$omp& reduction(default, +:r) in_reduction(+:r2)
+ do i = 1, 64
+ ll = ll + 1
+ end do
+ !$omp end taskgroup
+
+ !$omp taskgroup task_reduction (+:r2) !allocate (r2)
+ !$omp master taskloop simd &
+ !$omp& private (p) firstprivate (f) lastprivate (l) shared (s) default(shared) num_tasks (nta) &
+ !$omp& collapse(1) untied if(i1) final(fi) mergeable priority (pp) &
+ !$omp& safelen(8) simdlen(4) linear(ll: 1) aligned(q: 32) reduction(default, +:r) in_reduction(+:r2) nontemporal(ntm) &
+ !$omp& order(concurrent) !allocate (f)
+ do i = 1, 64
+ ll = ll + 1
+ end do
+ !$omp end taskgroup
+
+ !$omp parallel master taskloop &
+ !$omp& private (p) firstprivate (f) lastprivate (l) shared (s) default(shared) num_tasks (nta) &
+ !$omp& collapse(1) untied if(i1) final(fi) mergeable priority (pp) &
+ !$omp& reduction(default, +:r) num_threads (nth) proc_bind(spread) copyin(t) !allocate (f)
+ do i = 1, 64
+ ll = ll + 1
+ end do
+
+ !$omp parallel master taskloop simd &
+ !$omp& private (p) firstprivate (f) lastprivate (l) shared (s) default(shared) num_tasks (nta) &
+ !$omp& collapse(1) untied if(i1) final(fi) mergeable priority (pp) &
+ !$omp& safelen(8) simdlen(4) linear(ll: 1) aligned(q: 32) reduction(default, +:r) &
+ !$omp& nontemporal(ntm) num_threads (nth) proc_bind(spread) copyin(t) &
+ !$omp& order(concurrent) !allocate (f)
+ do i = 1, 64
+ ll = ll + 1
+ end do
+end
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [Patch] Fortran/OpenMP: Support (parallel) master taskloop (simd) [PR99928]
2021-06-01 9:53 [Patch] Fortran/OpenMP: Support (parallel) master taskloop (simd) [PR99928] Tobias Burnus
@ 2021-06-01 10:02 ` Jakub Jelinek
2021-06-01 13:18 ` Tobias Burnus
0 siblings, 1 reply; 3+ messages in thread
From: Jakub Jelinek @ 2021-06-01 10:02 UTC (permalink / raw)
To: Tobias Burnus; +Cc: gcc-patches, fortran
On Tue, Jun 01, 2021 at 11:53:12AM +0200, Tobias Burnus wrote:
> Fortran/OpenMP: Support (parallel) master taskloop (simd) [PR99928]
>
> PR middle-end/99928
>
> gcc/fortran/ChangeLog:
>
> * dump-parse-tree.c (show_omp_node, show_code_node): Handle
> (parallel) master taskloop (simd).
> * frontend-passes.c (gfc_code_walker): Set in_omp_workshare
> to false for parallel master taskloop (simd).
> * gfortran.h (enum gfc_statement):
> Add ST_OMP_(END_)(PARALLEL_)MASTER_TASKLOOP(_SIMD).
> (enum gfc_exec_op): EXEC_OMP_(PARALLEL_)MASTER_TASKLOOP(_SIMD).
> * match.h (gfc_match_omp_master_taskloop,
> gfc_match_omp_master_taskloop_simd,
> gfc_match_omp_parallel_master_taskloop,
> gfc_match_omp_parallel_master_taskloop_simd): New prototype.
> * openmp.c (gfc_match_omp_parallel_master_taskloop,
> gfc_match_omp_parallel_master_taskloop_simd,
> gfc_match_omp_master_taskloop,
> gfc_match_omp_master_taskloop_simd): New.
> (gfc_match_omp_taskloop_simd): Permit 'reduction' clause.
> (resolve_omp_clauses): Handle new combined directives; remove
> inscan-reduction check to reduce multiple errors; add
> task-reduction error for 'taskloop simd'.
> (gfc_resolve_omp_parallel_blocks,
> resolve_omp_do, omp_code_to_statement,
> gfc_resolve_omp_directive): Handle new combined constructs.
> * parse.c (decode_omp_directive, next_statement,
> gfc_ascii_statement, parse_omp_do, parse_omp_structured_block,
> parse_executable): Likewise.
> * resolve.c (gfc_resolve_blocks, gfc_resolve_code): Likewise.
> * st.c (gfc_free_statement): Likewise.
> * trans.c (trans_code): Likewise.
> * trans-openmp.c (gfc_split_omp_clauses,
> gfc_trans_omp_directive): Likewise.
> (gfc_trans_omp_parallel_master): Move after gfc_trans_omp_master_taskloop;
> handle parallel master taskloop (simd) as well.
> (gfc_trans_omp_taskloop): Take gfc_exec_op as arg.
> (gfc_trans_omp_master_taskloop): New.
>
> gcc/testsuite/ChangeLog:
>
> * gfortran.dg/gomp/reduction5.f90: Remove dg-error the issue is
> now diagnosed with error output less.
The above sentence is weird. I'd expect semicolon or comma before the
and less error output instead of error output less.
> * gfortran.dg/gomp/scan-1.f90: Likewise.
> * gfortran.dg/gomp/pr99928-3.f90: New test.
> * gfortran.dg/gomp/taskloop-1.f90: New test.
Otherwise LGTM.
Jakub
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [Patch] Fortran/OpenMP: Support (parallel) master taskloop (simd) [PR99928]
2021-06-01 10:02 ` Jakub Jelinek
@ 2021-06-01 13:18 ` Tobias Burnus
0 siblings, 0 replies; 3+ messages in thread
From: Tobias Burnus @ 2021-06-01 13:18 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: gcc-patches, fortran
[-- Attachment #1: Type: text/plain, Size: 560 bytes --]
On 01.06.21 12:02, Jakub Jelinek wrote:
>> * gfortran.dg/gomp/scan-1.f90: Likewise.
>> * gfortran.dg/gomp/pr99928-3.f90: New test.
>> * gfortran.dg/gomp/taskloop-1.f90: New test.
> Otherwise LGTM.
> Jakub
Thanks for the review! – I just saw that I missed to include another
testcase – unless there I comments, I will commit it later.
Tobias
-----------------
Mentor Graphics (Deutschland) GmbH, Arnulfstrasse 201, 80634 München Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Frank Thürauf
[-- Attachment #2: omp-taskloop.diff --]
[-- Type: text/x-patch, Size: 3688 bytes --]
Fortran/OpenMP: Add gfortran.dg/gomp/taskloop-2.f90 [PR99928]
PR middle-end/99928
gcc/testsuite/ChangeLog
* gfortran.dg/gomp/taskloop-2.f90: New.
diff --git a/gcc/testsuite/gfortran.dg/gomp/taskloop-2.f90 b/gcc/testsuite/gfortran.dg/gomp/taskloop-2.f90
new file mode 100644
index 00000000000..21427623584
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/gomp/taskloop-2.f90
@@ -0,0 +1,72 @@
+subroutine foo()
+implicit none
+integer :: i, r
+!$omp taskloop reduction(task, +: r) ! { dg-error "Only DEFAULT permitted as reduction-modifier in REDUCTION clause" }
+do i = 1, 64
+end do
+!$omp taskloop simd reduction(task, +: r) ! { dg-error "Only DEFAULT permitted as reduction-modifier in REDUCTION clause" }
+do i = 1, 64
+end do
+!$omp master taskloop reduction(task, +: r) ! { dg-error "Only DEFAULT permitted as reduction-modifier in REDUCTION clause" }
+do i = 1, 64
+end do
+!$omp master taskloop simd reduction(task, +: r) ! { dg-error "Only DEFAULT permitted as reduction-modifier in REDUCTION clause" }
+do i = 1, 64
+end do
+!$omp parallel master taskloop reduction(task, +: r) ! { dg-error "Only DEFAULT permitted as reduction-modifier in REDUCTION clause" }
+do i = 1, 64
+end do
+!$omp parallel master taskloop simd reduction(task, +: r) ! { dg-error "Only DEFAULT permitted as reduction-modifier in REDUCTION clause" }
+do i = 1, 64
+end do
+
+!$omp taskloop reduction(inscan, +: r) ! { dg-error "'inscan' REDUCTION clause on construct other than DO, SIMD, DO SIMD, PARALLEL DO, PARALLEL DO SIMD" }
+do i = 1, 64 ! { dg-error "OMP SCAN between two structured-block-sequences" "" { target *-*-* } .-1 }
+end do
+!$omp taskloop simd reduction(inscan, +: r) ! { dg-error "'inscan' REDUCTION clause on construct other than DO, SIMD, DO SIMD, PARALLEL DO, PARALLEL DO SIMD" }
+do i = 1, 64 ! { dg-error "OMP SCAN between two structured-block-sequences" "" { target *-*-* } .-1 }
+end do
+!$omp master taskloop reduction(inscan, +: r) ! { dg-error "'inscan' REDUCTION clause on construct other than DO, SIMD, DO SIMD, PARALLEL DO, PARALLEL DO SIMD" }
+do i = 1, 64
+end do
+!$omp master taskloop simd reduction(inscan, +: r) ! { dg-error "'inscan' REDUCTION clause on construct other than DO, SIMD, DO SIMD, PARALLEL DO, PARALLEL DO SIMD" }
+do i = 1, 64
+end do
+!$omp parallel master taskloop reduction(inscan, +: r) ! { dg-error "'inscan' REDUCTION clause on construct other than DO, SIMD, DO SIMD, PARALLEL DO, PARALLEL DO SIMD" }
+do i = 1, 64 ! { dg-error "OMP SCAN between two structured-block-sequences" "" { target *-*-* } .-1 }
+end do
+!$omp parallel master taskloop simd reduction(inscan, +: r) ! { dg-error "'inscan' REDUCTION clause on construct other than DO, SIMD, DO SIMD, PARALLEL DO, PARALLEL DO SIMD" }
+do i = 1, 64 ! { dg-error "OMP SCAN between two structured-block-sequences" "" { target *-*-* } .-1 }
+end do
+end
+
+
+subroutine bar()
+implicit none
+integer :: i, r
+r = 0
+!$omp parallel reduction(+:r)
+ !$omp master taskloop in_reduction(+:r)
+ do i = 1, 64
+ end do
+ !$omp master taskloop simd in_reduction(+:r)
+ do i = 1, 64
+ end do
+ !$omp master
+ !$omp taskloop in_reduction(+:r)
+ do i = 1, 64
+ end do
+ !$omp taskloop simd in_reduction(+:r)
+ do i = 1, 64
+ end do
+ !$omp end master
+!$omp end parallel
+
+!$omp parallel master taskloop in_reduction(+:r) ! { dg-error "Failed to match clause" }
+ do i = 1, 64
+ end do
+
+!$omp parallel master taskloop simd in_reduction(+:r) ! { dg-error "Failed to match clause" }
+ do i = 1, 64
+ end do
+end
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2021-06-01 13:18 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-06-01 9:53 [Patch] Fortran/OpenMP: Support (parallel) master taskloop (simd) [PR99928] Tobias Burnus
2021-06-01 10:02 ` Jakub Jelinek
2021-06-01 13:18 ` Tobias Burnus
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).