OpenMP: Allocate directive for static vars, clean up
For the 'allocate' directive, remove the sorry for static variables and
just keep using normal memory, but honor the requested alignment and set
a DECL_ATTRIBUTE in case a target may want to make use of this later on.
The documentation is updated accordingly.
The C diagnostic to check for predefined allocators in this case failed
to accept GCC's ompx_gnu_... allocator, now fixed. (Fortran was already
okay; but both now use new common #defined value for checking.)
And while Fortran common block variables are still rejected, the check
has been improved as before the sorry diagnostic did not work for
common blocks in modules.
Finally, for 'allocate' clause on the target/task/taskloop directives,
there is now a warning for omp_thread_mem_alloc (i.e. predefined allocator
with access = thread), which is undefined behavior according to the
OpenMP specification.
And, last, testing showed that var decl + static_assert sets TREE_USED
but does not produce a statement list in C, which did run into an assert
in gimplify. This special case is now also handled.
gcc/c/ChangeLog:
* c-parser.cc (c_parser_omp_allocate): Set alignment for alignof;
accept static variables and fix predef allocator check.
gcc/fortran/ChangeLog:
* openmp.cc (is_predefined_allocator): Use gomp-constants.h consts.
* trans-common.cc (translate_common): Reject OpenMP allocate directives.
* trans-decl.cc (gfc_finish_var_decl): Handle allocate directive
for static variables.
(gfc_trans_deferred_vars): Update for the latter.
gcc/ChangeLog:
* gimplify.cc (gimplify_bind_expr): Fix corner case for OpenMP
allocate directive.
(gimplify_scan_omp_clauses): Warn if omp_thread_mem_alloc is used
as allocator with the target/task/taskloop directive.
include/ChangeLog:
* gomp-constants.h (GOMP_OMP_PREDEF_ALLOC_MAX,
GOMP_OMPX_PREDEF_ALLOC_MIN, GOMP_OMPX_PREDEF_ALLOC_MAX,
GOMP_OMP_PREDEF_ALLOC_THREADS): New defines.
libgomp/ChangeLog:
* allocator.c: Add static asserts for news
GOMP_OMP{,X}_PREDEF_ALLOC_{MIN,MAX} range values.
* libgomp.texi (OpenMP Impl. Status): Allocate directive for
static vars is now supported. Refer to PR for allocate clause.
(Memory allocation): Update for static vars; minor word tweaking.
gcc/testsuite/ChangeLog:
* c-c++-common/gomp/allocate-9.c: Update for removed sorry.
* gfortran.dg/gomp/allocate-15.f90: Likewise.
* gfortran.dg/gomp/allocate-pinned-1.f90: Likewise.
* gfortran.dg/gomp/allocate-4.f90: Likewise; add dg-error for
previously missing diagnostic.
* c-c++-common/gomp/allocate-18.c: New test.
* c-c++-common/gomp/allocate-19.c: New test.
* gfortran.dg/gomp/allocate-clause.f90: New test.
* gfortran.dg/gomp/allocate-static-2.f90: New test.
* gfortran.dg/gomp/allocate-static.f90: New test.
gcc/c/c-parser.cc | 29 +++--
gcc/fortran/openmp.cc | 9 +-
gcc/fortran/trans-common.cc | 4 +
gcc/fortran/trans-decl.cc | 131 +++++++++++----------
gcc/gimplify.cc | 22 +++-
gcc/testsuite/c-c++-common/gomp/allocate-18.c | 59 ++++++++++
gcc/testsuite/c-c++-common/gomp/allocate-19.c | 69 +++++++++++
gcc/testsuite/c-c++-common/gomp/allocate-9.c | 43 +++----
gcc/testsuite/gfortran.dg/gomp/allocate-15.f90 | 2 +-
gcc/testsuite/gfortran.dg/gomp/allocate-4.f90 | 6 +-
gcc/testsuite/gfortran.dg/gomp/allocate-clause.f90 | 61 ++++++++++
.../gfortran.dg/gomp/allocate-pinned-1.f90 | 2 +-
.../gfortran.dg/gomp/allocate-static-2.f90 | 52 ++++++++
gcc/testsuite/gfortran.dg/gomp/allocate-static.f90 | 62 ++++++++++
include/gomp-constants.h | 8 ++
libgomp/allocator.c | 9 ++
libgomp/libgomp.texi | 15 +--
17 files changed, 469 insertions(+), 114 deletions(-)
@@ -20967,20 +20967,22 @@ c_parser_omp_allocate (c_parser *parser)
if (TREE_STATIC (var))
{
if (allocator == NULL_TREE && allocator_loc == UNKNOWN_LOCATION)
- error_at (loc, "%<allocator%> clause required for "
- "static variable %qD", var);
+ {
+ error_at (loc,
+ "%<allocator%> clause required for "
+ "static variable %qD", var);
+ continue;
+ }
else if (allocator
&& (wi::to_widest (allocator) < 1
- || wi::to_widest (allocator) > 8))
- /* 8 = largest predefined memory allocator. */
- error_at (allocator_loc,
- "%<allocator%> clause requires a predefined allocator as "
- "%qD is static", var);
- else
- sorry_at (OMP_CLAUSE_LOCATION (nl),
- "%<#pragma omp allocate%> for static variables like "
- "%qD not yet supported", var);
- continue;
+ || wi::to_widest (allocator) > GOMP_OMP_PREDEF_ALLOC_MAX)
+ && (wi::to_widest (allocator) < GOMP_OMPX_PREDEF_ALLOC_MIN
+ || wi::to_widest (allocator) > GOMP_OMPX_PREDEF_ALLOC_MAX))
+ {
+ error_at (allocator_loc,
+ "%<allocator%> clause requires a predefined allocator as "
+ "%qD is static", var);
+ }
}
if (allocator)
{
@@ -20988,6 +20990,9 @@ c_parser_omp_allocate (c_parser *parser)
= {EXPR_LOC_OR_LOC (allocator, OMP_CLAUSE_LOCATION (nl)), var};
walk_tree (&allocator, c_check_omp_allocate_allocator_r, &data, NULL);
}
+ if (alignment)
+ SET_DECL_ALIGN (var, BITS_PER_UNIT * MAX (tree_to_uhwi (alignment),
+ DECL_ALIGN_UNIT (var)));
DECL_ATTRIBUTES (var) = tree_cons (get_identifier ("omp allocate"),
build_tree_list (allocator, alignment),
DECL_ATTRIBUTES (var));
@@ -7857,9 +7857,12 @@ is_predefined_allocator (gfc_expr *expr)
&& expr->ts.kind == gfc_c_intptr_kind
&& expr->expr_type == EXPR_CONSTANT
&& ((mpz_sgn (expr->value.integer) > 0
- && mpz_cmp_si (expr->value.integer, 8) <= 0)
- || (mpz_cmp_si (expr->value.integer, 200) >= 0
- && mpz_cmp_si (expr->value.integer, 200) <= 0)));
+ && mpz_cmp_si (expr->value.integer,
+ GOMP_OMP_PREDEF_ALLOC_MAX) <= 0)
+ || (mpz_cmp_si (expr->value.integer,
+ GOMP_OMPX_PREDEF_ALLOC_MIN) >= 0
+ && mpz_cmp_si (expr->value.integer,
+ GOMP_OMPX_PREDEF_ALLOC_MAX) <= 0)));
}
/* Resolve declarative ALLOCATE statement. Note: Common block vars only appear
@@ -1219,6 +1219,10 @@ translate_common (gfc_common_head *common, gfc_symbol *var_list)
align = 1;
saw_equiv = false;
+ if (var_list->attr.omp_allocate)
+ gfc_error ("Sorry, !$OMP allocate for COMMON block variable %qs at %L "
+ "not supported", common->name, &common->where);
+
/* Add symbols to the segment. */
for (sym = var_list; sym; sym = sym->common_next)
{
@@ -821,6 +821,23 @@ gfc_finish_var_decl (tree decl, gfc_symbol * sym)
&& (TREE_STATIC (decl) || DECL_EXTERNAL (decl)))
set_decl_tls_model (decl, decl_default_tls_model (decl));
+ if (sym->attr.omp_allocate && TREE_STATIC (decl))
+ {
+ struct gfc_omp_namelist *n;
+ for (n = sym->ns->omp_allocate; n; n = n->next)
+ if (n->sym == sym)
+ break;
+ tree alloc = gfc_conv_constant_to_tree (n->u2.allocator);
+ tree align = (n->u.align ? gfc_conv_constant_to_tree (n->u.align)
+ : NULL_TREE);
+ if (align != NULL_TREE)
+ SET_DECL_ALIGN (decl, MAX (tree_to_uhwi (align),
+ DECL_ALIGN_UNIT (decl)) * BITS_PER_UNIT);
+ DECL_ATTRIBUTES (decl)
+ = tree_cons (get_identifier ("omp allocate"),
+ build_tree_list (alloc, align), DECL_ATTRIBUTES (decl));
+ }
+
/* Mark weak variables. */
if (sym->attr.ext_attr & (1 << EXT_ATTR_WEAK))
declare_weak (decl);
@@ -5251,71 +5268,55 @@ gfc_trans_deferred_vars (gfc_symbol * proc_sym, gfc_wrapped_block * block)
se.expr = NULL_TREE;
for (struct gfc_omp_namelist *n = omp_ns ? omp_ns->omp_allocate : NULL;
n; n = n->next)
- if (!TREE_STATIC (n->sym->backend_decl))
- {
- tree align = (n->u.align ? gfc_conv_constant_to_tree (n->u.align)
- : NULL_TREE);
- if (last_allocator != n->u2.allocator)
- {
- location_t loc = input_location;
- gfc_init_se (&se, NULL);
- if (n->u2.allocator)
- {
- input_location = gfc_get_location (&n->u2.allocator->where);
- gfc_conv_expr (&se, n->u2.allocator);
- }
- /* We need to evalulate non-constants - also to find the location
- after which the GOMP_alloc has to be added to - also as BLOCK
- does not yield a new BIND_EXPR_BODY. */
- if (n->u2.allocator
- && (!(CONSTANT_CLASS_P (se.expr) && DECL_P (se.expr))
- || se.pre.head || se.post.head))
- {
- stmtblock_t tmpblock;
- gfc_init_block (&tmpblock);
- se.expr = gfc_evaluate_now (se.expr, &tmpblock);
- /* First post then pre because the new code is inserted
- at the top. */
- gfc_add_init_cleanup (block, gfc_finish_block (&se.post), NULL);
- gfc_add_init_cleanup (block, gfc_finish_block (&tmpblock),
- NULL);
- gfc_add_init_cleanup (block, gfc_finish_block (&se.pre), NULL);
- }
- last_allocator = n->u2.allocator;
- input_location = loc;
- }
-
- /* 'omp allocate( {purpose: allocator, value: align},
- {purpose: init-stmtlist, value: cleanup-stmtlist},
- {purpose: size-var, value: last-size-expr}}
- where init-stmt/cleanup-stmt is the STATEMENT list to find the
- try-final block; last-size-expr is to find the location after
- which to add the code and 'size-var' is for the proper size, cf.
- gfc_trans_auto_array_allocation - either or both of the latter
- can be NULL. */
- tree tmp = lookup_attribute ("omp allocate",
- DECL_ATTRIBUTES (n->sym->backend_decl));
- tmp = TREE_VALUE (tmp);
- TREE_PURPOSE (tmp) = se.expr;
- TREE_VALUE (tmp) = align;
- TREE_PURPOSE (TREE_CHAIN (tmp)) = init_stmtlist;
- TREE_VALUE (TREE_CHAIN (tmp)) = cleanup_stmtlist;
- }
- else if (n->sym->attr.in_common)
- {
- gfc_error ("Sorry, !$OMP allocate for COMMON block variable %qs at %L "
- "not supported", n->sym->common_block->name,
- &n->sym->common_block->where);
- break;
- }
- else
- {
- gfc_error ("Sorry, !$OMP allocate for variable %qs at %L with SAVE "
- "attribute not yet implemented", n->sym->name,
- &n->sym->declared_at);
- /* FIXME: Remember to handle last_allocator. */
- break;
- }
+ {
+ tree align = (n->u.align ? gfc_conv_constant_to_tree (n->u.align) : NULL_TREE);
+ if (last_allocator != n->u2.allocator)
+ {
+ location_t loc = input_location;
+ gfc_init_se (&se, NULL);
+ if (n->u2.allocator)
+ {
+ input_location = gfc_get_location (&n->u2.allocator->where);
+ gfc_conv_expr (&se, n->u2.allocator);
+ }
+ /* We need to evalulate non-constants - also to find the location
+ after which the GOMP_alloc has to be added to - also as BLOCK
+ does not yield a new BIND_EXPR_BODY. */
+ if (n->u2.allocator
+ && (!(CONSTANT_CLASS_P (se.expr) && DECL_P (se.expr))
+ || se.pre.head || se.post.head))
+ {
+ stmtblock_t tmpblock;
+ gfc_init_block (&tmpblock);
+ se.expr = gfc_evaluate_now (se.expr, &tmpblock);
+ /* First post then pre because the new code is inserted
+ at the top. */
+ gfc_add_init_cleanup (block, gfc_finish_block (&se.post), NULL);
+ gfc_add_init_cleanup (block, gfc_finish_block (&tmpblock),
+ NULL);
+ gfc_add_init_cleanup (block, gfc_finish_block (&se.pre), NULL);
+ }
+ last_allocator = n->u2.allocator;
+ input_location = loc;
+ }
+ if (TREE_STATIC (n->sym->backend_decl))
+ continue;
+ /* 'omp allocate( {purpose: allocator, value: align},
+ {purpose: init-stmtlist, value: cleanup-stmtlist},
+ {purpose: size-var, value: last-size-expr}}
+ where init-stmt/cleanup-stmt is the STATEMENT list to find the
+ try-final block; last-size-expr is to find the location after
+ which to add the code and 'size-var' is for the proper size, cf.
+ gfc_trans_auto_array_allocation - either or both of the latter
+ can be NULL. */
+ tree tmp = lookup_attribute ("omp allocate",
+ DECL_ATTRIBUTES (n->sym->backend_decl));
+ tmp = TREE_VALUE (tmp);
+ TREE_PURPOSE (tmp) = se.expr;
+ TREE_VALUE (tmp) = align;
+ TREE_PURPOSE (TREE_CHAIN (tmp)) = init_stmtlist;
+ TREE_VALUE (TREE_CHAIN (tmp)) = cleanup_stmtlist;
+ }
gfc_init_block (&tmpblock);
@@ -1396,6 +1396,7 @@ gimplify_bind_expr (tree *expr_p, gimple_seq *pre_p)
if (flag_openmp
&& !is_global_var (t)
+ && !TREE_STATIC (t)
&& DECL_CONTEXT (t) == current_function_decl
&& TREE_USED (t)
&& (attr = lookup_attribute ("omp allocate", DECL_ATTRIBUTES (t)))
@@ -1427,11 +1428,17 @@ gimplify_bind_expr (tree *expr_p, gimple_seq *pre_p)
"%<allocate%> directive for %qD inside a target "
"region must specify an %<allocator%> clause", t);
/* Skip for omp_default_mem_alloc (= 1),
- unless align is present. */
+ unless align is present. For C/C++, there should be always a
+ statement list following if TREE_USED, except for, e.g., using
+ this decl in a static_assert; in that case, only a single
+ DECL_EXPR remains, which can be skipped here. */
else if (!errorcount
&& (align != NULL_TREE
|| alloc == NULL_TREE
- || !integer_onep (alloc)))
+ || !integer_onep (alloc))
+ && (lang_GNU_Fortran ()
+ || (TREE_CODE (BIND_EXPR_BODY (bind_expr))
+ != DECL_EXPR)))
{
/* Fortran might already use a pointer type internally;
use that pointer except for type(C_ptr) and type(C_funptr);
@@ -13329,6 +13336,17 @@ gimplify_scan_omp_clauses (tree *list_p, gimple_seq *pre_p,
break;
case OMP_CLAUSE_ALLOCATE:
+ decl = OMP_CLAUSE_ALLOCATE_ALLOCATOR (c);
+ if (decl
+ && TREE_CODE (decl) == INTEGER_CST
+ && wi::eq_p (wi::to_widest (decl), GOMP_OMP_PREDEF_ALLOC_THREADS)
+ && (code == OMP_TARGET || code == OMP_TASK || code == OMP_TASKLOOP))
+ warning_at (OMP_CLAUSE_LOCATION (c), OPT_Wopenmp,
+ "allocator with access trait set to %<thread%> "
+ "results in undfined behavior for %qs directive",
+ code == OMP_TARGET ? "target"
+ : (code == OMP_TASK
+ ? "task" : "taskloop"));
decl = OMP_CLAUSE_DECL (c);
if (error_operand_p (decl))
{
new file mode 100644
@@ -0,0 +1,59 @@
+typedef enum omp_allocator_handle_t
+#if __cplusplus >= 201103L
+: __UINTPTR_TYPE__
+#endif
+{
+ omp_null_allocator = 0,
+ omp_default_mem_alloc = 1,
+ omp_large_cap_mem_alloc = 2,
+ omp_const_mem_alloc = 3,
+ omp_high_bw_mem_alloc = 4,
+ omp_low_lat_mem_alloc = 5,
+ omp_cgroup_mem_alloc = 6,
+ omp_pteam_mem_alloc = 7,
+ omp_thread_mem_alloc = 8,
+ __omp_allocator_handle_t_max__ = __UINTPTR_MAX__
+} omp_allocator_handle_t;
+
+void test0 ()
+{
+ int A1[5];
+ #pragma omp allocate(A1) align(128) allocator(omp_default_mem_alloc)
+ /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+
+ #ifndef __cplusplus
+ _Static_assert (_Alignof(A1) == 128, "wrong alignment");
+ #elif __cplusplus >= 201103L
+ static_assert (alignof(A1) == 128, "wrong alignment"); /* { dg-bogus "static assertion failed: wrong alignment" "" { xfail c++ } } */
+ #endif
+}
+
+void
+test1 ()
+{
+ int x[5];
+ #pragma omp parallel allocate(omp_thread_mem_alloc: x) firstprivate(x)
+ x[0] = 1;
+
+ #pragma omp target allocate(omp_thread_mem_alloc: x) firstprivate(x) /* uses_allocators(omp_thread_mem_alloc) */
+ /* { dg-warning "allocator with access trait set to 'thread' results in undfined behavior for 'target' directive \\\[-Wopenmp\\\]" "" { target *-*-* } .-1 } */
+ x[0] = 1;
+
+ #pragma omp taskloop allocate(omp_thread_mem_alloc: x) firstprivate(x)
+ /* { dg-warning "allocator with access trait set to 'thread' results in undfined behavior for 'taskloop' directive \\\[-Wopenmp\\\]" "" { target *-*-* } .-1 } */
+ for (int i = 0; i < 5; i++)
+ x[i] = i;
+
+ #pragma omp parallel master taskloop simd allocate(omp_thread_mem_alloc: x) firstprivate(x)
+ /* { dg-warning "allocator with access trait set to 'thread' results in undfined behavior for 'taskloop' directive \\\[-Wopenmp\\\]" "" { target *-*-* } .-1 } */
+ for (int i = 0; i < 5; i++)
+ x[i] = i;
+
+ #pragma omp parallel
+ #pragma omp masked
+ {
+ #pragma omp task allocate(omp_thread_mem_alloc: x) firstprivate(x)
+ /* { dg-warning "allocator with access trait set to 'thread' results in undfined behavior for 'task' directive \\\[-Wopenmp\\\]" "" { target *-*-* } .-1 } */
+ x[0] = 1;
+ }
+}
new file mode 100644
@@ -0,0 +1,69 @@
+typedef enum omp_allocator_handle_t
+#if __cplusplus >= 201103L
+: __UINTPTR_TYPE__
+#endif
+{
+ omp_null_allocator = 0,
+ omp_default_mem_alloc = 1,
+ omp_large_cap_mem_alloc = 2,
+ omp_const_mem_alloc = 3,
+ omp_high_bw_mem_alloc = 4,
+ omp_low_lat_mem_alloc = 5,
+ omp_cgroup_mem_alloc = 6,
+ omp_pteam_mem_alloc = 7,
+ omp_thread_mem_alloc = 8,
+ ompx_gnu_pinned_bogus_1 = 9,
+ ompx_gnu_pinned_bogus_2 = 199,
+ ompx_gnu_pinned_mem_alloc = 200,
+ ompx_gnu_pinned_bogus_3 = 2001,
+ __omp_allocator_handle_t_max__ = __UINTPTR_MAX__
+} omp_allocator_handle_t;
+
+static int A1[5] = {1,2,3,4,5};
+#pragma omp allocate(A1) align(128) allocator(omp_default_mem_alloc)
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+
+#ifndef __cplusplus
+_Static_assert (_Alignof(A1) == 128, "wrong alignment");
+#elif __cplusplus >= 201103L
+static_assert (alignof(A1) == 128, "wrong alignment"); /* { dg-bogus "static assertion failed: wrong alignment" "" { xfail c++ } } */
+#endif
+
+
+static int *ptr;
+#pragma omp allocate(ptr) align(2) allocator(omp_default_mem_alloc)
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+
+#ifndef __cplusplus
+_Static_assert (_Alignof(ptr) == _Alignof(int*), "wrong alignment");
+#elif __cplusplus >= 201103L
+static_assert (alignof(ptr) == alignof(int*), "wrong alignment");
+#endif
+
+
+int *
+get ()
+{
+ static int q = 0;
+ #pragma omp allocate(q) align(1024) allocator(omp_default_mem_alloc)
+ /* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+
+#ifndef __cplusplus
+ _Static_assert (_Alignof(q) == 1024, "wrong alignment");
+#elif __cplusplus >= 201103L
+ static_assert (alignof(q) == 1024, "wrong alignment"); /* { dg-bogus "static assertion failed: wrong alignment" "" { xfail c++ } } */
+#endif
+
+ q += 1;
+ return &A1[q];
+}
+
+static int invalid1, okay1, invalid2, invalid3;
+#pragma omp allocate(invalid1) align(128) allocator(ompx_gnu_pinned_bogus_1) /* { dg-error "'allocator' clause requires a predefined allocator as 'invalid1' is static" "" { xfail c++ } } */
+#pragma omp allocate(okay1) align(128) allocator(ompx_gnu_pinned_mem_alloc) /* Okay */
+#pragma omp allocate(invalid2) align(128) allocator(ompx_gnu_pinned_bogus_2) /* { dg-error "'allocator' clause requires a predefined allocator as 'invalid2' is static" "" { xfail c++ } } */
+#pragma omp allocate(invalid3) align(128) allocator(ompx_gnu_pinned_bogus_3) /* { dg-error "'allocator' clause requires a predefined allocator as 'invalid3' is static" "" { xfail c++ } } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-4 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-4 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-4 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-4 } */
@@ -18,7 +18,11 @@ typedef enum omp_allocator_handle_t
static int A[5] = {1,2,3,4,5};
-int B, C, D;
+static int A2[5] = {1,2,3,4,5};
+static int A3[5] = {1,2,3,4,5};
+static int A4[5] = {1,2,3,4,5};
+static int A5[5] = {1,2,3,4,5};
+int B, C, C2, D;
/* If the following fails because of added predefined allocators, please update
- c/c-parser.c's c_parser_omp_allocate
@@ -30,46 +34,45 @@ int B, C, D;
#pragma omp allocate(A) align(32) allocator((omp_allocator_handle_t) 9) /* { dg-error "'allocator' clause requires a predefined allocator as 'A' is static" "" { xfail c++ } } */
/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
-
// typo in allocator name:
-#pragma omp allocate(A) allocator(omp_low_latency_mem_alloc)
+#pragma omp allocate(A2) allocator(omp_low_latency_mem_alloc)
/* { dg-error "'omp_low_latency_mem_alloc' undeclared here \\(not in a function\\); did you mean 'omp_low_lat_mem_alloc'\\?" "" { target c } .-1 } */
/* { dg-error "'omp_low_latency_mem_alloc' was not declared in this scope; did you mean 'omp_low_lat_mem_alloc'\\?" "" { target c++ } .-2 } */
-/* { dg-error "'allocator' clause required for static variable 'A'" "" { target c } .-3 } */
+/* { dg-error "'allocator' clause required for static variable 'A2'" "" { target c } .-3 } */
/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-4 } */
/* align be const multiple of 2 */
-#pragma omp allocate(A) align(31) allocator(omp_default_mem_alloc) /* { dg-error "'align' clause argument needs to be positive constant power of two integer expression" } */
-/* { dg-message "sorry, unimplemented: '#pragma omp allocate' for static variables like 'A' not yet supported" "" { target c } .-1 } */
-/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+#pragma omp allocate(A3) align(31) allocator(omp_default_mem_alloc) /* { dg-error "'align' clause argument needs to be positive constant power of two integer expression" } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+
/* allocator missing (required as A is static) */
-#pragma omp allocate(A) align(32) /* { dg-error "'allocator' clause required for static variable 'A'" "" { xfail c++ } } */
+#pragma omp allocate(A4) align(32) /* { dg-error "'allocator' clause required for static variable 'A4'" "" { xfail c++ } } */
/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
/* "expression in the clause must be a constant expression that evaluates to one of the
predefined memory allocator values -> omp_low_lat_mem_alloc" */
#pragma omp allocate(B) allocator((omp_allocator_handle_t) (omp_high_bw_mem_alloc+1)) align(32) /* OK: omp_low_lat_mem_alloc */
-/* { dg-message "sorry, unimplemented: '#pragma omp allocate' for static variables like 'B' not yet supported" "" { target c } .-1 } */
-/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+
#pragma omp allocate(C) allocator((omp_allocator_handle_t) 2) /* OK: omp_large_cap_mem_alloc */
-/* { dg-message "sorry, unimplemented: '#pragma omp allocate' for static variables like 'C' not yet supported" "" { target c } .-1 } */
-/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
-#pragma omp allocate(A) align(32) allocator(omp_null_allocator) /* { dg-error "'allocator' clause requires a predefined allocator as 'A' is static" "" { xfail c++ } } */
+
+#pragma omp allocate(A5) align(32) allocator(omp_null_allocator) /* { dg-error "'allocator' clause requires a predefined allocator as 'A5' is static" "" { xfail c++ } } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+
+#pragma omp allocate(C2) align(32) allocator(omp_large_cap_mem_alloc)
/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
-#pragma omp allocate(C) align(32) allocator(omp_large_cap_mem_alloc) /* { dg-error "'C' already appeared as list item in an 'allocate' directive" "" { xfail *-*-* } } */
-/* { dg-message "sorry, unimplemented: '#pragma omp allocate' for static variables like 'C' not yet supported" "" { target c } .-1 } */
-/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
// allocate directive in same TU
int f()
{
#pragma omp allocate(D) align(32) allocator(omp_large_cap_mem_alloc) /* { dg-error "'allocate' directive must be in the same scope as 'D'" "" { xfail c++ } } */
-/* { dg-note "declared here" "" { target c } 21 } */
-/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+/* { dg-note "declared here" "" { target c } 25 } */
return A[0];
}
@@ -83,8 +86,8 @@ int g()
{
int c2=3;
#pragma omp allocate(c2, b2) /* { dg-error "'allocate' directive must be in the same scope as 'b2'" "" { xfail c++ } } */
-/* { dg-note "declared here" "" { target c } .-8 } */
-/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-2 } */
+/* { dg-message "sorry, unimplemented: '#pragma omp allocate' not yet supported" "" { target c++ } .-1 } */
+/* { dg-note "declared here" "" { target c } .-9 } */
return c2+a2+b2;
}
}
@@ -30,7 +30,7 @@ end
integer function allocators() result(res)
use m
- integer, save :: a(5) = [1,2,3,4,5] ! { dg-error "Sorry, !.OMP allocate for variable 'a' at .1. with SAVE attribute not yet implemented" }
+ integer, save :: a(5) = [1,2,3,4,5]
!$omp allocate(a) allocator(omp_high_bw_mem_alloc)
res = a(4)
end
@@ -39,7 +39,7 @@ integer :: a,b,c(n),d(5),e(2)
!$omp allocate( e ) allocator( omp_high_bw_mem_alloc )
!saved vars
-integer, save :: k,l,m(5),r(2) ! { dg-error "Sorry, !.OMP allocate for variable 'k' at .1. with SAVE attribute not yet implemented" }
+integer, save :: k,l,m(5),r(2)
!$omp allocate(k) align(16) , allocator (omp_large_cap_mem_alloc)
!$omp allocate ( l ) allocator (omp_large_cap_mem_alloc) , align ( 32)
!$omp allocate (m) align( 128 ),allocator( omp_high_bw_mem_alloc )
@@ -47,8 +47,8 @@ integer, save :: k,l,m(5),r(2) ! { dg-error "Sorry, !.OMP allocate for variable
!common /block/
integer :: q,x,y(2),z(5)
-common /com1/ q,x
-common /com2/ y,z
+common /com1/ q,x ! { dg-error "Sorry, !.OMP allocate for COMMON block variable 'com1' at .1. not supported" }
+common /com2/ y,z ! { dg-error "Sorry, !.OMP allocate for COMMON block variable 'com2' at .1. not supported" }
!$omp allocate ( / com1/) align( 128 ) allocator( omp_high_bw_mem_alloc )
!$omp allocate(/com2 / ) allocator( omp_high_bw_mem_alloc )
end
new file mode 100644
@@ -0,0 +1,61 @@
+module omp_lib_kinds
+ use iso_c_binding, only: c_int, c_intptr_t
+ implicit none
+ private :: c_int, c_intptr_t
+ integer, parameter :: omp_allocator_handle_kind = c_intptr_t
+
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_null_allocator = 0
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_default_mem_alloc = 1
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_large_cap_mem_alloc = 2
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_const_mem_alloc = 3
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_high_bw_mem_alloc = 4
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_low_lat_mem_alloc = 5
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_cgroup_mem_alloc = 6
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_pteam_mem_alloc = 7
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_thread_mem_alloc = 8
+end module
+
+subroutine test1 ()
+ use omp_lib_kinds
+ implicit none
+ integer :: x(5), i
+
+ !$omp parallel allocate(omp_thread_mem_alloc: x) firstprivate(x)
+ x(1) = 1
+ !$omp end parallel
+
+ !$omp target allocate(omp_thread_mem_alloc: x) firstprivate(x) ! uses_allocators(omp_thread_mem_alloc)
+ ! { dg-warning "allocator with access trait set to 'thread' results in undfined behavior for 'target' directive \\\[-Wopenmp\\\]" "" { target *-*-* } .-1 }
+ x(1) = 1
+ !$omp end target
+
+ !$omp taskloop allocate(omp_thread_mem_alloc: x) firstprivate(x)
+ ! { dg-warning "allocator with access trait set to 'thread' results in undfined behavior for 'taskloop' directive \\\[-Wopenmp\\\]" "" { target *-*-* } .-1 }
+ do i = 1, 5
+ x(i) = i
+ end do
+
+ !$omp parallel master taskloop simd allocate(omp_thread_mem_alloc: x) firstprivate(x)
+ ! { dg-warning "allocator with access trait set to 'thread' results in undfined behavior for 'taskloop' directive \\\[-Wopenmp\\\]" "" { target *-*-* } .-1 }
+ do i = 1, 5
+ x(i) = i
+ end do
+
+ !$omp parallel
+ !$omp masked
+ !$omp task allocate(omp_thread_mem_alloc: x) firstprivate(x)
+ ! { dg-warning "allocator with access trait set to 'thread' results in undfined behavior for 'task' directive \\\[-Wopenmp\\\]" "" { target *-*-* } .-1 }
+ x(1) = 1
+ !$omp end task
+ !$omp end masked
+ !$omp end parallel
+end
@@ -11,6 +11,6 @@ subroutine f ()
use m
implicit none
! The "Sorry" is here temporarily only to avoid excess error failures.
- integer, save :: i ! { dg-error "Sorry, !.OMP allocate for variable 'i' at .1. with SAVE attribute not yet implemented" }
+ integer, save :: i
!$omp allocate(i) allocator(ompx_gnu_pinned_mem_alloc)
end
new file mode 100644
@@ -0,0 +1,52 @@
+module omp_lib_kinds
+ use iso_c_binding, only: c_int, c_intptr_t
+ implicit none
+ private :: c_int, c_intptr_t
+ integer, parameter :: omp_allocator_handle_kind = c_intptr_t
+
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_null_allocator = 0
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_default_mem_alloc = 1
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_large_cap_mem_alloc = 2
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_const_mem_alloc = 3
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_high_bw_mem_alloc = 4
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_low_lat_mem_alloc = 5
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_cgroup_mem_alloc = 6
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_pteam_mem_alloc = 7
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_thread_mem_alloc = 8
+end module
+
+module m
+use omp_lib_kinds, only: omp_default_mem_alloc
+implicit none
+integer a,b
+common /foo/ a,b ! { dg-error "Sorry, !.OMP allocate for COMMON block variable 'foo' at .1. not supported" }
+!$omp allocate(/foo/) align(128) allocator(omp_default_mem_alloc)
+end
+
+subroutine sub
+use omp_lib_kinds
+implicit none
+integer a,b
+common /foo/ a,b ! { dg-error "Sorry, !.OMP allocate for COMMON block variable 'foo' at .1. not supported" }
+!$omp allocate(/foo/) align(128) allocator(omp_default_mem_alloc)
+end
+
+subroutine outer
+contains
+subroutine inner
+use omp_lib_kinds
+implicit none
+integer a,b
+common /foo/ a,b ! { dg-error "Sorry, !.OMP allocate for COMMON block variable 'foo' at .1. not supported" }
+!$omp allocate(/foo/) align(128) allocator(omp_default_mem_alloc)
+end
+end
new file mode 100644
@@ -0,0 +1,62 @@
+! { dg-do run }
+
+module omp_lib_kinds
+ use iso_c_binding, only: c_int, c_intptr_t
+ implicit none
+ private :: c_int, c_intptr_t
+ integer, parameter :: omp_allocator_handle_kind = c_intptr_t
+
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_null_allocator = 0
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_default_mem_alloc = 1
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_large_cap_mem_alloc = 2
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_const_mem_alloc = 3
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_high_bw_mem_alloc = 4
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_low_lat_mem_alloc = 5
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_cgroup_mem_alloc = 6
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_pteam_mem_alloc = 7
+ integer (kind=omp_allocator_handle_kind), &
+ parameter :: omp_thread_mem_alloc = 8
+end module
+
+module m
+ use iso_c_binding, only: c_intptr_t
+ use omp_lib_kinds, only: omp_default_mem_alloc
+ implicit none (type, external)
+
+ integer(c_intptr_t) :: intptr
+
+ integer :: A(4) = [1,2,3,4]
+ !$omp allocate(A) align(128) allocator(omp_default_mem_alloc)
+contains
+ subroutine f()
+ integer :: B(4) = [1,2,3,4]
+ !$omp allocate(B) align(256) allocator(omp_default_mem_alloc)
+
+ if (mod (transfer (loc (A), intptr), 128_c_intptr_t) /= 0) stop 1
+ if (mod (transfer (loc (B), intptr), 256_c_intptr_t) /= 0) stop 2
+
+ call inner()
+ contains
+ subroutine inner()
+ integer :: C(4) = [1,2,3,4]
+ !$omp allocate(C) align(1024) allocator(omp_default_mem_alloc)
+ if (mod (transfer (loc (A), intptr), 128_c_intptr_t) /= 0) stop 3
+ if (mod (transfer (loc (B), intptr), 256_c_intptr_t) /= 0) stop 4
+ if (mod (transfer (loc (C), intptr), 1024_c_intptr_t) /= 0) stop 5
+ end
+ end
+end
+
+use m
+implicit none (type, external)
+if (mod (transfer (loc (A), intptr), 128_c_intptr_t) /= 0) stop 6
+call f()
+end
@@ -382,6 +382,14 @@ enum gomp_map_kind
#define GOMP_DEPEND_MUTEXINOUTSET 4
#define GOMP_DEPEND_INOUTSET 5
+/* Predefined allocator value ranges. */
+#define GOMP_OMP_PREDEF_ALLOC_MAX 8
+#define GOMP_OMPX_PREDEF_ALLOC_MIN 200
+#define GOMP_OMPX_PREDEF_ALLOC_MAX 200
+
+/* Predefined allocator with access == thread. */
+#define GOMP_OMP_PREDEF_ALLOC_THREADS 8
+
/* Flag values for OpenMP 'requires' directive features. */
// compiler use only: OMP_REQUIRES_ATOMIC_DEFAULT_MEM_ORDER 0xf
#define GOMP_REQUIRES_UNIFIED_ADDRESS 0x10
@@ -102,6 +102,15 @@ GOMP_is_alloc (void *ptr)
#define ompx_gnu_min_predefined_alloc ompx_gnu_pinned_mem_alloc
#define ompx_gnu_max_predefined_alloc ompx_gnu_pinned_mem_alloc
+_Static_assert (GOMP_OMP_PREDEF_ALLOC_MAX == omp_thread_mem_alloc,
+ "GOMP_OMP_PREDEF_ALLOC_MAX == omp_thread_mem_alloc");
+_Static_assert (GOMP_OMPX_PREDEF_ALLOC_MIN == ompx_gnu_min_predefined_alloc,
+ "GOMP_OMP_PREDEF_ALLOC_MAX == omp_thread_mem_alloc");
+_Static_assert (GOMP_OMPX_PREDEF_ALLOC_MAX == ompx_gnu_max_predefined_alloc,
+ "GOMP_OMP_PREDEF_ALLOC_MAX == omp_thread_mem_alloc");
+_Static_assert (GOMP_OMP_PREDEF_ALLOC_THREADS == omp_thread_mem_alloc,
+ "GOMP_OMP_PREDEF_ALLOC_THREADS == omp_thread_mem_alloc");
+
/* These macros may be overridden in config/<target>/allocator.c.
The defaults (no override) are to return NULL for pinned memory requests
and pass through to the regular OS calls otherwise.
@@ -233,9 +233,9 @@ The OpenMP 4.5 specification is fully supported.
@tab Y @tab See also @ref{Memory allocation}
@item Memory management routines @tab Y @tab
@item @code{allocate} directive @tab P
- @tab Only C for stack/automatic and Fortran for stack/automatic
- and allocatable/pointer variables
-@item @code{allocate} clause @tab P @tab Initial support
+ @tab C++ unsupported; see also @ref{Memory allocation}
+@item @code{allocate} clause @tab P @tab Clause has no effect on @code{target}
+ (@uref{https://gcc.gnu.org/PR113436,PR113436})
@item @code{use_device_addr} clause on @code{target data} @tab Y @tab
@item @code{ancestor} modifier on @code{device} clause @tab Y @tab
@item Implicit declare target directive @tab Y @tab
@@ -306,7 +306,7 @@ The OpenMP 4.5 specification is fully supported.
@item @code{strict} modifier in the @code{grainsize} and @code{num_tasks}
clauses of the @code{taskloop} construct @tab Y @tab
@item @code{align} clause in @code{allocate} directive @tab P
- @tab Only C and Fortran (and not for static variables)
+ @tab Only C and Fortran
@item @code{align} modifier in @code{allocate} clause @tab Y @tab
@item @code{thread_limit} clause to @code{target} construct @tab Y @tab
@item @code{has_device_addr} clause to @code{target} construct @tab Y @tab
@@ -6414,14 +6414,14 @@ The description below applies to:
constant expression with value @code{omp_default_mem_alloc} and no
@code{align} modifier has been specified. (In that case, the normal
@code{malloc} allocation is used.)
+@item The @code{allocate} directive for variables in static memory; while
+ the alignment is honored, the normal static memory is used.
@item Using the @code{allocate} directive for automatic/stack variables, except
when the @code{allocator} clause is a constant expression with value
@code{omp_default_mem_alloc} and no @code{align} clause has been
specified. (In that case, the normal allocation is used: stack allocation
and, sometimes for Fortran, also @code{malloc} [depending on flags such as
@option{-fstack-arrays}].)
-@item Using the @code{allocate} directive for variable in static memory is
- currently not supported (compile time error).
@item In Fortran, the @code{allocators} directive and the executable
@code{allocate} directive for Fortran pointers and allocatables is
supported, but requires that files containing those directives has to be
@@ -6433,7 +6433,8 @@ The description below applies to:
For the available predefined allocators and, as applicable, their associated
predefined memory spaces and for the available traits and their default values,
see @ref{OMP_ALLOCATOR}. Predefined allocators without an associated memory
-space use the @code{omp_default_mem_space} memory space.
+space use the @code{omp_default_mem_space} memory space. See additionally
+@ref{Offload-Target Specifics}.
For the memory spaces, the following applies:
@itemize
'omp allocate' permits to use a different (specified) allocator and alignment for both stack/automatic and static/saved variables; the latter takes only predefined allocators. Currently, only C and Fortran are support for stack/automatic variables; static variables are rejected before the attached patch. (For them, only predefined allocators are permitted.) * * * I happened to look at the 'allocate' directive recently and, doing so, I stumbled over a couple of issues, which the attached patch addresses (missing diagnostics for corner cases, not updated checks, unhelpful documentation ['allocate' *clause*], ...). Doing so, I wondered whether: Shouldn't we just accept 'omp allocate' for static variables by just honoring the aligning and ignoring the actually requested allocator? - First, we do already the same for actual allocations as not all traits are supported. And for the host this seems to be the most sensible to do in any case. [For some use cases, pointers + allocation in the constructor would be better, but in general, not adding an indirection seems to be better and has fewer corner-case usability issue.] I guess we later want to honor the requested memory for nvptx and/or gcn; at least Nvidia GPUs could make use for constant memory (having advantages for reading the same memory by many threads/broadcasting it). I guess OpenACC 2.7's 'readonly' modifier serves a similar purpose. For now we don't, but the attribute is passed on to the backends, which could make use of them, if desired. ('groupprivate' directive vs. cgroup/thread allocators are similar device-only features.) As mentioned, this patch also fixes a few other issues here and there, see commit log and source code for details. Code comments? Suggestions or remarks? - Before I apply this patch? Tobias PS: I am aware that C++ support is lacking. There is a pending patch that needs to be updated for this patch, probably some bitrotting, and in particular for the review comments, cf. https://gcc.gnu.org/pipermail/gcc-patches/2023-October/633782.html and https://gcc.gnu.org/pipermail/gcc-patches/2023-December/639929.html