re PR c++/26943 ([gomp] firstprivate + lastprivate uses inefficient barrier)

PR c++/26943
	* omp-low.c (maybe_lookup_decl_in_outer_ctx): New function.
	(build_outer_var_ref): Use maybe_lookup_decl_in_outer_ctx
	to find if var will be a global variable even in the nested context.
	(omp_copy_decl): Only check for global variable at the end, it might
	be overridden in outer contexts.
	(scan_sharing_clauses): For global variables don't create a field.
	(lower_rec_input_clauses): Do nothing for global shared variables.
	Emit a barrier at the end of ILIST if there were any decls in both
	firstprivate and lastprivate clauses.
	(lower_send_clauses): Do nothing for global variables except for
	COPYIN.

	* testsuite/libgomp.c/pr26943-1.c: New test.
	* testsuite/libgomp.c/pr26943-2.c: New test.
	* testsuite/libgomp.c/pr26943-3.c: New test.
	* testsuite/libgomp.c/pr26943-4.c: New test.
	* testsuite/libgomp.c++/pr27337.C: Remove barrier.
	* testsuite/libgomp.c++/pr26943.C: New test.

From-SVN: r113483
This commit is contained in:
Jakub Jelinek 2006-05-02 22:03:38 +02:00 committed by Jakub Jelinek
parent a3428e0701
commit 8ca5b2a2d4
9 changed files with 330 additions and 16 deletions

View File

@ -1,3 +1,18 @@
2006-05-02 Jakub Jelinek <jakub@redhat.com>
PR c++/26943
* omp-low.c (maybe_lookup_decl_in_outer_ctx): New function.
(build_outer_var_ref): Use maybe_lookup_decl_in_outer_ctx
to find if var will be a global variable even in the nested context.
(omp_copy_decl): Only check for global variable at the end, it might
be overridden in outer contexts.
(scan_sharing_clauses): For global variables don't create a field.
(lower_rec_input_clauses): Do nothing for global shared variables.
Emit a barrier at the end of ILIST if there were any decls in both
firstprivate and lastprivate clauses.
(lower_send_clauses): Do nothing for global variables except for
COPYIN.
2006-05-02 Zdenek Dvorak <dvorakz@suse.cz>
* tree.c (unsigned_type_for, signed_type_for): Make sure a type

View File

@ -112,6 +112,8 @@ struct omp_region *root_omp_region;
static void scan_omp (tree *, omp_context *);
static void lower_omp (tree *, omp_context *);
static tree lookup_decl_in_outer_ctx (tree, omp_context *);
static tree maybe_lookup_decl_in_outer_ctx (tree, omp_context *);
/* Find an OpenMP clause of type KIND within CLAUSES. */
@ -560,7 +562,7 @@ build_outer_var_ref (tree var, omp_context *ctx)
{
tree x;
if (is_global_var (var))
if (is_global_var (maybe_lookup_decl_in_outer_ctx (var, ctx)))
x = var;
else if (is_variable_sized (var))
{
@ -674,9 +676,6 @@ omp_copy_decl (tree var, copy_body_data *cb)
omp_context *ctx = (omp_context *) cb;
tree new_var;
if (is_global_var (var) || decl_function_context (var) != ctx->cb.src_fn)
return var;
if (TREE_CODE (var) == LABEL_DECL)
{
new_var = create_artificial_label ();
@ -695,6 +694,9 @@ omp_copy_decl (tree var, copy_body_data *cb)
return new_var;
}
if (is_global_var (var) || decl_function_context (var) != ctx->cb.src_fn)
return var;
return error_mark_node;
}
@ -937,6 +939,10 @@ scan_sharing_clauses (tree clauses, omp_context *ctx)
decl = OMP_CLAUSE_DECL (c);
gcc_assert (!is_variable_sized (decl));
by_ref = use_pointer_for_field (decl, true);
/* Global variables don't need to be copied,
the receiver side will use them directly. */
if (is_global_var (maybe_lookup_decl_in_outer_ctx (decl, ctx)))
break;
if (! TREE_READONLY (decl)
|| TREE_ADDRESSABLE (decl)
|| by_ref
@ -963,7 +969,9 @@ scan_sharing_clauses (tree clauses, omp_context *ctx)
do_private:
if (is_variable_sized (decl))
break;
else if (is_parallel_ctx (ctx))
else if (is_parallel_ctx (ctx)
&& ! is_global_var (maybe_lookup_decl_in_outer_ctx (decl,
ctx)))
{
by_ref = use_pointer_for_field (decl, false);
install_var_field (decl, by_ref, ctx);
@ -1029,7 +1037,8 @@ scan_sharing_clauses (tree clauses, omp_context *ctx)
case OMP_CLAUSE_SHARED:
decl = OMP_CLAUSE_DECL (c);
fixup_remapped_decl (decl, ctx, false);
if (! is_global_var (maybe_lookup_decl_in_outer_ctx (decl, ctx)))
fixup_remapped_decl (decl, ctx, false);
break;
case OMP_CLAUSE_COPYPRIVATE:
@ -1415,6 +1424,23 @@ lookup_decl_in_outer_ctx (tree decl, omp_context *ctx)
}
/* Similar to lookup_decl_in_outer_ctx, but return DECL if not found
in outer contexts. */
static tree
maybe_lookup_decl_in_outer_ctx (tree decl, omp_context *ctx)
{
tree t = NULL;
omp_context *up;
if (ctx->is_nested)
for (up = ctx->outer, t = NULL; up && t == NULL; up = up->outer)
t = maybe_lookup_decl (decl, up);
return t ? t : decl;
}
/* Construct the initialization value for reduction CLAUSE. */
tree
@ -1493,6 +1519,7 @@ lower_rec_input_clauses (tree clauses, tree *ilist, tree *dlist,
tree_stmt_iterator diter;
tree c, dtor, copyin_seq, x, args, ptr;
bool copyin_by_ref = false;
bool lastprivate_firstprivate = false;
int pass;
*dlist = alloc_stmt_list ();
@ -1518,14 +1545,22 @@ lower_rec_input_clauses (tree clauses, tree *ilist, tree *dlist,
continue;
break;
case OMP_CLAUSE_SHARED:
if (maybe_lookup_decl (OMP_CLAUSE_DECL (c), ctx) == NULL)
{
gcc_assert (is_global_var (OMP_CLAUSE_DECL (c)));
continue;
}
case OMP_CLAUSE_FIRSTPRIVATE:
case OMP_CLAUSE_COPYIN:
case OMP_CLAUSE_REDUCTION:
break;
case OMP_CLAUSE_LASTPRIVATE:
if (pass != 0
&& OMP_CLAUSE_LASTPRIVATE_FIRSTPRIVATE (c))
continue;
if (OMP_CLAUSE_LASTPRIVATE_FIRSTPRIVATE (c))
{
lastprivate_firstprivate = true;
if (pass != 0)
continue;
}
break;
default:
continue;
@ -1611,6 +1646,9 @@ lower_rec_input_clauses (tree clauses, tree *ilist, tree *dlist,
switch (OMP_CLAUSE_CODE (c))
{
case OMP_CLAUSE_SHARED:
/* Shared global vars are just accessed directly. */
if (is_global_var (new_var))
break;
/* Set up the DECL_VALUE_EXPR for shared variables now. This
needs to be delayed until after fixup_child_record_type so
that we get the correct type during the dereference. */
@ -1700,8 +1738,10 @@ lower_rec_input_clauses (tree clauses, tree *ilist, tree *dlist,
/* If any copyin variable is passed by reference, we must ensure the
master thread doesn't modify it before it is copied over in all
threads. */
if (copyin_by_ref)
threads. Similarly for variables in both firstprivate and
lastprivate clauses we need to ensure the lastprivate copying
happens after firstprivate copying in all threads. */
if (copyin_by_ref || lastprivate_firstprivate)
build_omp_barrier (ilist);
}
@ -1919,6 +1959,9 @@ lower_send_clauses (tree clauses, tree *ilist, tree *olist, omp_context *ctx)
if (ctx->is_nested)
var = lookup_decl_in_outer_ctx (val, ctx);
if (OMP_CLAUSE_CODE (c) != OMP_CLAUSE_COPYIN
&& is_global_var (var))
continue;
if (is_variable_sized (val))
continue;
by_ref = use_pointer_for_field (val, false);

View File

@ -1,3 +1,13 @@
2006-05-02 Jakub Jelinek <jakub@redhat.com>
PR c++/26943
* testsuite/libgomp.c/pr26943-1.c: New test.
* testsuite/libgomp.c/pr26943-2.c: New test.
* testsuite/libgomp.c/pr26943-3.c: New test.
* testsuite/libgomp.c/pr26943-4.c: New test.
* testsuite/libgomp.c++/pr27337.C: Remove barrier.
* testsuite/libgomp.c++/pr26943.C: New test.
2006-05-02 Jakub Jelinek <jakub@redhat.com>
PR middle-end/27337

View File

@ -0,0 +1,62 @@
// PR c++/26943
// { dg-do run }
#include <assert.h>
#include <unistd.h>
struct S
{
public:
int x;
S () : x(-1) { }
S (const S &);
S& operator= (const S &);
void test ();
};
static volatile int hold;
S::S (const S &s)
{
#pragma omp master
sleep (1);
assert (s.x == -1);
x = 0;
}
S&
S::operator= (const S& s)
{
assert (s.x == 1);
x = 2;
return *this;
}
void
S::test ()
{
assert (x == 0);
x = 1;
}
static S x;
void
foo ()
{
#pragma omp sections firstprivate(x) lastprivate(x)
{
x.test();
}
}
int
main ()
{
#pragma omp parallel num_threads(2)
foo();
assert (x.x == 2);
return 0;
}

View File

@ -48,11 +48,7 @@ foo ()
#pragma omp parallel for firstprivate (ret) lastprivate (ret) \
schedule (static, 1) num_threads (4)
for (i = 0; i < 4; i++)
{
ret.i += omp_get_thread_num ();
// FIXME: The following barrier should be unnecessary.
#pragma omp barrier
}
ret.i += omp_get_thread_num ();
return ret;
}

View File

@ -0,0 +1,24 @@
/* PR c++/26943 */
/* { dg-do run } */
extern void abort (void);
extern void omp_set_dynamic (int);
int n = 6;
int
main (void)
{
int i, x = 0;
omp_set_dynamic (0);
#pragma omp parallel for num_threads (16) firstprivate (n) lastprivate (n) \
schedule (static, 1) reduction (+: x)
for (i = 0; i < 16; i++)
{
if (n != 6)
++x;
n = i;
}
if (x || n != 15)
abort ();
return 0;
}

View File

@ -0,0 +1,47 @@
/* PR c++/26943 */
/* { dg-do run } */
extern int omp_set_dynamic (int);
extern void abort (void);
int a = 8, b = 12, c = 16, d = 20, j = 0;
char e[10] = "a", f[10] = "b", g[10] = "c", h[10] = "d";
int
main (void)
{
int i;
omp_set_dynamic (0);
#pragma omp parallel for shared (a, e) firstprivate (b, f) \
lastprivate (c, g) private (d, h) \
schedule (static, 1) num_threads (4) \
reduction (+:j)
for (i = 0; i < 4; i++)
{
if (a != 8 || b != 12 || e[0] != 'a' || f[0] != 'b')
j++;
#pragma omp barrier
#pragma omp atomic
a += i;
b += i;
c = i;
d = i;
#pragma omp atomic
e[0] += i;
f[0] += i;
g[0] = 'g' + i;
h[0] = 'h' + i;
#pragma omp barrier
if (a != 8 + 6 || b != 12 + i || c != i || d != i)
j += 8;
if (e[0] != 'a' + 6 || f[0] != 'b' + i || g[0] != 'g' + i)
j += 64;
if (h[0] != 'h' + i)
j += 512;
}
if (j || a != 8 + 6 || b != 12 || c != 3 || d != 20)
abort ();
if (e[0] != 'a' + 6 || f[0] != 'b' || g[0] != 'g' + 3 || h[0] != 'd')
abort ();
return 0;
}

View File

@ -0,0 +1,56 @@
/* PR c++/26943 */
/* { dg-do run } */
extern int omp_set_dynamic (int);
extern int omp_get_thread_num (void);
extern void abort (void);
int a = 8, b = 12, c = 16, d = 20, j = 0, l = 0;
char e[10] = "a", f[10] = "b", g[10] = "c", h[10] = "d";
volatile int k;
int
main (void)
{
int i;
omp_set_dynamic (0);
omp_set_nested (1);
#pragma omp parallel num_threads (2) reduction (+:l)
if (k == omp_get_thread_num ())
{
#pragma omp parallel for shared (a, e) firstprivate (b, f) \
lastprivate (c, g) private (d, h) \
schedule (static, 1) num_threads (4) \
reduction (+:j)
for (i = 0; i < 4; i++)
{
if (a != 8 || b != 12 || e[0] != 'a' || f[0] != 'b')
j++;
#pragma omp barrier
#pragma omp atomic
a += i;
b += i;
c = i;
d = i;
#pragma omp atomic
e[0] += i;
f[0] += i;
g[0] = 'g' + i;
h[0] = 'h' + i;
#pragma omp barrier
if (a != 8 + 6 || b != 12 + i || c != i || d != i)
j += 8;
if (e[0] != 'a' + 6 || f[0] != 'b' + i || g[0] != 'g' + i)
j += 64;
if (h[0] != 'h' + i)
j += 512;
}
if (j || a != 8 + 6 || b != 12 || c != 3 || d != 20)
++l;
if (e[0] != 'a' + 6 || f[0] != 'b' || g[0] != 'g' + 3 || h[0] != 'd')
l += 8;
}
if (l)
abort ();
return 0;
}

View File

@ -0,0 +1,61 @@
/* PR c++/26943 */
/* { dg-do run } */
extern int omp_set_dynamic (int);
extern int omp_get_thread_num (void);
extern void abort (void);
int a = 8, b = 12, c = 16, d = 20, j = 0, l = 0;
char e[10] = "a", f[10] = "b", g[10] = "c", h[10] = "d";
volatile int k;
int
main (void)
{
int i;
omp_set_dynamic (0);
omp_set_nested (1);
#pragma omp parallel num_threads (2) reduction (+:l) \
firstprivate (a, b, c, d, e, f, g, h, j)
if (k == omp_get_thread_num ())
{
#pragma omp parallel for shared (a, e) firstprivate (b, f) \
lastprivate (c, g) private (d, h) \
schedule (static, 1) num_threads (4) \
reduction (+:j)
for (i = 0; i < 4; i++)
{
if (a != 8 || b != 12 || e[0] != 'a' || f[0] != 'b')
j++;
#pragma omp barrier
#pragma omp atomic
a += i;
b += i;
c = i;
d = i;
#pragma omp atomic
e[0] += i;
f[0] += i;
g[0] = 'g' + i;
h[0] = 'h' + i;
#pragma omp barrier
if (a != 8 + 6 || b != 12 + i || c != i || d != i)
j += 8;
if (e[0] != 'a' + 6 || f[0] != 'b' + i || g[0] != 'g' + i)
j += 64;
if (h[0] != 'h' + i)
j += 512;
}
if (j || a != 8 + 6 || b != 12 || c != 3 || d != 20)
++l;
if (e[0] != 'a' + 6 || f[0] != 'b' || g[0] != 'g' + 3 || h[0] != 'd')
l += 8;
}
if (l)
abort ();
if (a != 8 || b != 12 || c != 16 || d != 20)
abort ();
if (e[0] != 'a' || f[0] != 'b' || g[0] != 'c' || h[0] != 'd')
abort ();
return 0;
}