FreeChainXenon/gcc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

tree-core.h (enum omp_clause_code): Add OMP_CLAUSE__SCANTEMP_ clause.

Browse source 
* tree-core.h (enum omp_clause_code): Add OMP_CLAUSE__SCANTEMP_
	clause.
	* tree.h (OMP_CLAUSE_DECL): Use OMP_CLAUSE__SCANTEMP_ instead of
	OMP_CLAUSE__CONDTEMP_ as range's upper bound.
	(OMP_CLAUSE__SCANTEMP__ALLOC, OMP_CLAUSE__SCANTEMP__CONTROL): Define.
	* tree.c (omp_clause_num_ops, omp_clause_code_name): Add
	OMP_CLAUSE__SCANTEMP_ entry.
	(walk_tree_1): Handle OMP_CLAUSE__SCANTEMP_.
	* tree-pretty-print.c (dump_omp_clause): Likewise.
	* tree-nested.c (convert_nonlocal_omp_clauses,
	convert_local_omp_clauses): Likewise.
	* omp-general.h (struct omp_for_data): Add have_scantemp and
	have_nonctrl_scantemp members.
	* omp-general.c (omp_extract_for_data): Initialize them.
	* omp-low.c (struct omp_context): Add scan_exclusive member.
	(scan_omp_1_stmt): Don't unnecessarily mask gimple_omp_for_kind
	result again with GF_OMP_FOR_KIND_MASK.  Initialize also
	ctx->scan_exclusive.
	(lower_rec_simd_input_clauses): Use ctx->scan_exclusive instead
	of !ctx->scan_inclusive.
	(lower_rec_input_clauses): Simplify gimplification of dtors using
	gimplify_and_add.  For non-is_simd test OMP_CLAUSE_REDUCTION_INSCAN
	rather than rvarp.  Handle OMP_CLAUSE_REDUCTION_INSCAN in worksharing
	loops.  Don't add barrier for reduction_omp_orig_ref if
	ctx->scan_??xclusive.
	(lower_reduction_clauses): Don't do anything for ctx->scan_??xclusive.
	(lower_omp_scan): Use ctx->scan_exclusive instead
	of !ctx->scan_inclusive.  Handle worksharing loops with inscan
	reductions.  Use new_vard != new_var instead of repeated
	omp_is_reference calls.
	(omp_find_scan, lower_omp_for_scan): New functions.
	(lower_omp_for): Call lower_omp_for_scan for worksharing loops with
	inscan reductions.
	* omp-expand.c (expand_omp_scantemp_alloc): New function.
	(expand_omp_for_static_nochunk): Handle fd->have_nonctrl_scantemp
	and fd->have_scantemp.

	* c-c++-common/gomp/scan-3.c (f1): Don't expect a sorry message.
	* c-c++-common/gomp/scan-5.c (foo): Likewise.

	* testsuite/libgomp.c++/scan-1.C: New test.
	* testsuite/libgomp.c++/scan-2.C: New test.
	* testsuite/libgomp.c++/scan-3.C: New test.
	* testsuite/libgomp.c++/scan-4.C: New test.
	* testsuite/libgomp.c++/scan-5.C: New test.
	* testsuite/libgomp.c++/scan-6.C: New test.
	* testsuite/libgomp.c++/scan-7.C: New test.
	* testsuite/libgomp.c++/scan-8.C: New test.
	* testsuite/libgomp.c/scan-1.c: New test.
	* testsuite/libgomp.c/scan-2.c: New test.
	* testsuite/libgomp.c/scan-3.c: New test.
	* testsuite/libgomp.c/scan-4.c: New test.
	* testsuite/libgomp.c/scan-5.c: New test.
	* testsuite/libgomp.c/scan-6.c: New test.
	* testsuite/libgomp.c/scan-7.c: New test.
	* testsuite/libgomp.c/scan-8.c: New test.

From-SVN: r272958
This commit is contained in:
Jakub Jelinek 2019-07-03 07:03:58 +02:00 • committed by Jakub Jelinek
parent 83eb952208
commit 2f6bb511d1
30 changed files with 3264 additions and 142 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

37
gcc/ChangeLog
View file

@ -1,5 +1,42 @@
2019-07-03 Jakub Jelinek <jakub@redhat.com>
* tree-core.h (enum omp_clause_code): Add OMP_CLAUSE__SCANTEMP_
clause.
* tree.h (OMP_CLAUSE_DECL): Use OMP_CLAUSE__SCANTEMP_ instead of
OMP_CLAUSE__CONDTEMP_ as range's upper bound.
(OMP_CLAUSE__SCANTEMP__ALLOC, OMP_CLAUSE__SCANTEMP__CONTROL): Define.
* tree.c (omp_clause_num_ops, omp_clause_code_name): Add
OMP_CLAUSE__SCANTEMP_ entry.
(walk_tree_1): Handle OMP_CLAUSE__SCANTEMP_.
* tree-pretty-print.c (dump_omp_clause): Likewise.
* tree-nested.c (convert_nonlocal_omp_clauses,
convert_local_omp_clauses): Likewise.
* omp-general.h (struct omp_for_data): Add have_scantemp and
have_nonctrl_scantemp members.
* omp-general.c (omp_extract_for_data): Initialize them.
* omp-low.c (struct omp_context): Add scan_exclusive member.
(scan_omp_1_stmt): Don't unnecessarily mask gimple_omp_for_kind
result again with GF_OMP_FOR_KIND_MASK. Initialize also
ctx->scan_exclusive.
(lower_rec_simd_input_clauses): Use ctx->scan_exclusive instead
of !ctx->scan_inclusive.
(lower_rec_input_clauses): Simplify gimplification of dtors using
gimplify_and_add. For non-is_simd test OMP_CLAUSE_REDUCTION_INSCAN
rather than rvarp. Handle OMP_CLAUSE_REDUCTION_INSCAN in worksharing
loops. Don't add barrier for reduction_omp_orig_ref if
ctx->scan_??xclusive.
(lower_reduction_clauses): Don't do anything for ctx->scan_??xclusive.
(lower_omp_scan): Use ctx->scan_exclusive instead
of !ctx->scan_inclusive. Handle worksharing loops with inscan
reductions. Use new_vard != new_var instead of repeated
omp_is_reference calls.
(omp_find_scan, lower_omp_for_scan): New functions.
(lower_omp_for): Call lower_omp_for_scan for worksharing loops with
inscan reductions.
* omp-expand.c (expand_omp_scantemp_alloc): New function.
(expand_omp_for_static_nochunk): Handle fd->have_nonctrl_scantemp
and fd->have_scantemp.
* gimplify.c (gimplify_scan_omp_clauses): For inscan reductions
on worksharing loop propagate it as shared clause to containing
combined parallel.

292
gcc/omp-expand.c
View file

@ -3502,6 +3502,98 @@ expand_omp_for_generic (struct omp_region *region,
}
}
/* Helper function for expand_omp_for_static_nochunk. If PTR is NULL,
compute needed allocation size. If !ALLOC of team allocations,
if ALLOC of thread allocation. SZ is the initial needed size for
other purposes, ALLOC_ALIGN guaranteed alignment of allocation in bytes,
CNT number of elements of each array, for !ALLOC this is
omp_get_num_threads (), for ALLOC number of iterations handled by the
current thread. If PTR is non-NULL, it is the start of the allocation
and this routine shall assign to OMP_CLAUSE_DECL (c) of those _scantemp_
clauses pointers to the corresponding arrays. */
static tree
expand_omp_scantemp_alloc (tree clauses, tree ptr, unsigned HOST_WIDE_INT sz,
unsigned HOST_WIDE_INT alloc_align, tree cnt,
gimple_stmt_iterator *gsi, bool alloc)
{
tree eltsz = NULL_TREE;
unsigned HOST_WIDE_INT preval = 0;
if (ptr && sz)
ptr = fold_build2 (POINTER_PLUS_EXPR, TREE_TYPE (ptr),
ptr, size_int (sz));
for (tree c = clauses; c; c = OMP_CLAUSE_CHAIN (c))
if (OMP_CLAUSE_CODE (c) == OMP_CLAUSE__SCANTEMP_
&& !OMP_CLAUSE__SCANTEMP__CONTROL (c)
&& (!OMP_CLAUSE__SCANTEMP__ALLOC (c)) != alloc)
{
tree pointee_type = TREE_TYPE (TREE_TYPE (OMP_CLAUSE_DECL (c)));
unsigned HOST_WIDE_INT al = TYPE_ALIGN_UNIT (pointee_type);
if (tree_fits_uhwi_p (TYPE_SIZE_UNIT (pointee_type)))
{
unsigned HOST_WIDE_INT szl
= tree_to_uhwi (TYPE_SIZE_UNIT (pointee_type));
szl = least_bit_hwi (szl);
if (szl)
al = MIN (al, szl);
}
if (ptr == NULL_TREE)
{
if (eltsz == NULL_TREE)
eltsz = TYPE_SIZE_UNIT (pointee_type);
else
eltsz = size_binop (PLUS_EXPR, eltsz,
TYPE_SIZE_UNIT (pointee_type));
}
if (preval == 0 && al <= alloc_align)
{
unsigned HOST_WIDE_INT diff = ROUND_UP (sz, al) - sz;
sz += diff;
if (diff && ptr)
ptr = fold_build2 (POINTER_PLUS_EXPR, TREE_TYPE (ptr),
ptr, size_int (diff));
}
else if (al > preval)
{
if (ptr)
{
ptr = fold_convert (pointer_sized_int_node, ptr);
ptr = fold_build2 (PLUS_EXPR, pointer_sized_int_node, ptr,
build_int_cst (pointer_sized_int_node,
al - 1));
ptr = fold_build2 (BIT_AND_EXPR, pointer_sized_int_node, ptr,
build_int_cst (pointer_sized_int_node,
-(HOST_WIDE_INT) al));
ptr = fold_convert (ptr_type_node, ptr);
}
else
sz += al - 1;
}
if (tree_fits_uhwi_p (TYPE_SIZE_UNIT (pointee_type)))
preval = al;
else
preval = 1;
if (ptr)
{
expand_omp_build_assign (gsi, OMP_CLAUSE_DECL (c), ptr, false);
ptr = OMP_CLAUSE_DECL (c);
ptr = fold_build2 (POINTER_PLUS_EXPR, TREE_TYPE (ptr), ptr,
size_binop (MULT_EXPR, cnt,
TYPE_SIZE_UNIT (pointee_type)));
}
}
if (ptr == NULL_TREE)
{
eltsz = size_binop (MULT_EXPR, eltsz, cnt);
if (sz)
eltsz = size_binop (PLUS_EXPR, eltsz, size_int (sz));
return eltsz;
}
else
return ptr;
}
/* A subroutine of expand_omp_for. Generate code for a parallel
loop with static schedule and no specified chunk size. Given
parameters:
@ -3544,11 +3636,12 @@ expand_omp_for_static_nochunk (struct omp_region *region,
struct omp_for_data *fd,
gimple *inner_stmt)
{
tree n, q, s0, e0, e, t, tt, nthreads, threadid;
tree n, q, s0, e0, e, t, tt, nthreads = NULL_TREE, threadid;
tree type, itype, vmain, vback;
basic_block entry_bb, second_bb, third_bb, exit_bb, seq_start_bb;
basic_block body_bb, cont_bb, collapse_bb = NULL;
basic_block fin_bb;
basic_block fin_bb, fourth_bb = NULL, fifth_bb = NULL, sixth_bb = NULL;
basic_block exit1_bb = NULL, exit2_bb = NULL, exit3_bb = NULL;
gimple_stmt_iterator gsi, gsip;
edge ep;
bool broken_loop = region->cont == NULL;
@ -3650,7 +3743,9 @@ expand_omp_for_static_nochunk (struct omp_region *region,
c = omp_find_clause (OMP_CLAUSE_CHAIN (c), OMP_CLAUSE__CONDTEMP_);
cond_var = OMP_CLAUSE_DECL (c);
}
if (fd->have_reductemp || fd->have_pointer_condtemp)
if (fd->have_reductemp
|| fd->have_pointer_condtemp
|| fd->have_nonctrl_scantemp)
{
tree t1 = build_int_cst (long_integer_type_node, 0);
tree t2 = build_int_cst (long_integer_type_node, 1);
@ -3660,8 +3755,11 @@ expand_omp_for_static_nochunk (struct omp_region *region,
gimple_stmt_iterator gsi2 = gsi_none ();
gimple *g = NULL;
tree mem = null_pointer_node, memv = NULL_TREE;
unsigned HOST_WIDE_INT condtemp_sz = 0;
unsigned HOST_WIDE_INT alloc_align = 0;
if (fd->have_reductemp)
{
gcc_assert (!fd->have_nonctrl_scantemp);
tree c = omp_find_clause (clauses, OMP_CLAUSE__REDUCTEMP_);
reductions = OMP_CLAUSE_DECL (c);
gcc_assert (TREE_CODE (reductions) == SSA_NAME);
@ -3678,16 +3776,40 @@ expand_omp_for_static_nochunk (struct omp_region *region,
gsi2 = gsip;
reductions = null_pointer_node;
}
if (fd->have_pointer_condtemp)
if (fd->have_pointer_condtemp || fd->have_nonctrl_scantemp)
{
tree type = TREE_TYPE (condtemp);
tree type;
if (fd->have_pointer_condtemp)
type = TREE_TYPE (condtemp);
else
type = ptr_type_node;
memv = create_tmp_var (type);
TREE_ADDRESSABLE (memv) = 1;
unsigned HOST_WIDE_INT sz
= tree_to_uhwi (TYPE_SIZE_UNIT (TREE_TYPE (type)));
sz *= fd->lastprivate_conditional;
expand_omp_build_assign (&gsi2, memv, build_int_cst (type, sz),
false);
unsigned HOST_WIDE_INT sz = 0;
tree size = NULL_TREE;
if (fd->have_pointer_condtemp)
{
sz = tree_to_uhwi (TYPE_SIZE_UNIT (TREE_TYPE (type)));
sz *= fd->lastprivate_conditional;
condtemp_sz = sz;
}
if (fd->have_nonctrl_scantemp)
{
nthreads = builtin_decl_explicit (BUILT_IN_OMP_GET_NUM_THREADS);
gimple *g = gimple_build_call (nthreads, 0);
nthreads = create_tmp_var (integer_type_node);
gimple_call_set_lhs (g, nthreads);
gsi_insert_before (&gsi2, g, GSI_SAME_STMT);
nthreads = fold_convert (sizetype, nthreads);
alloc_align = TYPE_ALIGN_UNIT (long_long_integer_type_node);
size = expand_omp_scantemp_alloc (clauses, NULL_TREE, sz,
alloc_align, nthreads, NULL,
false);
size = fold_convert (type, size);
}
else
size = build_int_cst (type, sz);
expand_omp_build_assign (&gsi2, memv, size, false);
mem = build_fold_addr_expr (memv);
}
tree t
@ -3698,6 +3820,12 @@ expand_omp_for_static_nochunk (struct omp_region *region,
true, GSI_SAME_STMT);
if (fd->have_pointer_condtemp)
expand_omp_build_assign (&gsi2, condtemp, memv, false);
if (fd->have_nonctrl_scantemp)
{
tree ptr = fd->have_pointer_condtemp ? condtemp : memv;
expand_omp_scantemp_alloc (clauses, ptr, condtemp_sz,
alloc_align, nthreads, &gsi2, false);
}
if (fd->have_reductemp)
{
gsi_remove (&gsi2, true);
@ -3788,6 +3916,72 @@ expand_omp_for_static_nochunk (struct omp_region *region,
gsi = gsi_last_nondebug_bb (third_bb);
gcc_assert (gimple_code (gsi_stmt (gsi)) == GIMPLE_OMP_FOR);
if (fd->have_nonctrl_scantemp)
{
tree clauses = gimple_omp_for_clauses (fd->for_stmt);
tree controlp = NULL_TREE, controlb = NULL_TREE;
for (tree c = clauses; c; c = OMP_CLAUSE_CHAIN (c))
if (OMP_CLAUSE_CODE (c) == OMP_CLAUSE__SCANTEMP_
&& OMP_CLAUSE__SCANTEMP__CONTROL (c))
{
if (TREE_TYPE (OMP_CLAUSE_DECL (c)) == boolean_type_node)
controlb = OMP_CLAUSE_DECL (c);
else
controlp = OMP_CLAUSE_DECL (c);
if (controlb && controlp)
break;
}
gcc_assert (controlp && controlb);
tree cnt = create_tmp_var (sizetype);
gimple *g = gimple_build_assign (cnt, NOP_EXPR, q);
gsi_insert_before (&gsi, g, GSI_SAME_STMT);
unsigned HOST_WIDE_INT alloc_align = TYPE_ALIGN_UNIT (ptr_type_node);
tree sz = expand_omp_scantemp_alloc (clauses, NULL_TREE, 0,
alloc_align, cnt, NULL, true);
tree size = create_tmp_var (sizetype);
expand_omp_build_assign (&gsi, size, sz, false);
tree cmp = fold_build2 (GT_EXPR, boolean_type_node,
size, size_int (16384));
expand_omp_build_assign (&gsi, controlb, cmp);
g = gimple_build_cond (NE_EXPR, controlb, boolean_false_node,
NULL_TREE, NULL_TREE);
gsi_insert_before (&gsi, g, GSI_SAME_STMT);
fourth_bb = split_block (third_bb, g)->dest;
gsi = gsi_last_nondebug_bb (fourth_bb);
/* FIXME: Once we have allocators, this should use allocator. */
g = gimple_build_call (builtin_decl_explicit (BUILT_IN_MALLOC), 1, size);
gimple_call_set_lhs (g, controlp);
gsi_insert_before (&gsi, g, GSI_SAME_STMT);
expand_omp_scantemp_alloc (clauses, controlp, 0, alloc_align, cnt,
&gsi, true);
gsi_prev (&gsi);
g = gsi_stmt (gsi);
fifth_bb = split_block (fourth_bb, g)->dest;
gsi = gsi_last_nondebug_bb (fifth_bb);
g = gimple_build_call (builtin_decl_implicit (BUILT_IN_STACK_SAVE), 0);
gimple_call_set_lhs (g, controlp);
gsi_insert_before (&gsi, g, GSI_SAME_STMT);
tree alloca_decl = builtin_decl_explicit (BUILT_IN_ALLOCA_WITH_ALIGN);
for (tree c = clauses; c; c = OMP_CLAUSE_CHAIN (c))
if (OMP_CLAUSE_CODE (c) == OMP_CLAUSE__SCANTEMP_
&& OMP_CLAUSE__SCANTEMP__ALLOC (c))
{
tree tmp = create_tmp_var (sizetype);
tree pointee_type = TREE_TYPE (TREE_TYPE (OMP_CLAUSE_DECL (c)));
g = gimple_build_assign (tmp, MULT_EXPR, cnt,
TYPE_SIZE_UNIT (pointee_type));
gsi_insert_before (&gsi, g, GSI_SAME_STMT);
g = gimple_build_call (alloca_decl, 2, tmp,
size_int (TYPE_ALIGN (pointee_type)));
gimple_call_set_lhs (g, OMP_CLAUSE_DECL (c));
gsi_insert_before (&gsi, g, GSI_SAME_STMT);
}
sixth_bb = split_block (fifth_bb, g)->dest;
gsi = gsi_last_nondebug_bb (sixth_bb);
}
t = build2 (MULT_EXPR, itype, q, threadid);
t = build2 (PLUS_EXPR, itype, t, tt);
s0 = force_gimple_operand_gsi (&gsi, t, true, NULL_TREE, true, GSI_SAME_STMT);
@ -4018,7 +4212,9 @@ expand_omp_for_static_nochunk (struct omp_region *region,
if (!gimple_omp_return_nowait_p (gsi_stmt (gsi)))
{
t = gimple_omp_return_lhs (gsi_stmt (gsi));
if (fd->have_reductemp || fd->have_pointer_condtemp)
if (fd->have_reductemp
|| ((fd->have_pointer_condtemp || fd->have_scantemp)
&& !fd->have_nonctrl_scantemp))
{
tree fn;
if (t)
@ -4045,6 +4241,38 @@ expand_omp_for_static_nochunk (struct omp_region *region,
gcall *g = gimple_build_call (fn, 0);
gsi_insert_after (&gsi, g, GSI_SAME_STMT);
}
if (fd->have_scantemp && !fd->have_nonctrl_scantemp)
{
tree clauses = gimple_omp_for_clauses (fd->for_stmt);
tree controlp = NULL_TREE, controlb = NULL_TREE;
for (tree c = clauses; c; c = OMP_CLAUSE_CHAIN (c))
if (OMP_CLAUSE_CODE (c) == OMP_CLAUSE__SCANTEMP_
&& OMP_CLAUSE__SCANTEMP__CONTROL (c))
{
if (TREE_TYPE (OMP_CLAUSE_DECL (c)) == boolean_type_node)
controlb = OMP_CLAUSE_DECL (c);
else
controlp = OMP_CLAUSE_DECL (c);
if (controlb && controlp)
break;
}
gcc_assert (controlp && controlb);
gimple *g = gimple_build_cond (NE_EXPR, controlb, boolean_false_node,
NULL_TREE, NULL_TREE);
gsi_insert_before (&gsi, g, GSI_SAME_STMT);
exit1_bb = split_block (exit_bb, g)->dest;
gsi = gsi_after_labels (exit1_bb);
g = gimple_build_call (builtin_decl_explicit (BUILT_IN_FREE), 1,
controlp);
gsi_insert_before (&gsi, g, GSI_SAME_STMT);
exit2_bb = split_block (exit1_bb, g)->dest;
gsi = gsi_after_labels (exit2_bb);
g = gimple_build_call (builtin_decl_implicit (BUILT_IN_STACK_RESTORE), 1,
controlp);
gsi_insert_before (&gsi, g, GSI_SAME_STMT);
exit3_bb = split_block (exit2_bb, g)->dest;
gsi = gsi_after_labels (exit3_bb);
}
gsi_remove (&gsi, true);
/* Connect all the blocks. */
@ -4053,8 +4281,34 @@ expand_omp_for_static_nochunk (struct omp_region *region,
ep = find_edge (entry_bb, second_bb);
ep->flags = EDGE_TRUE_VALUE;
ep->probability = profile_probability::guessed_always ().apply_scale (1, 4);
find_edge (third_bb, seq_start_bb)->flags = EDGE_FALSE_VALUE;
find_edge (third_bb, fin_bb)->flags = EDGE_TRUE_VALUE;
if (fourth_bb)
{
ep = make_edge (third_bb, fifth_bb, EDGE_FALSE_VALUE);
ep->probability
= profile_probability::guessed_always ().apply_scale (1, 2);
ep = find_edge (third_bb, fourth_bb);
ep->flags = EDGE_TRUE_VALUE;
ep->probability
= profile_probability::guessed_always ().apply_scale (1, 2);
ep = find_edge (fourth_bb, fifth_bb);
redirect_edge_and_branch (ep, sixth_bb);
}
else
sixth_bb = third_bb;
find_edge (sixth_bb, seq_start_bb)->flags = EDGE_FALSE_VALUE;
find_edge (sixth_bb, fin_bb)->flags = EDGE_TRUE_VALUE;
if (exit1_bb)
{
ep = make_edge (exit_bb, exit2_bb, EDGE_FALSE_VALUE);
ep->probability
= profile_probability::guessed_always ().apply_scale (1, 2);
ep = find_edge (exit_bb, exit1_bb);
ep->flags = EDGE_TRUE_VALUE;
ep->probability
= profile_probability::guessed_always ().apply_scale (1, 2);
ep = find_edge (exit1_bb, exit2_bb);
redirect_edge_and_branch (ep, exit3_bb);
}
if (!broken_loop)
{
@ -4082,12 +4336,22 @@ expand_omp_for_static_nochunk (struct omp_region *region,
set_immediate_dominator (CDI_DOMINATORS, second_bb, entry_bb);
set_immediate_dominator (CDI_DOMINATORS, third_bb, entry_bb);
set_immediate_dominator (CDI_DOMINATORS, seq_start_bb, third_bb);
if (fourth_bb)
{
set_immediate_dominator (CDI_DOMINATORS, fifth_bb, third_bb);
set_immediate_dominator (CDI_DOMINATORS, sixth_bb, third_bb);
}
set_immediate_dominator (CDI_DOMINATORS, seq_start_bb, sixth_bb);
set_immediate_dominator (CDI_DOMINATORS, body_bb,
recompute_dominator (CDI_DOMINATORS, body_bb));
set_immediate_dominator (CDI_DOMINATORS, fin_bb,
recompute_dominator (CDI_DOMINATORS, fin_bb));
if (exit1_bb)
{
set_immediate_dominator (CDI_DOMINATORS, exit2_bb, exit_bb);
set_immediate_dominator (CDI_DOMINATORS, exit3_bb, exit_bb);
}
struct loop *loop = body_bb->loop_father;
if (loop != entry_bb->loop_father)

8
gcc/omp-general.c
View file

@ -169,6 +169,8 @@ omp_extract_for_data (gomp_for *for_stmt, struct omp_for_data *fd,
fd->have_ordered = false;
fd->have_reductemp = false;
fd->have_pointer_condtemp = false;
fd->have_scantemp = false;
fd->have_nonctrl_scantemp = false;
fd->lastprivate_conditional = 0;
fd->tiling = NULL_TREE;
fd->collapse = 1;
@ -231,6 +233,12 @@ omp_extract_for_data (gomp_for *for_stmt, struct omp_for_data *fd,
if (POINTER_TYPE_P (TREE_TYPE (OMP_CLAUSE_DECL (t))))
fd->have_pointer_condtemp = true;
break;
case OMP_CLAUSE__SCANTEMP_:
fd->have_scantemp = true;
if (!OMP_CLAUSE__SCANTEMP__ALLOC (t)
&& !OMP_CLAUSE__SCANTEMP__CONTROL (t))
fd->have_nonctrl_scantemp = true;
break;
default:
break;
}

2
gcc/omp-general.h
View file

@ -63,7 +63,7 @@ struct omp_for_data
int collapse; /* Collapsed loops, 1 for a non-collapsed loop. */
int ordered;
bool have_nowait, have_ordered, simd_schedule, have_reductemp;
bool have_pointer_condtemp;
bool have_pointer_condtemp, have_scantemp, have_nonctrl_scantemp;
int lastprivate_conditional;
unsigned char sched_modifiers;
enum omp_clause_schedule_kind sched_kind;

939
gcc/omp-low.c
View file

File diff suppressed because it is too large Load diff

3
gcc/testsuite/ChangeLog
View file

@ -1,5 +1,8 @@
2019-07-03 Jakub Jelinek <jakub@redhat.com>
* c-c++-common/gomp/scan-3.c (f1): Don't expect a sorry message.
* c-c++-common/gomp/scan-5.c (foo): Likewise.
* c-c++-common/gomp/scan-5.c: New test.
* c-c++-common/gomp/lastprivate-conditional-5.c: New test.

2
gcc/testsuite/c-c++-common/gomp/scan-3.c
View file

@ -8,7 +8,7 @@ f1 (int *c, int *d)
for (i = 0; i < 64; i++)
{
d[i] = a;
#pragma omp scan inclusive (a) /* { dg-message "sorry, unimplemented: '#pragma omp scan' not supported yet" } */
#pragma omp scan inclusive (a)
a += c[i];
}
}

2
gcc/testsuite/c-c++-common/gomp/scan-5.c
View file

@ -6,7 +6,7 @@ foo (int *a, int *b)
for (int i = 0; i < 64; i++)
{
r += a[i];
#pragma omp scan inclusive (r) /* { dg-message "sorry, unimplemented: '#pragma omp scan' not supported yet" } */
#pragma omp scan inclusive (r)
b[i] = r;
}
return r;

3
gcc/tree-core.h
View file

@ -352,6 +352,9 @@ enum omp_clause_code {
/* Internal clause: temporary for lastprivate(conditional:). */
OMP_CLAUSE__CONDTEMP_,
/* Internal clause: temporary for inscan reductions. */
OMP_CLAUSE__SCANTEMP_,
/* OpenACC/OpenMP clause: if (scalar-expression). */
OMP_CLAUSE_IF,

2
gcc/tree-nested.c
View file

@ -1349,6 +1349,7 @@ convert_nonlocal_omp_clauses (tree *pclauses, struct walk_stmt_info *wi)
case OMP_CLAUSE_IF_PRESENT:
case OMP_CLAUSE_FINALIZE:
case OMP_CLAUSE__CONDTEMP_:
case OMP_CLAUSE__SCANTEMP_:
break;
/* The following clause belongs to the OpenACC cache directive, which
@ -2078,6 +2079,7 @@ convert_local_omp_clauses (tree *pclauses, struct walk_stmt_info *wi)
case OMP_CLAUSE_IF_PRESENT:
case OMP_CLAUSE_FINALIZE:
case OMP_CLAUSE__CONDTEMP_:
case OMP_CLAUSE__SCANTEMP_:
break;
/* The following clause belongs to the OpenACC cache directive, which

3
gcc/tree-pretty-print.c
View file

@ -483,6 +483,9 @@ dump_omp_clause (pretty_printer *pp, tree clause, int spc, dump_flags_t flags)
case OMP_CLAUSE__CONDTEMP_:
name = "_condtemp_";
goto print_remap;
case OMP_CLAUSE__SCANTEMP_:
name = "_scantemp_";
goto print_remap;
case OMP_CLAUSE_TO_DECLARE:
name = "to";
goto print_remap;

3
gcc/tree.c
View file

@ -311,6 +311,7 @@ unsigned const char omp_clause_num_ops[] =
1, /* OMP_CLAUSE__LOOPTEMP_ */
1, /* OMP_CLAUSE__REDUCTEMP_ */
1, /* OMP_CLAUSE__CONDTEMP_ */
1, /* OMP_CLAUSE__SCANTEMP_ */
1, /* OMP_CLAUSE_IF */
1, /* OMP_CLAUSE_NUM_THREADS */
1, /* OMP_CLAUSE_SCHEDULE */
@ -391,6 +392,7 @@ const char * const omp_clause_code_name[] =
"_looptemp_",
"_reductemp_",
"_condtemp_",
"_scantemp_",
"if",
"num_threads",
"schedule",
@ -12316,6 +12318,7 @@ walk_tree_1 (tree *tp, walk_tree_fn func, void *data,
case OMP_CLAUSE__LOOPTEMP_:
case OMP_CLAUSE__REDUCTEMP_:
case OMP_CLAUSE__CONDTEMP_:
case OMP_CLAUSE__SCANTEMP_:
case OMP_CLAUSE__SIMDUID_:
WALK_SUBTREE (OMP_CLAUSE_OPERAND (*tp, 0));
/* FALLTHRU */

13
gcc/tree.h
View file

@ -1449,7 +1449,7 @@ class auto_suppress_location_wrappers
#define OMP_CLAUSE_DECL(NODE) \
OMP_CLAUSE_OPERAND (OMP_CLAUSE_RANGE_CHECK (OMP_CLAUSE_CHECK (NODE), \
OMP_CLAUSE_PRIVATE, \
OMP_CLAUSE__CONDTEMP_), 0)
OMP_CLAUSE__SCANTEMP_), 0)
#define OMP_CLAUSE_HAS_LOCATION(NODE) \
(LOCATION_LOCUS ((OMP_CLAUSE_CHECK (NODE))->omp_clause.locus) \
!= UNKNOWN_LOCATION)
@ -1761,6 +1761,17 @@ class auto_suppress_location_wrappers
#define OMP_CLAUSE__CONDTEMP__ITER(NODE) \
(OMP_CLAUSE_SUBCODE_CHECK (NODE, OMP_CLAUSE__CONDTEMP_)->base.public_flag)
/* _SCANTEMP_ holding temporary with pointer to thread's local array;
allocation. */
#define OMP_CLAUSE__SCANTEMP__ALLOC(NODE) \
(OMP_CLAUSE_SUBCODE_CHECK (NODE, OMP_CLAUSE__SCANTEMP_)->base.public_flag)
/* _SCANTEMP_ holding temporary with a control variable for deallocation;
one boolean_type_node for test whether alloca was used, another one
to pass to __builtin_stack_restore or free. */
#define OMP_CLAUSE__SCANTEMP__CONTROL(NODE) \
TREE_PRIVATE (OMP_CLAUSE_SUBCODE_CHECK (NODE, OMP_CLAUSE__SCANTEMP_))
/* SSA_NAME accessors. */
/* Whether SSA_NAME NODE is a virtual operand. This simply caches the

19
libgomp/ChangeLog
View file

@ -1,3 +1,22 @@
2019-07-03 Jakub Jelinek <jakub@redhat.com>
* testsuite/libgomp.c++/scan-1.C: New test.
* testsuite/libgomp.c++/scan-2.C: New test.
* testsuite/libgomp.c++/scan-3.C: New test.
* testsuite/libgomp.c++/scan-4.C: New test.
* testsuite/libgomp.c++/scan-5.C: New test.
* testsuite/libgomp.c++/scan-6.C: New test.
* testsuite/libgomp.c++/scan-7.C: New test.
* testsuite/libgomp.c++/scan-8.C: New test.
* testsuite/libgomp.c/scan-1.c: New test.
* testsuite/libgomp.c/scan-2.c: New test.
* testsuite/libgomp.c/scan-3.c: New test.
* testsuite/libgomp.c/scan-4.c: New test.
* testsuite/libgomp.c/scan-5.c: New test.
* testsuite/libgomp.c/scan-6.c: New test.
* testsuite/libgomp.c/scan-7.c: New test.
* testsuite/libgomp.c/scan-8.c: New test.
2019-06-18 Thomas Schwinge <thomas@codesourcery.com>
* testsuite/libgomp.oacc-c++/firstprivate-mappings-1.C: New file.

151
libgomp/testsuite/libgomp.c++/scan-1.C Normal file
View file

@ -0,0 +1,151 @@
// { dg-require-effective-target size32plus }
extern "C" void abort ();
struct S {
inline S ();
inline ~S ();
inline S (const S &);
inline S & operator= (const S &);
int s;
};
S::S () : s (0)
{
}
S::~S ()
{
}
S::S (const S &x)
{
s = x.s;
}
S &
S::operator= (const S &x)
{
s = x.s;
return *this;
}
static inline void
ini (S &x)
{
x.s = 0;
}
S r, a[1024], b[1024];
#pragma omp declare reduction (+: S: omp_out.s += omp_in.s)
#pragma omp declare reduction (plus: S: omp_out.s += omp_in.s) initializer (ini (omp_priv))
__attribute__((noipa)) void
foo (S *a, S *b)
{
#pragma omp for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
r.s += a[i].s;
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) S
bar (void)
{
S s;
#pragma omp parallel
#pragma omp for reduction (inscan, plus:s)
for (int i = 0; i < 1024; i++)
{
s.s += 2 * a[i].s;
#pragma omp scan inclusive(s)
b[i] = s;
}
return S (s);
}
__attribute__((noipa)) void
baz (S *a, S *b)
{
#pragma omp parallel for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
r.s += a[i].s;
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) S
qux (void)
{
S s;
#pragma omp parallel for reduction (inscan, plus:s)
for (int i = 0; i < 1024; i++)
{
s.s += 2 * a[i].s;
#pragma omp scan inclusive(s)
b[i] = s;
}
return S (s);
}
int
main ()
{
S s;
for (int i = 0; i < 1024; ++i)
{
a[i].s = i;
b[i].s = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b);
if (r.s != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
s.s += i;
if (b[i].s != s.s)
abort ();
else
b[i].s = 25;
}
if (bar ().s != 1024 * 1023)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
s.s += 2 * i;
if (b[i].s != s.s)
abort ();
}
r.s = 0;
baz (a, b);
if (r.s != 1024 * 1023 / 2)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
s.s += i;
if (b[i].s != s.s)
abort ();
else
b[i].s = 25;
}
if (qux ().s != 1024 * 1023)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
s.s += 2 * i;
if (b[i].s != s.s)
abort ();
}
return 0;
}

116
libgomp/testsuite/libgomp.c++/scan-2.C Normal file
View file

@ -0,0 +1,116 @@
// { dg-require-effective-target size32plus }
extern "C" void abort ();
int r, a[1024], b[1024], q;
__attribute__((noipa)) void
foo (int *a, int *b, int &r)
{
#pragma omp for reduction (inscan, +:r) nowait
for (int i = 0; i < 1024; i++)
{
r += a[i];
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) int
bar (void)
{
int &s = q;
q = 0;
#pragma omp parallel
#pragma omp for reduction (inscan, +:s) nowait
for (int i = 0; i < 1024; i++)
{
s += 2 * a[i];
#pragma omp scan inclusive(s)
b[i] = s;
}
return s;
}
__attribute__((noipa)) void
baz (int *a, int *b, int &r)
{
#pragma omp parallel for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
r += a[i];
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) int
qux (void)
{
int &s = q;
q = 0;
#pragma omp parallel for reduction (inscan, +:s)
for (int i = 0; i < 1024; i++)
{
s += 2 * a[i];
#pragma omp scan inclusive(s)
b[i] = s;
}
return s;
}
int
main ()
{
int s = 0;
for (int i = 0; i < 1024; ++i)
{
a[i] = i;
b[i] = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b, r);
if (r != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
s += i;
if (b[i] != s)
abort ();
else
b[i] = 25;
}
if (bar () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += 2 * i;
if (b[i] != s)
abort ();
else
b[i] = -1;
}
r = 0;
baz (a, b, r);
if (r != 1024 * 1023 / 2)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += i;
if (b[i] != s)
abort ();
else
b[i] = -25;
}
if (qux () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += 2 * i;
if (b[i] != s)
abort ();
}
}

119
libgomp/testsuite/libgomp.c++/scan-3.C Normal file
View file

@ -0,0 +1,119 @@
// { dg-require-effective-target size32plus }
extern "C" void abort ();
int r, a[1024], b[1024], q;
#pragma omp declare reduction (foo: int: omp_out += omp_in) initializer (omp_priv = 0)
__attribute__((noipa)) void
foo (int *a, int *b, int &r)
{
#pragma omp for reduction (inscan, foo:r)
for (int i = 0; i < 1024; i++)
{
r += a[i];
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) int
bar (void)
{
int &s = q;
q = 0;
#pragma omp parallel
#pragma omp for reduction (inscan, foo:s)
for (int i = 0; i < 1024; i++)
{
s += 2 * a[i];
#pragma omp scan inclusive(s)
b[i] = s;
}
return s;
}
__attribute__((noipa)) void
baz (int *a, int *b, int &r)
{
#pragma omp parallel for reduction (inscan, foo:r)
for (int i = 0; i < 1024; i++)
{
r += a[i];
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) int
qux (void)
{
int &s = q;
q = 0;
#pragma omp parallel for reduction (inscan, foo:s)
for (int i = 0; i < 1024; i++)
{
s += 2 * a[i];
#pragma omp scan inclusive(s)
b[i] = s;
}
return s;
}
int
main ()
{
int s = 0;
for (int i = 0; i < 1024; ++i)
{
a[i] = i;
b[i] = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b, r);
if (r != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
s += i;
if (b[i] != s)
abort ();
else
b[i] = 25;
}
if (bar () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += 2 * i;
if (b[i] != s)
abort ();
else
b[i] = -1;
}
r = 0;
baz (a, b, r);
if (r != 1024 * 1023 / 2)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += i;
if (b[i] != s)
abort ();
else
b[i] = -25;
}
if (qux () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += 2 * i;
if (b[i] != s)
abort ();
}
return 0;
}

150
libgomp/testsuite/libgomp.c++/scan-4.C Normal file
View file

@ -0,0 +1,150 @@
// { dg-require-effective-target size32plus }
extern "C" void abort ();
struct S {
inline S ();
inline ~S ();
inline S (const S &);
inline S & operator= (const S &);
int s;
};
S::S () : s (0)
{
}
S::~S ()
{
}
S::S (const S &x)
{
s = x.s;
}
S &
S::operator= (const S &x)
{
s = x.s;
return *this;
}
static inline void
ini (S &x)
{
x.s = 0;
}
S r, a[1024], b[1024];
#pragma omp declare reduction (+: S: omp_out.s += omp_in.s)
#pragma omp declare reduction (plus: S: omp_out.s += omp_in.s) initializer (ini (omp_priv))
__attribute__((noipa)) void
foo (S *a, S *b, S &r)
{
#pragma omp for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
r.s += a[i].s;
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) S
bar ()
{
S s;
#pragma omp parallel
#pragma omp for reduction (inscan, plus:s)
for (int i = 0; i < 1024; i++)
{
s.s += 2 * a[i].s;
#pragma omp scan inclusive(s)
b[i] = s;
}
return s;
}
__attribute__((noipa)) void
baz (S *a, S *b, S &r)
{
#pragma omp parallel for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
r.s += a[i].s;
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) S
qux ()
{
S s;
#pragma omp parallel for reduction (inscan, plus:s)
for (int i = 0; i < 1024; i++)
{
s.s += 2 * a[i].s;
#pragma omp scan inclusive(s)
b[i] = s;
}
return s;
}
int
main ()
{
S s;
for (int i = 0; i < 1024; ++i)
{
a[i].s = i;
b[i].s = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b, r);
if (r.s != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
s.s += i;
if (b[i].s != s.s)
abort ();
else
b[i].s = 25;
}
if (bar ().s != 1024 * 1023)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
s.s += 2 * i;
if (b[i].s != s.s)
abort ();
}
r.s = 0;
baz (a, b, r);
if (r.s != 1024 * 1023 / 2)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
s.s += i;
if (b[i].s != s.s)
abort ();
else
b[i].s = 25;
}
if (qux ().s != 1024 * 1023)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
s.s += 2 * i;
if (b[i].s != s.s)
abort ();
}
}

158
libgomp/testsuite/libgomp.c++/scan-5.C Normal file
View file

@ -0,0 +1,158 @@
// { dg-require-effective-target size32plus }
extern "C" void abort ();
template <typename T>
struct S {
inline S ();
inline ~S ();
inline S (const S &);
inline S & operator= (const S &);
T s;
};
template <typename T>
S<T>::S () : s (0)
{
}
template <typename T>
S<T>::~S ()
{
}
template <typename T>
S<T>::S (const S &x)
{
s = x.s;
}
template <typename T>
S<T> &
S<T>::operator= (const S &x)
{
s = x.s;
return *this;
}
template <typename T>
static inline void
ini (S<T> &x)
{
x.s = 0;
}
S<int> r, a[1024], b[1024];
#pragma omp declare reduction (+: S<int>: omp_out.s += omp_in.s)
#pragma omp declare reduction (plus: S<int>: omp_out.s += omp_in.s) initializer (ini (omp_priv))
template <typename T>
__attribute__((noipa)) void
foo (S<T> *a, S<T> *b)
{
#pragma omp for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r.s += a[i].s;
}
}
template <typename T>
__attribute__((noipa)) S<T>
bar (void)
{
S<T> s;
#pragma omp parallel
#pragma omp for reduction (inscan, plus:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s.s += 2 * a[i].s;
}
return S<T> (s);
}
__attribute__((noipa)) void
baz (S<int> *a, S<int> *b)
{
#pragma omp parallel for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r.s += a[i].s;
}
}
__attribute__((noipa)) S<int>
qux (void)
{
S<int> s;
#pragma omp parallel for reduction (inscan, plus:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s.s += 2 * a[i].s;
}
return S<int> (s);
}
int
main ()
{
S<int> s;
for (int i = 0; i < 1024; ++i)
{
a[i].s = i;
b[i].s = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b);
if (r.s != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
else
b[i].s = 25;
s.s += i;
}
if (bar<int> ().s != 1024 * 1023)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
s.s += 2 * i;
}
r.s = 0;
baz (a, b);
if (r.s != 1024 * 1023 / 2)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
else
b[i].s = 25;
s.s += i;
}
if (qux ().s != 1024 * 1023)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
s.s += 2 * i;
}
}

120
libgomp/testsuite/libgomp.c++/scan-6.C Normal file
View file

@ -0,0 +1,120 @@
// { dg-require-effective-target size32plus }
extern "C" void abort ();
int r, a[1024], b[1024], q;
template <typename T, typename U>
__attribute__((noipa)) void
foo (T a, T b, U r)
{
#pragma omp for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r += a[i];
}
}
template <typename T>
__attribute__((noipa)) T
bar ()
{
T &s = q;
q = 0;
#pragma omp parallel
#pragma omp for reduction (inscan, +:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s += 2 * a[i];
}
return s;
}
template <typename T>
__attribute__((noipa)) void
baz (T *a, T *b, T &r)
{
#pragma omp parallel for reduction (inscan, +:r)
for (T i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r += a[i];
}
}
template <typename T>
__attribute__((noipa)) int
qux ()
{
T s = q;
q = 0;
#pragma omp parallel for reduction (inscan, +:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s += 2 * a[i];
}
return s;
}
int
main ()
{
int s = 0;
for (int i = 0; i < 1024; ++i)
{
a[i] = i;
b[i] = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo<int *, int &> (a, b, r);
if (r != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = 25;
s += i;
}
if (bar<int> () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = -1;
s += 2 * i;
}
r = 0;
baz<int> (a, b, r);
if (r != 1024 * 1023 / 2)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = -25;
s += i;
}
if (qux<int &> () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
s += 2 * i;
}
}

118
libgomp/testsuite/libgomp.c++/scan-7.C Normal file
View file

@ -0,0 +1,118 @@
// { dg-require-effective-target size32plus }
extern "C" void abort ();
int r, a[1024], b[1024], q;
#pragma omp declare reduction (foo: int: omp_out += omp_in) initializer (omp_priv = 0)
__attribute__((noipa)) void
foo (int *a, int *b, int &r)
{
#pragma omp for reduction (inscan, foo:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r += a[i];
}
}
__attribute__((noipa)) int
bar (void)
{
int &s = q;
q = 0;
#pragma omp parallel
#pragma omp for reduction (inscan, foo:s) nowait
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s += 2 * a[i];
}
return s;
}
__attribute__((noipa)) void
baz (int *a, int *b, int &r)
{
#pragma omp parallel for reduction (inscan, foo:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r += a[i];
}
}
__attribute__((noipa)) int
qux (void)
{
int &s = q;
q = 0;
#pragma omp parallel for reduction (inscan, foo:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s += 2 * a[i];
}
return s;
}
int
main ()
{
int s = 0;
for (int i = 0; i < 1024; ++i)
{
a[i] = i;
b[i] = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b, r);
if (r != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = 25;
s += i;
}
if (bar () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = -1;
s += 2 * i;
}
r = 0;
baz (a, b, r);
if (r != 1024 * 1023 / 2)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = -25;
s += i;
}
if (qux () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
s += 2 * i;
}
}

150
libgomp/testsuite/libgomp.c++/scan-8.C Normal file
View file

@ -0,0 +1,150 @@
// { dg-require-effective-target size32plus }
extern "C" void abort ();
struct S {
inline S ();
inline ~S ();
inline S (const S &);
inline S & operator= (const S &);
int s;
};
S::S () : s (0)
{
}
S::~S ()
{
}
S::S (const S &x)
{
s = x.s;
}
S &
S::operator= (const S &x)
{
s = x.s;
return *this;
}
static inline void
ini (S &x)
{
x.s = 0;
}
S r, a[1024], b[1024];
#pragma omp declare reduction (+: S: omp_out.s += omp_in.s)
#pragma omp declare reduction (plus: S: omp_out.s += omp_in.s) initializer (ini (omp_priv))
__attribute__((noipa)) void
foo (S *a, S *b, S &r)
{
#pragma omp for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r.s += a[i].s;
}
}
__attribute__((noipa)) S
bar (void)
{
S s;
#pragma omp parallel
#pragma omp for reduction (inscan, plus:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s.s += 2 * a[i].s;
}
return s;
}
__attribute__((noipa)) void
baz (S *a, S *b, S &r)
{
#pragma omp parallel for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r.s += a[i].s;
}
}
__attribute__((noipa)) S
qux (void)
{
S s;
#pragma omp parallel for reduction (inscan, plus:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s.s += 2 * a[i].s;
}
return s;
}
int
main ()
{
S s;
for (int i = 0; i < 1024; ++i)
{
a[i].s = i;
b[i].s = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b, r);
if (r.s != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
else
b[i].s = 25;
s.s += i;
}
if (bar ().s != 1024 * 1023)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
s.s += 2 * i;
}
r.s = 0;
baz (a, b, r);
if (r.s != 1024 * 1023 / 2)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
else
b[i].s = 25;
s.s += i;
}
if (qux ().s != 1024 * 1023)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
s.s += 2 * i;
}
}

115
libgomp/testsuite/libgomp.c/scan-1.c Normal file
View file

@ -0,0 +1,115 @@
/* { dg-require-effective-target size32plus } */
extern void abort (void);
int r, a[1024], b[1024];
__attribute__((noipa)) void
foo (int *a, int *b)
{
#pragma omp for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
r += a[i];
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) int
bar (void)
{
int s = 0;
#pragma omp parallel
#pragma omp for reduction (inscan, +:s)
for (int i = 0; i < 1024; i++)
{
s += 2 * a[i];
#pragma omp scan inclusive(s)
b[i] = s;
}
return s;
}
__attribute__((noipa)) void
baz (int *a, int *b)
{
#pragma omp parallel for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
r += a[i];
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) int
qux (void)
{
int s = 0;
#pragma omp parallel for reduction (inscan, +:s)
for (int i = 0; i < 1024; i++)
{
s += 2 * a[i];
#pragma omp scan inclusive(s)
b[i] = s;
}
return s;
}
int
main ()
{
int s = 0;
for (int i = 0; i < 1024; ++i)
{
a[i] = i;
b[i] = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b);
if (r != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
s += i;
if (b[i] != s)
abort ();
else
b[i] = 25;
}
if (bar () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += 2 * i;
if (b[i] != s)
abort ();
else
b[i] = -1;
}
r = 0;
baz (a, b);
if (r != 1024 * 1023 / 2)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += i;
if (b[i] != s)
abort ();
else
b[i] = -25;
}
if (qux () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += 2 * i;
if (b[i] != s)
abort ();
}
return 0;
}

117
libgomp/testsuite/libgomp.c/scan-2.c Normal file
View file

@ -0,0 +1,117 @@
/* { dg-require-effective-target size32plus } */
extern void abort (void);
int r, a[1024], b[1024];
#pragma omp declare reduction (foo: int: omp_out += omp_in) initializer (omp_priv = 0)
__attribute__((noipa)) void
foo (int *a, int *b)
{
#pragma omp for reduction (inscan, foo:r)
for (int i = 0; i < 1024; i++)
{
r += a[i];
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) int
bar (void)
{
int s = 0;
#pragma omp parallel
#pragma omp for reduction (inscan, foo:s)
for (int i = 0; i < 1024; i++)
{
s += 2 * a[i];
#pragma omp scan inclusive(s)
b[i] = s;
}
return s;
}
__attribute__((noipa)) void
baz (int *a, int *b)
{
#pragma omp parallel for reduction (inscan, foo:r)
for (int i = 0; i < 1024; i++)
{
r += a[i];
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) int
qux (void)
{
int s = 0;
#pragma omp parallel for reduction (inscan, foo:s)
for (int i = 0; i < 1024; i++)
{
s += 2 * a[i];
#pragma omp scan inclusive(s)
b[i] = s;
}
return s;
}
int
main ()
{
int s = 0;
for (int i = 0; i < 1024; ++i)
{
a[i] = i;
b[i] = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b);
if (r != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
s += i;
if (b[i] != s)
abort ();
else
b[i] = 25;
}
if (bar () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += 2 * i;
if (b[i] != s)
abort ();
else
b[i] = -1;
}
r = 0;
baz (a, b);
if (r != 1024 * 1023 / 2)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += i;
if (b[i] != s)
abort ();
else
b[i] = -25;
}
if (qux () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += 2 * i;
if (b[i] != s)
abort ();
}
return 0;
}

88
libgomp/testsuite/libgomp.c/scan-3.c Normal file
View file

@ -0,0 +1,88 @@
/* { dg-require-effective-target size32plus } */
extern void abort (void);
float r = 1.0f, a[1024], b[1024];
__attribute__((noipa)) void
foo (float *a, float *b)
{
#pragma omp for reduction (inscan, *:r)
for (int i = 0; i < 1024; i++)
{
r *= a[i];
#pragma omp scan inclusive(r)
b[i] = r;
}
}
__attribute__((noipa)) float
bar (void)
{
float s = -__builtin_inff ();
#pragma omp parallel for reduction (inscan, max:s)
for (int i = 0; i < 1024; i++)
{
s = s > a[i] ? s : a[i];
#pragma omp scan inclusive(s)
b[i] = s;
}
return s;
}
int
main ()
{
float s = 1.0f;
for (int i = 0; i < 1024; ++i)
{
if (i < 80)
a[i] = (i & 1) ? 0.25f : 0.5f;
else if (i < 200)
a[i] = (i % 3) == 0 ? 2.0f : (i % 3) == 1 ? 4.0f : 1.0f;
else if (i < 280)
a[i] = (i & 1) ? 0.25f : 0.5f;
else if (i < 380)
a[i] = (i % 3) == 0 ? 2.0f : (i % 3) == 1 ? 4.0f : 1.0f;
else
switch (i % 6)
{
case 0: a[i] = 0.25f; break;
case 1: a[i] = 2.0f; break;
case 2: a[i] = -1.0f; break;
case 3: a[i] = -4.0f; break;
case 4: a[i] = 0.5f; break;
case 5: a[i] = 1.0f; break;
default: a[i] = 0.0f; break;
}
b[i] = -19.0f;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b);
if (r * 16384.0f != 0.125f)
abort ();
float m = -175.25f;
for (int i = 0; i < 1024; ++i)
{
s *= a[i];
if (b[i] != s)
abort ();
else
{
a[i] = m - ((i % 3) == 1 ? 2.0f : (i % 3) == 2 ? 4.0f : 0.0f);
b[i] = -231.75f;
m += 0.75f;
}
}
if (bar () != 592.0f)
abort ();
s = -__builtin_inff ();
for (int i = 0; i < 1024; ++i)
{
if (s < a[i])
s = a[i];
if (b[i] != s)
abort ();
}
return 0;
}

179
libgomp/testsuite/libgomp.c/scan-4.c Normal file
View file

@ -0,0 +1,179 @@
/* { dg-require-effective-target size32plus } */
extern void abort (void);
int r, a[1024], b[1024];
unsigned short r2, b2[1024];
unsigned char r3, b3[1024];
__attribute__((noipa)) void
foo (int *a, int *b, unsigned short *b2, unsigned char *b3)
{
#pragma omp for reduction (inscan, +:r, r2, r3)
for (int i = 0; i < 1024; i++)
{
{ r += a[i]; r2 += a[i]; r3 += a[i]; }
#pragma omp scan inclusive(r, r2, r3)
{
b[i] = r;
b2[i] = r2;
b3[i] = r3;
}
}
}
__attribute__((noipa)) int
bar (unsigned short *s2p, unsigned char *s3p)
{
int s = 0;
unsigned short s2 = 0;
unsigned char s3 = 0;
#pragma omp parallel
#pragma omp for reduction (inscan, +:s, s2, s3)
for (int i = 0; i < 1024; i++)
{
{
s += 2 * a[i];
s2 += 2 * a[i];
s3 += 2 * a[i];
}
#pragma omp scan inclusive(s, s2, s3)
{ b[i] = s; b2[i] = s2; b3[i] = s3; }
}
*s2p = s2;
*s3p = s3;
return s;
}
__attribute__((noipa)) void
baz (int *a, int *b, unsigned short *b2, unsigned char *b3)
{
#pragma omp parallel for reduction (inscan, +:r, r2, r3)
for (int i = 0; i < 1024; i++)
{
{
r += a[i];
r2 += a[i];
r3 += a[i];
}
#pragma omp scan inclusive(r, r2, r3)
{
b[i] = r;
b2[i] = r2;
b3[i] = r3;
}
}
}
__attribute__((noipa)) int
qux (unsigned short *s2p, unsigned char *s3p)
{
int s = 0;
unsigned short s2 = 0;
unsigned char s3 = 0;
#pragma omp parallel for reduction (inscan, +:s, s2, s3)
for (int i = 0; i < 1024; i++)
{
{ s += 2 * a[i]; s2 += 2 * a[i]; s3 += 2 * a[i]; }
#pragma omp scan inclusive(s, s2, s3)
{ b[i] = s; b2[i] = s2; b3[i] = s3; }
}
*s2p = s2;
*s3p = s3;
return s;
}
int
main ()
{
int s = 0;
unsigned short s2;
unsigned char s3;
for (int i = 0; i < 1024; ++i)
{
a[i] = i;
b[i] = -1;
b2[i] = -1;
b3[i] = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b, b2, b3);
if (r != 1024 * 1023 / 2
|| r2 != (unsigned short) r
|| r3 != (unsigned char) r)
abort ();
for (int i = 0; i < 1024; ++i)
{
s += i;
if (b[i] != s
|| b2[i] != (unsigned short) s
|| b3[i] != (unsigned char) s)
abort ();
else
{
b[i] = 25;
b2[i] = 24;
b3[i] = 26;
}
}
if (bar (&s2, &s3) != 1024 * 1023)
abort ();
if (s2 != (unsigned short) (1024 * 1023)
|| s3 != (unsigned char) (1024 * 1023))
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += 2 * i;
if (b[i] != s
|| b2[i] != (unsigned short) s
|| b3[i] != (unsigned char) s)
abort ();
else
{
b[i] = -1;
b2[i] = -1;
b3[i] = -1;
}
}
r = 0;
r2 = 0;
r3 = 0;
baz (a, b, b2, b3);
if (r != 1024 * 1023 / 2
|| r2 != (unsigned short) r
|| r3 != (unsigned char) r)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += i;
if (b[i] != s
|| b2[i] != (unsigned short) s
|| b3[i] != (unsigned char) s)
abort ();
else
{
b[i] = 25;
b2[i] = 24;
b3[i] = 26;
}
}
s2 = 0;
s3 = 0;
if (qux (&s2, &s3) != 1024 * 1023)
abort ();
if (s2 != (unsigned short) (1024 * 1023)
|| s3 != (unsigned char) (1024 * 1023))
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
s += 2 * i;
if (b[i] != s
|| b2[i] != (unsigned short) s
|| b3[i] != (unsigned char) s)
abort ();
}
return 0;
}

115
libgomp/testsuite/libgomp.c/scan-5.c Normal file
View file

@ -0,0 +1,115 @@
/* { dg-require-effective-target size32plus } */
extern void abort (void);
int r, a[1024], b[1024];
__attribute__((noipa)) void
foo (int *a, int *b)
{
#pragma omp for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r += a[i];
}
}
__attribute__((noipa)) int
bar (void)
{
int s = 0;
#pragma omp parallel
#pragma omp for reduction (inscan, +:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s += 2 * a[i];
}
return s;
}
__attribute__((noipa)) void
baz (int *a, int *b)
{
#pragma omp parallel for reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r += a[i];
}
}
__attribute__((noipa)) int
qux (void)
{
int s = 0;
#pragma omp parallel for reduction (inscan, +:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s += 2 * a[i];
}
return s;
}
int
main ()
{
int s = 0;
for (int i = 0; i < 1024; ++i)
{
a[i] = i;
b[i] = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b);
if (r != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = 25;
s += i;
}
if (bar () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = -1;
s += 2 * i;
}
r = 0;
baz (a, b);
if (r != 1024 * 1023 / 2)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = -25;
s += i;
}
if (qux () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
s += 2 * i;
}
return 0;
}

117
libgomp/testsuite/libgomp.c/scan-6.c Normal file
View file

@ -0,0 +1,117 @@
/* { dg-require-effective-target size32plus } */
extern void abort (void);
int r, a[1024], b[1024];
#pragma omp declare reduction (foo: int: omp_out += omp_in) initializer (omp_priv = 0)
__attribute__((noipa)) void
foo (int *a, int *b)
{
#pragma omp for reduction (inscan, foo:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r += a[i];
}
}
__attribute__((noipa)) int
bar (void)
{
int s = 0;
#pragma omp parallel
#pragma omp for reduction (inscan, foo:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s += 2 * a[i];
}
return s;
}
__attribute__((noipa)) void
baz (int *a, int *b)
{
#pragma omp parallel for reduction (inscan, foo:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r += a[i];
}
}
__attribute__((noipa)) int
qux (void)
{
int s = 0;
#pragma omp parallel for reduction (inscan, foo:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s += 2 * a[i];
}
return s;
}
int
main ()
{
int s = 0;
for (int i = 0; i < 1024; ++i)
{
a[i] = i;
b[i] = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b);
if (r != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = 25;
s += i;
}
if (bar () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = -1;
s += 2 * i;
}
r = 0;
baz (a, b);
if (r != 1024 * 1023 / 2)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = -25;
s += i;
}
if (qux () != 1024 * 1023)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
s += 2 * i;
}
return 0;
}

86
libgomp/testsuite/libgomp.c/scan-7.c Normal file
View file

@ -0,0 +1,86 @@
/* { dg-require-effective-target size32plus } */
extern void abort (void);
float r = 1.0f, a[1024], b[1024];
__attribute__((noipa)) void
foo (float *a, float *b)
{
#pragma omp for reduction (inscan, *:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r *= a[i];
}
}
__attribute__((noipa)) float
bar (void)
{
float s = -__builtin_inff ();
#pragma omp parallel for reduction (inscan, max:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s = s > a[i] ? s : a[i];
}
return s;
}
int
main ()
{
float s = 1.0f;
for (int i = 0; i < 1024; ++i)
{
if (i < 80)
a[i] = (i & 1) ? 0.25f : 0.5f;
else if (i < 200)
a[i] = (i % 3) == 0 ? 2.0f : (i % 3) == 1 ? 4.0f : 1.0f;
else if (i < 280)
a[i] = (i & 1) ? 0.25f : 0.5f;
else if (i < 380)
a[i] = (i % 3) == 0 ? 2.0f : (i % 3) == 1 ? 4.0f : 1.0f;
else
switch (i % 6)
{
case 0: a[i] = 0.25f; break;
case 1: a[i] = 2.0f; break;
case 2: a[i] = -1.0f; break;
case 3: a[i] = -4.0f; break;
case 4: a[i] = 0.5f; break;
case 5: a[i] = 1.0f; break;
default: a[i] = 0.0f; break;
}
b[i] = -19.0f;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b);
if (r * 16384.0f != 0.125f)
abort ();
float m = -175.25f;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
else
b[i] = -231.75f;
s *= a[i];
a[i] = m - ((i % 3) == 1 ? 2.0f : (i % 3) == 2 ? 4.0f : 0.0f);
m += 0.75f;
}
if (bar () != 592.0f)
abort ();
s = -__builtin_inff ();
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s)
abort ();
if (s < a[i])
s = a[i];
}
return 0;
}

179
libgomp/testsuite/libgomp.c/scan-8.c Normal file
View file

@ -0,0 +1,179 @@
/* { dg-require-effective-target size32plus } */
extern void abort (void);
int r, a[1024], b[1024];
unsigned short r2, b2[1024];
unsigned char r3, b3[1024];
__attribute__((noipa)) void
foo (int *a, int *b, unsigned short *b2, unsigned char *b3)
{
#pragma omp for reduction (inscan, +:r, r2, r3)
for (int i = 0; i < 1024; i++)
{
{
b[i] = r;
b2[i] = r2;
b3[i] = r3;
}
#pragma omp scan exclusive(r, r2, r3)
{ r += a[i]; r2 += a[i]; r3 += a[i]; }
}
}
__attribute__((noipa)) int
bar (unsigned short *s2p, unsigned char *s3p)
{
int s = 0;
unsigned short s2 = 0;
unsigned char s3 = 0;
#pragma omp parallel
#pragma omp for reduction (inscan, +:s, s2, s3)
for (int i = 0; i < 1024; i++)
{
{ b[i] = s; b2[i] = s2; b3[i] = s3; }
#pragma omp scan exclusive(s, s2, s3)
{
s += 2 * a[i];
s2 += 2 * a[i];
s3 += 2 * a[i];
}
}
*s2p = s2;
*s3p = s3;
return s;
}
__attribute__((noipa)) void
baz (int *a, int *b, unsigned short *b2, unsigned char *b3)
{
#pragma omp parallel for reduction (inscan, +:r, r2, r3)
for (int i = 0; i < 1024; i++)
{
{
b[i] = r;
b2[i] = r2;
b3[i] = r3;
}
#pragma omp scan exclusive(r, r2, r3)
{
r += a[i];
r2 += a[i];
r3 += a[i];
}
}
}
__attribute__((noipa)) int
qux (unsigned short *s2p, unsigned char *s3p)
{
int s = 0;
unsigned short s2 = 0;
unsigned char s3 = 0;
#pragma omp parallel for reduction (inscan, +:s, s2, s3)
for (int i = 0; i < 1024; i++)
{
{ b[i] = s; b2[i] = s2; b3[i] = s3; }
#pragma omp scan exclusive(s, s2, s3)
{ s += 2 * a[i]; s2 += 2 * a[i]; s3 += 2 * a[i]; }
}
*s2p = s2;
*s3p = s3;
return s;
}
int
main ()
{
int s = 0;
unsigned short s2;
unsigned char s3;
for (int i = 0; i < 1024; ++i)
{
a[i] = i;
b[i] = -1;
b2[i] = -1;
b3[i] = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b, b2, b3);
if (r != 1024 * 1023 / 2
|| r2 != (unsigned short) r
|| r3 != (unsigned char) r)
abort ();
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s
|| b2[i] != (unsigned short) s
|| b3[i] != (unsigned char) s)
abort ();
else
{
b[i] = 25;
b2[i] = 24;
b3[i] = 26;
}
s += i;
}
if (bar (&s2, &s3) != 1024 * 1023)
abort ();
if (s2 != (unsigned short) (1024 * 1023)
|| s3 != (unsigned char) (1024 * 1023))
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s
|| b2[i] != (unsigned short) s
|| b3[i] != (unsigned char) s)
abort ();
else
{
b[i] = -1;
b2[i] = -1;
b3[i] = -1;
}
s += 2 * i;
}
r = 0;
r2 = 0;
r3 = 0;
baz (a, b, b2, b3);
if (r != 1024 * 1023 / 2
|| r2 != (unsigned short) r
|| r3 != (unsigned char) r)
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s
|| b2[i] != (unsigned short) s
|| b3[i] != (unsigned char) s)
abort ();
else
{
b[i] = 25;
b2[i] = 24;
b3[i] = 26;
}
s += i;
}
s2 = 0;
s3 = 0;
if (qux (&s2, &s3) != 1024 * 1023)
abort ();
if (s2 != (unsigned short) (1024 * 1023)
|| s3 != (unsigned char) (1024 * 1023))
abort ();
s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i] != s
|| b2[i] != (unsigned short) s
|| b3[i] != (unsigned char) s)
abort ();
s += 2 * i;
}
return 0;
}