2017-01-27 Cesar Philippidis <cesar@codesourcery.com>
gcc/fortran/
* trans-openmp.c (gfc_omp_finish_clause): Use GOMP_MAP_POINTER
for POINTER_TYPE decls.
(gfc_trans_omp_clauses_1): Likewise.
gcc/
* gimplify.c (demote_firstprivate_pointer): New function.
(gimplify_scan_omp_clauses): Enable target_map_pointers_as_0len_arrays
and target_map_scalars_firstprivate in OpenACC and gfortran.
(gimplify_adjust_omp_clauses): Demote FIRSTPRIVATE_POINTERS for OpenACC
retuction variables.
* omp-low.c (lower_omp_target): Adjust receiver reference of decls for
fortran dummy arguments.
gcc/testsuite/
* gfortran.dg/goacc/kernels-loop-n.f95: Xfail test.
libgomp/
* testsuite/libgomp.oacc-fortran/deviceptr-1.f90: Add -foffload-force.
* testsuite/libgomp.oacc-fortran/non-scalar-data.f90: Likewise.
@@ -1042,7 +1042,7 @@ gfc_omp_finish_clause (tree c, gimple_seq *pre_p)
return;
tree orig_decl = decl;
c4 = build_omp_clause (OMP_CLAUSE_LOCATION (c), OMP_CLAUSE_MAP);
- OMP_CLAUSE_SET_MAP_KIND (c4, GOMP_MAP_POINTER);
+ OMP_CLAUSE_SET_MAP_KIND (c4, GOMP_MAP_FIRSTPRIVATE_POINTER);
OMP_CLAUSE_DECL (c4) = decl;
OMP_CLAUSE_SIZE (c4) = size_int (0);
decl = build_fold_indirect_ref (decl);
@@ -2005,9 +2005,12 @@ gfc_trans_omp_clauses_1 (stmtblock_t *block, gfc_omp_clauses *clauses,
(TREE_TYPE (TREE_TYPE (decl)))))
{
tree orig_decl = decl;
+ enum gomp_map_kind gmk = GOMP_MAP_FIRSTPRIVATE_POINTER;
+ if (n->u.map_op == OMP_MAP_FORCE_DEVICEPTR)
+ gmk = GOMP_MAP_POINTER;
node4 = build_omp_clause (input_location,
OMP_CLAUSE_MAP);
- OMP_CLAUSE_SET_MAP_KIND (node4, GOMP_MAP_POINTER);
+ OMP_CLAUSE_SET_MAP_KIND (node4, gmk);
OMP_CLAUSE_DECL (node4) = decl;
OMP_CLAUSE_SIZE (node4) = size_int (0);
decl = build_fold_indirect_ref (decl);
@@ -6589,6 +6589,37 @@ find_decl_expr (tree *tp, int *walk_subtrees, void *data)
return NULL_TREE;
}
+static void
+demote_firstprivate_pointer (tree decl, gimplify_omp_ctx *ctx)
+{
+ if (!lang_GNU_Fortran ())
+ return;
+
+ while (ctx)
+ {
+ if (ctx->region_type == ORT_ACC_PARALLEL
+ || ctx->region_type == ORT_ACC_KERNELS)
+ break;
+ ctx = ctx->outer_context;
+ }
+
+ if (ctx == NULL)
+ return;
+
+ tree clauses = ctx->clauses;
+
+ for (tree c = clauses; c; c = OMP_CLAUSE_CHAIN (c))
+ {
+ if (OMP_CLAUSE_CODE (c) == OMP_CLAUSE_MAP
+ && OMP_CLAUSE_MAP_KIND (c) == GOMP_MAP_FIRSTPRIVATE_POINTER
+ && OMP_CLAUSE_DECL (c) == decl)
+ {
+ OMP_CLAUSE_SET_MAP_KIND (c, GOMP_MAP_POINTER);
+ return;
+ }
+ }
+}
+
/* Scan the OMP clauses in *LIST_P, installing mappings into a new
and previous omp contexts. */
@@ -6605,11 +6636,10 @@ gimplify_scan_omp_clauses (tree *list_p, gimple_seq *pre_p,
ctx = new_omp_context (region_type);
ctx->clauses = *list_p;
outer_ctx = ctx->outer_context;
- if (code == OMP_TARGET && !lang_GNU_Fortran ())
+ if (code == OMP_TARGET && !(lang_GNU_Fortran () && !(region_type & ORT_ACC)))
{
- ctx->target_map_pointers_as_0len_arrays = true;
- /* FIXME: For Fortran we want to set this too, when
- the Fortran FE is updated to OpenMP 4.5. */
+ if (!lang_GNU_Fortran () || region_type & ORT_ACC)
+ ctx->target_map_pointers_as_0len_arrays = true;
ctx->target_map_scalars_firstprivate = true;
}
if (!lang_GNU_Fortran ())
@@ -6717,6 +6747,7 @@ gimplify_scan_omp_clauses (tree *list_p, gimple_seq *pre_p,
if (!(region_type & ORT_ACC))
check_non_private = "reduction";
decl = OMP_CLAUSE_DECL (c);
+ demote_firstprivate_pointer (decl, ctx->outer_context);
if (TREE_CODE (decl) == MEM_REF)
{
tree type = TREE_TYPE (decl);
@@ -8254,11 +8285,16 @@ gimplify_adjust_omp_clauses (gimple_seq *pre_p, gimple_seq body, tree *list_p,
&& kind != GOMP_MAP_FORCE_PRESENT
&& kind != GOMP_MAP_POINTER)
{
- warning_at (OMP_CLAUSE_LOCATION (c), 0,
- "incompatible data clause with reduction "
- "on %qE; promoting to present_or_copy",
- DECL_NAME (t));
- OMP_CLAUSE_SET_MAP_KIND (c, GOMP_MAP_TOFROM);
+ if (lang_hooks.decls.omp_privatize_by_reference (decl))
+ OMP_CLAUSE_SET_MAP_KIND (c, GOMP_MAP_POINTER);
+ else
+ {
+ warning_at (OMP_CLAUSE_LOCATION (c), 0,
+ "incompatible data clause with reduction "
+ "on %qE; promoting to present_or_copy",
+ DECL_NAME (t));
+ OMP_CLAUSE_SET_MAP_KIND (c, GOMP_MAP_TOFROM);
+ }
}
}
}
@@ -17547,7 +17547,8 @@ lower_omp_target (gimple_stmt_iterator *gsi_p, omp_context *ctx)
}
else
is_ref = is_reference (var);
- if (OMP_CLAUSE_MAP_KIND (c) == GOMP_MAP_FIRSTPRIVATE_REFERENCE)
+ if (OMP_CLAUSE_MAP_KIND (c) == GOMP_MAP_FIRSTPRIVATE_REFERENCE
+ || (lang_GNU_Fortran () && TREE_CODE (var) == PARM_DECL))
is_ref = false;
bool ref_to_array = false;
if (is_ref)
@@ -37,4 +37,6 @@ end module test
! Check that the loop has been split off into a function.
! { dg-final { scan-tree-dump-times "(?n);; Function __test_MOD_foo._omp_fn.0 " 1 "optimized" } }
-! { dg-final { scan-tree-dump-times "(?n)oacc function \\(0," 1 "parloops1" } }
+! This failure was introduced with the GOMP_MAP_POINTER ->
+! GOMP_MAP_FIRSTPRIVATE_POINTER conversion.
+! { dg-final { scan-tree-dump-times "(?n)oacc function \\(0," 1 "parloops1" { xfail *-*-* } } }
@@ -3,6 +3,7 @@
! the deviceptr variable is implied.
! { dg-do run }
+! { dg-additional-options "-foffload-force" }
subroutine subr1 (a, b)
implicit none
@@ -2,6 +2,7 @@
! offloaded regions are properly mapped using present_or_copy.
! { dg-do run }
+! { dg-additional-options "-foffload-force" }
program main
implicit none
This patch partially enables GOMP_MAP_FIRSTPRIVATE_POINTER in gfortran. gfortran still falls back to GOMP_MAP_POINTER for arrays with descriptors and derived types. The limitation on derived types is there because we don't have much test coverage for it, and this patch series was more exploratory for performance enhancements. With that in mind, there are a couple of shortcomings with this patch. 1) Dummy reduction variables fallback to GOMP_MAP_POINTER because of a pointer dereferencing bug. The state of debugging such problems on PTX targets leaves something to be desired, especially since print isn't working on nvptx targets currently. 2) Apparently, firstprivate pointers negatively affects the alias analysis used by ACC KERNELS and parloops, so a couple of more execution tests fail to generate offloaded code. I plan to resolve issue 1) in a follow up patch later on (but maybe not in the immediate future). Regarding 2), ACC KERNELS are eventually going to need a significant rework, but that's not going to happen in the near future either. I've been pushing to get the performance of ACC PARALLEL regions on par to other OpenACC compilers first, and hopefully that won't be too far way. With this patch, I'm observing an approximate 0.6s reduction in CloverLeaf's original 0.9s execution time (it takes approximate 0.9s after the GOMP_MAP_FIRSTPRIVATE_INT and GOMP_MAP_TO_PSET patches), to yield a final execution time somewhere in the neighborhood of 0.3s. That's about a one second savings from the unpatched version of GCC. There's still quite a bit of room left for improvement, with respect to data movement. Specifically, the nvptx runtime populates the struct containing the addresses of all of the remapped offloaded variables, and uploads it immediately before launching the PTX kernel. Clearly this works, however is requires a synchronous host2dev data copy. And while those synchronous barriers are generally quick, if you have thousands of them, like CloverLeaf, they end up becoming a bottleneck. One way to get around this is to pass in the remapped decl pointers as arguments to cuLaunchKernel directly. Maybe we can do that during oaccdevlow for nvptx targets. The next big thing, however, probably should be updating the default launch geometry again. This patch has been committed to gomp-4_0-branch. Cesar