For partitioned loops, we're currently calling library functions (in
libgcc) to determine the cardinality of the set of threads a particular
loop is distributed over (given a set of gang/worker/vector toggles),
and the index of the current thread within that set.
This patch reimplements those two functions in terms of the
(PTX-specific!) builtins that Bernd has recently added in order to
implement vector-single/worker-single predication, which expand
directly to machine instructions on the target (or to constant zero/one
on the host). It also makes use of the same "gwv" bitfields that are set
up by that new code.
The previous BUILT_IN_GOACC_GET_THREAD_NUM and
BUILT_IN_GOACC_GET_NUM_THREADS builtins are removed entirely.
This works reasonably well, but there are some regressions caused by
middle-end optimisers having extra freedom to manipulate the CFG in
ways that PTX cannot support without the "optimisation barrier" of the
calls to the thread builtins being present. This will be addressed by a
follow-on patch.
Pre-approved for gomp4, but I'll wait for comments on the follow-on
patch before applying so as not to leave the branch in a "broken" state.
Thanks,
Julian
ChangeLog
gcc/
* builtins.c (expand_oacc_builtin): Return const1_rtx for
ntid/nctaid builtins when the associated patterns are not present.
* omp-builtins.def (BUILT_IN_GOACC_GET_THREAD_NUM)
(BUILT_IN_GOACC_GET_NUM_THREADS): Remove.
* omp-low.c (struct omp_for_data): Remove gang, worker, vector
fields.
(extract_omp_for_data): Don't initialise deleted gang, worker,
vector fields.
(expand_oacc_get_num_threads, expand_oacc_get_thread_num): New
functions.
(lower_reduction_clauses): Use above functions.
(expand_omp_for_static_nochunk): Likewise.
(expand_omp_for_static_chunk): Likewise.
commit 1be8ada44a9f91d2eba16ef1f81243707647f237
Author: Julian Brown <jul...@codesourcery.com>
Date: Fri May 15 03:20:42 2015 -0700
Inlined OpenACC thread builtins.
diff --git a/gcc/builtins.c b/gcc/builtins.c
index ebd4b4a..cd51821 100644
--- a/gcc/builtins.c
+++ b/gcc/builtins.c
@@ -5964,8 +5964,8 @@ expand_oacc_builtin (enum built_in_function fcode, tree exp, rtx target)
case BUILT_IN_GOACC_NTID:
#ifdef HAVE_oacc_ntid
icode = CODE_FOR_oacc_ntid;
- result = const1_rtx;
#endif
+ result = const1_rtx;
break;
case BUILT_IN_GOACC_TID:
#ifdef HAVE_oacc_tid
@@ -5975,8 +5975,8 @@ expand_oacc_builtin (enum built_in_function fcode, tree exp, rtx target)
case BUILT_IN_GOACC_NCTAID:
#ifdef HAVE_oacc_nctaid
icode = CODE_FOR_oacc_nctaid;
- result = const1_rtx;
#endif
+ result = const1_rtx;
break;
case BUILT_IN_GOACC_CTAID:
#ifdef HAVE_oacc_ctaid
diff --git a/gcc/omp-builtins.def b/gcc/omp-builtins.def
index ac1f802..47d9e45 100644
--- a/gcc/omp-builtins.def
+++ b/gcc/omp-builtins.def
@@ -69,10 +69,6 @@ DEF_GOACC_BUILTIN (BUILT_IN_GOACC_NCTAID, "GOACC_nctaid",
BT_FN_UINT_UINT, ATTR_CONST_NOTHROW_LEAF_LIST)
DEF_GOACC_BUILTIN (BUILT_IN_GOACC_CTAID, "GOACC_ctaid",
BT_FN_UINT_UINT, ATTR_CONST_NOTHROW_LEAF_LIST)
-DEF_GOACC_BUILTIN (BUILT_IN_GOACC_GET_THREAD_NUM, "GOACC_get_thread_num",
- BT_FN_INT_INT_INT_INT, ATTR_NOTHROW_LEAF_LIST)
-DEF_GOACC_BUILTIN (BUILT_IN_GOACC_GET_NUM_THREADS, "GOACC_get_num_threads",
- BT_FN_INT_INT_INT_INT, ATTR_NOTHROW_LEAF_LIST)
DEF_GOACC_BUILTIN (BUILT_IN_GOACC_GET_GANGLOCAL_PTR, "GOACC_get_ganglocal_ptr",
BT_FN_PTR, ATTR_NOTHROW_LEAF_LIST)
DEF_GOACC_BUILTIN (BUILT_IN_GOACC_DEVICEPTR, "GOACC_deviceptr",
diff --git a/gcc/omp-low.c b/gcc/omp-low.c
index b114887..f82247b 100644
--- a/gcc/omp-low.c
+++ b/gcc/omp-low.c
@@ -263,7 +263,6 @@ struct omp_for_data
tree chunk_size;
gomp_for *for_stmt;
tree pre, iter_type;
- tree gang, worker, vector;
int collapse;
bool have_nowait, have_ordered;
enum omp_clause_schedule_kind sched_kind;
@@ -749,16 +748,6 @@ extract_omp_for_data (gomp_for *for_stmt, struct omp_for_data *fd,
gcc_assert (fd->chunk_size == NULL_TREE);
fd->chunk_size = build_int_cst (TREE_TYPE (fd->loop.v), 1);
}
-
- /* Extract the OpenACC gang, worker and vector clauses. */
- t = find_omp_clause (gimple_omp_for_clauses (for_stmt), OMP_CLAUSE_GANG);
- fd->gang = (t == NULL_TREE) ? integer_zero_node : integer_one_node;
-
- t = find_omp_clause (gimple_omp_for_clauses (for_stmt), OMP_CLAUSE_WORKER);
- fd->worker = (t == NULL_TREE) ? integer_zero_node : integer_one_node;
-
- t = find_omp_clause (gimple_omp_for_clauses (for_stmt), OMP_CLAUSE_VECTOR);
- fd->vector = (t == NULL_TREE) ? integer_zero_node : integer_one_node;
}
@@ -4919,6 +4908,159 @@ is_atomic_compatible_reduction (tree var, omp_context *ctx)
return true;
}
+
+/* Find the total number of threads used by a region partitioned by
+ GWV_BITS. Setup code required for the calculation is added to SEQ. Note
+ that this is currently used from both OMP-lowering and OMP-expansion phases,
+ and uses builtins specific to NVidia PTX: this will need refactoring into a
+ generic interface when support for other targets is added. */
+
+static tree
+expand_oacc_get_num_threads (gimple_seq *seq, int gwv_bits)
+{
+ tree res = NULL_TREE;
+ tree u0 = fold_convert (unsigned_type_node, integer_zero_node);
+ tree u1 = fold_convert (unsigned_type_node, integer_one_node);
+
+ if (gwv_bits & MASK_GANG)
+ {
+ tree decl = builtin_decl_explicit (BUILT_IN_GOACC_NCTAID);
+ tree gang_count = create_tmp_var (unsigned_type_node);
+ gimple call = gimple_build_call (decl, 1, u0);
+ gimple_call_set_lhs (call, gang_count);
+ gimple_seq_add_stmt (seq, call);
+ res = gang_count;
+ }
+
+ if (gwv_bits & MASK_WORKER)
+ {
+ tree decl = builtin_decl_explicit (BUILT_IN_GOACC_NTID);
+ tree worker_count = create_tmp_var (unsigned_type_node);
+ gimple call = gimple_build_call (decl, 1, u1);
+ gimple_call_set_lhs (call, worker_count);
+ gimple_seq_add_stmt (seq, call);
+ if (res != NULL_TREE)
+ res = fold_build2 (MULT_EXPR, unsigned_type_node, res, worker_count);
+ else
+ res = worker_count;
+ }
+
+ if (gwv_bits & MASK_VECTOR)
+ {
+ tree decl = builtin_decl_explicit (BUILT_IN_GOACC_NTID);
+ tree vector_length = create_tmp_var (unsigned_type_node);
+ gimple call = gimple_build_call (decl, 1, u0);
+ gimple_call_set_lhs (call, vector_length);
+ gimple_seq_add_stmt (seq, call);
+ if (res != NULL_TREE)
+ res = fold_build2 (MULT_EXPR, unsigned_type_node, res, vector_length);
+ else
+ res = vector_length;
+ }
+
+ if (res == NULL_TREE)
+ res = u1;
+
+ return res;
+}
+
+
+/* Find the current thread number to use within a region partitioned by
+ GWV_BITS. Setup code required for the calculation is added to SEQ. See
+ note for expand_oacc_get_num_threads above re: builtin usage. */
+
+static tree
+expand_oacc_get_thread_num (gimple_seq *seq, int gwv_bits)
+{
+ tree res = NULL_TREE;
+ tree u0 = fold_convert (unsigned_type_node, integer_zero_node);
+ tree u1 = fold_convert (unsigned_type_node, integer_one_node);
+ tree vector_count = NULL_TREE;
+ tree tid_decl = builtin_decl_explicit (BUILT_IN_GOACC_TID);
+ tree ntid_decl = builtin_decl_explicit (BUILT_IN_GOACC_NTID);
+
+ if (gwv_bits & MASK_VECTOR)
+ {
+ tree vector_id = create_tmp_var (unsigned_type_node);
+ gimple call = gimple_build_call (tid_decl, 1, u0);
+ gimple_call_set_lhs (call, vector_id);
+ gimple_seq_add_stmt (seq, call);
+ res = vector_id;
+ }
+
+ if (gwv_bits & MASK_WORKER)
+ {
+ tree worker_id = create_tmp_var (unsigned_type_node);
+ gimple call = gimple_build_call (tid_decl, 1, u1);
+ gimple_call_set_lhs (call, worker_id);
+ gimple_seq_add_stmt (seq, call);
+ if (res != NULL_TREE)
+ {
+ vector_count = create_tmp_var (unsigned_type_node);
+ call = gimple_build_call (ntid_decl, 1, u0);
+ gimple_call_set_lhs (call, vector_count);
+ gimple_seq_add_stmt (seq, call);
+ res = fold_build2 (PLUS_EXPR, unsigned_type_node,
+ fold_build2 (MULT_EXPR, unsigned_type_node,
+ vector_count, worker_id), res);
+ }
+ else
+ res = worker_id;
+ }
+
+ if (gwv_bits & MASK_GANG)
+ {
+ tree worker_count;
+ tree ctaid_decl = builtin_decl_explicit (BUILT_IN_GOACC_CTAID);
+ tree gang_id = create_tmp_var (unsigned_type_node);
+ gimple call = gimple_build_call (ctaid_decl, 1, u0);
+ gimple_call_set_lhs (call, gang_id);
+ gimple_seq_add_stmt (seq, call);
+
+ if (gwv_bits & MASK_WORKER)
+ {
+ worker_count = create_tmp_var (unsigned_type_node);
+ call = gimple_build_call (ntid_decl, 1, u1);
+ gimple_call_set_lhs (call, worker_count);
+ gimple_seq_add_stmt (seq, call);
+ }
+ else
+ worker_count = u1;
+
+ if (gwv_bits & MASK_VECTOR)
+ {
+ if (vector_count == NULL_TREE)
+ {
+ vector_count = create_tmp_var (unsigned_type_node);
+ call = gimple_build_call (ntid_decl, 1, u0);
+ gimple_call_set_lhs (call, vector_count);
+ gimple_seq_add_stmt (seq, call);
+ }
+ }
+ else
+ vector_count = u1;
+
+ if (gwv_bits & (MASK_WORKER | MASK_VECTOR))
+ {
+ gcc_assert (res != NULL_TREE);
+ res = fold_build2 (PLUS_EXPR, unsigned_type_node,
+ fold_build2 (MULT_EXPR, unsigned_type_node,
+ fold_build2 (MULT_EXPR, unsigned_type_node,
+ worker_count, vector_count),
+ gang_id),
+ res);
+ }
+ else
+ res = gang_id;
+ }
+
+ if (res == NULL_TREE)
+ res = u0;
+
+ return res;
+}
+
+
static void
oacc_finalize_reduction_data (tree clauses, tree nthreads, tree tid,
gimple_seq *stmt_seqp, omp_context *ctx,
@@ -4936,7 +5078,6 @@ lower_reduction_clauses (tree clauses, gimple_seq *stmt_seqp, omp_context *ctx)
gimple stmt;
tree x, c, tid = NULL_TREE;
int count = 0;
- tree gang = NULL_TREE, worker = NULL_TREE, vector = NULL_TREE;
/* SIMD reductions are handled in lower_rec_input_clauses. */
if (gimple_code (ctx->stmt) == GIMPLE_OMP_FOR
@@ -4966,21 +5107,8 @@ lower_reduction_clauses (tree clauses, gimple_seq *stmt_seqp, omp_context *ctx)
&& oacc_needs_global_memory (clauses, ctx))
&& !is_oacc_parallel (ctx))
{
- x = find_omp_clause (clauses, OMP_CLAUSE_GANG);
- gang = (x == NULL_TREE) ? integer_zero_node : integer_one_node;
-
- x = find_omp_clause (clauses, OMP_CLAUSE_WORKER);
- worker = (x == NULL_TREE) ? integer_zero_node : integer_one_node;
-
- x = find_omp_clause (clauses, OMP_CLAUSE_VECTOR);
- vector = (x == NULL_TREE) ? integer_zero_node : integer_one_node;
-
/* Get the current thread id. */
- tree call = builtin_decl_explicit (BUILT_IN_GOACC_GET_THREAD_NUM);
- tid = create_tmp_var (TREE_TYPE (TREE_TYPE (call)));
- gimple stmt = gimple_build_call (call, 3, gang, worker, vector);
- gimple_call_set_lhs (stmt, tid);
- gimple_seq_add_stmt (stmt_seqp, stmt);
+ tid = expand_oacc_get_thread_num (stmt_seqp, ctx->gwv_this);
}
for (c = clauses; c ; c = OMP_CLAUSE_CHAIN (c))
@@ -5055,16 +5183,7 @@ lower_reduction_clauses (tree clauses, gimple_seq *stmt_seqp, omp_context *ctx)
/* Find the total number of threads. A reduction clause
only appears inside a loop construction or a combined
parallel and loop construct. */
-
- t = create_tmp_var (integer_type_node, NULL);
- call = builtin_decl_explicit (BUILT_IN_GOACC_GET_NUM_THREADS);
- stmt = gimple_build_call (call, 3, gang, worker, vector);
- gimple_call_set_lhs (stmt, t);
- gimple_seq_add_stmt (stmt_seqp, stmt);
-
- nthreads = create_tmp_var (sizetype, NULL);
- gimplify_assign (nthreads, fold_build1 (NOP_EXPR, sizetype, t),
- stmt_seqp);
+ nthreads = expand_oacc_get_num_threads (stmt_seqp, ctx->gwv_this);
/* Now insert the partial reductions into the array. */
@@ -7200,12 +7319,12 @@ expand_omp_for_static_nochunk (struct omp_region *region,
threadid = build_call_expr (threadid, 0);
break;
case GF_OMP_FOR_KIND_OACC_LOOP:
- nthreads = builtin_decl_explicit (BUILT_IN_GOACC_GET_NUM_THREADS);
- nthreads = build_call_expr (nthreads, 3, fd->gang, fd->worker,
- fd->vector);
- threadid = builtin_decl_explicit (BUILT_IN_GOACC_GET_THREAD_NUM);
- threadid = build_call_expr (threadid, 3, fd->gang, fd->worker,
- fd->vector);
+ {
+ gimple_seq seq = NULL;
+ nthreads = expand_oacc_get_num_threads (&seq, region->gwv_this);
+ threadid = expand_oacc_get_thread_num (&seq, region->gwv_this);
+ gsi_insert_seq_before (&gsi, seq, GSI_SAME_STMT);
+ }
break;
default:
gcc_unreachable ();
@@ -7616,12 +7735,12 @@ expand_omp_for_static_chunk (struct omp_region *region,
threadid = build_call_expr (threadid, 0);
break;
case GF_OMP_FOR_KIND_OACC_LOOP:
- nthreads = builtin_decl_explicit (BUILT_IN_GOACC_GET_NUM_THREADS);
- nthreads = build_call_expr (nthreads, 3, fd->gang, fd->worker,
- fd->vector);
- threadid = builtin_decl_explicit (BUILT_IN_GOACC_GET_THREAD_NUM);
- threadid = build_call_expr (threadid, 3, fd->gang, fd->worker,
- fd->vector);
+ {
+ gimple_seq seq = NULL;
+ nthreads = expand_oacc_get_num_threads (&seq, region->gwv_this);
+ threadid = expand_oacc_get_thread_num (&seq, region->gwv_this);
+ gsi_insert_seq_before (&gsi, seq, GSI_SAME_STMT);
+ }
break;
default:
gcc_unreachable ();
