[OpenMP] Codegen for the 'target parallel' directive on the NVPTX device.
This patch adds codegen for the 'target parallel' directive on the NVPTX
device. We term offload OpenMP directives such as 'target parallel' and
'target teams distribute parallel for' as SPMD constructs. SPMD constructs,
in contrast to Generic ones like the plain 'target', can never contain
a serial region.
SPMD constructs can be handled more efficiently on the GPU and do not
require the Warp Loop of the Generic codegen scheme. This patch adds
SPMD codegen support for 'target parallel' on the NVPTX device and can
be reused for other SPMD constructs.
Reviewers: ABataev
Differential Revision: https://reviews.llvm.org/D28755
llvm-svn: 292428
diff --git a/clang/test/OpenMP/nvptx_target_codegen.cpp b/clang/test/OpenMP/nvptx_target_codegen.cpp
index 59c4d5b..dab0c5a 100644
--- a/clang/test/OpenMP/nvptx_target_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_target_codegen.cpp
@@ -8,6 +8,14 @@
#ifndef HEADER
#define HEADER
+// Check that the execution mode of all 6 target regions is set to Generic Mode.
+// CHECK-DAG: {{@__omp_offloading_.+l98}}_exec_mode = weak constant i8 1
+// CHECK-DAG: {{@__omp_offloading_.+l175}}_exec_mode = weak constant i8 1
+// CHECK-DAG: {{@__omp_offloading_.+l284}}_exec_mode = weak constant i8 1
+// CHECK-DAG: {{@__omp_offloading_.+l321}}_exec_mode = weak constant i8 1
+// CHECK-DAG: {{@__omp_offloading_.+l339}}_exec_mode = weak constant i8 1
+// CHECK-DAG: {{@__omp_offloading_.+l304}}_exec_mode = weak constant i8 1
+
template<typename tx, typename ty>
struct TT{
tx X;
@@ -23,7 +31,7 @@
double cn[5][n];
TT<long long, char> d;
- // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+foo.+l90}}_worker()
+ // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+foo.+l98}}_worker()
// CHECK-DAG: [[OMP_EXEC_STATUS:%.+]] = alloca i8,
// CHECK-DAG: [[OMP_WORK_FN:%.+]] = alloca i8*,
// CHECK: store i8* null, i8** [[OMP_WORK_FN]],
@@ -54,7 +62,7 @@
// CHECK: [[EXIT]]
// CHECK: ret void
- // CHECK: define {{.*}}void [[T1:@__omp_offloading_.+foo.+l90]]()
+ // CHECK: define {{.*}}void [[T1:@__omp_offloading_.+foo.+l98]]()
// CHECK-DAG: [[TID:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
// CHECK-DAG: [[NTH:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.ntid.x()
// CHECK-DAG: [[WS:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
@@ -96,7 +104,7 @@
{
}
- // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+foo.+l167}}_worker()
+ // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+foo.+l175}}_worker()
// CHECK-DAG: [[OMP_EXEC_STATUS:%.+]] = alloca i8,
// CHECK-DAG: [[OMP_WORK_FN:%.+]] = alloca i8*,
// CHECK: store i8* null, i8** [[OMP_WORK_FN]],
@@ -127,7 +135,7 @@
// CHECK: [[EXIT]]
// CHECK: ret void
- // CHECK: define {{.*}}void [[T2:@__omp_offloading_.+foo.+l167]](i[[SZ:32|64]] [[ARG1:%[a-zA-Z_]+]])
+ // CHECK: define {{.*}}void [[T2:@__omp_offloading_.+foo.+l175]](i[[SZ:32|64]] [[ARG1:%[a-zA-Z_]+]])
// CHECK: [[AA_ADDR:%.+]] = alloca i[[SZ]],
// CHECK: store i[[SZ]] [[ARG1]], i[[SZ]]* [[AA_ADDR]],
// CHECK: [[AA_CADDR:%.+]] = bitcast i[[SZ]]* [[AA_ADDR]] to i16*
@@ -169,7 +177,7 @@
aa += 1;
}
- // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+foo.+l276}}_worker()
+ // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+foo.+l284}}_worker()
// CHECK-DAG: [[OMP_EXEC_STATUS:%.+]] = alloca i8,
// CHECK-DAG: [[OMP_WORK_FN:%.+]] = alloca i8*,
// CHECK: store i8* null, i8** [[OMP_WORK_FN]],
@@ -200,7 +208,7 @@
// CHECK: [[EXIT]]
// CHECK: ret void
- // CHECK: define {{.*}}void [[T3:@__omp_offloading_.+foo.+l276]](i[[SZ]]
+ // CHECK: define {{.*}}void [[T3:@__omp_offloading_.+foo.+l284]](i[[SZ]]
// Create local storage for each capture.
// CHECK: [[LOCAL_A:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_B:%.+]] = alloca [10 x float]*
@@ -353,7 +361,7 @@
return a;
}
- // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+static.+313}}_worker()
+ // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+static.+321}}_worker()
// CHECK-DAG: [[OMP_EXEC_STATUS:%.+]] = alloca i8,
// CHECK-DAG: [[OMP_WORK_FN:%.+]] = alloca i8*,
// CHECK: store i8* null, i8** [[OMP_WORK_FN]],
@@ -384,7 +392,7 @@
// CHECK: [[EXIT]]
// CHECK: ret void
- // CHECK: define {{.*}}void [[T4:@__omp_offloading_.+static.+l313]](i[[SZ]]
+ // CHECK: define {{.*}}void [[T4:@__omp_offloading_.+static.+l321]](i[[SZ]]
// Create local storage for each capture.
// CHECK: [[LOCAL_A:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_AA:%.+]] = alloca i[[SZ]]
@@ -439,7 +447,7 @@
- // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+S1.+l331}}_worker()
+ // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+S1.+l339}}_worker()
// CHECK-DAG: [[OMP_EXEC_STATUS:%.+]] = alloca i8,
// CHECK-DAG: [[OMP_WORK_FN:%.+]] = alloca i8*,
// CHECK: store i8* null, i8** [[OMP_WORK_FN]],
@@ -470,7 +478,7 @@
// CHECK: [[EXIT]]
// CHECK: ret void
- // CHECK: define {{.*}}void [[T5:@__omp_offloading_.+S1.+l331]](
+ // CHECK: define {{.*}}void [[T5:@__omp_offloading_.+S1.+l339]](
// Create local storage for each capture.
// CHECK: [[LOCAL_THIS:%.+]] = alloca [[S1:%struct.*]]*
// CHECK: [[LOCAL_B:%.+]] = alloca i[[SZ]]
@@ -529,7 +537,7 @@
- // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+template.+l296}}_worker()
+ // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+template.+l304}}_worker()
// CHECK-DAG: [[OMP_EXEC_STATUS:%.+]] = alloca i8,
// CHECK-DAG: [[OMP_WORK_FN:%.+]] = alloca i8*,
// CHECK: store i8* null, i8** [[OMP_WORK_FN]],
@@ -560,7 +568,7 @@
// CHECK: [[EXIT]]
// CHECK: ret void
- // CHECK: define {{.*}}void [[T6:@__omp_offloading_.+template.+l296]](i[[SZ]]
+ // CHECK: define {{.*}}void [[T6:@__omp_offloading_.+template.+l304]](i[[SZ]]
// Create local storage for each capture.
// CHECK: [[LOCAL_A:%.+]] = alloca i[[SZ]]
// CHECK: [[LOCAL_AA:%.+]] = alloca i[[SZ]]