[OPENMP][NVPTX]Run combined constructs with if clause in SPMD mode.

All target-parallel-based constructs can be run in SPMD mode from now
on. Even if num_threads clauses or if clauses are used, such constructs
can be executed in SPMD mode.

llvm-svn: 358595

5 files changed, 43 insertions, 47 deletions

diff --git a/clang/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp b/clang/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp
index d9056eeff5c..9470aa7972a 100644
--- a/clang/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp
@@ -24,29 +24,22 @@ int main(int argc, char **argv) {
 // CHECK: [[MEM_TY:%.+]] = type { [128 x i8] }
 // CHECK-DAG: [[SHARED_GLOBAL_RD:@.+]] = common addrspace(3) global [[MEM_TY]] zeroinitializer
 // CHECK-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* null
-// CHECK-DAG: [[KERNEL_SIZE:@.+]] = internal unnamed_addr constant i{{64|32}} 84
-// CHECK-DAG: @__omp_offloading_{{.*}}_main_l17_exec_mode = weak constant i8 1
-
-// CHECK-LABEL: define internal void @__omp_offloading_{{.*}}_main_l17_worker(
+// CHECK-DAG: [[KERNEL_SIZE:@.+]] = internal unnamed_addr constant i{{64|32}} 40
+// CHECK-DAG: @__omp_offloading_{{.*}}_main_l17_exec_mode = weak constant i8 0
 
 // CHECK: define weak void @__omp_offloading_{{.*}}_main_l17([10 x i32]* dereferenceable(40) %{{.+}}, [10 x i32]* dereferenceable(40) %{{.+}}, i32* dereferenceable(4) %{{.+}}, i{{64|32}} %{{.+}}, [10 x i32]* dereferenceable(40) %{{.+}})
-// CHECK: call void @__kmpc_get_team_static_memory(i16 0, i8* addrspacecast (i8 addrspace(3)* getelementptr inbounds ([[MEM_TY]], [[MEM_TY]] addrspace(3)* [[SHARED_GLOBAL_RD]], i32 0, i32 0, i32 0) to i8*), i{{64|32}} 84, i16 1, i8** addrspacecast (i8* addrspace(3)* [[KERNEL_PTR]] to i8**))
+// CHECK: call void @__kmpc_get_team_static_memory(i16 1, i8* addrspacecast (i8 addrspace(3)* getelementptr inbounds ([[MEM_TY]], [[MEM_TY]] addrspace(3)* [[SHARED_GLOBAL_RD]], i32 0, i32 0, i32 0) to i8*), i{{64|32}} 40, i16 1, i8** addrspacecast (i8* addrspace(3)* [[KERNEL_PTR]] to i8**))
 // CHECK: [[PTR:%.+]] = load i8*, i8* addrspace(3)* [[KERNEL_PTR]],
 // CHECK: [[STACK:%.+]] = bitcast i8* [[PTR]] to %struct._globalized_locals_ty*
-// CHECK: [[ARGC:%.+]] = load i32, i32* %{{.+}}, align
-// CHECK: [[ARGC_ADDR:%.+]] = getelementptr inbounds %struct._globalized_locals_ty, %struct._globalized_locals_ty* [[STACK]], i{{32|64}} 0, i{{32|64}} 0
-// CHECK: store i32 [[ARGC]], i32* [[ARGC_ADDR]],
-// CHECK: getelementptr inbounds %struct._globalized_locals_ty, %struct._globalized_locals_ty* [[STACK]], i{{32|64}} 0, i{{32|64}} 1
-// CHECK: getelementptr inbounds %struct._globalized_locals_ty, %struct._globalized_locals_ty* [[STACK]], i{{32|64}} 0, i{{32|64}} 2
+// CHECK: getelementptr inbounds %struct._globalized_locals_ty, %struct._globalized_locals_ty* [[STACK]], i{{32|64}} 0, i{{32|64}} 0
+// CHECK-NOT: getelementptr inbounds %struct._globalized_locals_ty, %struct._globalized_locals_ty* [[STACK]],
 // CHECK: call void @__kmpc_for_static_init_4(
 
-// CHECK: call void @__kmpc_serialized_parallel(
 // CHECK: call void [[PARALLEL:@.+]](
-// CHECK: call void @__kmpc_end_serialized_parallel(
 
 // CHECK: call void @__kmpc_for_static_fini(%struct.ident_t* @
 
-// CHECK: call void @__kmpc_restore_team_static_memory(i16 0, i16 1)
+// CHECK: call void @__kmpc_restore_team_static_memory(i16 1, i16 1)
 
 // CHECK: define internal void [[PARALLEL]](
 // CHECK-NOT: call i8* @__kmpc_data_sharing_push_stack(
diff --git a/clang/test/OpenMP/nvptx_target_simd_codegen.cpp b/clang/test/OpenMP/nvptx_target_simd_codegen.cpp
index 89ea173addc..ce4a4238f81 100644
--- a/clang/test/OpenMP/nvptx_target_simd_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_target_simd_codegen.cpp
@@ -9,10 +9,10 @@
 #define HEADER
 
 // Check that the execution mode of all 2 target regions on the gpu is set to NonSPMD Mode.
-// CHECK-DAG: {{@__omp_offloading_.+l25}}_exec_mode = weak constant i8 1
-// CHECK-DAG: {{@__omp_offloading_.+l30}}_exec_mode = weak constant i8 1
-// CHECK-DAG: {{@__omp_offloading_.+l35}}_exec_mode = weak constant i8 1
-// CHECK-DAG: {{@__omp_offloading_.+l40}}_exec_mode = weak constant i8 1
+// CHECK-DAG: {{@__omp_offloading_.+l25}}_exec_mode = weak constant i8 0
+// CHECK-DAG: {{@__omp_offloading_.+l30}}_exec_mode = weak constant i8 0
+// CHECK-DAG: {{@__omp_offloading_.+l35}}_exec_mode = weak constant i8 0
+// CHECK-DAG: {{@__omp_offloading_.+l40}}_exec_mode = weak constant i8 0
 
 #define N 1000
 
@@ -54,33 +54,33 @@ int bar(int n){
 }
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l25}}(
-// CHECK: call void @__kmpc_kernel_init(i32 %{{.+}}, i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_init(i32 %{{.+}}, i16 1, i16 0)
 // CHECK-NOT: call void @__kmpc_for_static_init
 // CHECK-NOT: call void @__kmpc_for_static_fini
-// CHECK: call void @__kmpc_kernel_deinit(i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1)
 // CHECK: ret void
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l30}}(
-// CHECK: call void @__kmpc_kernel_init(i32 %{{.+}}, i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_init(i32 %{{.+}}, i16 1, i16 0)
 // CHECK-NOT: call void @__kmpc_for_static_init
 // CHECK-NOT: call void @__kmpc_for_static_fini
-// CHECK: call void @__kmpc_kernel_deinit(i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1)
 // CHECK: ret void
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l35}}(
-// CHECK: call void @__kmpc_kernel_init(i32 %{{.+}}, i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_init(i32 %{{.+}}, i16 1, i16 0)
 // CHECK-NOT: call void @__kmpc_for_static_init
 // CHECK-NOT: call void @__kmpc_for_static_fini
-// CHECK: call void @__kmpc_kernel_deinit(i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1)
 // CHECK: ret void
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l40}}(
-// CHECK: call void @__kmpc_kernel_init(i32 %{{.+}}, i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_init(i32 %{{.+}}, i16 1, i16 0)
 // CHECK-NOT: call void @__kmpc_for_static_init
 // CHECK-NOT: call void @__kmpc_for_static_fini
 // CHECK-NOT: call i32 @__kmpc_nvptx_simd_reduce_nowait(
 // CHECK-NOT: call void @__kmpc_nvptx_end_reduce_nowait(
-// CHECK: call void @__kmpc_kernel_deinit(i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1)
 // CHECK: ret void
 
 
diff --git a/clang/test/OpenMP/nvptx_target_teams_distribute_simd_codegen.cpp b/clang/test/OpenMP/nvptx_target_teams_distribute_simd_codegen.cpp
index 48f314785c8..f86b456b3b0 100644
--- a/clang/test/OpenMP/nvptx_target_teams_distribute_simd_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_target_teams_distribute_simd_codegen.cpp
@@ -9,10 +9,10 @@
 #define HEADER
 
 // Check that the execution mode of all 2 target regions on the gpu is set to NonSPMD Mode.
-// CHECK-DAG: {{@__omp_offloading_.+l30}}_exec_mode = weak constant i8 1
-// CHECK-DAG: {{@__omp_offloading_.+l36}}_exec_mode = weak constant i8 1
-// CHECK-DAG: {{@__omp_offloading_.+l41}}_exec_mode = weak constant i8 1
-// CHECK-DAG: {{@__omp_offloading_.+l46}}_exec_mode = weak constant i8 1
+// CHECK-DAG: {{@__omp_offloading_.+l30}}_exec_mode = weak constant i8 0
+// CHECK-DAG: {{@__omp_offloading_.+l36}}_exec_mode = weak constant i8 0
+// CHECK-DAG: {{@__omp_offloading_.+l41}}_exec_mode = weak constant i8 0
+// CHECK-DAG: {{@__omp_offloading_.+l46}}_exec_mode = weak constant i8 0
 
 #define N 1000
 #define M 10
@@ -22,7 +22,7 @@ tx ftemplate(int n) {
   tx a[N];
   short aa[N];
   tx b[10];
-  tx c[M][M];  
+  tx c[M][M];
   tx f = n;
   tx l;
   int k;
@@ -47,7 +47,7 @@ tx ftemplate(int n) {
   for(int i = 0; i < M; i++) {
     for(int j = 0; j < M; j++) {
       k = M;
-      c[i][j] = i+j*f+k;      
+      c[i][j] = i + j * f + k;
     }
   }
 
@@ -63,33 +63,33 @@ int bar(int n){
 }
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+}}_l30(
-// CHECK: call void @__kmpc_kernel_init(i32 %{{.+}}, i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_init(i32 %{{.+}}, i16 1, i16 0)
 // CHECK: call void @__kmpc_for_static_init_4({{.+}}, {{.+}}, {{.+}} 91,
 // CHECK: call void @__kmpc_for_static_fini(
-// CHECK: call void @__kmpc_kernel_deinit(i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1)
 // CHECK: ret void
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+}}_l36(
-// CHECK: call void @__kmpc_kernel_init(i32 %{{.+}}, i16 1)
-// CHECK: call void @__kmpc_for_static_init_4({{.+}}, {{.+}}, {{.+}} 92,
+// CHECK: call void @__kmpc_spmd_kernel_init(i32 %{{.+}}, i16 1, i16 0)
+// CHECK: call void @__kmpc_for_static_init_4({{.+}}, {{.+}}, {{.+}} 91,
 // CHECK: call void @__kmpc_for_static_fini(
-// CHECK: call void @__kmpc_kernel_deinit(i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1)
 // CHECK: ret void
 
 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+}}_l41(
-// CHECK: call void @__kmpc_kernel_init(i32 %{{.+}}, i16 1)
-// CHECK: call void @__kmpc_for_static_init_4({{.+}}, {{.+}}, {{.+}} 92,
+// CHECK: call void @__kmpc_spmd_kernel_init(i32 %{{.+}}, i16 1, i16 0)
+// CHECK: call void @__kmpc_for_static_init_4({{.+}}, {{.+}}, {{.+}} 91,
 // CHECK: call void @__kmpc_for_static_fini(
-// CHECK: call void @__kmpc_kernel_deinit(i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1)
 // CHECK: ret void
 
 // CHECK: define {{.*}}void {{@__omp_offloading_.+}}_l46({{.+}}, i{{32|64}} [[F_IN:%.+]])
 // CHECK: store {{.+}} [[F_IN]], {{.+}}* {{.+}},
-// CHECK: call void @__kmpc_kernel_init(i32 %{{.+}}, i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_init(i32 %{{.+}}, i16 1, i16 0)
 // CHECK: store {{.+}} 99, {{.+}}* [[COMB_UB:%.+]], align
-// CHECK: call void @__kmpc_for_static_init_4({{.+}}, {{.+}}, {{.+}} 92, {{.+}}, {{.+}}, {{.+}}* [[COMB_UB]],
+// CHECK: call void @__kmpc_for_static_init_4({{.+}}, {{.+}}, {{.+}} 91, {{.+}}, {{.+}}, {{.+}}* [[COMB_UB]],
 // CHECK: call void @__kmpc_for_static_fini(
-// CHECK: call void @__kmpc_kernel_deinit(i16 1)
+// CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1)
 // CHECK: ret void
 
 #endif
diff --git a/clang/test/OpenMP/target_parallel_if_codegen.cpp b/clang/test/OpenMP/target_parallel_if_codegen.cpp
index 03e8c344450..a4fdaa62f1d 100644
--- a/clang/test/OpenMP/target_parallel_if_codegen.cpp
+++ b/clang/test/OpenMP/target_parallel_if_codegen.cpp
@@ -163,7 +163,8 @@ int bar(int n){
 // CHECK:       store i8 [[FB]], i8* [[CONV]], align
 // CHECK:       [[ARG:%.+]] = load i[[SZ]], i[[SZ]]* [[CAPEC_ADDR]], align
 //
-// CHECK-DAG:   [[RET:%.+]] = call i32 @__tgt_target_teams(i64 -1, i8* @{{[^,]+}}, i32 4, {{.*}}, i32 1, i32 0)
+// CHECK-DAG:   [[RET:%.+]] = call i32 @__tgt_target_teams(i64 -1, i8* @{{[^,]+}}, i32 4, {{.*}}, i32 1, i32 [[NT:%.+]])
+// CHECK-DAG:   [[NT]] = select i1 %{{.+}}, i32 0, i32 1
 // CHECK:       [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
 // CHECK:       br i1 [[ERROR]], label %[[FAIL:.+]], label %[[END:[^,]+]]
 //
@@ -189,7 +190,8 @@ int bar(int n){
 // CHECK:       br i1 [[CMP]], label {{%?}}[[IF_THEN:.+]], label {{%?}}[[IF_ELSE:.+]]
 //
 // CHECK:       [[IF_THEN]]
-// CHECK-DAG:   [[RET:%.+]] = call i32 @__tgt_target_teams(i64 -1, i8* @{{[^,]+}}, i32 3, {{.*}}, i32 1, i32 0)
+// CHECK-DAG:   [[RET:%.+]] = call i32 @__tgt_target_teams(i64 -1, i8* @{{[^,]+}}, i32 3, {{.*}}, i32 1, i32 [[NT:%.+]])
+// CHECK-DAG:   [[NT]] = select i1 %{{.+}}, i32 0, i32 1
 // CHECK:       [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
 // CHECK-NEXT:  br i1 [[ERROR]], label %[[FAIL:.+]], label %[[END:[^,]+]]
 // CHECK:       [[FAIL]]
@@ -221,7 +223,8 @@ int bar(int n){
 // CHECK:       br i1 [[TB]], label {{%?}}[[IF_THEN:.+]], label {{%?}}[[IF_ELSE:.+]]
 //
 // CHECK:       [[IF_THEN]]
-// CHECK-DAG:   [[RET:%.+]] = call i32 @__tgt_target_teams(i64 -1, i8* @{{[^,]+}}, i32 1, {{.*}}, i32 1, i32 0)
+// CHECK-DAG:   [[RET:%.+]] = call i32 @__tgt_target_teams(i64 -1, i8* @{{[^,]+}}, i32 1, {{.*}}, i32 1, i32 [[NT:%.+]])
+// CHECK-DAG:   [[NT]] = select i1 %{{.+}}, i32 0, i32 1
 // CHECK:       [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
 // CHECK-NEXT:  br i1 [[ERROR]], label %[[FAIL:.+]], label %[[END:[^,]+]]
 // CHECK:       [[FAIL]]
@@ -263,7 +266,7 @@ int bar(int n){
 //
 // CHECK: define {{.*}}[[FTEMPLATE]]
 //
-// CHECK-DAG:   [[RET:%.+]] = call i32 @__tgt_target_teams(i64 -1, i8* @{{[^,]+}}, i32 1, {{.*}}, i32 1, i32 0)
+// CHECK-DAG:   [[RET:%.+]] = call i32 @__tgt_target_teams(i64 -1, i8* @{{[^,]+}}, i32 1, {{.*}}, i32 1, i32 1)
 // CHECK-NEXT:  [[ERROR:%.+]] = icmp ne i32 [[RET]], 0
 // CHECK-NEXT:  br i1 [[ERROR]], label %[[FAIL:.+]], label %[[END:[^,]+]]
 //
diff --git a/clang/test/OpenMP/target_parallel_num_threads_codegen.cpp b/clang/test/OpenMP/target_parallel_num_threads_codegen.cpp
index 7a39a3fd1c0..41a779ced35 100644
--- a/clang/test/OpenMP/target_parallel_num_threads_codegen.cpp
+++ b/clang/test/OpenMP/target_parallel_num_threads_codegen.cpp
@@ -263,7 +263,7 @@ int bar(int n){
 // CHECK:       store i16 [[CEV]], i16* [[CONV]], align
 // CHECK:       [[ARG:%.+]] = load i[[SZ]], i[[SZ]]* [[CAPEC_ADDR]], align
 // CHECK:       [[T:%.+]] = load i16, i16* [[CAPE_ADDR]], align
-// CHECK:       [[THREADS:%.+]] = sext i16 [[T]] to i32
+// CHECK:       [[THREADS:%.+]] = zext i16 [[T]] to i32
 //
 // CHECK-DAG:   [[RET:%.+]] = call i32 @__tgt_target_teams(i64 -1, i8* @{{[^,]+}}, i32 3, {{.*}}, i32 1, i32 [[THREADS]])
 // CHECK:       [[ERROR:%.+]] = icmp ne i32 [[RET]], 0