diff options
| author | Alexey Bataev <a.bataev@hotmail.com> | 2018-10-12 16:04:20 +0000 |
|---|---|---|
| committer | Alexey Bataev <a.bataev@hotmail.com> | 2018-10-12 16:04:20 +0000 |
| commit | 9bfe91da3d2c2eb3f0cba9dc587a51ddbbf4d8ac (patch) | |
| tree | 6a8da57be99cd61b82cbe3160d91fc0fcd578744 /clang/test/OpenMP/nvptx_target_codegen.cpp | |
| parent | c046b6856ec91e65b1112e66af0fa6735c20bc7d (diff) | |
| download | bcm5719-llvm-9bfe91da3d2c2eb3f0cba9dc587a51ddbbf4d8ac.tar.gz bcm5719-llvm-9bfe91da3d2c2eb3f0cba9dc587a51ddbbf4d8ac.zip | |
[OPENMP][NVPTX]Reduce memory usage in orphaned functions.
if the function has globalized variables and called in context of
target/teams/distribute regions, it does not need to globalize 32
copies of the same variables for memory coalescing, it is enough to
have just one copy, because there is parallel region.
Patch does this by adding call for `__kmpc_parallel_level` function and
checking its return value. If the code sees that the parallel level is
0, then only one variable is allocated, not 32.
llvm-svn: 344356
Diffstat (limited to 'clang/test/OpenMP/nvptx_target_codegen.cpp')
| -rw-r--r-- | clang/test/OpenMP/nvptx_target_codegen.cpp | 12 |
1 files changed, 9 insertions, 3 deletions
diff --git a/clang/test/OpenMP/nvptx_target_codegen.cpp b/clang/test/OpenMP/nvptx_target_codegen.cpp index 69c338995b7..3525d2e2054 100644 --- a/clang/test/OpenMP/nvptx_target_codegen.cpp +++ b/clang/test/OpenMP/nvptx_target_codegen.cpp @@ -557,20 +557,26 @@ int baz(int f, double &a) { // CHECK: alloca i32, // CHECK: [[LOCAL_F_PTR:%.+]] = alloca i32, // CHECK: [[ZERO_ADDR:%.+]] = alloca i32, - // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* // CHECK: store i32 0, i32* [[ZERO_ADDR]] + // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* + // CHECK: [[PAR_LEVEL:%.+]] = call i16 @__kmpc_parallel_level(%struct.ident_t* @0, i32 [[GTID]]) + // CHECK: [[IS_TTD:%.+]] = icmp eq i16 %1, 0 // CHECK: [[RES:%.+]] = call i8 @__kmpc_is_spmd_exec_mode() // CHECK: [[IS_SPMD:%.+]] = icmp ne i8 [[RES]], 0 // CHECK: br i1 [[IS_SPMD]], label // CHECK: br label - // CHECK: [[PTR:%.+]] = call i8* @__kmpc_data_sharing_push_stack(i{{64|32}} 128, i16 0) + // CHECK: [[SIZE:%.+]] = select i1 [[IS_TTD]], i{{64|32}} 4, i{{64|32}} 128 + // CHECK: [[PTR:%.+]] = call i8* @__kmpc_data_sharing_push_stack(i{{64|32}} [[SIZE]], i16 0) // CHECK: [[REC_ADDR:%.+]] = bitcast i8* [[PTR]] to [[GLOBAL_ST:%.+]]* // CHECK: br label // CHECK: [[ITEMS:%.+]] = phi [[GLOBAL_ST]]* [ null, {{.+}} ], [ [[REC_ADDR]], {{.+}} ] + // CHECK: [[TTD_ITEMS:%.+]] = bitcast [[GLOBAL_ST]]* [[ITEMS]] to [[SEC_GLOBAL_ST:%.+]]* // CHECK: [[F_PTR_ARR:%.+]] = getelementptr inbounds [[GLOBAL_ST]], [[GLOBAL_ST]]* [[ITEMS]], i32 0, i32 0 // CHECK: [[TID:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x() // CHECK: [[LID:%.+]] = and i32 [[TID]], 31 - // CHECK: [[GLOBAL_F_PTR:%.+]] = getelementptr inbounds [32 x i32], [32 x i32]* [[F_PTR_ARR]], i32 0, i32 [[LID]] + // CHECK: [[GLOBAL_F_PTR_PAR:%.+]] = getelementptr inbounds [32 x i32], [32 x i32]* [[F_PTR_ARR]], i32 0, i32 [[LID]] + // CHECK: [[GLOBAL_F_PTR_TTD:%.+]] = getelementptr inbounds [[SEC_GLOBAL_ST]], [[SEC_GLOBAL_ST]]* [[TTD_ITEMS]], i32 0, i32 0 + // CHECK: [[GLOBAL_F_PTR:%.+]] = select i1 [[IS_TTD]], i32* [[GLOBAL_F_PTR_TTD]], i32* [[GLOBAL_F_PTR_PAR]] // CHECK: [[F_PTR:%.+]] = select i1 [[IS_SPMD]], i32* [[LOCAL_F_PTR]], i32* [[GLOBAL_F_PTR]] // CHECK: store i32 %{{.+}}, i32* [[F_PTR]], |
