[Openmp-commits] [openmp] 2c941fa - [OpenMP][deviceRTLs] Fix wrong return value of `__kmpc_is_spmd_exec_mode`

[Shilei Tian via Openmp-commits]

Author: Shilei Tian
Date: 2021-10-16T12:58:29-04:00
New Revision: 2c941fa2f9b9a93b42bdddc810a817b02a937b55

URL: https://github.com/llvm/llvm-project/commit/2c941fa2f9b9a93b42bdddc810a817b02a937b55
DIFF: https://github.com/llvm/llvm-project/commit/2c941fa2f9b9a93b42bdddc810a817b02a937b55.diff

LOG: [OpenMP][deviceRTLs] Fix wrong return value of `__kmpc_is_spmd_exec_mode`

D110279 introduced a bug to the device runtime. In `__kmpc_parallel_51`, we detect
whether we are already in parallel region by `__kmpc_parallel_level() > __kmpc_is_spmd_exec_mode()`.
It is based on the assumption that:
- In SPMD mode, parallel level is initialized to 1.
- In generic mode, parallel level is initialized to 0.
- `__kmpc_is_spmd_exec_mode` returns `1` for SPMD mode, 0 otherwise.

Because the return value type of `__kmpc_is_spmd_exec_mode` is `int8_t`, there
was an implicit cast from `bool` to `int8_t`. We can make sure it is either 0 or
1 since C++14. In D110279, the return value is the result of an `and` operation,
which is 2 in SPMD mode. This breaks the assumption in `__kmpc_parallel_51`.

Reviewed By: carlo.bertolli, dpalermo

Differential Revision: https://reviews.llvm.org/D111905

Added: 
    

Modified: 
    openmp/libomptarget/deviceRTLs/common/src/omptarget.cu

Removed: 
    


################################################################################
diff  --git a/openmp/libomptarget/deviceRTLs/common/src/omptarget.cu b/openmp/libomptarget/deviceRTLs/common/src/omptarget.cu
index 78bd5d7cd62b1..d0be541b87953 100644
--- a/openmp/libomptarget/deviceRTLs/common/src/omptarget.cu
+++ b/openmp/libomptarget/deviceRTLs/common/src/omptarget.cu
@@ -160,8 +160,12 @@ static void __kmpc_spmd_kernel_deinit(bool RequiresFullRuntime) {
 }
 
 // Return true if the current target region is executed in SPMD mode.
+// NOTE: This function has to return 1 for SPMD mode, and 0 for generic mode.
+// That's because `__kmpc_parallel_51` checks if it's already in parallel region
+// by comparision between the parallel level and the return value of this
+// function.
 EXTERN int8_t __kmpc_is_spmd_exec_mode() {
-  return execution_param & OMP_TGT_EXEC_MODE_SPMD;
+  return (execution_param & OMP_TGT_EXEC_MODE_SPMD) == OMP_TGT_EXEC_MODE_SPMD;
 }
 
 EXTERN int8_t __kmpc_is_generic_main_thread(kmp_int32 Tid) {