Initialise workFn to zero explicitly

[JonChesterfield]

Summary

Initialise workFn to zero explicitly

It may be implicitly zero initialised by some toolchains.
This change would remove an out of tree patch for amdgcn.

[JonChesterfield]

Tiny change ofc. I'm looking through the diff between in tree nvptx and out of tree amdgcn and this stuck out.

[ABataev]

Why do we need this?

[JonChesterfield]

I think it needs to be initialized to zero to handle the case where there is no work, though I'm not the original author of the patch.

Iirc cuda/nvcc will implicitly zero initialize this and hip/clang does not.

[ABataev]

In D68367#1693116, @JonChesterfield wrote:

I think it needs to be initialized to zero to handle the case where there is no work, though I'm not the original author of the patch.

Iirc cuda/nvcc will implicitly zero initialize this and hip/clang does not.

cuda does not initialize it to zero. If there is no work, __kmpc_kernel_deinit will set it to 0 and terminate the workers.

[JonChesterfield]

Interesting. I'll have to put more thought into this, thanks.

openmp/libomptarget/deviceRTLs/nvptx/src/omptarget-nvptx.cu

EXTERN void __kmpc_kernel_init(int ThreadLimit, int16_t RequiresOMPRuntime) {
   omptarget_nvptx_threadPrivateContext->SetTopLevelTaskDescr(
       threadId, currTeamDescr.LevelZeroTaskDescr());
 
   // set number of threads and thread limit in team to started value
   omptarget_nvptx_TaskDescr *currTaskDescr =
       omptarget_nvptx_threadPrivateContext->GetTopLevelTaskDescr(threadId);
   nThreads = GetNumberOfWorkersInTeam();
   threadLimit = ThreadLimit;
+  omptarget_nvptx_workFn = 0;
 }
 
 EXTERN void __kmpc_kernel_deinit(int16_t IsOMPRuntimeInitialized) {
   PRINT0(LD_IO, "call to __kmpc_kernel_deinit\n");
   ASSERT0(LT_FUSSY, IsOMPRuntimeInitialized,
           "Generic always requires initialized runtime.");
   // Enqueue omp state object for use by another team.
   int slot = usedSlotIdx;