[llvm] [Offload][OMPX] Add the runtime support for multi-dim grid and block (PR #118042)

[Shilei Tian via llvm-commits]

================
@@ -616,38 +622,36 @@ KernelLaunchParamsTy GenericKernelTy::prepareArgs(
   return KernelLaunchParamsTy{sizeof(void *) * NumArgs, &Args[0], &Ptrs[0]};
 }
 
-uint32_t GenericKernelTy::getNumThreads(GenericDeviceTy &GenericDevice,
-                                        uint32_t ThreadLimitClause[3]) const {
-  assert(ThreadLimitClause[1] == 0 && ThreadLimitClause[2] == 0 &&
-         "Multi dimensional launch not supported yet.");
-
-  if (IsBareKernel && ThreadLimitClause[0] > 0)
-    return ThreadLimitClause[0];
+void GenericKernelTy::getNumThreads(GenericDeviceTy &GenericDevice,
+                                    uint32_t ThreadLimitClause[3]) const {
+  if (IsBareKernel)
----------------
shiltian wrote:

No, the logic is different because we directly use whatever users provide for us in bare mode.

https://github.com/llvm/llvm-project/pull/118042