[libomptarget] Disable test bug49334 on amdgpu

[JonChesterfield]

Summary

[libomptarget] Disable test bug49334 on amdgpu

Hangs on amdgpu, do not know why. Disable to unblock build.

[protze.joachim]

You can just use UNSUPPORTED.

Can you try first, whether reducing the matrix size helps with your hang? I planned to modify the value anyways. The test just runs too long.
Reducing the size also triggers the race in data mapping more reliably.

openmp/libomptarget/test/offloading/bug49334.cpp

1–5

73–74

[JonChesterfield]

It terminated with the smaller parameters, which suggests it might (eventually) terminate with the larger ones. Trying to run it under time now, but it keeps segfaulting in __kmp_push_task

[JonChesterfield]

• use unsupported

[JonChesterfield]

@protze.joachim can we disable this for now? the hang / excessive runtime makes a mess of build/debug cycles and will cause CI problems when it tries to go online

[ye-luo]

This test annoyed me again.

openmp/libomptarget/test/offloading/bug49334.cpp

 // RUN: %libomptarget-compilexx-run-and-check-generic
 
+// Currently hangs on amdgpu
+// UNSUPPORTED: amdgcn-amd-amdhsa
+
 #include <cassert>
 #include <iostream>
 #include <memory>
 #include <vector>
 
 class BlockMatrix {
 private:
   const int rowsPerBlock;
public:
   }
 
   float *GetBlock(int i, int j) const {
     assert(i < nBlocksPerCol && j < nBlocksPerRow && "Accessing outside block");
     return Blocks[i][j].get();
   }
 };
 
 constexpr const int BS = 256;
 constexpr const int N = 1024;
 
 int BlockMatMul_TargetNowait(BlockMatrix &A, BlockMatrix &B, BlockMatrix &C) {
 #pragma omp parallel
 #pragma omp master
   for (int i = 0; i < N / BS; ++i)
     for (int j = 0; j < N / BS; ++j) {
       float *BlockC = C.GetBlock(i, j);
       for (int k = 0; k < N / BS; ++k) {