add sync after gemm

Benchmark/cholesky_gemm/gemm.cpp

@@ -116,6 +116,7 @@ auto gemm(const int NbLoops, SpBlas::Block blocksC[], const SpBlas::Block blocks
                                     inBlockDim, inBlockDim, inBlockDim, &alphaBeta, (const double*)paramA.getRawPtr(), inBlockDim,
                                     (const double*)paramB.getRawPtr(), inBlockDim,
                                     &alphaBeta, (double*)paramC.getRawPtr(), inBlockDim ) );
+                            CUDA_ASSERT(cudaStreamSynchronize(SpCudaUtils::GetCurrentStream()));
                         })
                 #endif
                 #ifdef SPECX_COMPILE_WITH_HIP