Searched refs:MDOT_WORKGROUP_SIZE (Results 1 – 1 of 1) sorted by relevance
1033 #define MDOT_WORKGROUP_SIZE MDOT_WORKGROUP_NUM macro1052 PETSC_SHAREDMEM_DECL PetscScalar shmem[N * MDOT_WORKGROUP_SIZE]; in MDot_kernel()1075 for (auto i = 0; i < N; ++i) shmem[tx + i * MDOT_WORKGROUP_SIZE] = sumlocal[i]; in MDot_kernel()1082 …or (auto i = 0; i < N; ++i) shmem[tx + i * MDOT_WORKGROUP_SIZE] += shmem[tx + stride + i * MDOT_WO… in MDot_kernel()1090 if (tx < N) results[bx + tx * gdx] = shmem[tx * MDOT_WORKGROUP_SIZE]; in MDot_kernel()1119 …for (auto j = i * MDOT_WORKGROUP_SIZE; j < (i + 1) * MDOT_WORKGROUP_SIZE; ++j) z_sum += results[j]; in sum_kernel()1124 if (size >= MDOT_WORKGROUP_SIZE) __syncthreads(); in sum_kernel()1161 MDOT_WORKGROUP_NUM, MDOT_WORKGROUP_SIZE, 0, stream, in MDot_kernel_dispatch_()1263 #undef MDOT_WORKGROUP_SIZE