gpu - update gen non-tensor block strategy

Author: jeremylt

backends/cuda-gen/ceed-cuda-gen-operator.c

@@ -282,8 +282,15 @@ static int CeedOperatorApplyAdd_Cuda_gen(CeedOperator op, CeedVector input_vec,
   CeedCallCuda(ceed, cuOccupancyMaxPotentialBlockSize(&min_grid_size, &max_threads_per_block, data->op, dynamicSMemSize, 0, 0x10000));
   int block[3] = {thread_1d, ((!is_tensor || dim == 1) ? 1 : thread_1d), -1};
 
-  CeedCallBackend(BlockGridCalculate(num_elem, min_grid_size / cuda_data->device_prop.multiProcessorCount, max_threads_per_block,
-                                     cuda_data->device_prop.maxThreadsDim[2], cuda_data->device_prop.warpSize, block, &grid));
+  if (is_tensor) {
+    CeedCallBackend(BlockGridCalculate(num_elem, min_grid_size / cuda_data->device_prop.multiProcessorCount, max_threads_per_block,
+                                       cuda_data->device_prop.maxThreadsDim[2], cuda_data->device_prop.warpSize, block, &grid));
+  } else {
+    CeedInt elems_per_block = CeedIntMin(cuda_data->device_prop.maxThreadsDim[2], CeedIntMax(512 / thread_1d, 1));
+
+    grid     = num_elem / elems_per_block + (num_elem % elems_per_block > 0);
+    block[2] = elems_per_block;
+  }
   CeedInt shared_mem = block[0] * block[1] * block[2] * sizeof(CeedScalar);
 
   CeedCallBackend(CeedRunKernelDimShared_Cuda(ceed, data->op, grid, block[0], block[1], block[2], shared_mem, opargs));

backends/hip-gen/ceed-hip-gen-operator.c

@@ -203,10 +203,18 @@ static int CeedOperatorApplyAdd_Hip_gen(CeedOperator op, CeedVector input_vec, C
   const CeedInt Q_1d      = data->Q_1d;
   const CeedInt P_1d      = data->max_P_1d;
   const CeedInt thread_1d = CeedIntMax(Q_1d, P_1d);
-  CeedInt       block_sizes[3];
 
   CeedCallBackend(CeedOperatorHasTensorBases(op, &is_tensor));
-  CeedCallBackend(BlockGridCalculate_Hip_gen(is_tensor ? dim : 1, num_elem, P_1d, Q_1d, block_sizes));
+  CeedInt block_sizes[3] = {thread_1d, ((!is_tensor || dim == 1) ? 1 : thread_1d), -1};
+
+  if (is_tensor) {
+    CeedCallBackend(BlockGridCalculate_Hip_gen(is_tensor ? dim : 1, num_elem, P_1d, Q_1d, block_sizes));
+  } else {
+    CeedInt elems_per_block = 64 * thread_1d > 256 ? 256 / thread_1d : 64;
+
+    elems_per_block = elems_per_block > 0 ? elems_per_block : 1;
+    block_sizes[2]  = elems_per_block;
+  }
   if (dim == 1 || !is_tensor) {
     CeedInt grid      = num_elem / block_sizes[2] + ((num_elem / block_sizes[2] * block_sizes[2] < num_elem) ? 1 : 0);
     CeedInt sharedMem = block_sizes[2] * thread_1d * sizeof(CeedScalar);