| /external/tensorflow/tensorflow/compiler/xla/service/gpu/ |
| D | partition_assignment.cc | 38 launch_dims.threads_per_block()); in operator <<()
43 int64 threads_per_block = device_desc.threads_per_block_limit(); in ThreadsPerBlockLimit() local
44 if (threads_per_block <= 0) { in ThreadsPerBlockLimit()
52 threads_per_block = device_desc.threads_per_warp(); in ThreadsPerBlockLimit()
53 if (threads_per_block == 0) { in ThreadsPerBlockLimit()
55 threads_per_block = 32; in ThreadsPerBlockLimit()
58 return threads_per_block; in ThreadsPerBlockLimit()
84 int64 threads_per_block = ThreadsPerBlockLimit(device_desc); in CalculateLaunchDimensions() local
85 if (num_elements < threads_per_block) { in CalculateLaunchDimensions()
86 threads_per_block = num_elements; in CalculateLaunchDimensions()
[all …]
|
| D | partition_assignment.h | 41 LaunchDimensions(int64 block_count, int64 threads_per_block) in LaunchDimensions() argument
42 : block_count_(block_count), threads_per_block_(threads_per_block) {} in LaunchDimensions()
49 int64 threads_per_block() const { return threads_per_block_; } in threads_per_block() function
50 int64 launch_bound() const { return block_count() * threads_per_block(); } in launch_bound()
|
| D | parallel_loop_emitter.cc | 87 llvm_ir::AddRangeMetadata(0, launch_dimensions_.threads_per_block(), in EmitIndexAndSetExitBasicBlock()
94 index_type, launch_dimensions_.threads_per_block()), in EmitIndexAndSetExitBasicBlock()
112 launch_dimensions_.threads_per_block() * in EmitIndexAndSetExitBasicBlock()
|
| D | kernel_thunk.cc | 103 stream, se::ThreadDim(launch_dimensions.threads_per_block()), in ExecuteOnStream()
|
| D | elemental_ir_emitter.cc | 321 llvm::Value* threads_per_block = in EmitThreadId() local
325 return NSWAdd(NSWMul(block_id, threads_per_block), thread_id_in_block); in EmitThreadId()
|
| D | ir_emitter_unnested.cc | 162 launch_dims.threads_per_block()); in UpdateLaunchDimensions()
|
| /external/tensorflow/tensorflow/core/kernels/ |
| D | adjust_saturation_op_gpu.cu.cc | 36 const int threads_per_block = config.thread_per_block; in operator ()() local
38 (number_of_elements + threads_per_block - 1) / threads_per_block; in operator ()()
40 block_count, threads_per_block, 0, stream, in operator ()()
|
| D | adjust_hue_op_gpu.cu.cc | 35 const int threads_per_block = config.thread_per_block; in operator ()() local
37 (number_of_elements + threads_per_block - 1) / threads_per_block; in operator ()()
39 block_count, threads_per_block, 0, stream, in operator ()()
|
| D | reduction_gpu_kernels.cu.h | 578 const int threads_per_block = 128;
579 const int warps_per_block = threads_per_block / 32;
583 threads_per_block, 0, cu_stream, in, out,
709 int threads_per_block = 128;
710 int num_blocks = Eigen::divup(extent_y, threads_per_block);
713 num_blocks, threads_per_block, 0, cu_stream,
722 int threads_per_block = 128;
724 (extent_x * extent_z + threads_per_block - 1) / threads_per_block;
729 num_blocks, threads_per_block, 0, cu_stream, in,
|
| /external/tensorflow/tensorflow/contrib/tensorrt/custom_plugin_examples/ |
| D | inc_op_kernel.cu.cc | 39 int threads_per_block = 256; in IncrementKernel() local
40 int blocks_per_grid = (count + threads_per_block - 1) / threads_per_block; in IncrementKernel()
42 TF_CHECK_OK(CudaLaunchKernel(VecInc, threads_per_block, blocks_per_grid, 0, in IncrementKernel()
|
| /external/tensorflow/tensorflow/stream_executor/ |
| D | device_description.cc | 151 int64 element_count, int64 *threads_per_block, in CalculateDimensionality() argument
153 *threads_per_block = device_description.threads_per_block_limit(); in CalculateDimensionality()
154 *block_count = port::MathUtil::CeilOfRatio(element_count, *threads_per_block); in CalculateDimensionality()
156 CHECK_LE(element_count, *threads_per_block); in CalculateDimensionality()
157 *threads_per_block = element_count; in CalculateDimensionality()
|
| D | device_description.h | 343 int64 element_count, int64 *threads_per_block,
|
| /external/mesa3d/src/gallium/drivers/radeonsi/ |
| D | si_get.c | 785 unsigned threads_per_block = get_max_threads_per_block(sscreen, ir_type); in si_get_compute_param() local
786 block_size[0] = threads_per_block; in si_get_compute_param()
787 block_size[1] = threads_per_block; in si_get_compute_param()
788 block_size[2] = threads_per_block; in si_get_compute_param()
|
| /external/mesa3d/src/gallium/drivers/r600/ |
| D | r600_pipe_common.c | 1041 unsigned threads_per_block = get_max_threads_per_block(rscreen, ir_type); in r600_get_compute_param() local
1042 block_size[0] = threads_per_block; in r600_get_compute_param()
1043 block_size[1] = threads_per_block; in r600_get_compute_param()
1044 block_size[2] = threads_per_block; in r600_get_compute_param()
|
| /external/tensorflow/tensorflow/stream_executor/gpu/ |
| D | gpu_driver.h | 497 GpuContext* context, GpuFunctionHandle kernel, int threads_per_block,
|
| /external/tensorflow/tensorflow/stream_executor/cuda/ |
| D | cuda_driver.cc | 1613 GpuContext* context, CUfunction kernel, int threads_per_block, in GetMaxOccupiedBlocksPerCore() argument
1619 &max_blocks, kernel, threads_per_block, dynamic_shared_memory_bytes); in GetMaxOccupiedBlocksPerCore()
|
| /external/tensorflow/tensorflow/stream_executor/rocm/ |
| D | rocm_driver.cc | 1356 GpuContext* context, hipFunction_t kernel, int threads_per_block, in GetMaxOccupiedBlocksPerCore() argument
|