| /external/tensorflow/tensorflow/compiler/xla/service/gpu/ |
| D | launch_dimensions.cc | 38 int64 threads_per_block = gpu_device_info.threads_per_block_limit; in ThreadsPerBlockLimit() local
39 if (threads_per_block <= 0) { in ThreadsPerBlockLimit()
47 threads_per_block = gpu_device_info.threads_per_warp; in ThreadsPerBlockLimit()
48 if (threads_per_block == 0) { in ThreadsPerBlockLimit()
50 threads_per_block = 32; in ThreadsPerBlockLimit()
53 return threads_per_block; in ThreadsPerBlockLimit()
79 int64 threads_per_block = ThreadsPerBlockLimit(gpu_device_info); in CalculateLaunchDimensions() local
84 threads_per_block = in CalculateLaunchDimensions()
85 RoundUpToNearest(threads_per_block / unroll_factor, int64{32}); in CalculateLaunchDimensions()
86 if (num_elements < threads_per_block) { in CalculateLaunchDimensions()
[all …]
|
| D | ir_emitter_unnested.h | 632 int threads_per_block);
639 int threads_per_block);
703 llvm::Value* EmitThreadId(int64 threads_per_block, llvm::Type* index_ty);
709 ThreadIdInfo EmitThreadIdInfo(int64 threads_per_block, llvm::Type* index_ty,
|
| D | elemental_ir_emitter.cc | 313 llvm::Value* threads_per_block = IntCast( in EmitThreadId() local
316 return NSWAdd(NSWMul(block_id, threads_per_block), thread_id_in_block); in EmitThreadId()
|
| D | ir_emitter_unnested.cc | 4321 int threads_per_block) { in EmitFullWarpShuffleDownLoopForAllReduces() argument
4326 partial_result_addresses[i], threads_per_block); in EmitFullWarpShuffleDownLoopForAllReduces()
4332 llvm::Value* partial_result_address, int threads_per_block) { in EmitFullWarpShuffleDownLoopForReduce() argument
4334 CHECK_EQ(threads_per_block % 32, 0); in EmitFullWarpShuffleDownLoopForReduce()
4669 llvm::Value* IrEmitterUnnested::EmitThreadId(int64 threads_per_block, in EmitThreadId() argument
4675 llvm_ir::AddRangeMetadata(0, threads_per_block, thread_id_raw); in EmitThreadId()
4681 int64 threads_per_block, llvm::Type* index_ty, int64 num_threads_x) { in EmitThreadIdInfo() argument
4685 llvm::Value* thread_id = EmitThreadId(threads_per_block, index_ty); in EmitThreadIdInfo()
|
| /external/tensorflow/tensorflow/core/kernels/image/ |
| 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 | 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 ()()
|
| /external/tensorflow/tensorflow/stream_executor/ |
| D | device_description.cc | 155 int64 element_count, int64 *threads_per_block, in CalculateDimensionality() argument
157 *threads_per_block = device_description.threads_per_block_limit(); in CalculateDimensionality()
158 *block_count = port::MathUtil::CeilOfRatio(element_count, *threads_per_block); in CalculateDimensionality()
160 CHECK_LE(element_count, *threads_per_block); in CalculateDimensionality()
161 *threads_per_block = element_count; in CalculateDimensionality()
|
| D | device_description.h | 349 int64 element_count, int64 *threads_per_block,
|
| /external/tensorflow/tensorflow/core/kernels/ |
| D | reduction_gpu_kernels.cu.h | 721 const int threads_per_block = 128;
722 const int warps_per_block = threads_per_block / TF_RED_WARPSIZE;
726 threads_per_block, 0, cu_stream, in, out,
862 int threads_per_block = 128;
863 int num_blocks = Eigen::divup(extent_y, threads_per_block);
866 num_blocks, threads_per_block, 0, cu_stream, in,
875 int threads_per_block = 128;
877 (extent_x * extent_z + threads_per_block - 1) / threads_per_block;
882 num_blocks, threads_per_block, 0, cu_stream, in,
890 int threads_per_block = 128;
[all …]
|
| /external/tensorflow/tensorflow/stream_executor/gpu/ |
| D | redzone_allocator.cc | 223 int64 threads_per_block = std::min( in RunRedzoneChecker() local
226 tensorflow::MathUtil::CeilOfRatio(num_elements, threads_per_block); in RunRedzoneChecker()
228 stream->ThenLaunch(ThreadDim(threads_per_block), BlockDim(block_count), in RunRedzoneChecker()
|
| D | gpu_driver.h | 569 GpuContext* context, GpuFunctionHandle kernel, int threads_per_block,
|
| /external/mesa3d/src/gallium/drivers/radeonsi/ |
| D | si_get.c | 731 unsigned threads_per_block = get_max_threads_per_block(sscreen, ir_type); in si_get_compute_param() local
732 block_size[0] = threads_per_block; in si_get_compute_param()
733 block_size[1] = threads_per_block; in si_get_compute_param()
734 block_size[2] = threads_per_block; in si_get_compute_param()
|
| /external/mesa3d/src/gallium/drivers/r600/ |
| D | r600_pipe_common.c | 958 unsigned threads_per_block = get_max_threads_per_block(rscreen, ir_type); in r600_get_compute_param() local
959 block_size[0] = threads_per_block; in r600_get_compute_param()
960 block_size[1] = threads_per_block; in r600_get_compute_param()
961 block_size[2] = threads_per_block; in r600_get_compute_param()
|
| /external/tensorflow/tensorflow/stream_executor/cuda/ |
| D | cuda_driver.cc | 1624 GpuContext* context, CUfunction kernel, int threads_per_block, in GetMaxOccupiedBlocksPerCore() argument
1631 &max_blocks, kernel, threads_per_block, dynamic_shared_memory_bytes), in GetMaxOccupiedBlocksPerCore()
|
| /external/tensorflow/tensorflow/stream_executor/rocm/ |
| D | rocm_driver.cc | 1308 GpuContext* context, hipFunction_t kernel, int threads_per_block, in GetMaxOccupiedBlocksPerCore() argument
|