| /aosp_15_r20/external/tensorflow/tensorflow/core/profiler/utils/ |
| H A D | hardware_type_utils.cc | 35 switch (device_cap.compute_capability().major()) { in GetFmaMaxThroughputPerSMPerCycle()
50 if (device_cap.compute_capability().minor() > 0) { in GetFmaMaxThroughputPerSMPerCycle()
65 if (device_cap.compute_capability().minor() >= 6) { in GetFmaMaxThroughputPerSMPerCycle()
85 if (device_cap.compute_capability().major() <= 9) { in GetFmaMaxThroughputPerSMPerCycle()
108 switch (device_cap.compute_capability().major()) { in GpuModelName()
118 if (device_cap.compute_capability().minor() < 5) { in GpuModelName()
129 switch (device_cap.compute_capability().major()) { in GpuModelName()
|
| H A D | device_caps_utils.cc | 45 caps.compute_capability().major()); in SetDeviceCaps()
48 caps.compute_capability().minor()); in SetDeviceCaps()
|
| /aosp_15_r20/external/tensorflow/tensorflow/compiler/xla/service/gpu/llvm_gpu_backend/ |
| H A D | gpu_backend_lib.cc | 89 static std::string GetSmName(se::CudaComputeCapability compute_capability) { in GetSmName() argument
91 compute_capability.major * 10 + compute_capability.minor; in GetSmName()
111 << compute_capability.ToString() in GetSmName()
360 llvm::Triple target_triple, se::CudaComputeCapability compute_capability, in NVPTXGetTargetMachine() argument
364 return GetTargetMachine(target_triple, GetSmName(compute_capability), in NVPTXGetTargetMachine()
535 auto compute_capability = in CompileToPtx() local
537 if (!compute_capability) { in CompileToPtx()
545 default_target_triple, *compute_capability, hlo_module_config); in CompileToPtx()
793 auto compute_capability = in AMDGPUTargetModuleLinker() local
795 if (!compute_capability) { in AMDGPUTargetModuleLinker()
[all …]
|
| /aosp_15_r20/external/tensorflow/tensorflow/compiler/xla/service/gpu/ |
| H A D | cudnn_support_utils.cc | 28 const se::CudaComputeCapability& compute_capability, in CudnnSupportsOptimizedIntegerConvolution() argument
45 if ((vector_size == 32 && !compute_capability.IsAtLeast(7, 5)) || in CudnnSupportsOptimizedIntegerConvolution()
46 !compute_capability.IsAtLeast(6, 1)) { in CudnnSupportsOptimizedIntegerConvolution()
47 VLOG(3) << "Compute capability " << compute_capability.ToString() in CudnnSupportsOptimizedIntegerConvolution()
|
| H A D | cudnn_vectorize_convolutions.h | 49 se::CudaComputeCapability compute_capability) in CudnnVectorizeConvolutions() argument
50 : compute_capability_(compute_capability) {} in CudnnVectorizeConvolutions()
|
| H A D | cudnn_pad_for_convolutions.h | 34 explicit CudnnPadForConvolutions(se::CudaComputeCapability compute_capability) in CudnnPadForConvolutions() argument
35 : compute_capability_(compute_capability) {} in CudnnPadForConvolutions()
|
| H A D | cudnn_vectorize_convolutions.cc | 261 const se::CudaComputeCapability& compute_capability, in TryRevectorizeConv() argument
297 compute_capability, *conv, vect_size)); in TryRevectorizeConv()
399 const se::CudaComputeCapability& compute_capability, in TryVectorizeConv() argument
424 compute_capability, *conv, vect_size)); in TryVectorizeConv()
|
| H A D | cudnn_simplify_padding_test.cc | 46 StatusOr<bool> RunEndToEnd(std::pair<int, int> compute_capability, in RunEndToEnd() argument
48 se::CudaComputeCapability cc{compute_capability.first, in RunEndToEnd()
49 compute_capability.second}; in RunEndToEnd()
|
| H A D | cudnn_vectorize_convolutions_test.cc | 37 StatusOr<bool> Run(std::pair<int, int> compute_capability, in Run() argument
40 compute_capability.first, compute_capability.second}); in Run()
|
| H A D | cudnn_support_utils.h | 35 const se::CudaComputeCapability& compute_capability,
|
| H A D | cudnn_pad_for_convolutions.cc | 295 int pad_to, const se::CudaComputeCapability& compute_capability, in TryResolvePaddedShapesForIntegerConvolution() argument
351 compute_capability, *conv, pad_to)); in TryResolvePaddedShapesForIntegerConvolution()
|
| /aosp_15_r20/external/pytorch/c10/cuda/ |
| H A D | CUDADeviceAssertionHost.cpp | 50 int compute_capability = -1; in dsa_get_device_compute_capability() local
52 &compute_capability, cudaDevAttrComputeCapabilityMajor, device_num)); in dsa_get_device_compute_capability()
53 return compute_capability; in dsa_get_device_compute_capability()
|
| /aosp_15_r20/external/tensorflow/tensorflow/core/grappler/optimizers/ |
| H A D | generic_layout_optimizer.cc | 63 double compute_capability = 0.0; in GetNumGPUs() local
64 if (absl::SimpleAtod(compute_capability_it->second, &compute_capability) && in GetNumGPUs()
65 compute_capability >= 7.0) { in GetNumGPUs()
|
| H A D | remapper.cc | 423 double compute_capability = 0.0; in RuntimeFusionEnabled() local
424 if (absl::SimpleAtod(cc_it->second, &compute_capability) && in RuntimeFusionEnabled()
425 compute_capability >= 8.0) { in RuntimeFusionEnabled()
|
| /aosp_15_r20/external/tensorflow/ |
| H A D | configure.py | 912 for compute_capability in tf_cuda_compute_capabilities.split(','):
913 m = re.match('[0-9]+.[0-9]+', compute_capability)
917 compute_capability)
919 print('Invalid compute capability: %s' % compute_capability)
|
| /aosp_15_r20/external/tensorflow/tensorflow/core/profiler/protobuf/ |
| H A D | hardware_types.proto | 29 GPUComputeCapability compute_capability = 5; field
|
| /aosp_15_r20/external/tensorflow/tensorflow/core/protobuf/ |
| H A D | autotuning.proto | 98 ComputeCapability compute_capability = 4; field
|
| /aosp_15_r20/external/tensorflow/tensorflow/core/kernels/ |
| H A D | depthwise_conv_op_gpu.h | 677 se::CudaComputeCapability compute_capability =
680 return compute_capability.IsAtLeast(5, 3) &&
681 compute_capability != se::CudaComputeCapability{6, 1};
|
| /aosp_15_r20/external/tensorflow/tensorflow/python/framework/ |
| H A D | test_util.py | 1889 cc = gpu_info.compute_capability or (0, 0)
|