| /external/tensorflow/tensorflow/core/kernels/ |
| D | constant_op.cc | 105 REGISTER_KERNEL(GPU, Eigen::half);
106 REGISTER_KERNEL(GPU, bfloat16);
107 REGISTER_KERNEL(GPU, float);
108 REGISTER_KERNEL(GPU, double);
109 REGISTER_KERNEL(GPU, uint8);
110 REGISTER_KERNEL(GPU, int8);
111 REGISTER_KERNEL(GPU, qint8);
112 REGISTER_KERNEL(GPU, uint16);
113 REGISTER_KERNEL(GPU, int16);
114 REGISTER_KERNEL(GPU, qint16);
[all …]
|
| D | depthwise_conv_ops_test.cc | 38 enum class Device { CPU, GPU }; enumerator
42 if (device == Device::GPU) { in Run()
106 TEST_F(DepthwiseConvOpTest, DepthwiseConvFloatGpu) { Run<float>(Device::GPU); } in TEST_F()
108 Run<double>(Device::GPU); in TEST_F()
111 Run<Eigen::half>(Device::GPU); in TEST_F()
|
| D | cwise_op_reciprocal.cc | 22 REGISTER4(UnaryOp, GPU, "Inv", functor::inverse, float, Eigen::half, double,
29 REGISTER3(SimpleBinaryOp, GPU, "InvGrad", functor::inverse_grad, float,
43 REGISTER4(UnaryOp, GPU, "Reciprocal", functor::inverse, float, Eigen::half,
53 REGISTER3(SimpleBinaryOp, GPU, "ReciprocalGrad", functor::inverse_grad, float,
|
| D | cwise_op_igammas.cc | 24 REGISTER2(BinaryOp, GPU, "Igamma", functor::igamma, float, double);
25 REGISTER2(BinaryOp, GPU, "IgammaGradA", functor::igamma_grad_a, float, double);
26 REGISTER2(BinaryOp, GPU, "Igammac", functor::igammac, float, double);
|
| D | cwise_op_div.cc | 32 REGISTER9(BinaryOp, GPU, "Div", functor::div, float, Eigen::half, double, uint8,
34 REGISTER4(BinaryOp, GPU, "TruncateDiv", functor::div, uint8, uint16, int16,
36 REGISTER5(BinaryOp, GPU, "RealDiv", functor::div, float, Eigen::half, double,
38 REGISTER5(BinaryOp, GPU, "DivNoNan", functor::div_no_nan, Eigen::half, float,
|
| /external/tensorflow/tensorflow/compiler/tests/ |
| D | lstm_layer_inference.pbtxt | 6 device: "/device:GPU:*"
31 device: "/device:GPU:*"
53 device: "/device:GPU:*"
76 device: "/device:GPU:*"
107 device: "/device:GPU:*"
120 device: "/device:GPU:*"
133 device: "/device:GPU:*"
144 device: "/device:GPU:*"
182 device: "/device:GPU:*"
214 device: "/device:GPU:*"
[all …]
|
| /external/perfetto/test/trace_processor/ |
| D | gpu_log.out | 2 "gpu_log","GPU Log",1,0,"VERBOSE","message","message0"
3 "gpu_log","GPU Log",1,0,"VERBOSE","tag","tag0"
4 "gpu_log","GPU Log",2,0,"DEBUG","message","message1"
5 "gpu_log","GPU Log",2,0,"DEBUG","tag","tag0"
6 "gpu_log","GPU Log",3,0,"INFO","message","message2"
7 "gpu_log","GPU Log",3,0,"INFO","tag","tag0"
8 "gpu_log","GPU Log",4,0,"ERROR","message","message4"
9 "gpu_log","GPU Log",4,0,"ERROR","tag","tag0"
10 "gpu_log","GPU Log",4,0,"WARNING","message","message3"
11 "gpu_log","GPU Log",4,0,"WARNING","tag","tag0"
[all …]
|
| /external/tensorflow/tensorflow/compiler/jit/tests/ |
| D | keras_imagenet_main.pbtxt | 4 device: "/job:localhost/replica:0/task:0/device:GPU:0"
29 device: "/job:localhost/replica:0/task:0/device:GPU:0"
51 device: "/job:localhost/replica:0/task:0/device:GPU:0"
88 device: "/job:localhost/replica:0/task:0/device:GPU:0"
128 device: "/job:localhost/replica:0/task:0/device:GPU:0"
153 device: "/job:localhost/replica:0/task:0/device:GPU:0"
176 device: "/job:localhost/replica:0/task:0/device:GPU:0"
213 device: "/job:localhost/replica:0/task:0/device:GPU:0"
250 device: "/job:localhost/replica:0/task:0/device:GPU:0"
296 device: "/job:localhost/replica:0/task:0/device:GPU:0"
[all …]
|
| D | keras_imagenet_main_graph_mode.pbtxt | 4 device: "/job:localhost/replica:0/task:0/device:GPU:0"
54 device: "/job:localhost/replica:0/task:0/device:GPU:0"
98 device: "/job:localhost/replica:0/task:0/device:GPU:0"
139 device: "/job:localhost/replica:0/task:0/device:GPU:0"
180 device: "/job:localhost/replica:0/task:0/device:GPU:0"
221 device: "/job:localhost/replica:0/task:0/device:GPU:0"
262 device: "/job:localhost/replica:0/task:0/device:GPU:0"
303 device: "/job:localhost/replica:0/task:0/device:GPU:0"
344 device: "/job:localhost/replica:0/task:0/device:GPU:0"
385 device: "/job:localhost/replica:0/task:0/device:GPU:0"
[all …]
|
| D | opens2s_gnmt_mixed_precision.pbtxt.gz | |
| /external/skia/src/gpu/vk/ |
| D | GrVkUtil.h | 24 #define GR_VK_CALL_RESULT(GPU, RESULT, X) \ argument
26 (RESULT) = GR_VK_CALL(GPU->vkInterface(), X); \
28 if (RESULT != VK_SUCCESS && !GPU->isDeviceLost()) { \
32 GPU->setDeviceLost(); \
36 #define GR_VK_CALL_RESULT_NOCHECK(GPU, RESULT, X) \ argument
38 (RESULT) = GR_VK_CALL(GPU->vkInterface(), X); \
40 GPU->setDeviceLost(); \
45 #define GR_VK_CALL_ERRCHECK(GPU, X) \ argument
47 GR_VK_CALL_RESULT(GPU, SK_MACRO_APPEND_LINE(ret), X) \
|
| /external/llvm/lib/Target/AMDGPU/ |
| D | AMDGPUSubtarget.cpp | 38 StringRef GPU, StringRef FS) { in initializeSubtargetDependencies() argument
53 ParseSubtargetFeatures(GPU, FullFS); in initializeSubtargetDependencies()
70 AMDGPUSubtarget::AMDGPUSubtarget(const Triple &TT, StringRef GPU, StringRef FS, in AMDGPUSubtarget() argument
72 : AMDGPUGenSubtargetInfo(TT, GPU, FS), in AMDGPUSubtarget()
119 InstrItins(getInstrItineraryForCPU(GPU)) { in AMDGPUSubtarget()
120 initializeSubtargetDependencies(TT, GPU, FS); in AMDGPUSubtarget()
181 R600Subtarget::R600Subtarget(const Triple &TT, StringRef GPU, StringRef FS, in R600Subtarget() argument
183 AMDGPUSubtarget(TT, GPU, FS, TM), in R600Subtarget()
188 SISubtarget::SISubtarget(const Triple &TT, StringRef GPU, StringRef FS, in SISubtarget() argument
190 AMDGPUSubtarget(TT, GPU, FS, TM), in SISubtarget()
|
| D | AMDGPUTargetMachine.cpp | 125 static StringRef getGPUOrDefault(const Triple &TT, StringRef GPU) { in getGPUOrDefault() argument
126 if (!GPU.empty()) in getGPUOrDefault()
127 return GPU; in getGPUOrDefault()
185 StringRef GPU = getGPUName(F); in getSubtargetImpl() local
188 SmallString<128> SubtargetKey(GPU); in getSubtargetImpl()
197 I = llvm::make_unique<R600Subtarget>(TargetTriple, GPU, FS, *this); in getSubtargetImpl()
226 StringRef GPU = getGPUName(F); in getSubtargetImpl() local
229 SmallString<128> SubtargetKey(GPU); in getSubtargetImpl()
238 I = llvm::make_unique<SISubtarget>(TargetTriple, GPU, FS, *this); in getSubtargetImpl()
|
| /external/tensorflow/tensorflow/lite/g3doc/performance/ |
| D | gpu.md | 1 # TensorFlow Lite GPU delegate
4 accelerators. This document describes how to use the GPU backend using the
11 resulting in lower latency. In the best scenario, inference on the GPU may now
17 Another benefit with GPU inference is its power efficiency. GPUs carry out the
23 The easiest way to try out the GPU delegate is to follow the below tutorials,
24 which go through building our classification demo applications with GPU support.
25 The GPU code is only binary for now; it will be open-sourced soon. Once you
32 [GPU Delegate for Android](https://youtu.be/Xkhgre8r5G0) video.
42 #### Step 2. Edit `app/build.gradle` to use the nightly GPU AAR
58 enabling the GPU. Change from quantized to a float model and then click GPU to
[all …]
|
| D | gpu_advanced.md | 1 # TensorFlow Lite on GPU
4 hardware accelerators. This document describes how to use the GPU backend using
8 ## Benefits of GPU Acceleration
17 on the GPU may run fast enough to become suitable for real-time applications
25 neural network on a GPU may eliminate this concern.
29 Another benefit that comes with GPU inference is its power efficiency. A GPU
35 TensorFlow Lite on GPU supports the following ops in 16-bit and 32-bit float
62 Run TensorFlow Lite on GPU with `TfLiteDelegate`. In Java, you can specify the
66 // NEW: Prepare GPU delegate.
84 For C/C++ usage of TensorFlow Lite GPU on Android, the GPU delegate can be
[all …]
|
| /external/perfetto/docs/data-sources/ |
| D | gpu.md | 1 # GPU chapter
5 ## GPU Frequency
7 GPU frequency can be included in the trace by adding the ftrace category.
20 ## GPU Counters
22 GPU counters can be configured by adding the data source to the trace config as follows:
|
| /external/tensorflow/tensorflow/lite/delegates/gpu/ |
| D | README.md | 1 # TFLite on GPU
4 describes how to use the GPU backend using the TFLite delegate APIs on Android
11 resulting in lower latency. In the best scenario, inference on the GPU may now
19 net models on the GPU.
21 Another benefit that comes with GPU inference is its power efficiency. GPUs
26 TFLite on GPU supports the following ops in 16-bit and 32-bit float precision:
51 Using TFLite on GPU is as simple as getting the GPU delegate via
65 // NEW: Prepare GPU delegate.
89 TFLite GPU backend uses OpenGL ES 3.1 compute shaders or OpenCL.
104 There are GPU options that can be set and passed on to
[all …]
|
| /external/tensorflow/tensorflow/python/eager/ |
| D | benchmarks_test.py | 67 GPU = "/device:GPU:0" variable
195 if device == GPU:
294 self._benchmark_create_tensor([[3.0]], dtypes.float32.as_datatype_enum, GPU)
301 GPU)
307 self._benchmark_create_tensor([[3]], dtypes.int32.as_datatype_enum, GPU)
314 np.array([[3]], dtype=np.int32), dtypes.int32.as_datatype_enum, GPU)
352 with context.device(GPU):
364 with context.device(GPU):
578 with context.device(GPU):
586 with context.device(GPU):
[all …]
|
| /external/angle/src/feature_support_util/ |
| D | feature_support_util.cpp | 425 class GPU class
428 GPU(const std::string vendor, uint32_t deviceId, const Version &version) in GPU() function in angle::GPU
431 GPU(const std::string vendor, uint32_t deviceId) in GPU() function in angle::GPU
434 GPU(const std::string vendor) : mVendor(vendor), mDeviceId(), mVersion(), mWildcard(false) {} in GPU() function in angle::GPU
435 GPU() : mVendor(), mDeviceId(), mVersion(), mWildcard(true) {} in GPU() function in angle::GPU
436 bool match(const GPU &toCheck) const in match()
452 ~GPU() {} in ~GPU()
454 static GPU *CreateGpuFromJson(const Json::Value &jObject) in CreateGpuFromJson()
456 GPU *gpu = nullptr; in CreateGpuFromJson()
469 gpu = new GPU(vendor, deviceId, *version); in CreateGpuFromJson()
[all …]
|
| /external/tensorflow/tensorflow/core/protobuf/ |
| D | config.proto | 19 // Fraction of the available GPU memory to allocate for each process.
20 // 1 means to allocate all of the GPU memory, 0.5 means the process
21 // allocates up to ~50% of the available GPU memory.
23 // GPU memory is pre-allocated unless the allow_growth option is enabled.
26 // the amount of memory available on the GPU device by using host memory as a
39 // GPU memory region, instead starting small and growing as needed.
42 // The type of GPU allocation strategy to use.
57 // A comma-separated list of GPU ids that determines the 'visible'
58 // to 'virtual' mapping of GPU devices. For example, if TensorFlow
59 // can see 8 GPU devices in the process, and one wanted to map
[all …]
|
| /external/tensorflow/tensorflow/core/api_def/base_api/ |
| D | api_def_Copy.pbtxt | 26 summary: "Copy a tensor from CPU-to-CPU or GPU-to-GPU."
28 Performs CPU-to-CPU or GPU-to-GPU deep-copying of tensor, depending on the
|
| /external/tensorflow/tensorflow/compiler/xla/g3doc/ |
| D | custom_call.md | 57 ## Custom-call on GPU
59 The GPU custom call framework is somewhat different than that on the CPU. Here
85 Notice first that the GPU custom call function *is still a function executed on
87 on the GPU. Here it launches a CUDA kernel, but it could also do something else,
91 contains points to device (i.e. GPU) memory. The parameters come first, followed
102 operands to our custom call, their values would live in GPU memory. We'd then
117 serialized proto inside of `opaque` and deserialize it within your GPU
145 On both CPU and GPU, a tuple is represented in memory as an array of pointers.
150 // and GPU.
166 Although the in-memory representation of tuples is the same in CPU and GPU, they
[all …]
|
| /external/mesa3d/docs/drivers/ |
| D | vc4.rst | 5 Broadcom's VideoCore IV GPU. It is notably used in the Raspberry Pi 0
16 GPU block from Linux.
76 For GPU hangs, if you can get a short apitrace that produces the
99 (GPU) Initialize the per-tile draw call lists to empty.
100 (GPU) Run all draw calls collecting vertex data
101 (GPU) For each tile covered by a draw call's primitive.
117 from memory, in favor of a cheap GPU-side ``memset()`` of the tile
155 * Increasing GPU memory Increase CMA pool size
192 * Step 2: CPU vs GPU
195 application, the CPU is busy in the GL driver, the GPU is waiting for
[all …]
|
| /external/angle/extensions/ |
| D | EGL_ANGLE_power_preference.txt | 36 This extension allows selection of the high- or low-power GPU on
37 dual-GPU systems, specifically on macOS.
65 created on the integrated (low-power) or discrete (high-power) GPU
66 on dual-GPU systems. EGL_POWER_PREFERENCE_ANGLE is only a legal
|
| /external/swiftshader/third_party/llvm-7.0/llvm/lib/Target/AMDGPU/ |
| D | AMDGPUSubtarget.cpp | 48 StringRef GPU, StringRef FS) { in initializeSubtargetDependencies() argument
51 ParseSubtargetFeatures(GPU, FullFS); in initializeSubtargetDependencies()
68 StringRef GPU, StringRef FS) { in initializeSubtargetDependencies() argument
94 ParseSubtargetFeatures(GPU, FullFS); in initializeSubtargetDependencies()
145 GCNSubtarget::GCNSubtarget(const Triple &TT, StringRef GPU, StringRef FS, in GCNSubtarget() argument
147 AMDGPUGenSubtargetInfo(TT, GPU, FS), in GCNSubtarget()
211 InstrInfo(initializeSubtargetDependencies(TT, GPU, FS)), in GCNSubtarget()
447 R600Subtarget::R600Subtarget(const Triple &TT, StringRef GPU, StringRef FS, in R600Subtarget() argument
449 R600GenSubtargetInfo(TT, GPU, FS), in R600Subtarget()
462 TLInfo(TM, initializeSubtargetDependencies(TT, GPU, FS)), in R600Subtarget()
[all …]
|