| /external/tensorflow/tensorflow/compiler/xla/service/gpu/ |
| fft_thunk.h | 20 #include "tensorflow/compiler/xla/service/gpu/buffer_allocations.h"
21 #include "tensorflow/compiler/xla/service/gpu/gpu_executable.h"
22 #include "tensorflow/compiler/xla/service/gpu/thunk.h"
31 namespace gpu { namespace in namespace:xla
95 } // namespace gpu
|
| gpu_executable.h | 25 #include "tensorflow/compiler/xla/service/gpu/buffer_allocations.h"
26 #include "tensorflow/compiler/xla/service/gpu/stream_assignment.h"
27 #include "tensorflow/compiler/xla/service/gpu/thunk.h"
28 #include "tensorflow/compiler/xla/service/gpu/thunk_schedule.h"
41 namespace gpu { namespace in namespace:xla
43 // GPU-targeting implementation of the XLA Executable interface.
51 // compilation up to the GPU driver.
72 // empty, in which case compilation is left up to the GPU driver.
87 // TODO(b/62952745) Implement equality test on GPU executable.
88 return Unimplemented("Equality test on GPU executable is not implemented.")
[all...] |
| infeed_manager.h | 17 // buffer that are used by the GPU runtime to transfer buffers into an
18 // executing GPU computation, e.g., to feed data into a while loop.
31 namespace gpu { namespace in namespace:xla
33 // TODO(b/30467474) Once GPU infeed implementation settles, consider
34 // folding back the cpu and gpu infeed implementations into a generic
40 // * Buffer space on GPU is allocated on every infeed enqueue request,
130 // Singleton creator-or-accessor: Returns the GPU infeed manager.
133 } // namespace gpu
|
| ir_emission_utils.h | 29 namespace gpu { namespace in namespace:xla
136 } // namespace gpu
|
| partition_assignment.cc | 16 #include "tensorflow/compiler/xla/service/gpu/partition_assignment.h"
35 namespace gpu { namespace in namespace:xla
55 // smallest block size that allows the GPU to reach full occupancy (assuming
73 LOG(WARNING) << "Attempting to calculate launch dimensions for GPU "
100 } // namespace gpu
|
| thunk_schedule.h | 24 #include "tensorflow/compiler/xla/service/gpu/stream_assignment.h"
25 #include "tensorflow/compiler/xla/service/gpu/thunk.h"
30 namespace gpu { namespace in namespace:xla
90 } // namespace gpu
|
| while_thunk.cc | 16 #include "tensorflow/compiler/xla/service/gpu/while_thunk.h"
23 namespace gpu { namespace in namespace:xla
74 } // namespace gpu
|
| /external/tensorflow/tensorflow/compiler/xla/service/llvm_ir/ |
| ops.cc | 17 #include "tensorflow/compiler/xla/service/gpu/parallel_loop_emitter.h"
18 #include "tensorflow/compiler/xla/service/gpu/partition_assignment.h"
42 const gpu::LaunchDimensions* launch_dimensions,
78 return gpu::ParallelLoopEmitter(loop_body_emitter, update_shape,
119 const gpu::LaunchDimensions* launch_dimensions,
173 const gpu::LaunchDimensions& launch_dimensions,
|
| ops.h | 21 #include "tensorflow/compiler/xla/service/gpu/partition_assignment.h"
87 const gpu::LaunchDimensions& launch_dimensions,
|
| /external/tensorflow/tensorflow/core/common_runtime/gpu/ |
| gpu_util.h | 30 namespace gpu = ::perftools::gputools;
34 // "tensor" is GPU-local. "dev" is the hosting GPU.
35 // "device_context" should be the context of the GPU "_Send" op
38 // bytes from GPU to CPU RAM. "is_dead" indicates that the
60 // corresponding GPU device at the time of call have completed.
66 // (up to a limit). "device" can be either a CPU or a GPU device.
104 // Deep-copying of GPU tensor on the same device.
|
| gpu_util.cc | 16 #include "tensorflow/core/common_runtime/gpu/gpu_util.h"
21 #include "tensorflow/core/common_runtime/gpu/gpu_event_mgr.h"
22 #include "tensorflow/core/common_runtime/gpu/process_state.h"
63 namespace gpu = ::perftools::gputools;
68 gpu::Stream** stream) {
82 return errors::Internal("No gpu stream is available.");
104 return errors::Internal("GPU copy from non-DMA ",
123 gpu::Stream* send_stream = nullptr;
135 done(errors::Internal("No send gpu copy-out-stream is available."));
141 // Tensor values need to be copied from GPU to CPU ram so tha
[all...] |
| /external/tensorflow/tensorflow/core/kernels/ |
| adjust_contrast_op_benchmark_test.cc | 52 // The GPU test ran with -c opt --config=cuda --copt=-mavx, CPU ran without
55 // GPU: Tesla K40m
60 BM_AdjustContrastDev(gpu, 32, 299, 299);
|
| multinomial_op_test.cc | 52 BM_MultinomialDev(gpu, B, C, S);
|
| /frameworks/native/vulkan/tools/ |
| vkinfo.cpp | 106 void EnumerateDeviceExtensions(VkPhysicalDevice gpu,
112 vkEnumerateDeviceExtensionProperties(gpu, layer_name, &count, nullptr);
117 result = vkEnumerateDeviceExtensionProperties(gpu, layer_name, &count,
124 void GatherGpuInfo(VkPhysicalDevice gpu,
130 vkGetPhysicalDeviceProperties(gpu, &info.properties);
131 vkGetPhysicalDeviceMemoryProperties(gpu, &info.memory);
132 vkGetPhysicalDeviceFeatures(gpu, &info.features);
134 vkGetPhysicalDeviceQueueFamilyProperties(gpu, &count, nullptr);
136 vkGetPhysicalDeviceQueueFamilyProperties(gpu, &count,
139 result = vkEnumerateDeviceLayerProperties(gpu, &count, nullptr)
[all...] |
| /external/skia/src/gpu/ |
| GrBufferAllocPool.cpp | 34 TRACE_EVENT_INSTANT1("skia.gpu", \
42 GrBufferAllocPool::GrBufferAllocPool(GrGpu* gpu, GrBufferType bufferType, size_t blockSize)
45 fGpu = SkRef(gpu);
53 fBufferMapThreshold = gpu->caps()->bufferMapThreshold();
381 GrVertexBufferAllocPool::GrVertexBufferAllocPool(GrGpu* gpu)
382 : GrBufferAllocPool(gpu, kVertex_GrBufferType, MIN_VERTEX_BUFFER_SIZE) {
436 GrIndexBufferAllocPool::GrIndexBufferAllocPool(GrGpu* gpu)
437 : GrBufferAllocPool(gpu, kIndex_GrBufferType, MIN_INDEX_BUFFER_SIZE) {
|
| GrDrawingManager.cpp | 158 GrGpu* gpu = fContext->contextPriv().getGpu(); local
160 GrOpFlushState flushState(gpu, fContext->contextPriv().resourceProvider(),
232 GrSemaphoresSubmitted result = gpu->finishFlush(numSemaphores, backendSemaphores);
279 // Upload all data to the GPU
342 GrGpu* gpu = fContext->contextPriv().getGpu(); local
345 if (gpu && surface->asRenderTarget()) {
346 gpu->resolveRenderTarget(surface->asRenderTarget());
|
| /external/skqp/src/gpu/ |
| GrBufferAllocPool.cpp | 34 TRACE_EVENT_INSTANT1("skia.gpu", \
42 GrBufferAllocPool::GrBufferAllocPool(GrGpu* gpu, GrBufferType bufferType, size_t blockSize)
45 fGpu = SkRef(gpu);
53 fBufferMapThreshold = gpu->caps()->bufferMapThreshold();
381 GrVertexBufferAllocPool::GrVertexBufferAllocPool(GrGpu* gpu)
382 : GrBufferAllocPool(gpu, kVertex_GrBufferType, MIN_VERTEX_BUFFER_SIZE) {
436 GrIndexBufferAllocPool::GrIndexBufferAllocPool(GrGpu* gpu)
437 : GrBufferAllocPool(gpu, kIndex_GrBufferType, MIN_INDEX_BUFFER_SIZE) {
|
| GrDrawingManager.cpp | 126 GrGpu* gpu = fContext->contextPriv().getGpu(); local
128 GrOpFlushState flushState(gpu, fContext->contextPriv().resourceProvider(),
207 GrSemaphoresSubmitted result = gpu->finishFlush(numSemaphores, backendSemaphores);
249 // Upload all data to the GPU
312 GrGpu* gpu = fContext->contextPriv().getGpu(); local
315 if (gpu && surface->asRenderTarget()) {
316 gpu->resolveRenderTarget(surface->asRenderTarget(), proxy->origin());
|
| /external/skia/src/gpu/vk/ |
| GrVkResourceProvider.cpp | 25 GrVkResourceProvider::GrVkResourceProvider(GrVkGpu* gpu)
26 : fGpu(gpu)
28 fPipelineStateCache = new PipelineStateCache(gpu);
445 const GrVkGpu* gpu,
449 fRenderPasses[0]->initSimple(gpu, target);
461 const GrVkGpu* gpu,
472 renderPass->init(gpu, *this->getCompatibleRenderPass(), colorOps, stencilOps);
477 void GrVkResourceProvider::CompatibleRenderPassSet::releaseResources(const GrVkGpu* gpu) {
480 fRenderPasses[i]->unref(gpu);
|
| /external/skqp/src/gpu/vk/ |
| GrVkResourceProvider.cpp | 25 GrVkResourceProvider::GrVkResourceProvider(GrVkGpu* gpu)
26 : fGpu(gpu)
28 fPipelineStateCache = new PipelineStateCache(gpu);
445 const GrVkGpu* gpu,
449 fRenderPasses[0]->initSimple(gpu, target);
461 const GrVkGpu* gpu,
472 renderPass->init(gpu, *this->getCompatibleRenderPass(), colorOps, stencilOps);
477 void GrVkResourceProvider::CompatibleRenderPassSet::releaseResources(const GrVkGpu* gpu) {
480 fRenderPasses[i]->unref(gpu);
|
| /external/skia/tests/ |
| VkHeapTests.cpp | 8 // This is a GPU-backend specific test. It relies on static intializers to work
23 GrVkGpu* gpu = static_cast<GrVkGpu*>(context->contextPriv().getGpu()); local
26 GrVkSubHeap heap(gpu, 0, 0, 64 * 1024, 32);
120 GrVkGpu* gpu = static_cast<GrVkGpu*>(context->contextPriv().getGpu()); local
123 GrVkHeap heap(gpu, GrVkHeap::kSubAlloc_Strategy, 64 * 1024);
179 GrVkGpu* gpu = static_cast<GrVkGpu*>(context->contextPriv().getGpu()); local
182 GrVkHeap heap(gpu, GrVkHeap::kSingleAlloc_Strategy, 64 * 1024);
|
| /external/skqp/tests/ |
| VkHeapTests.cpp | 8 // This is a GPU-backend specific test. It relies on static intializers to work
23 GrVkGpu* gpu = static_cast<GrVkGpu*>(context->contextPriv().getGpu()); local
26 GrVkSubHeap heap(gpu, 0, 0, 64 * 1024, 32);
120 GrVkGpu* gpu = static_cast<GrVkGpu*>(context->contextPriv().getGpu()); local
123 GrVkHeap heap(gpu, GrVkHeap::kSubAlloc_Strategy, 64 * 1024);
179 GrVkGpu* gpu = static_cast<GrVkGpu*>(context->contextPriv().getGpu()); local
182 GrVkHeap heap(gpu, GrVkHeap::kSingleAlloc_Strategy, 64 * 1024);
|
| /hardware/qcom/display/msm8084/libgralloc/ |
| gpu.cpp | 25 #include "gpu.h"
359 gpu_context_t* gpu = reinterpret_cast<gpu_context_t*>(dev); local
360 return gpu->alloc_impl(w, h, format, usage, pHandle, pStride, 0);
370 gpu_context_t* gpu = reinterpret_cast<gpu_context_t*>(dev); local
371 return gpu->alloc_impl(w, h, format, usage, pHandle, pStride, bufferSize);
382 gpu_context_t* gpu = reinterpret_cast<gpu_context_t*>(dev); local
383 return gpu->free_impl(hnd);
|
| /hardware/qcom/display/msm8226/libgralloc/ |
| gpu.cpp | 25 #include "gpu.h"
362 gpu_context_t* gpu = reinterpret_cast<gpu_context_t*>(dev); local
363 return gpu->alloc_impl(w, h, format, usage, pHandle, pStride, 0);
373 gpu_context_t* gpu = reinterpret_cast<gpu_context_t*>(dev); local
374 return gpu->alloc_impl(w, h, format, usage, pHandle, pStride, bufferSize);
385 gpu_context_t* gpu = reinterpret_cast<gpu_context_t*>(dev); local
386 return gpu->free_impl(hnd);
|
| /hardware/qcom/display/msm8909/libgralloc/ |
| gpu.cpp | 28 #include "gpu.h"
225 // As GPU needs ION FD, the private handle is created
383 gpu_context_t* gpu = reinterpret_cast<gpu_context_t*>(dev); local
384 return gpu->alloc_impl(w, h, format, usage, pHandle, pStride, 0);
394 gpu_context_t* gpu = reinterpret_cast<gpu_context_t*>(dev); local
395 return gpu->alloc_impl(w, h, format, usage, pHandle, pStride, bufferSize);
406 gpu_context_t* gpu = reinterpret_cast<gpu_context_t*>(dev); local
407 return gpu->free_impl(hnd);
|
