Searched refs:num_devices (Results 1 - 25 of 37) sorted by relevance

12

/external/tensorflow/tensorflow/contrib/all_reduce/python/
H A Dall_reduce.py196 In the basic ring reduction algorithm there are size(T)/num_devices
317 num_devices = len(input_tensors)
318 if num_devices == 0:
320 if num_devices == 1:
325 num_chunks = num_devices * num_subchunks
326 num_ticks = num_devices - 1
330 for d in range(0, num_devices):
339 for d in range(0, num_devices):
343 seg_index = (rank + num_devices - (2 + tick)) % num_devices
[all...]
H A Dall_reduce_test.py121 num_devices = len(dev_list)
123 for d in range(0, num_devices):
143 num_devices = num_workers * num_gpus
145 for _ in range(num_devices)]
149 x, constant_op.constant(num_devices, dtype=types_pb2.DT_FLOAT))
/external/tensorflow/tensorflow/contrib/nccl/kernels/
H A Dnccl_manager.cc70 : num_devices(members.size()), members(std::move(members)) {}
72 const int num_devices; member in struct:tensorflow::NcclManager::Communicator
139 ncclRedOp_t reduction_op_in, int num_devices)
143 remaining_participants(num_devices) {
144 participants.reserve(num_devices);
187 const int num_devices = collective->participants.size(); local
211 if (comm->num_devices == num_devices) {
213 for (i = 0; i < num_devices; ++i) {
219 if (i == num_devices) retur
138 Collective(DataType data_type_in, CollectiveType type_in, ncclRedOp_t reduction_op_in, int num_devices) argument
299 AddToAllReduce(int num_devices, const string& key, ncclRedOp_t reduction_op, perftools::gputools::StreamExecutor* executor, int gpu_device_id, EventMgr* event_mgr, perftools::gputools::Stream* tensor_stream, const Tensor* in_t, Tensor* out_t, const DoneCallback& done_callback) argument
313 AddBroadcastSend( int num_devices, const string& key, perftools::gputools::StreamExecutor* executor, int gpu_device_id, EventMgr* event_mgr, perftools::gputools::Stream* tensor_stream, const Tensor* in_t, DoneCallback done_callback) argument
326 AddBroadcastRecv( int num_devices, const string& key, perftools::gputools::StreamExecutor* executor, int gpu_device_id, EventMgr* event_mgr, perftools::gputools::Stream* tensor_stream, Tensor* out_t, DoneCallback done_callback) argument
338 AddReduceSend(int num_devices, const string& key, ncclRedOp_t reduction_op, perftools::gputools::StreamExecutor* executor, int gpu_device_id, EventMgr* event_mgr, perftools::gputools::Stream* tensor_stream, const Tensor* in_t, DoneCallback done_callback) argument
352 AddReduceRecv(int num_devices, const string& key, ncclRedOp_t reduction_op, perftools::gputools::StreamExecutor* executor, int gpu_device_id, EventMgr* event_mgr, perftools::gputools::Stream* tensor_stream, const Tensor* in_t, Tensor* out_t, DoneCallback done_callback) argument
367 AddParticipant(int num_devices, const string& key, std::unique_ptr<Participant> participant, DataType data_type, CollectiveType collective_type, ncclRedOp_t reduction_op) argument
[all...]
H A Dnccl_manager.h57 void AddToAllReduce(int num_devices, const string& key,
67 void AddBroadcastSend(int num_devices, const string& key,
72 void AddBroadcastRecv(int num_devices, const string& key,
80 void AddReduceSend(int num_devices, const string& key,
86 void AddReduceRecv(int num_devices, const string& key,
108 void AddParticipant(int num_devices, const string& key,
H A Dnccl_ops.cc42 OP_REQUIRES_OK(c, c->GetAttr("num_devices", &num_devices_));
52 int num_devices() const { return num_devices_; } function in class:tensorflow::__anon26086::NcclAsyncOpBase
106 num_devices(), GetCollectiveKey(c), reduction_op(),
131 num_devices(), GetCollectiveKey(c), reduction_op(),
160 num_devices(), GetCollectiveKey(c), reduction_op(),
188 num_devices(), GetCollectiveKey(c), compute_stream->parent(),
220 num_devices(), GetCollectiveKey(c), compute_stream->parent(),
H A Dnccl_rewrite.cc34 int num_devices = node->num_inputs(); local
39 .Attr("num_devices", num_devices)
119 int num_devices = 0; // Number of distinct devices, incremented below. local
134 recv_index_map[dst_dev] = num_devices;
135 ++num_devices;
140 if (num_devices <= 1) {
162 .Attr("num_devices", num_devices)
/external/mesa3d/src/gallium/state_trackers/clover/api/
H A Dcontext.cpp71 cl_uint num_devices; local
77 ret = clGetDeviceIDs(d_platform, type, 0, NULL, &num_devices);
80 devs.resize(num_devices);
81 ret = clGetDeviceIDs(d_platform, type, num_devices, devs.data(), 0);
85 return clCreateContext(d_props, num_devices, devs.data(), pfn_notify,
H A Ddispatch.hpp56 cl_uint *num_devices);
67 cl_uint num_devices,
204 cl_uint num_devices,
219 cl_uint num_devices,
630 cl_uint *num_devices);
648 cl_uint *num_devices);
666 cl_uint num_devices,
673 cl_uint num_devices,
684 cl_uint num_devices,
/external/eigen/unsupported/Eigen/CXX11/src/Tensor/
H A DTensorDeviceCuda.h60 int num_devices; local
61 cudaError_t status = cudaGetDeviceCount(&num_devices);
68 m_deviceProperties = new cudaDeviceProp[num_devices];
69 for (int i = 0; i < num_devices; ++i) {
119 int num_devices; local
120 cudaError_t err = cudaGetDeviceCount(&num_devices);
123 assert(device < num_devices);
/external/mesa3d/src/loader/
H A Dloader.c163 int i, num_devices, fd; local
182 num_devices = drmGetDevices(devices, MAX_DRM_DEVICES);
183 if (num_devices < 0)
195 for (i = 0; i < num_devices; i++) {
204 for (i = 0; i < num_devices; i++) {
215 drmFreeDevices(devices, num_devices);
220 drmFreeDevices(devices, num_devices);
/external/webrtc/webrtc/tools/e2e_quality/audio/
H A Daudio_e2e_harness.cc70 int num_devices = 0; local
71 ASSERT_EQ(0, hardware->GetNumOfPlayoutDevices(num_devices));
76 for (device_index = 0; device_index < num_devices; device_index++) {
/external/kernel-headers/original/uapi/linux/
H A Dpktcdvd.h101 __u32 num_devices; /* out: Largest device index + 1 */ member in struct:pkt_ctrl_command
/external/tensorflow/tensorflow/contrib/nccl/python/ops/
H A Dnccl_ops.py68 num_devices = op.get_attr('num_devices')
75 num_devices=num_devices,
227 num_devices=len(tensors),
/external/webrtc/webrtc/tools/agc/
H A Dagc_harness.cc158 int num_devices = 0; local
161 RTC_CHECK_EQ(0, hardware_->GetNumOfRecordingDevices(num_devices));
163 for (int i = 0; i < num_devices; i++) {
167 RTC_CHECK_EQ(0, hardware_->GetNumOfPlayoutDevices(num_devices));
169 for (int i = 0; i < num_devices; i++) {
/external/webrtc/talk/media/devices/
H A Dmacdevicemanager.cc121 size_t num_devices = propsize / sizeof(AudioDeviceID); local
123 new AudioDeviceID[num_devices]);
133 for (size_t i = 0; i < num_devices; ++i) {
/external/tensorflow/tensorflow/compiler/xla/service/
H A Dhlo_sharding.cc163 Status HloSharding::ValidateTuple(const Shape& shape, int64 num_devices) const {
187 ShapeUtil::GetSubshape(shape, index_to_sharding.first), num_devices);
199 Status HloSharding::Validate(const Shape& shape, int64 num_devices) const {
200 Status status = IsTuple() ? ValidateTuple(shape, num_devices)
201 : ValidateNonTuple(shape, num_devices);
211 int64 num_devices) const {
228 if (core >= num_devices) {
230 "core ", core, " > ", num_devices, " in tile assignment"));
H A Dhlo_sharding.h101 Status Validate(const Shape& shape, int64 num_devices) const;
237 Status ValidateTuple(const Shape& shape, int64 num_devices) const;
239 Status ValidateNonTuple(const Shape& shape, int64 num_devices) const;
/external/mesa3d/include/CL/
H A Dcl_dx9_media_sharing.h96 cl_uint * num_devices) CL_API_SUFFIX__VERSION_1_2;
H A Dcl_d3d10.h83 cl_uint * num_devices) CL_API_SUFFIX__VERSION_1_0;
H A Dcl_d3d11.h83 cl_uint * num_devices) CL_API_SUFFIX__VERSION_1_2;
/external/syslinux/extlinux/
H A Dbtrfs.h70 uint64_t num_devices; member in struct:btrfs_super_block
168 __u64 num_devices; /* out */ member in struct:btrfs_ioctl_fs_info_args
/external/ImageMagick/MagickCore/
H A Dopencl-private.h77 cl_uint *num_devices) CL_API_SUFFIX__VERSION_1_0;
88 const cl_context_properties *properties,cl_uint num_devices,
136 cl_uint num_devices,const cl_device_id *device_list,const size_t *lengths,
146 cl_uint num_devices,const cl_device_id *device_list,const char *options,
/external/e2fsprogs/misc/
H A Dfsck.c102 static int num_devices, num_args; variable
1108 num_devices = 0;
1119 if (num_devices >= MAX_DEVICES) {
1148 devices[num_devices++] = dev ? dev : string_copy(arg);
1296 if ((num_devices == 1) || (serialize))
1303 if (num_devices == 0) {
1308 for (i = 0 ; i < num_devices; i++) {
/external/linux-kselftest/tools/testing/selftests/zram/
H A Dzram_lib.sh68 modprobe zram num_devices=$dev_num
/external/tensorflow/tensorflow/c/eager/
H A Dc_api_test.cc103 const int num_devices = TF_DeviceListCount(devices); local
104 for (int i = 0; i < num_devices; ++i) {
180 const int num_devices = TF_DeviceListCount(devices); local
181 EXPECT_GE(num_devices, 1) << "At least one CPU device should exist";
182 for (int i = 0; i < num_devices; ++i) {
222 const int num_devices = TF_DeviceListCount(devices); local
225 for (int i = 0; i < num_devices; ++i) {
291 const int num_devices = TF_DeviceListCount(devices); local
294 if (num_devices < 3) {

Completed in 470 milliseconds

12