Commits

Breadcrumbs

History for

onr0.9

User selector

All users

All time

Commit History

Commits on Aug 8, 2016

• Hard-code libcuda version number to "1". Fixes #2865. (#3576)

  Show description for 0685152
  

  cancan101

  authored and

  Vijay Vasudevan

  committed
  0685152

  Copy full SHA for 0685152

Commits on Jun 3, 2016

• cleans up warning/errors tensorflow/stream_executor (#2555)

  

  Mistobaan

  authored and

  Vijay Vasudevan

  committed
  4c789e3

  Copy full SHA for 4c789e3

Commits on Jun 2, 2016

• Update copyright for 3p/tf.

  Show description for 122cdce
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  122cdce

  Copy full SHA for 122cdce

Commits on May 26, 2016

• Merge changes from github.

  Show description for 8cc567b
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  8cc567b

  Copy full SHA for 8cc567b

Commits on May 25, 2016

• StreamExecutor add CUDA support for cudnnConvolutionBackwardBias

  Show description for a9f3979
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  a9f3979

  Copy full SHA for a9f3979

Commits on May 23, 2016

• Add macros for mutex_lock and shared_lock, so compilation fails if

  Show description for 81af087
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  81af087

  Copy full SHA for 81af087

Commits on May 17, 2016

• Extend the StreamExecutor interface with ...WithAlgorithm() calls for fp16;

  Show description for 00bc81a
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  00bc81a

  Copy full SHA for 00bc81a

Commits on May 16, 2016

• Add Autotune support for Winograd on Cudnn R5.

  Show description for 75b1f2f
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  75b1f2f

  Copy full SHA for 75b1f2f

Commits on May 13, 2016

• In the StreamExecutor, make lack of CUDA 7.5 a non-fatal error for SGEMM

  Show description for cfa7ffb
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  cfa7ffb

  Copy full SHA for cfa7ffb

• Maintain current cuda context in CUDADriver by checking against a threadlocal

  Show description for 069547b
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  069547b

  Copy full SHA for 069547b

Commits on May 11, 2016

• Add fp16 cuDNN convolution support to StreamExecutor. (TensorFlow ops will

  Show description for 61b12f5
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  61b12f5

  Copy full SHA for 61b12f5

• Add fp16 matrix multiplication (GEMM) support to StreamExecutor, gated on

  Show description for 5230554
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  5230554

  Copy full SHA for 5230554

• Add the autotune version for the backward passes. It is currently

  Show description for 939ede0
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  939ede0

  Copy full SHA for 939ede0

Commits on May 10, 2016

• fp16 support for BiasAdd. Includes support for atomic adds for Eigen::half,

  Show description for aec09b6
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  aec09b6

  Copy full SHA for aec09b6

Commits on May 9, 2016

• Adding autotune to the Cudnn conv algorithm selection.

  Show description for 6ff265e
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  6ff265e

  Copy full SHA for 6ff265e

Commits on May 5, 2016

• Merge changes from github.

  Show description for 8bf6ef1
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  8bf6ef1

  Copy full SHA for 8bf6ef1

Commits on May 4, 2016

• Change CUDAStream::GetOrCreateCompletedEvent from lazy initialization the

  Show description for 47916d4
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  47916d4

  Copy full SHA for 47916d4

Commits on Apr 27, 2016

• Switch CuDNN avg pooling to not use padding when computing the average.

  Show description for afcbe82
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  afcbe82

  Copy full SHA for afcbe82

• Generalize the stream executor interface to support N-d operations.

  Show description for 57b4b9b
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  57b4b9b

  Copy full SHA for 57b4b9b

Commits on Apr 22, 2016

• Calling Cudnn calls in the current thread, instead of dispatching it into its

  Show description for b68860c
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  b68860c

  Copy full SHA for b68860c

Commits on Apr 19, 2016

• Merge changes from github.

  Show description for 5c9bc51
  

  ilblackdragon

  authored and

  tensorflower-gardener

  committed
  5c9bc51

  Copy full SHA for 5c9bc51

Commits on Mar 18, 2016

• Support ScratchAllocator in BLAS Batched GEMM

  Show description for 05ea40f
  

  A. Unique TensorFlower

  authored and

  tensorflower-gardener

  committed
  05ea40f

  Copy full SHA for 05ea40f

Commits on Feb 18, 2016

• TensorFlow: change cuda-diagnostics to search for so.1

  Show description for 8174381
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  8174381

  Copy full SHA for 8174381

• TensorFlow: switch dsoloader path for libcuda to end in .so.1

  Show description for d0a822f
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  d0a822f

  Copy full SHA for d0a822f

• TensorFlow: fix missing GetCudaVersion() for libcuda.so. h/t to @3XX0.

  Show description for 07fa552
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  07fa552

  Copy full SHA for 07fa552

Commits on Feb 17, 2016

• Merge changes from github.

  Show description for fe056f0
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  fe056f0

  Copy full SHA for fe056f0

• Prepare to take another snapshot from stream-executor.

  Show description for d134099
  

  zheng-xq

  authored and

  tensorflower-gardener

  committed
  d134099

  Copy full SHA for d134099

Commits on Feb 6, 2016

• Deleted unused code

  Show description for 3a30eb9
  

  benoitsteiner

  authored and

  Vijay Vasudevan

  committed
  3a30eb9

  Copy full SHA for 3a30eb9

Commits on Feb 3, 2016

• Add assert_same_float_dtype.

  Show description for 1e14183
  

  A. Unique TensorFlower

  authored and

  Vijay Vasudevan

  committed
  1e14183

  Copy full SHA for 1e14183

Commits on Jan 30, 2016

• Use cc_binary rather than cc_library to reduce size of native library in APK from 5.5mb to 3.2mb (compressed).

  Show description for 8a59748
  

  A. Unique TensorFlower

  authored and

  Vijay Vasudevan

  committed
  8a59748

  Copy full SHA for 8a59748

Commits on Jan 26, 2016

• Global search & replace to move to the new location for

  Show description for b2f0bc2
  

  josh11b

  authored and

  Vijay Vasudevan

  committed
  b2f0bc2

  Copy full SHA for b2f0bc2

Commits on Jan 22, 2016

• Make cuda_dnn work with R4.

  Show description for 22ebf0a
  

  zheng-xq

  authored and

  keveman

  committed
  22ebf0a

  Copy full SHA for 22ebf0a

Commits on Jan 7, 2016

• #include "tensorflow/core/platform/mutex.h"

  Show description for ded6170
  

  josh11b

  authored and

  Vijay Vasudevan

  committed
  ded6170

  Copy full SHA for ded6170

Commits on Jan 6, 2016

• #include third_party/tensorflow/core/platform/macros.h

  Show description for 252bb90
  

  josh11b

  authored and

  Vijay Vasudevan

  committed
  252bb90

  Copy full SHA for 252bb90

Commits on Jan 5, 2016

• Added 'logging' import to control_flow_ops which is used in the file but not imported.

  Show description for 1c57936
  

  A. Unique TensorFlower

  authored and

  Vijay Vasudevan

  committed
  1c57936

  Copy full SHA for 1c57936