projects / platform / upstream / tensorflow.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 143f358) | patch
[XLA:CPU] Fix the parallel task assignment to not parallelize dot operations.
authorBixia Zheng <bixia@google.com>
Thu, 15 Mar 2018 18:22:58 +0000 (11:22 -0700)
committerTensorFlower Gardener <gardener@tensorflow.org>
Thu, 15 Mar 2018 18:29:16 +0000 (11:29 -0700)
commitf465d7467ef5488aca1b4e620c27d4016c774ef8
tree718b5018d5d7ef2f311c95ea935f5fb753f828b1tree | snapshot
parent143f3585ca78380891cb862880c14c4ca9d7b9fdcommit | diff
[XLA:CPU] Fix the parallel task assignment to not parallelize dot operations.

The IR emitter currently generates incorrect code for parallelized dot
operations.

Add test cases to check for dot operation parallelization.

PiperOrigin-RevId: 189216963
tensorflow/compiler/xla/service/cpu/BUILD diff | blob | history
tensorflow/compiler/xla/service/cpu/dot_op_emitter.cc diff | blob | history
tensorflow/compiler/xla/service/cpu/parallel_task_assignment.cc diff | blob | history
tensorflow/compiler/xla/service/cpu/parallel_task_assignment_test.cc [new file with mode: 0644] blob
Domain: Machine Learning / ML Framework;