projects / platform / upstream / tensorflow.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 0e79074) | patch
[XLA:GPU] Fix a problem in calculating the address of the memory used to
authorA. Unique TensorFlower <gardener@tensorflow.org>
Wed, 17 Jan 2018 20:20:20 +0000 (12:20 -0800)
committerTensorFlower Gardener <gardener@tensorflow.org>
Wed, 17 Jan 2018 20:23:52 +0000 (12:23 -0800)
commitaabe9698daf5004ec90f04e2da5b71b51b010b6d
tree634ce169eea0e699484cf9036f8bb2d91b3d8051tree | snapshot
parent0e79074cf19c0b09ac7f0dd832d1f16d8ca78c25commit | diff
[XLA:GPU] Fix a problem in calculating the address of the memory used to
implement small data type atomic operations.

To calculate the address of the enclosing 4 byte memory,
output_address&(-2) was used while it should be output_address&(-4).

Add a test case.

PiperOrigin-RevId: 182251760
tensorflow/compiler/xla/service/gpu/ir_emitter.cc diff | blob | history
tensorflow/compiler/xla/tests/reduce_test.cc diff | blob | history
Domain: Machine Learning / ML Framework;