projects / platform / upstream / tensorflow.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 834093d) | patch
[XLA:GPU] Mark bitcasts as eligible for fusion.
authorBenjamin Kramer <kramerb@google.com>
Tue, 6 Mar 2018 11:31:45 +0000 (03:31 -0800)
committerTensorFlower Gardener <gardener@tensorflow.org>
Tue, 6 Mar 2018 11:35:51 +0000 (03:35 -0800)
commitbec6e47cf93ce3fad041580de4d922f30190b1c7
tree276bea9872e881659ef2d54bd86028e6bf41cccbtree | snapshot
parent834093de427445b4ed49729146e69b05786f4d1dcommit | diff
[XLA:GPU] Mark bitcasts as eligible for fusion.

Currently this never happens because we only turn rehaspes into bitcasts after
layout assignment. This changes when layout assignment runs before fusion. Once
layouts are available the pipeline turns reshapes into bitcasts, which would be
left unfused without this change.

PiperOrigin-RevId: 187999864
tensorflow/compiler/xla/service/elemental_ir_emitter.cc diff | blob | history
tensorflow/compiler/xla/service/gpu/BUILD diff | blob | history
tensorflow/compiler/xla/service/gpu/instruction_fusion.cc diff | blob | history
tensorflow/compiler/xla/service/gpu/instruction_fusion_test.cc diff | blob | history
tensorflow/compiler/xla/tests/llvm_irgen_test_base.cc diff | blob | history
Domain: Machine Learning / ML Framework;