diff options
author | Yunxing Dai <yunxing@google.com> | 2018-06-07 18:42:30 -0700 |
---|---|---|
committer | TensorFlower Gardener <gardener@tensorflow.org> | 2018-06-07 18:45:05 -0700 |
commit | a9ddfe50eee83b2f18293241ab96f0a1e2b4b05b (patch) | |
tree | 849afa106cb882d918cfa0a633134bb89f7f014f /tensorflow/compiler/xla/service/hlo_parser_test.cc | |
parent | 2f41346cbc0c8ecb915983a1f8711fd0d0ccc50e (diff) |
[DataFlowAnalysis] Be less conservative on loop fusion nodes when reusing buffer.
- Previously, we say we cannot reuse operand buffer for a loop fusion node if any of the fusion's inputs is a broadcast or reshape. That's too conservative since in theory we can still reuse the operand's buffer if all the users of that particular operand are elementwise. This CL implements that.
- Also fixed a bug in previous code where a dynamic update fusion node that ends with convert (added for bf16) is not caught by the if condition currectly.
PiperOrigin-RevId: 199731488
Diffstat (limited to 'tensorflow/compiler/xla/service/hlo_parser_test.cc')
-rw-r--r-- | tensorflow/compiler/xla/service/hlo_parser_test.cc | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/tensorflow/compiler/xla/service/hlo_parser_test.cc b/tensorflow/compiler/xla/service/hlo_parser_test.cc index 08068dc504..1c5a47c875 100644 --- a/tensorflow/compiler/xla/service/hlo_parser_test.cc +++ b/tensorflow/compiler/xla/service/hlo_parser_test.cc @@ -765,7 +765,7 @@ add_F32.v3 { ENTRY MapBinaryAdder.v3 { param0 = f32[4]{0} parameter(0) param1 = f32[4]{0} parameter(1) - ROOT map = f32[4]{0} map(param0, param1), to_apply=add_F32.v3 + ROOT map = f32[4]{0} map(param0, param1), dimensions={0}, to_apply=add_F32.v3 } )" |