experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[XLA:CPU] Turn off MLIR matmul codegen while we investigate an upstream regression

Browse Source 
PiperOrigin-RevId: 354616434
Change-Id: Ic77fda05a548a9c8735d47b691060a62110b2388
This commit is contained in:
Benjamin Kramer 2021-01-29 14:52:33 -08:00 committed by TensorFlower Gardener
parent dad06db8f5
commit b2e90ecf2d
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tensorflow/compiler/xla/service/cpu/dot_op_emitter.cc
View File

@ -1028,7 +1028,7 @@ DotImplementationStrategy GetDotImplementationStrategy(
if (IsAlignedGemm(dot_info, target_machine_features)) {
if (CanEmitTiledLlvmIrGemm(config, dot_info, target_machine_features)) {
return DotImplementationStrategy::kLinalgMatmul;
return DotImplementationStrategy::kTiledLlvmIrGemm;
}
return DotImplementationStrategy::kEigen;
}