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

Set xla_gpu_ftz to true in xla DebugOptions.

Browse Source 
This is necessary to be compatible with Tensorflow.

PiperOrigin-RevId: 347829756
Change-Id: I8a933de74634181d32bc35c70a8053d2c92909eb
This commit is contained in:
A. Unique TensorFlower 2020-12-16 08:45:05 -08:00 committed by TensorFlower Gardener
parent 76e17494d6
commit ca42aa829c
1 changed files with 2 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
tensorflow/compiler/mlir/tools/kernel_gen/transforms/gpu_kernel_to_blob_pass.cc
View File

@ -99,9 +99,7 @@ class GpuKernelToBlobPass
llvmModule->setModuleIdentifier("acme");
xla::HloModuleConfig config;
xla::DebugOptions options = xla::GetDebugOptionsFromFlags();
options.set_xla_gpu_ftz(true);
config.set_debug_options(options);
config.set_debug_options(xla::GetDebugOptionsFromFlags());
using AmdGpuHsaco = std::vector<tensorflow::uint8>;
std::vector<tensorflow::se::HsacoImage> images;
@ -150,9 +148,7 @@ class GpuKernelToBlobPass
llvmModule->setDataLayout(xla::gpu::nvptx::kDataLayout);
xla::HloModuleConfig config;
xla::DebugOptions options = xla::GetDebugOptionsFromFlags();
options.set_xla_gpu_ftz(true);
config.set_debug_options(options);
config.set_debug_options(xla::GetDebugOptionsFromFlags());
auto enable_fusion = [](llvm::TargetMachine* target) {
target->Options.AllowFPOpFusion = llvm::FPOpFusion::FPOpFusionMode::Fast;