diff options
| author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2019-10-31 18:50:30 -0700 |
|---|---|---|
| committer | Matt Arsenault <arsenm2@gmail.com> | 2019-11-19 19:55:43 +0530 |
| commit | db0ed3e429b55d1730d1ecc253b0643de7fca099 (patch) | |
| tree | 33f725b9778863f7c737075b4ca9d9e6c43ceb08 /llvm/lib/Target/AMDGPU/R600ISelLowering.cpp | |
| parent | ea23b6428b88ed50a2cfc91b783f627fa139bb36 (diff) | |
| download | bcm5719-llvm-db0ed3e429b55d1730d1ecc253b0643de7fca099.tar.gz bcm5719-llvm-db0ed3e429b55d1730d1ecc253b0643de7fca099.zip | |
AMDGPU: Refactor treatment of denormal mode
Start moving towards treating this as a property of the calling
convention, and not the subtarget. The default denormal mode should
not be part of the subtarget, and be moved into a separate function
attribute.
This patch is still NFC. The denormal mode remains as a subtarget
feature for now, but make the necessary changes to switch to using an
attribute.
Diffstat (limited to 'llvm/lib/Target/AMDGPU/R600ISelLowering.cpp')
| -rw-r--r-- | llvm/lib/Target/AMDGPU/R600ISelLowering.cpp | 6 |
1 files changed, 2 insertions, 4 deletions
diff --git a/llvm/lib/Target/AMDGPU/R600ISelLowering.cpp b/llvm/lib/Target/AMDGPU/R600ISelLowering.cpp index 6196be8ec83..ee24022c65f 100644 --- a/llvm/lib/Target/AMDGPU/R600ISelLowering.cpp +++ b/llvm/lib/Target/AMDGPU/R600ISelLowering.cpp @@ -226,10 +226,8 @@ R600TargetLowering::R600TargetLowering(const TargetMachine &TM, setOperationAction(ISD::FMA, MVT::f64, Expand); } - // FIXME: This was moved from AMDGPUTargetLowering, I'm not sure if we - // need it for R600. - if (!Subtarget->hasFP32Denormals()) - setOperationAction(ISD::FMAD, MVT::f32, Legal); + // FIXME: May need no denormals check + setOperationAction(ISD::FMAD, MVT::f32, Legal); if (!Subtarget->hasBFI()) { // fcopysign can be done in a single instruction with BFI. |
