Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Commit 2fa99dc

Browse files
arsenmarsenm
committed
AMDGPU: Move softPromoteHalfType override to R600 only
As expected the code is much worse, but more correct. We could do a better job with source modifier management around fp16_to_fp/fp_to_fp16.
1 parent 6fc8028 commit 2fa99dc

122 files changed

Lines changed: 94997 additions & 123445 deletions

File tree

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

‎llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h‎

Lines changed: 0 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -405,8 +405,6 @@ class AMDGPUTargetLowering : public TargetLowering {
405405
// are using vector compares until that is fixed.
406406
return true;
407407
}
408-
409-
bool softPromoteHalfType() const override { return false; }
410408
};
411409

412410
} // End namespace llvm

‎llvm/lib/Target/AMDGPU/R600ISelLowering.h‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -117,6 +117,8 @@ class R600TargetLowering final : public AMDGPUTargetLowering {
117117

118118
TargetLowering::AtomicExpansionKind
119119
shouldExpandAtomicRMWInIR(const AtomicRMWInst *RMW) const override;
120+
121+
bool softPromoteHalfType() const override { return false; }
120122
};
121123

122124
} // End namespace llvm;

‎llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll‎

Lines changed: 17895 additions & 22432 deletions
Large diffs are not rendered by default.

0 commit comments

Comments
 (0)