diff options
| author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2020-03-20 17:16:46 -0400 |
|---|---|---|
| committer | Matt Arsenault <arsenm2@gmail.com> | 2020-03-23 17:00:41 -0400 |
| commit | 66073953a5cb7ecac32bcb897033fdc337c56b5e (patch) | |
| tree | 6a6ec427ebdbd432574f02633c0a0d3b26c1a159 /llvm/lib/Support/LockFileManager.cpp | |
| parent | 5896e2df45dab7680be321ee63903431a673817b (diff) | |
| download | llvm-66073953a5cb7ecac32bcb897033fdc337c56b5e.zip llvm-66073953a5cb7ecac32bcb897033fdc337c56b5e.tar.gz llvm-66073953a5cb7ecac32bcb897033fdc337c56b5e.tar.bz2 | |
AMDGPU: Allow vectorization of round intrinsic
There seems to be a small benefit to the legalized sequence for v2f16
round with packed instructions, so allow vectorizing it by reducing
the cost.
An unintended side effect is vectorization of f32 round also
happens. The current FMA logic seems off to me, and isn't checking for
packed instructions.
Diffstat (limited to 'llvm/lib/Support/LockFileManager.cpp')
0 files changed, 0 insertions, 0 deletions