| commit | 4263b2ecf8c4b9b62094a731bb92c501197531b0 | |
| author | peterbell10 <peterbell10@live.co.uk> | |
| Fri, 17 Nov 2023 17:51:50 +0000 (17 17:51 +0000) | ||
| committer | GitHub <noreply@github.com> | |
| Fri, 17 Nov 2023 17:51:50 +0000 (17 09:51 -0800) | ||
| tree | 4532ed395d2d32d5fc983c898687f316d1620a4b | treesnapshot (tar.gz zip) |
| parent | bfbfd1caa4da70774547c1c298e482661822a137 | commitdiff |
[NVPTX] Expand EXTLOAD for v8f16 and v8bf16 (#72672)
In openai/triton#2483 I've encountered a bug in the NVPTX codegen. Given
`load<8 x half>` followed by `fpext to <8 x float>` we get
```
ld.shared.v4.b16 {%f1, %f2, %f3, %f4}, [%r15+8];
ld.shared.v4.b16 {%f5, %f6, %f7, %f8}, [%r15];
```
Which loads float16 values into float registers without any conversion
and the result is simply garbage.
This PR brings `v8f16` and `v8bf16` into line with the other vector
types by expanding it to load + cvt.
cc @manman-ren @Artem-B @jlebar
In openai/triton#2483 I've encountered a bug in the NVPTX codegen. Given
`load<8 x half>` followed by `fpext to <8 x float>` we get
```
ld.shared.v4.b16 {%f1, %f2, %f3, %f4}, [%r15+8];
ld.shared.v4.b16 {%f5, %f6, %f7, %f8}, [%r15];
```
Which loads float16 values into float registers without any conversion
and the result is simply garbage.
This PR brings `v8f16` and `v8bf16` into line with the other vector
types by expanding it to load + cvt.
cc @manman-ren @Artem-B @jlebar
| llvm/lib/Target/NVPTX/NVPTXISelLowering.cpp | diffblobblamehistory | |
| llvm/test/CodeGen/NVPTX/bf16-instructions.ll | diffblobblamehistory | |
| llvm/test/CodeGen/NVPTX/vector-loads.ll | diffblobblamehistory |