Commit 1c2072f
committed
[AMDGPU] Fix AGPR_32 reg assign for mfma scale ops
In MFMA rewrite pass, prevent AGPR_32 reg class assignment for scale
operands, not permitted by instruction format.1 parent 76d614b commit 1c2072f
File tree
2 files changed
+7
-3
lines changed- llvm
- lib/Target/AMDGPU
- test/CodeGen/AMDGPU
2 files changed
+7
-3
lines changed| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
302 | 302 | | |
303 | 303 | | |
304 | 304 | | |
| 305 | + | |
| 306 | + | |
| 307 | + | |
| 308 | + | |
305 | 309 | | |
306 | 310 | | |
307 | 311 | | |
| |||
Lines changed: 3 additions & 3 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
1 | | - | |
2 | | - | |
3 | | - | |
| 1 | + | |
| 2 | + | |
| 3 | + | |
4 | 4 | | |
5 | 5 | | |
6 | 6 | | |
| |||
0 commit comments