Commit bae4f80
[AMDGPU] Fix AGPR_32 reg assign for mfma scale ops (llvm#168964)
In MFMA rewrite pass, prevent AGPR_32 reg class assignment for scale
operands, not permitted by instruction format.
---------
Co-authored-by: Matt Arsenault <[email protected]>1 parent 087a61c commit bae4f80
File tree
2 files changed
+17
-8
lines changed- llvm
- lib/Target/AMDGPU
- test/CodeGen/AMDGPU
2 files changed
+17
-8
lines changed| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
97 | 97 | | |
98 | 98 | | |
99 | 99 | | |
100 | | - | |
101 | | - | |
| 100 | + | |
| 101 | + | |
102 | 102 | | |
103 | 103 | | |
104 | 104 | | |
| |||
152 | 152 | | |
153 | 153 | | |
154 | 154 | | |
155 | | - | |
156 | | - | |
| 155 | + | |
| 156 | + | |
157 | 157 | | |
| 158 | + | |
| 159 | + | |
| 160 | + | |
| 161 | + | |
| 162 | + | |
| 163 | + | |
| 164 | + | |
158 | 165 | | |
159 | 166 | | |
160 | 167 | | |
| |||
Lines changed: 6 additions & 4 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
1 | | - | |
2 | | - | |
3 | | - | |
4 | | - | |
| 1 | + | |
| 2 | + | |
| 3 | + | |
| 4 | + | |
| 5 | + | |
| 6 | + | |
5 | 7 | | |
6 | 8 | | |
7 | 9 | | |
| |||
0 commit comments