[mlir][amdgpu] Add scaled_ext_packed{8,16} operations #159830
amd-eochoalo
merged
13 commits into
llvm:main
from
amd-eochoalo:eochoa/2025-09-19/cvt-amd-gpu
krzysz00
requested changes
on 2025-09-19
[mlir][amdgpu] Add scaled_ext_packed{8,16} operations
9c09c356
Use TypesMatchWith and make the scale a constant type
f8b11c4a
Add note about availability on gfx1250+
e71f8d8c
kuhar
approved these changes
on 2025-10-16
Add verifier for blockSize and firstScaleByte
4f83cd9a
Use ConfinedType
b7763efe
Only use AllOfType
d50b6fe5
krzysz00
commented
on 2025-10-16
Verify shape matches and better type constraint
3cdb1747
Added scale type to the assembly format
30dcfea0
kuhar
commented
on 2025-10-17
Use functional-type
67261e52
Use : source_ty, scale_ty -> res_ty
fadf035f
kuhar
approved these changes
on 2025-10-17
Adds examples and better verification
03831000
kuhar
commented
on 2025-10-17
kuhar
commented
on 2025-10-17
no else after return and remove global resolution
7aa61696
indentation and syntax highlighting
7a5fea59
Login to write a write a comment.
Login via GitHub