Fix incorrect K dim in CuTe MMA Atom doc. (#2544)
This commit is contained in:
@ -426,7 +426,7 @@ where we see 16 copies of the 64x8 tile.
|
||||
GMMA atoms that consume A and B sources directly from shared memory are a bit interesting. The GMMA Descriptor is constructed on an entire tile of A and/or B data in shared memory rather than being partitioned by threads. That is, every thread sees the entire tile of data and the tile is not reordered so that the descriptor can be constructed on it. In `ALayout` form, this can be expressed
|
||||
|
||||
```cpp
|
||||
// (T128,V64x8) -> (M64,K16)
|
||||
// (T128,V64x16) -> (M64,K16)
|
||||
using ALayout = Layout<Shape <_128, Shape <_64,_16>>,
|
||||
Stride< _0, Stride< _1,_64>>>;
|
||||
```
|
||||
|
||||
Reference in New Issue
Block a user