@@ -208,11 +208,11 @@ struct TiledCopy : Copy_Atom
208
208
// Tile a tensor or a layout from shape
209
209
// (M,N,...)
210
210
// to shape
211
- // ((ThrV,ThrX),FrgV ,(RestM,RestN,...))
211
+ // (Thr,(FrgV,FrgX) ,(RestM,RestN,...))
212
212
// where
213
- // ThrV: The threads local to a COPY_ATOM Src.
214
- // ThrX: The threads tiled across COPY_ATOMs Src.
213
+ // Thr: The logical threads within the tiled copy.
215
214
// FrgV: The values local to a COPY_ATOM Src.
215
+ // FrgX: The values tiled across COPY_ATOMs Src.
216
216
// RestM: The values tiled in M.
217
217
// RestN: The values tiled in N.
218
218
template <class STensor >
@@ -229,11 +229,11 @@ struct TiledCopy : Copy_Atom
229
229
// Tile a tensor or a layout from shape
230
230
// (M,N,...)
231
231
// to shape
232
- // ((ThrV,ThrX),FrgV ,(RestM,RestN,...))
232
+ // (Thr,(FrgV,FrgX) ,(RestM,RestN,...))
233
233
// where
234
- // ThrV: The threads local to a COPY_ATOM Dst.
235
- // ThrX: The threads tiled across COPY_ATOMs Dst.
234
+ // Thr: The logical threads within the tiled copy.
236
235
// FrgV: The values local to a COPY_ATOM Dst.
236
+ // FrgX: The values tiled across COPY_ATOMs Dst.
237
237
// RestM: The values tiled in M.
238
238
// RestN: The values tiled in N.
239
239
template <class DTensor >
@@ -250,7 +250,7 @@ struct TiledCopy : Copy_Atom
250
250
// Tile a tensor or a layout from shape
251
251
// ((TileM,TileN,...), (RestM,RestN,...))
252
252
// to shape
253
- // ((ThrV,ThrX),FrgV ,(RestM,RestN,...))
253
+ // (Thr,(FrgV,FrgX) ,(RestM,RestN,...))
254
254
template <class Tensor , class Ref2TrgLayout >
255
255
CUTE_HOST_DEVICE constexpr static
256
256
auto
0 commit comments