You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We may generate out-of-bounds reads when doing 2D tiling for inputs that are smaller in the outer dimension than the tile size. E.g. a standard matrix multiplication will have problems when given inputs of size [31][32] by [32][32].
The text was updated successfully, but these errors were encountered:
There is a straightforward fix for this. Unfortunately, it will reduce the performance of 2D block tiled kernels, like matrix multiplication. I think we just have to accept that for now, and see if we can recover the performance later.
We may generate out-of-bounds reads when doing 2D tiling for inputs that are smaller in the outer dimension than the tile size. E.g. a standard matrix multiplication will have problems when given inputs of size
[31][32]
by[32][32]
.The text was updated successfully, but these errors were encountered: