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
I encountered the following error when try split_k_slices(2) on V100. As #73 (comment) suggested, I use cudaGetErrorString to dump error message.
an illegal memory access was encountered
terminate called after throwing an instance of 'std::runtime_error'what(): Failed to copy equality flag to device.
Aborted
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
I encountered the following error when try split_k_slices(2) on V100. As #73 (comment) suggested, I use cudaGetErrorString to dump error message.
cutlass/tools/util/include/cutlass/util/reference/device/tensor_compare.h
Lines 188 to 195 in af332d4
M=N=K=1776, FP16, split_k_slices=2
split_k_slices=1 can pass, but split_k_slices=2 hit error. Decrease mnk size also can pass.
I'm confused why >1776 will hit IMA.
update:
I tried the dev branch and the problem disappeared. It appears that this issue only occurs in cutlass-2.5.
Beta Was this translation helpful? Give feedback.
All reactions