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
The dynamic dispatch PR for Merge Sort #3525 introduced a bug into how the tuning policy is selected. By making VSMemHelper a template of the DispatchMergeSort class that is instantiated with typename PolicyHub::MaxPolicy::MergeSortPolicy, this PR changed the original behavior, where we were instantiating it with the ActivePolicy passed from Invoke(), which would have selected the correct tuning policy. See this comment as well #3525 (comment).
How to Reproduce
There are differences in the SASS for pre-Pascal architectures before and after applying the changes in this PR.
Expected behavior
VSMemHelper should be instantiated with the correct policy.
Reproduction link
No response
Operating System
No response
nvidia-smi output
No response
NVCC version
No response
The text was updated successfully, but these errors were encountered:
Is this a duplicate?
Type of Bug
Something else
Component
CUB
Describe the bug
The dynamic dispatch PR for Merge Sort #3525 introduced a bug into how the tuning policy is selected. By making
VSMemHelper
a template of theDispatchMergeSort
class that is instantiated withtypename PolicyHub::MaxPolicy::MergeSortPolicy
, this PR changed the original behavior, where we were instantiating it with theActivePolicy
passed fromInvoke()
, which would have selected the correct tuning policy. See this comment as well #3525 (comment).How to Reproduce
There are differences in the SASS for pre-Pascal architectures before and after applying the changes in this PR.
Expected behavior
VSMemHelper
should be instantiated with the correct policy.Reproduction link
No response
Operating System
No response
nvidia-smi output
No response
NVCC version
No response
The text was updated successfully, but these errors were encountered: