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 try to use CLTune for tuning the XgemmDirect kernel which is part of the CLBlast library.
In CLBlast, the global size is set as ((1 + ((kSizeM - 1) / WGD))*WGD * MDIMCD) / WGD in dimension 1 and as ((1 + ((kSizeN - 1) / WGD))*WGD * NDIMCD) / WGD in dimension 2. Is it possible to set the same global in CLTune?
The text was updated successfully, but these errors were encountered:
Somehow I missed this issue, sorry for the late reply. Not sure what you mean though? CLBlast uses CLTune for tuning, so yes, everything that is done in CLBlast is possible in general with CLTune.
I try to use CLTune for tuning the
XgemmDirect
kernel which is part of the CLBlast library.In CLBlast, the global size is set as
((1 + ((kSizeM - 1) / WGD))*WGD * MDIMCD) / WGD
in dimension 1 and as((1 + ((kSizeN - 1) / WGD))*WGD * NDIMCD) / WGD
in dimension 2. Is it possible to set the same global in CLTune?The text was updated successfully, but these errors were encountered: