-
Notifications
You must be signed in to change notification settings - Fork 221
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[OCL][MI100][MI200] Fix iGemm ASM GTC XDLOPS failures with OCL backend (Staging 95b58f72f) (#1317) and Implement abstraction for multi-buffer workspace (#1326) #1327
Conversation
shaojiewang
commented
Dec 7, 2021
•
edited
Loading
edited
- fix Implement and use an abstraction that represents a multi-buffer workspace #1326 : abstract a class to hold the offsets and sizes of each sub buffer for transpose+asm_igemm solvers.
- fix [OCL][MI100][MI200] iGemm ASM GTC XDLOPS failures with OCL backend (Staging 95b58f72f) #1317 : fix mem fault on OCL BE
- add some test case for transpose+asm_igemm solvers.
This comment has been minimized.
This comment has been minimized.
@atamazov I can not run ctest on gfx90a with OCL BE at all. Could you please help to check if some code work not properly? |
@shaojiewang Yes, as soon I do some other urgent work. |
@atamazov @shaojiewang should we include this into the next round of staging too? Is it related to |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This should remove WORKAROUND_ISSUE_1317 from #1321
@junliume No, this is not needed in 5.0. |
Oops, looked into wrong diff. |
Review recommendations adopted. Please re-review.
@JehandadKhan |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This can't be merged as is due to #1349
@shaojiewang #1352 (which extends WORKAROUND_ISSUE_1317 to tests) will be merged soon. This PR should revert it. |
OK |
#1352 merged, now we can move forward with this one. Please re-review
@shaojiewang tried to resolve conflict, please double check :) |
Thanks a lot, I'll check very soon. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
@shaojiewang Please do not use hash sign in branch names anymore. It seems incompatible with some functions of GUI of our CI. |
Yes, sure. Sorry for that. |
No problem ;) |