-
Notifications
You must be signed in to change notification settings - Fork 75
[offload][plugin] add new field OMPX_AdjustNumTeamsForXteamRedSmallBl… #469
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
Conversation
…ockSize for MI2xx Recovers performance for CQE testing of 513.soma on mi2xx. Disavles adjustment for mi2xx and enables for all others. see: https://ontrack-internal.amd.com/browse/SWDEV-553832
| {"MI250X",{.OMPX_UseMultipleSdmaEngines = true, | ||
| .OMPX_AdjustNumTeamsForXteamRedSmallBlockSize=0}}, | ||
| {"MI250X/MI250",{ | ||
| .OMPX_UseMultipleSdmaEngines = true, | ||
| .OMPX_AdjustNumTeamsForXteamRedSmallBlockSize=0}}, |
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.
why 2 entries for mi250x?
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.
r11, a supermicro, returns the following
Marketing Name: AMD Instinct MI250X/MI250
whereas lockhart returns
Marketing Name: AMD Instinct MI250X
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.
Interesting, thanks. I did not expect the r11 GPUs to have mi250X in their names.
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.
to add to the confusion:
lockhart mi250X has 110 CUs
r11 mi250X/MI250 has 104 CUs
r9 mi210 has 104 CUs
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.
yeah, I looked up the specs.
lockhart has proper mi250X. That's top-of-the-line 110 CUs in OAM form factor.
mi250 is the same number of CUs (104) as mi210 but in OAM form factor.
mi210 is PCIe.
dhruvachak
left a comment
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.
…ockSize for MI2xx
Recovers performance for CQE testing of 513.soma on mi2xx. Disavles adjustment for mi2xx and enables for all others.
see:
https://ontrack-internal.amd.com/browse/SWDEV-553832