[Mandatory for Wormhole users] Update Hugepages settings by reinstalling it via latest main #4451
tt-rkim
announced in
Contributor announcements
Replies: 1 comment
-
cc: @tenstorrent-metal/developers @tenstorrent-metal/external-developers |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
As part of revamping fast dispatch for multi devices and WH, main now has some new changes to have a
CommandQueue
per device.This means we need to upgrade our machines for hugepages. Specifically, the deal with hugepages is now:
If you have a WH machine, you must have 2 1GB hugepages per MMIO-connected chip. If you have a GS machine, you may remain using only 1 1GB hugepage.
Therefore, this change is required for WH machines.
I have made these changes for CI machines.
I will be doing some maintenance myself on the shared WH cloud BMs soon.
I will not be doing such maintenance on WH VMs before the new year. If you would like me to, please send me an email with your machine name, IP, and confirmation of WH architecture by January 3rd, 2024, 9am Toronto time. I will ignore Slack messages regarding this topic, unless you have already sent me email regarding so. I will do the maintenance after that time.
If you use IRD, you will have to check that your machine meets the HP requirements yourself. I can't really help you there.
Let us know if you have questions.
Beta Was this translation helpful? Give feedback.
All reactions