revert pinned hostmem for pod comm#292
Open
AtlantaPepsi wants to merge 1 commit into
Open
Conversation
Contributor
There was a problem hiding this comment.
Pull request overview
This PR adjusts pod-communication (fabric-handle) memory allocation to better support exporting extended (HOST_NUMA) memory across pods, and avoids a failure mode where NUMA page validation caused one rank to error while other ranks hung in collectives.
Changes:
- Add a
GetMemLocation()helper to correctly selectDEVICEvsHOST_NUMAlocations for pod-comm VMM allocations and access descriptors. - Remove
CheckPages()/move_pages()validation for fabric-exportable host allocations (keep zeroing), with an explanatory comment. - Extend CUDA-compat macro aliases/undefs to include
hipMemLocationandhipMemLocationTypeHostNuma.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+1426
to
+1431
| // Determine location id | ||
| if (memDevice.memType == MEM_CPU_CLOSEST) { | ||
| location.id = GetClosestCpuNumaToGpu(memDevice.memIndex); | ||
| } else { | ||
| location.id = memDevice.memIndex; | ||
| } |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation
This is to add back the option of extended GPU memory in cross pod transfer.
Technical Details
Removed
CheckPages()insideAllocateMemory()for pinned host memory. This would cause a silent fail on the node where it's allocated, and other nodes will hang at broadcast insideExchangeMemory.It's also not tested yet on either Nvidia or AMD platforms.
Test Plan
Test Result
Submission Checklist