Revised Unified Virtual Memory(UVM) support#107
Open
yechen3 wants to merge 154 commits intoaccel-sim:devfrom
Open
Revised Unified Virtual Memory(UVM) support#107yechen3 wants to merge 154 commits intoaccel-sim:devfrom
yechen3 wants to merge 154 commits intoaccel-sim:devfrom
Conversation
- Full functional implementation; Performance/time simulation pending - CUDA APIs for cudaMallocManaged, cudaMemPrefetchAsync, and cudaDeviceSynchronize - Data-structures to maintain mapping between CPU and GPU side memory during cudaMallocManaged - Track mapping during cudaSetupArgument to override CPU memory pointer by GPU side memory pointer
- Same benchmark with cudaMalloc and cudaMemcpy, i.e., old APIs - Same benchmark with cudaMallocManaged, i.e., new UVM API
- Adding PCI-e latency as part of config (value subject to change based on architecture) - Parsing logic of latency and setting clock domain - Adding valid flag as part of page table implementation
This reverts commit f708eb3.
… gddr size constraint
- TLB look up - page table access/walk latency - multi-lane bidirectional PCI-E with far fetch latency
- Don't process if warp instruction's access queue is empty
- Change input size for bfs, nw, pathfinder, srad_v2 etc. to finish in reasonable time - Add pathfinder benchmark for managed code - Fixing BFS managed code
…ze, and a bollean flag to denote whether data is copied from cpu to gpu on first kernel launch
…llocations from GPU to CPU
… queue is ready to complete. Determine based on whether dispatched warps from all SMs are stalling for PCI-E transfer and no progress can be made by any component of the simulator.
…ion write back or device synchronize check for allocations which were copied at the first place during kernel launch from CPU to GPU
…nslated to managed code
- kernel by kernel, data structure wise basic block access - need to implement actual detection logic and policy engine
- remaining policy engine
- from experiments on smart adaptive runtime with dynamic migration pattern detection and memory management - along with validation results
tgrogers
requested changes
Mar 16, 2025
tgrogers
left a comment
tgrogers
left a comment
There was a problem hiding this comment.
@yechen3 - one weird side-effect of the integration changes is that you ended up re-adding a bunch of files that have since been deleted. Jenkinsfile, bitbucket*, all the old configs in the old folders outside of tested and deprecated.
Please clean this up. Make sure you are only adding files that are actually added for this to work. Otherwise, the directory structure should be the same as the mainline dev.
Author
|
@tgrogers Got it. I've already cleaned things up, so everything should be in order now. Let me know if there's anything else that needs attention. |
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
4 participants
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.
Same to PR#85, but this has a cleaner commit history.