mirror of
https://github.com/deepseek-ai/DeepEP.git
synced 2026-11-05 00:47:32 +00:00
[PR #588] [MERGED] Add Permute/Unpermute Fusion Code Path #647
Labels
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
deepseek-ai/DeepEP#647
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
📋 Pull Request Information
Original PR: https://github.com/deepseek-ai/DeepEP/pull/588
Author: @Autumn1998
Created: 3/3/2026
Status: ✅ Merged
Merged: 3/31/2026
Merged by: @jershi425
Base:
hybrid-ep← Head:tongliu_permute_fusion📝 Commits (10+)
ccd840arefator to ready for permute fusionb779dfatry to add new backend for fuse permute-dispatch57cc5eefinish adding fuse permute-dispatche21347afix inter-node bug090de83fix inter-node8f87560refactor, tune the fused permure APIc137775aufo configurer8838bc4rm fence.release.sys8f3b43crefactor benchmark36cac45minor change📊 Changes
25 files changed (+4504 additions, -1187 deletions)
View changed files
📝
csrc/hybrid_ep/allocator/allocator.cu(+9 -7)📝
csrc/hybrid_ep/backend/hybrid_ep_backend.cuh(+2610 -186)📝
csrc/hybrid_ep/buffer/internode.cu(+47 -23)📝
csrc/hybrid_ep/buffer/internode.cuh(+13 -10)📝
csrc/hybrid_ep/buffer/intranode.cu(+136 -18)📝
csrc/hybrid_ep/buffer/intranode.cuh(+21 -8)📝
csrc/hybrid_ep/config.cuh(+394 -5)➕
csrc/hybrid_ep/coordinator.cuh(+13 -0)📝
csrc/hybrid_ep/executor/executor.cu(+169 -101)📝
csrc/hybrid_ep/executor/executor.cuh(+50 -7)📝
csrc/hybrid_ep/extension/allgather.cu(+18 -5)📝
csrc/hybrid_ep/extension/allgather.cuh(+10 -6)📝
csrc/hybrid_ep/extension/permute.cu(+47 -73)📝
csrc/hybrid_ep/extension/permute.cuh(+24 -44)📝
csrc/hybrid_ep/hybrid_ep.cu(+119 -109)📝
csrc/hybrid_ep/hybrid_ep.cuh(+24 -29)📝
csrc/hybrid_ep/jit/compiler.cu(+70 -21)📝
csrc/hybrid_ep/jit/compiler.cuh(+16 -3)📝
csrc/hybrid_ep/pybind_hybrid_ep.cu(+84 -31)📝
csrc/hybrid_ep/utils.cuh(+2 -0)...and 5 more files
📄 Description
🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.