deepseek-ai/DeepEP
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepEP.git synced 2026-11-05 00:47:32 +00:00
Code Issues 308 Projects Packages Wiki Activity

[GH-ISSUE #186] [bug] Incorrect function implementation in csrc/deep_ep.cpp #159

New issue
Closed
opened 2026-10-26 09:55:21 +00:00 by bsakaguchi · 3 comments
bsakaguchi commented 2026-10-26 09:55:21 +00:00
Owner
Copy link

Originally created by @HypoopyH on GitHub (Jun 4, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/186

... 
\\ csrc/deep_ep.cpp:139
int Buffer::get_root_rdma_rank(bool global) const {
    return global ? nvl_rank : 0;
}

I'm confused, this functionality seemed not implemented correctly, the correct change should be:

... 
\\ csrc/deep_ep.cpp:139
int Buffer::get_root_rdma_rank(bool global) const {
    return global ? rdma_rank : 0;
}
Originally created by @HypoopyH on GitHub (Jun 4, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/186 ``` ... \\ csrc/deep_ep.cpp:139 int Buffer::get_root_rdma_rank(bool global) const { return global ? nvl_rank : 0; } ``` I'm confused, this functionality seemed not implemented correctly, the correct change should be: ``` ... \\ csrc/deep_ep.cpp:139 int Buffer::get_root_rdma_rank(bool global) const { return global ? rdma_rank : 0; } ```
bsakaguchi commented 2026-10-26 09:55:22 +00:00
Author
Owner
Copy link

@LyricZhao commented on GitHub (Jun 5, 2025):

No, this is correct. The RDMA domain of the current GPU covers all the GPUs with same device index, and the root rank is in node 0.

<!-- gh-comment-id:2942921978 --> @LyricZhao commented on GitHub (Jun 5, 2025): No, this is correct. The RDMA domain of the current GPU covers all the GPUs with same device index, and the root rank is in node 0.
bsakaguchi commented 2026-10-26 09:55:22 +00:00
Author
Owner
Copy link

@LyricZhao commented on GitHub (Jun 5, 2025):

e.g.

Node 0: 0 1 2 3 4 5 6 7
Node 1:  8 9 10 11 12 13 14 15

Rank 8 and rank 0 both have GPU index 0, so 0/8 is in the same RDMA domain, the root of this group is 0.
For rank 4 and 12, the root of this domain is 4.

<!-- gh-comment-id:2942925205 --> @LyricZhao commented on GitHub (Jun 5, 2025): e.g. ``` Node 0: 0 1 2 3 4 5 6 7 Node 1: 8 9 10 11 12 13 14 15 ``` Rank 8 and rank 0 both have GPU index 0, so 0/8 is in the same RDMA domain, the root of this group is 0. For rank 4 and 12, the root of this domain is 4.
bsakaguchi commented 2026-10-26 09:55:23 +00:00
Author
Owner
Copy link

@HypoopyH commented on GitHub (Jun 6, 2025):

well, i get it.

<!-- gh-comment-id:2947863770 --> @HypoopyH commented on GitHub (Jun 6, 2025): well, i get it.
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
hybrid-ep
chore/remove-clang-format-yapf
v2.5
nvDev
try-fix-cu128
mori-ep
antgroup-opt
compat-old-torch
add-linter
eager-rdma
v1.2.1
Labels
Clear labels   
pull-request
Mirrored from GitHub Pull Request

No labels
pull-request
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/DeepEP#159
No description provided.