Skip to content

[RFC][DSV4][ROCm][P4] Communication and parallelism #530

Open
Open
[RFC][DSV4][ROCm][P4] Communication and parallelism#530
Labels
DSv4deepseek-P4platform: rocmSpecific tasks specific to AMD graphics cards (such as CK, bpreshuffle/FA)type: designIssues requiring in-depth discussion of architecture design

Description

@inaniloquentee
inaniloquentee
opened on Oct 10, 2026

P1-P7 are source module IDs, not priorities.

1. Expert dispatch, buffers, and reverse traffic

Anchor RFC: Expert dispatch, buffers, and reverse traffic (P4-A)

2. Parallel collectives and overlap

Anchor RFC: Parallel collectives and overlap (P4-B)

Exit

Expert traffic and fixed-order collectives preserve bytes and provenance.

Activity

  1. inaniloquentee
    added
    platform: rocmSpecific tasks specific to AMD graphics cards (such as CK, bpreshuffle/FA)
    type: designIssues requiring in-depth discussion of architecture design
    DSv4
    deepseek-P4
    on Oct 10, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    DSv4deepseek-P4platform: rocmSpecific tasks specific to AMD graphics cards (such as CK, bpreshuffle/FA)type: designIssues requiring in-depth discussion of architecture design

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions