Enable subgroup float reductions on AMD GCN#12

Open

rasbid wants to merge 1 commit intomasterfrom

codex/split-use_subgroups-logic-for-amd_gcn

Owner

rasbid commented Oct 13, 2025

Summary

keep a separate float subgroup flag so AMD GCN devices can use subgroup arithmetic when available
forward the float subgroup setting through the mul_mat_vec pipelines while keeping quantized paths unchanged

Testing

not run (not available in container)

https://chatgpt.com/codex/tasks/task_e_68ebeecaa0a0833095175b630f0bf88c


          Enable subgroup float reductions on AMD GCN

bbf8916

rasbid added the codex label

— with

ChatGPT Codex Connector

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels