Add experimental Radeon 8065S (Halo/gfx1151) custom-workload support - #1468
Draft
powderluv wants to merge 1 commit into
Draft
Add experimental Radeon 8065S (Halo/gfx1151) custom-workload support#1468powderluv wants to merge 1 commit into
powderluv wants to merge 1 commit into
Conversation
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Radeon 8065S currently has no dispatch identity, and spaced product names such as
AMD Radeon 8065Sare not recognized. This adds experimental Halo custom-workload bring-up:radeon8065sresolves togfx1151/ 40 CUs, keeps its own script runner label, and is recognized through rocm-smi or PyTorch product names. GenericAMD Radeon Graphicson gfx1151 remains ambiguous and requires an explicit board selection.The change includes a standalone GPU attention smoke workload checked against an independent CPU reference, plus setup and validation limits. It does not add Instinct peak constants, serving runners, scheduler changes, provider changes, or dependency workarounds.
Closes #1467. Related platform work: #1390 adds broader Radeon 8060S support. This smaller PR covers the separately tested Radeon 8065S identity and ambiguous-name handling. External Magpie/GEAK/TraceLens bring-up remains outside this contribution; no full kernel-optimization support is claimed.
Validation:
ruff check .andruff format --check .passed.--python-version 3.14. Default repository-wide mypy stops in installed NumPy stubs because the configured Python target is older than their syntax.e3a0ef325) in this environment; they concern source-resolver paths, Python 3.14 behavior, and campaign/shadow-repo handling. The full suite is not claimed green.Physical end-to-end evidence is limited to the local custom Qwen3-0.6B BF16 benchmark: baseline approximately 67 tokens/s, configuration candidates exercised, no accepted gain. The 15-minute source/kernel trial exhausted its budget before GEAK executed.
Breaking changes: no. Single concern: initial Radeon 8065S platform identification and a reproducible custom-workload smoke test.