Skip to content

Bump gpu-kernels: 17a882640fa88 (its main), goldens re-recorded - #571

Merged
bong-water-water-bong merged 2 commits into
mainfrom
1bit/bump-gpu-kernels-17a882640
Oct 25, 2026
Merged

bong-water-water-bong merged 2 commits into
mainfrom
1bit/bump-gpu-kernels-17a882640

Conversation

@bong-water-water-bong

Copy link
Copy Markdown
Collaborator

Moves the private HIP add-on's pin to its main: config/gpu-kernels.config
sha=744bd713d968… → sha=17a882640fa88e78ec92af42703a0707a6b7a061 (one-token MoE decode kernels, on top of the
DFlash2 verify/selector and NVFP4/Q2_0 work that was already there). Checked first that the new pin still satisfies the
config's require=addon.cmake kernels — both are present.

Goldens re-recorded, because they had to move on two counts. tests/golden/hw-gate/manifest.json recorded the last
set at llama.cpp 8792923f, and #565 moved that pin to e7e5cd08; with the add-on moving too, both the kat goldens
(7 models × 9 cases) and the KLD bases are re-recorded here:

scripts/hw-gate.sh --phases build,kat --record
scripts/hw-gate.sh --phases kld --record
scripts/hw-gate.sh --phases build,kat,kld        # check back: 13/13 PASS
build PASS
kat 7/7 PASS — Qwen3-0.6B (+q8_0 K and K/V), ZAYA1-8B, Qwen3.8-27B, Qwen3-Coder-30B-A3B, Qwen3-4B-IQ4_XS
kld 5/5 PASS — mean KLD ≤ 1e-6, p99 ≤ 4e-5, top-1 agreement 100 %
run 344 s, Tctl 58 °C at end, 0 amdgpu ring timeouts

The five .kld bases are not in git (they live on the gate box; manifest.json pins their sha256). They are staged
in ~/wt/release-rc/hw-gate-store-pin/kld and get copied into ~/.cache/hw-gate/kld/ as soon as this merges —
which is also the moment this PR's own hw_gate kld rows stop reporting a base mismatch. Until then expect those rows
to fail with "base logits" and nothing else; it is not a real failure, and I will comment here when the bases are in
place.

Changed files: config/gpu-kernels.config, tests/golden/hw-gate/kat-*.json (7), tests/golden/hw-gate/manifest.json.
python3 tools/copyright.py --check rc=0. Not merged here.

🤖 Generated with Claude Code

config/gpu-kernels.config: 744bd713d968 -> 17a882640fa88e78ec92af42703a0707a6b7a061, the private HIP
add-on's main (one-token MoE decode kernels, plus the DFlash2 verify/selector and NVFP4/Q2_0 work already
on it; the add-on's addon.cmake and kernels/ are still there, which is what the config's `require` needs).

Goldens are recorded at the release pin (docs/hw-gate.md), and they had to move twice over: they were last
recorded at llama.cpp 8792923f, and #565 moved that pin to e7e5cd08, so the kat goldens and the KLD bases
are re-recorded here with the new add-on. Re-recorded with
`scripts/hw-gate.sh --phases build,kat --record` then `--phases kld --record` on Strix Halo, then checked
back the same way: build, kat (7 models, 9 cases each) and kld (5 models) all PASS, KLD means <= 1e-6 with
top-1 agreement 100%, 0 amdgpu ring timeouts.

The five .kld bases live on the gate box, not in git (manifest.json pins their sha256); they are copied into
~/.cache/hw-gate/kld right after this merges, which is also when this PR's own hw_gate kld rows stop
reporting a base mismatch.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
@github-actions

github-actions Bot commented Oct 25, 2026 •

Copy link
Copy Markdown

PR Reviewer Guide 🔍

(Review updated until commit 6e515e2)

Here are some key observations to aid the review process:

🎫 Ticket compliance analysis 🔶

565 - Partially compliant

Compliant requirements:

  • Bump HRX submodules to llama.cpp e7e5cd08f9bb and hrx-system 82804197bcbe
  • Move the private HIP add-on's pin to its main: config/gpu-kernels.config
  • Re-record goldens due to changes in kernel behavior

Non-compliant requirements:

  • Validate on Strix Halo with long-prompt correctness probe and test-backend-ops -b HRX0

Requires further human verification:

  • Validate on Strix Halo with long-prompt correctness probe and test-backend-ops -b HRX0
⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪
🧪 No relevant tests
🔒 No security concerns identified
⚡ Recommended focus areas for review

Pin Update

The gpu-kernels pin has been updated from 744bd713d968... to 17a882640fa88e78ec92af42703a0707a6b7a061. This change is part of the bump to the main branch of the HIP add-on, which includes one-token MoE decode kernels and other improvements. The update is correctly reflected in the configuration file.

sha=17a882640fa88e78ec92af42703a0707a6b7a061
SHA256 Mismatch

The base_sha256 values in manifest.json have changed, indicating that the golden files have been re-recorded. This is expected due to the kernel updates. However, the change in SHA256 values should be verified to ensure that the new golden files are indeed generated from the updated kernel configurations and not due to an unintended change in the model or data.

"base_sha256": "af6eb4309794905f9535be33cddc779e6ad1a749e03ee59fe2b0509fd971bada",

⚠️ Review coverage: The following files were not included in this review because of the token budget:

  • tests/golden/hw-gate/kat-qwen3-0.6b.json
  • tests/golden/hw-gate/kat-qwen3-0.6b-kvq8.json
  • tests/golden/hw-gate/kat-qwen3-0.6b-kq8.json
  • tests/golden/hw-gate/kat-qwen3-4b-iq4xs.json
  • tests/golden/hw-gate/kat-qwen3-coder-30b-a3b.json

@bong-water-water-bong
bong-water-water-bong enabled auto-merge (squash) October 25, 2026 14:46
@github-actions

Copy link
Copy Markdown

Persistent review updated to latest commit 6e515e2

@bong-water-water-bong
bong-water-water-bong merged commit 0d0c846 into main Oct 25, 2026
8 checks passed
@bong-water-water-bong
bong-water-water-bong deleted the 1bit/bump-gpu-kernels-17a882640 branch October 25, 2026 14:50
@bong-water-water-bong

Copy link
Copy Markdown
Collaborator Author

Merged as 0d0c846; the five .kld bases are now in the gate box shared store, verified against the manifest that shipped with this PR (all MATCH) and hardlinked so a wipe of ~/.cache/hw-gate cannot lose them again:

total 3850200
drwxr-xr-x 2 bcloud bcloud       4096 Oct 25 11:51 .
drwxr-xr-x 5 bcloud bcloud       4096 Oct 24 08:12 ..
-rw-rw-r-- 2 bcloud bcloud  619931604 Oct 25 11:06 qwen3-0.6b.kld
-rw-rw-r-- 2 bcloud bcloud  619931604 Oct 25 11:08 qwen3-4b-iq4xs.kld
-rw-rw-r-- 2 bcloud bcloud  619931604 Oct 25 11:07 qwen3-coder-30b-a3b.kld
-rw-rw-r-- 2 bcloud bcloud 1013178324 Oct 25 11:07 qwen3.8-27b.kld
-rw-rw-r-- 2 bcloud bcloud 1069596564 Oct 25 11:06 zaya1-8b.kld

So hw_gate on every PR from here checks against the new manifest and bases.

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

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant