PLAN: finding 38 with the Poco measured like the SM8850 - #8
Merged
Merged
Conversation
… had no long-prompt CPU run On Mali the CPU file decodes 2.2 to 3.7 times faster at every length and the GPU only reads a long prompt faster, 1.2 to 1.5 times; the study's raw files are in openweights. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Documentation only. Finding 38 and the Vulkan bullet in
docs/PLAN.mdnow carry the POCO's like-for-like CPU-vs-GPU run, measured through openweights' engine and ExecuServe. The Mali-G925 had no long-prompt CPU run before this.The raw logs and the runs CSV are in openweights
tools/eval/results/vulkan-2026-10-04/.pytestpasses.🤖 Generated with Claude Code
Need help on this PR? Tag
@codesmith-botwith what you need. Autofix is enabled.