Conversation
Contributor
|
👍 |
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This adds an opt-in CUDA experiment that shares the per-job prestate across a thread block and expresses the existing square/reduction sequence in one PTX block. It pairs that layout with 1024-thread mining blocks and a 640-block cap; diagnostic hashing stays at 256 threads. Default builds retain the current kernel and launch geometry.
The equivalent prototype on RTX 3080 Ti observed 363.89 → 384.07 MH/s (+5.55%) in an offline A/B/B/A engine run. Baseline drift was substantial, so this is not a statistically established gain or a claim for other NVIDIA cards. The fixed cap needs broader device validation. Full method and per-process results are in
docs/cuda-shared-square.md.The PR includes an offline benchmark and
QUANTUS_REQUIRE_CUDA=1to reject skipped device tests. CPU candidate verification, nonce coverage and rejection handling are preserved. This draft is intended for review and reproduction before considering wider enablement.Validation on RTX 3080 Ti: default release engine tests 10/10 passed; feature-enabled release engine tests 10/10 passed, both with QUANTUS_REQUIRE_CUDA=1 and one test thread. Feature-enabled all-targets cargo check passed. git diff --check passed.