cuda: add a cuda-batched runtime that micro-batches concurrent run() …#2421
Open
kali wants to merge 1 commit into
Open
cuda: add a cuda-batched runtime that micro-batches concurrent run() …#2421kali wants to merge 1 commit into
kali wants to merge 1 commit into
Conversation
…calls Independent caller threads each fire tiny per-token predictor/joiner calls; with many concurrent streams that is one serialized CPU<->GPU round-trip per token per stream. A per-model background worker drains the in-flight queue, concatenates inputs along the symbolic batch axis, runs the inner CUDA runnable once, and scatters the row-slices back to each caller. WIP parked for later.
|
🔴 Bench vs main — 2 speed regression(s) Reference: main nightly, latest 2026-06-29 (0d old) · PR Speed — evaltime · prefill · decode
🟢 1 improvement(s)
lower is better except prefill/decode (tok/s) · adaptive thresholds (max(floor, k×noise) vs the series' own history) · single-shot vs nightly reference · full report → run |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
…calls
Independent caller threads each fire tiny per-token predictor/joiner calls; with many concurrent streams that is one serialized CPU<->GPU round-trip per token per stream. A per-model background worker drains the in-flight queue, concatenates inputs along the symbolic batch axis, runs the inner CUDA runnable once, and scatters the row-slices back to each caller. WIP parked for later.