doc: prefill/decode performance technique survey (June 2026)#43
Closed
hvasconcelos wants to merge 1 commit into
Closed
doc: prefill/decode performance technique survey (June 2026)#43hvasconcelos wants to merge 1 commit into
hvasconcelos wants to merge 1 commit into
Conversation
Survey of 2024-2026 inference-performance techniques applicable to the engine, prioritized by Apple-Silicon evidence, effort, and compatibility with the exactness invariants. Covers wired-memory limits, sampler compilation, n-gram and MTP speculative decoding, token-budget chunked prefill, packed prefill, fused Metal kernels, and paged attention, plus MLX framework status (v0.31.2 is current; no fused quantized-KV SDPA upstream). https://claude.ai/code/session_0119yHPn3SDzSACP7Cy4V2kM
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Survey of 2024-2026 inference-performance techniques applicable to the
engine, prioritized by Apple-Silicon evidence, effort, and compatibility
with the exactness invariants. Covers wired-memory limits, sampler
compilation, n-gram and MTP speculative decoding, token-budget chunked
prefill, packed prefill, fused Metal kernels, and paged attention, plus
MLX framework status (v0.31.2 is current; no fused quantized-KV SDPA
upstream).
https://claude.ai/code/session_0119yHPn3SDzSACP7Cy4V2kM