Upstream/flash kv decode - #55
Open
DrJesseGlass wants to merge 2 commits into
Open
Conversation
Adds an f16 variant of the raw interleaved KV cache and matching CPU flash kernels: RawInterleavedKvCacheF16 (head-major, grow-on-demand) with causal_decode_f16kv_interleaved and causal_prefill_f16kv_headmajor. Storing K/V as f16 halves the bytes streamed per decoded token - decode is memory bound on CPU, so this is a direct throughput win - while accumulation stays in f32 via the existing widening dot. The optional f16-attn-dot cargo feature (default OFF, bit-exactness preserved by default) switches the q.k dot to native fp16 FMA where the target supports it. Measured on Qwen3-0.6B (Apple M1, CPU): decode 1.28x at 1 thread, 1.13x at 4 threads; greedy output byte-identical to the f32 cache. On Graviton2 the f16 cache underpins a 1.14x-1.17x decode win in the downstream stack. Tests: cpu_flash_attn suite under both default and f16-attn-dot builds.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
No description provided.