server : consolidate slot selection into get_available_slot - #24755
Merged
Merged
Conversation
Absorb get_slot_by_id logic into get_available_slot so slot selection is handled by a single function call. When a specific slot id is requested, the LCP similarity check still runs to enable proper prompt cache updates. Assisted-by: pi:llama.cpp/Qwen3.6-27B
ggerganov
marked this pull request as ready for review
June 19, 2026 06:21
adrianhoehne
pushed a commit
to adrianhoehne/llama.cpp
that referenced
this pull request
Jul 5, 2026
…#24755) Absorb get_slot_by_id logic into get_available_slot so slot selection is handled by a single function call. When a specific slot id is requested, the LCP similarity check still runs to enable proper prompt cache updates. Assisted-by: pi:llama.cpp/Qwen3.6-27B
zommiommy
pushed a commit
to zommiommy/llama.cpp
that referenced
this pull request
Aug 18, 2026
…#24755) Absorb get_slot_by_id logic into get_available_slot so slot selection is handled by a single function call. When a specific slot id is requested, the LCP similarity check still runs to enable proper prompt cache updates. Assisted-by: pi:llama.cpp/Qwen3.6-27B
zbrad
pushed a commit
to zbrad/llama.cpp
that referenced
this pull request
Sep 10, 2026
…#24755) Absorb get_slot_by_id logic into get_available_slot so slot selection is handled by a single function call. When a specific slot id is requested, the LCP similarity check still runs to enable proper prompt cache updates. Assisted-by: pi:llama.cpp/Qwen3.6-27B
marcospaulo
added a commit
to torad-labs/llama.cpp
that referenced
this pull request
Sep 26, 2026
…hen empty, and is left alone while busy Since ggml-org#24755 a request's id_slot goes through get_available_slot, and the cache update ran on the named slot as if the similarity loop had picked it: - An idle slot saved and cleared (--cache-idle-slots, the default, with a unified KV cache) holds no tokens, so f_keep was 0 / 0. NaN compares false against 0.5, the prompt cache was never consulted, and a conversation returning to its slot was processed again from its first token: 74.1 s for a 178K-token conversation on an RTX 5090 (8 x 786,432 unified). An empty named slot now always goes through the prompt cache, also with --slot-prompt-similarity 0. A request that names no slot takes the similarity or LRU path, which already did. - The similarity loop skips a named slot that is processing, so f_keep was 0 and the cache update ran on it: the running request's state was saved, and the best cached match for the new prompt was restored into its sequence mid-generation. A busy named slot is now returned as is, and the caller defers the task as it did before ggml-org#24755. test_kv_keep_only_active: a named slot that was saved and cleared restores from cache-ram, and a stream on slot 0 while another request asks for slot 0 generates the same text as alone.
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.
Overview
fix #24746
Absorb
get_slot_by_idlogic intoget_available_slotso slot selection is handled by a single function call. When a specific slot id is requested, the LCP similarity check still runs to enable proper prompt cache updates.Requirements