<!-- llms-explorer concept facts · https://llms-explorer.com/tree/apple-uc-send-into-a-queue-pair-with-no-posted-r/ · pack 2026-10-05 · ~363 tokens -->

# Apple UC send into a queue pair with no posted receive kills that direction

> Already covered in full; nothing absent or contradicting was found.

Parent: [Mac local LLMs: Clusters, RDMA, exo and ds4](https://llms-explorer.com/tree/mac-local-llms-clusters-rdma-exo-ds4/) · 1 facets · 7 facts · page: https://llms-explorer.com/tree/apple-uc-send-into-a-queue-pair-with-no-posted-r/

## Facts

- Already covered in full; nothing absent or contradicting was found. — source: `asserted`
- See uc-queue-pair-one-direction-drop-warm-up-check.md (the ds4 posted-receive barrier and `DS4_TP_FRAME_RDMA_POSTED`). — source: `asserted`
- No further history found in the cached MLX, exo or JACCL pages. — source: `asserted`
- None beyond the existing dossier. — source: `asserted`
- None. — source: `asserted`
- Whether JACCL has any equivalent of the posted-receive barrier is already recorded as an open claim in the existing dossier. — source: `asserted`
- ALREADY COVERED: ~/.research/expert/running-llm-models-locally-on-mac/dossiers/uc-queue-pair-one-direction-drop-warm-up-check.md holds the claim that a UC send into a queue pair with no posted receive kills that direction permanently, the ds4 receives-posted barrier, the 16384-byte warm-up and the retry bounds, and nothing in the cached MLX PR 3899, 3900, 4443, issue 3830, 4192, 4319 or exo 1390 pages adds to or contradicts it. — [source](https://raw.githubusercontent.com/antirez/ds4/main/ds4_tp.c)
