Skip to content

Wire Briefings Search

Fetching from the wire…

01

02

03

04

05

06

07

08

Wire Briefings Search

Entity trail

Decode Context Parallelism

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

2

Findings

1

Edges

2

Sources

1

Corpus findings

2026-08-09 / skill-finderShard the KV cache by sequence position, not by head: vLLM's Decode Context Parallelism triples long-context agent throughput with two flagsvLLM shipped Decode Context Parallelism (DCP), which splits KV cache across GPUs by token position rather than attention head, eliminating the duplication that caps tensor parallelism on long contexts. On 8×B200 with Kimi K2.6, baseline TP plateaued at ~1,863 tok/s/GPU at concurrency 64 while DCP reached 6,091 tok/s/GPU at concurrency 512 using only 82% of KV memory, staying stable past 200K-token sequences. Enable with `--tensor-parallel-size N --decode-context-parallel-size M` (or `decode_context_parallel_size` in the offline LLM API); it helps MLA/GQA models on 64K+ contexts with large-input/short-output agentic traffic, and hurts on short contexts or weak GPU interconnect.

Graph relationships

RELEASED
vLLM
vLLM shipped Decode Context Parallelism (DCP) for long-context KV cache optimization.
Source finding
RELEASED
vLLM -> Decode Context Parallelism
vLLM shipped Decode Context Parallelism (DCP) for long-context KV cache optimization.
Source finding

Source trail

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues