mpmc_result__1.12.6.md

Benchmark: Multi-producer multi-consumer (MPMC) performance of AtomicChannels.jl

Date: 2026-05-06 09:38:57
AtomicChannels.jl Version 1.0.2
Julia Version 1.12.6
Commit 15346901f00 (2026-04-09 19:20 UTC)
Build Info:
Official https://julialang.org release
Platform Info:
OS: Linux (x86_64-linux-gnu)
CPU: 128 × AMD Ryzen Threadripper PRO 7985WX 64-Cores
WORD_SIZE: 64
LLVM: libLLVM-18.1.7 (ORCJIT, znver4)
GC: Built with stock GC

Case 1: Low capacity to encourage contention

This benchmark evaluates the performance of data operations (put and take), without task switching (when threads>1).

threads	workers	capacity	items	Channel/ms	AtomicChannel/ms	speedup
64	64	4	100000	2779.6	82.24	33.8x
32	32	4	100000	1441.26	71.42	20.18x
16	16	4	100000	772.32	56.61	13.643x
8	8	1	100000	465.8	39.79	11.707x
4	4	1	100000	190.83	25.54	7.47x
2	2	1	100000	121.67	24.58	4.95x
1	2	1	100000	121.58	115.59	1.052x

Case 2: Higher capacity with minimal contention

This benchmark mimics a lowest contention scenario where the contention is only from task switching, not data operations.

threads	workers	capacity	items	Channel/ms	AtomicChannel/ms	speedup
64	64	256	100000	69.6	23.02	3.023x
32	32	256	100000	39.93	24.44	1.633x
16	16	256	100000	37.29	24.84	1.501x
8	8	256	100000	26.3	30.24	0.87x
4	4	256	100000	23.38	19.17	1.22x
2	2	256	100000	17.72	14.92	1.188x
1	2	256	100000	13.99	10.88	1.286x

Case 3: Varying worker (task) counts to mimic different levels of concurrency

This benchmark mimics varying concurrency levels on both data operations and task switching.

threads	workers	capacity	items	Channel/ms	AtomicChannel/ms	speedup
32	512	256	100000	620.24	26.56	23.351x
32	256	256	100000	732.27	26.49	27.638x
32	128	256	100000	696.89	27.65	25.201x
32	64	256	100000	678.93	22.85	29.715x
32	32	256	100000	40.13	18.56	2.162x
32	16	256	100000	31.29	22.38	1.398x
32	8	256	100000	26.36	18.23	1.446x