util(queue): explore different in-memory queue implementations

[csegarragonz]

No description provided.

[lgarithm]

...
[11:29:31] [41971] [E] Different message types (got: 4, expected: 8)
[11:29:31] [41970] [E] Different message types (got: 4, expected: 6)
[11:29:31] [41973] [E] Different message types (got: 4, expected: 8)
[11:29:31] [41972] [E] Different message types (got: 4, expected: 8)
...

[csegarragonz]

@lgarithm my bad.

could you try now with: 7a36d4a

[lgarithm]

now it improves both small and large message:

BGN ========================================bench_allreduce========================================
bench_all_reduce(np=4) took 0.0329s, total workload: 3.662MiB, rate: 0.109GiB/s
bench_all_reduce(np=4) took 0.0299s, total workload: 3.662MiB, rate: 0.119GiB/s
bench_all_reduce(np=4) took 0.0296s, total workload: 3.662MiB, rate: 0.121GiB/s
bench_all_reduce(np=4) took 0.0295s, total workload: 3.662MiB, rate: 0.121GiB/s
bench_all_reduce(np=4) took 0.0297s, total workload: 3.662MiB, rate: 0.121GiB/s
bench_all_reduce(np=4) took 0.0300s, total workload: 3.662MiB, rate: 0.119GiB/s
bench_all_reduce(np=4) took 0.0298s, total workload: 3.662MiB, rate: 0.120GiB/s
bench_all_reduce(np=4) took 0.0300s, total workload: 3.662MiB, rate: 0.119GiB/s
bench_all_reduce(np=4) took 0.0299s, total workload: 3.662MiB, rate: 0.119GiB/s
bench_all_reduce(np=4) took 0.0298s, total workload: 3.662MiB, rate: 0.120GiB/s
bench_all_reduce(np=4) took 0.2324s, total workload: 1.144GiB, rate: 4.922GiB/s
bench_all_reduce(np=4) took 0.2028s, total workload: 1.144GiB, rate: 5.639GiB/s
bench_all_reduce(np=4) took 0.2174s, total workload: 1.144GiB, rate: 5.259GiB/s
bench_all_reduce(np=4) took 0.2103s, total workload: 1.144GiB, rate: 5.437GiB/s
bench_all_reduce(np=4) took 0.2104s, total workload: 1.144GiB, rate: 5.434GiB/s
bench_all_reduce(np=4) took 0.2073s, total workload: 1.144GiB, rate: 5.518GiB/s
bench_all_reduce(np=4) took 0.2059s, total workload: 1.144GiB, rate: 5.556GiB/s
bench_all_reduce(np=4) took 0.2106s, total workload: 1.144GiB, rate: 5.430GiB/s
bench_all_reduce(np=4) took 0.2095s, total workload: 1.144GiB, rate: 5.459GiB/s
bench_all_reduce(np=4) took 0.2103s, total workload: 1.144GiB, rate: 5.439GiB/s
END ========================================bench_allreduce========================================