Linux Benchmarks

Generated 2026-10-02 from corosio e2de06fda069 — benchmark run. This page is fully generated; do not edit by hand (see the Benchmarks landing page).

Summary

Within noise means the relative difference is within twice the combined run-to-run noise (the root-sum-square of each side’s CV). Differences that small are indistinguishable from measurement jitter. See Methodology for how these figures are computed.

io_uring

44 faster · 25 within noise · 44 slower of 113 benchmarks — median -0.7% vs Boost.Asio (callbacks) on the same reactor.

Summary (io_uring) — Linux ← slower · within noise · faster → per benchmark, vs Boost.Asio (callbacks) local_socket_latency 3 benchmarks slower than asio (callbacks) 3 1 benchmark within noise than asio (callbacks) 8 benchmarks faster than asio (callbacks) 8 +20.3% socket_latency 2 benchmarks slower than asio (callbacks) 2 2 benchmarks within noise than asio (callbacks) 2 8 benchmarks faster than asio (callbacks) 8 +10.7% socket_throughput 3 benchmarks slower than asio (callbacks) 3 8 benchmarks within noise than asio (callbacks) 8 16 benchmarks faster than asio (callbacks) 16 +5.1% http_server 4 benchmarks slower than asio (callbacks) 4 2 benchmarks within noise than asio (callbacks) 2 5 benchmarks faster than asio (callbacks) 5 +1.8% local_socket_throughput 7 benchmarks slower than asio (callbacks) 7 10 benchmarks within noise than asio (callbacks) 10 7 benchmarks faster than asio (callbacks) 7 -4.1% fan_out 16 benchmarks slower than asio (callbacks) 16 2 benchmarks within noise than asio (callbacks) 2 -10.1% accept_churn 9 benchmarks slower than asio (callbacks) 9 -26.1%

epoll

33 faster · 32 within noise · 48 slower of 113 benchmarks — median -2.4% vs Boost.Asio (callbacks) on the same reactor.

Summary (epoll) — Linux ← slower · within noise · faster → per benchmark, vs Boost.Asio (callbacks) local_socket_latency 3 benchmarks slower than asio (callbacks) 3 3 benchmarks within noise than asio (callbacks) 3 6 benchmarks faster than asio (callbacks) 6 +11.1% socket_latency 1 benchmark slower than asio (callbacks) 3 benchmarks within noise than asio (callbacks) 3 8 benchmarks faster than asio (callbacks) 8 +5.8% socket_throughput 3 benchmarks slower than asio (callbacks) 3 11 benchmarks within noise than asio (callbacks) 11 13 benchmarks faster than asio (callbacks) 13 +2.4% local_socket_throughput 4 benchmarks slower than asio (callbacks) 4 14 benchmarks within noise than asio (callbacks) 14 6 benchmarks faster than asio (callbacks) 6 -0.8% http_server 10 benchmarks slower than asio (callbacks) 10 1 benchmark within noise than asio (callbacks) -6.0% fan_out 18 benchmarks slower than asio (callbacks) 18 -8.5% accept_churn 9 benchmarks slower than asio (callbacks) 9 -9.3%

Test Environment

CPU

AMD EPYC 7302P 16-Core Processor

Cores

4

RAM (GB)

31

OS

Linux 7.0.0-34-generic

Kernel/build

#34-Ubuntu SMP PREEMPT_DYNAMIC Wed Sep 2 14:29:37 UTC 2026

Compiler

c++ (Ubuntu 15.2.0-16ubuntu1) 15.2.0

CMake

cmake version 4.2.3

liburing

2.14-1

Boost commit

39fa1e9a3491bd099b570935b3f3422065f91b03

Asio commit

a7dc25b4cb6c49a6946d86ea20664f1027203225

Asio reactor

epoll and io_uring (matched per configuration)

Capy commit

a372a6b054261f29497ac0d19c2a9533f9eaad40

Corosio commit

e2de06fda069607f34bc957cbefcbd9ba12fddca

Corosio branch

pr/benchmark-report

Date (UTC)

2026-10-02

Iterations

7

Duration per benchmark (s)

2.0

Results

accept_churn

Rate of setting up and tearing down short-lived TCP connections: connect, accept, close.

io_uring

accept_churn (io_uring) — Linux ↑ faster · dotted line = asio (callbacks) · backend = io_uring corosio asio (coroutines) -60% -40% -20% +20% 0% burst/10 — N connects are fired at once, then all N are accepted before closing; N is the burst size. burst/10 — N connects are fired at once, then all N are accepted before closing; N is the burst size. burst/10 burst/100 — N connects are fired at once, then all N are accepted before closing; N is the burst size. burst/100 — N connects are fired at once, then all N are accepted before closing; N is the burst size. burst/100 burst_lockless/10 — Same as burst, with the context in single-threaded lockless mode; N is the burst size. burst_lockless/10 — Same as burst, with the context in single-threaded lockless mode; N is the burst size. burst_lockless/10 burst_lockless/100 — Same as burst, with the context in single-threaded lockless mode; N is the burst size. burst_lockless/100 — Same as burst, with the context in single-threaded lockless mode; N is the burst size. +0% burst_lockless/100 concurrent/1 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. concurrent/1 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. -3% concurrent/1 concurrent/16 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. -35% concurrent/16 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. concurrent/16 concurrent/4 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. concurrent/4 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. concurrent/4 sequential — Single connect/accept/close loop with a one-way one-byte transfer (client writes, server reads) per connection, one connection at a time. sequential — Single connect/accept/close loop with a one-way one-byte transfer (client writes, server reads) per connection, one connection at a time. sequential sequential_lockless — Same as sequential, with the context in single-threaded lockless mode. -22% sequential_lockless — Same as sequential, with the context in single-threaded lockless mode. sequential_lockless
Detailed results
Benchmark Implementation Median CV vs asio

burst/10

corosio (io_uring)

1.70K ops/s

1.19%

-29.3%

burst/10

asio (coroutines, io_uring)

2.38K ops/s

0.97%

-1.0%

burst/10

asio (callbacks, io_uring)

2.40K ops/s

1.23%

baseline (io_uring)

burst/100

corosio (io_uring)

173.5 ops/s

1.03%

-29.2%

burst/100

asio (coroutines, io_uring)

245.5 ops/s

0.95%

+0.2%

burst/100

asio (callbacks, io_uring)

244.9 ops/s

0.78%

baseline (io_uring)

burst_lockless/10

corosio (io_uring)

1.81K ops/s

1.38%

-25.0%

burst_lockless/10

asio (coroutines, io_uring)

2.39K ops/s

1.12%

-1.0%

burst_lockless/10

asio (callbacks, io_uring)

2.42K ops/s

1.20%

baseline (io_uring)

burst_lockless/100

corosio (io_uring)

185.0 ops/s

1.10%

-25.1%

burst_lockless/100

asio (coroutines, io_uring)

248.0 ops/s

0.88%

+0.4%

burst_lockless/100

asio (callbacks, io_uring)

246.9 ops/s

0.70%

baseline (io_uring)

concurrent/1

corosio (io_uring)

12.66K ops/s

1.30%

-26.1%

concurrent/1

asio (coroutines, io_uring)

16.65K ops/s

0.90%

-2.8%

concurrent/1

asio (callbacks, io_uring)

17.13K ops/s

0.93%

baseline (io_uring)

concurrent/16

corosio (io_uring)

12.44K ops/s

1.00%

-35.3%

concurrent/16

asio (coroutines, io_uring)

18.96K ops/s

0.89%

-1.4%

concurrent/16

asio (callbacks, io_uring)

19.22K ops/s

0.96%

baseline (io_uring)

concurrent/4

corosio (io_uring)

12.54K ops/s

1.01%

-31.8%

concurrent/4

asio (coroutines, io_uring)

18.04K ops/s

0.95%

-1.9%

concurrent/4

asio (callbacks, io_uring)

18.39K ops/s

0.93%

baseline (io_uring)

sequential

corosio (io_uring)

12.68K ops/s

1.32%

-25.7%

sequential

asio (coroutines, io_uring)

16.73K ops/s

0.55%

-2.0%

sequential

asio (callbacks, io_uring)

17.07K ops/s

0.88%

baseline (io_uring)

sequential_lockless

corosio (io_uring)

13.52K ops/s

1.24%

-21.7%

sequential_lockless

asio (coroutines, io_uring)

16.86K ops/s

0.91%

-2.4%

sequential_lockless

asio (callbacks, io_uring)

17.27K ops/s

0.87%

baseline (io_uring)

epoll

accept_churn (epoll) — Linux ↑ faster · dotted line = asio (callbacks) · backend = epoll corosio asio (coroutines) -15% -10% -5% +5% 0% burst/10 — N connects are fired at once, then all N are accepted before closing; N is the burst size. burst/10 — N connects are fired at once, then all N are accepted before closing; N is the burst size. -4% burst/10 burst/100 — N connects are fired at once, then all N are accepted before closing; N is the burst size. -13% burst/100 — N connects are fired at once, then all N are accepted before closing; N is the burst size. burst/100 burst_lockless/10 — Same as burst, with the context in single-threaded lockless mode; N is the burst size. burst_lockless/10 — Same as burst, with the context in single-threaded lockless mode; N is the burst size. burst_lockless/10 burst_lockless/100 — Same as burst, with the context in single-threaded lockless mode; N is the burst size. burst_lockless/100 — Same as burst, with the context in single-threaded lockless mode; N is the burst size. burst_lockless/100 concurrent/1 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. concurrent/1 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. concurrent/1 concurrent/16 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. concurrent/16 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. -2% concurrent/16 concurrent/4 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. concurrent/4 — N independent accept loops run on separate listeners at once, each doing a one-way one-byte transfer (client writes, server reads) per connection; N is the number of loops. concurrent/4 sequential — Single connect/accept/close loop with a one-way one-byte transfer (client writes, server reads) per connection, one connection at a time. -4% sequential — Single connect/accept/close loop with a one-way one-byte transfer (client writes, server reads) per connection, one connection at a time. sequential sequential_lockless — Same as sequential, with the context in single-threaded lockless mode. sequential_lockless — Same as sequential, with the context in single-threaded lockless mode. sequential_lockless
Detailed results
Benchmark Implementation Median CV vs asio

burst/10

corosio (epoll)

1.87K ops/s

0.71%

-12.5%

burst/10

asio (coroutines, epoll)

2.05K ops/s

0.26%

-4.2%

burst/10

asio (callbacks, epoll)

2.14K ops/s

0.84%

baseline (epoll)

burst/100

corosio (epoll)

188.4 ops/s

0.85%

-12.8%

burst/100

asio (coroutines, epoll)

208.7 ops/s

0.26%

-3.3%

burst/100

asio (callbacks, epoll)

215.9 ops/s

0.24%

baseline (epoll)

burst_lockless/10

corosio (epoll)

1.88K ops/s

0.86%

-12.3%

burst_lockless/10

asio (coroutines, epoll)

2.06K ops/s

0.91%

-4.1%

burst_lockless/10

asio (callbacks, epoll)

2.14K ops/s

0.24%

baseline (epoll)

burst_lockless/100

corosio (epoll)

189.9 ops/s

0.78%

-12.3%

burst_lockless/100

asio (coroutines, epoll)

209.8 ops/s

0.89%

-3.1%

burst_lockless/100

asio (callbacks, epoll)

216.5 ops/s

0.49%

baseline (epoll)

concurrent/1

corosio (epoll)

14.90K ops/s

0.38%

-4.7%

concurrent/1

asio (coroutines, epoll)

15.19K ops/s

0.35%

-2.9%

concurrent/1

asio (callbacks, epoll)

15.63K ops/s

1.13%

baseline (epoll)

concurrent/16

corosio (epoll)

14.91K ops/s

0.45%

-9.3%

concurrent/16

asio (coroutines, epoll)

16.12K ops/s

0.28%

-1.9%

concurrent/16

asio (callbacks, epoll)

16.43K ops/s

0.97%

baseline (epoll)

concurrent/4

corosio (epoll)

14.91K ops/s

0.40%

-7.0%

concurrent/4

asio (coroutines, epoll)

15.66K ops/s

0.40%

-2.3%

concurrent/4

asio (callbacks, epoll)

16.03K ops/s

0.94%

baseline (epoll)

sequential

corosio (epoll)

14.93K ops/s

0.48%

-3.7%

sequential

asio (coroutines, epoll)

15.20K ops/s

0.98%

-2.0%

sequential

asio (callbacks, epoll)

15.51K ops/s

0.58%

baseline (epoll)

sequential_lockless

corosio (epoll)

14.99K ops/s

0.51%

-4.7%

sequential_lockless

asio (coroutines, epoll)

15.34K ops/s

0.86%

-2.4%

sequential_lockless

asio (callbacks, epoll)

15.73K ops/s

0.83%

baseline (epoll)

fan_out

Fan-out/fan-in coroutine coordination: a parent starts concurrent sub-requests against echo servers and awaits their completion via a shared latch.

io_uring

fan_out (io_uring) — Linux, part 1 ↑ faster · dotted line = asio (callbacks) · backend = io_uring corosio asio (coroutines) -20% -15% -10% -5% +5% 0% concurrent_parents/1 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -13% concurrent_parents/1 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -4% concurrent_parents/1 concurrent_parents/16 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -13% concurrent_parents/16 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -3% concurrent_parents/16 concurrent_parents/4 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -13% concurrent_parents/4 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -3% concurrent_parents/4 concurrent_parents_lockless/1 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -8% concurrent_parents_lockless/1 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -4% concurrent_parents_lockless/1 concurrent_parents_lockless/16 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -9% concurrent_parents_lockless/16 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -3% concurrent_parents_lockless/16 concurrent_parents_lockless/4 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -9% concurrent_parents_lockless/4 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -3% concurrent_parents_lockless/4
Detailed results
Benchmark Implementation Median CV vs asio

concurrent_parents/1

corosio (io_uring)

3.43K ops/s

0.77%

-12.6%

concurrent_parents/1

asio (coroutines, io_uring)

3.78K ops/s

1.13%

-3.7%

concurrent_parents/1

asio (callbacks, io_uring)

3.92K ops/s

0.60%

baseline (io_uring)

concurrent_parents/16

corosio (io_uring)

3.32K ops/s

0.47%

-13.3%

concurrent_parents/16

asio (coroutines, io_uring)

3.71K ops/s

1.26%

-3.0%

concurrent_parents/16

asio (callbacks, io_uring)

3.82K ops/s

0.58%

baseline (io_uring)

concurrent_parents/4

corosio (io_uring)

3.34K ops/s

0.49%

-13.3%

concurrent_parents/4

asio (coroutines, io_uring)

3.73K ops/s

1.16%

-3.1%

concurrent_parents/4

asio (callbacks, io_uring)

3.85K ops/s

0.26%

baseline (io_uring)

concurrent_parents_lockless/1

corosio (io_uring)

3.65K ops/s

0.26%

-8.1%

concurrent_parents_lockless/1

asio (coroutines, io_uring)

3.83K ops/s

1.27%

-3.5%

concurrent_parents_lockless/1

asio (callbacks, io_uring)

3.97K ops/s

0.92%

baseline (io_uring)

concurrent_parents_lockless/16

corosio (io_uring)

3.53K ops/s

0.16%

-9.0%

concurrent_parents_lockless/16

asio (coroutines, io_uring)

3.75K ops/s

0.85%

-3.3%

concurrent_parents_lockless/16

asio (callbacks, io_uring)

3.88K ops/s

0.88%

baseline (io_uring)

concurrent_parents_lockless/4

corosio (io_uring)

3.55K ops/s

0.44%

-8.8%

concurrent_parents_lockless/4

asio (coroutines, io_uring)

3.77K ops/s

0.84%

-3.3%

concurrent_parents_lockless/4

asio (callbacks, io_uring)

3.89K ops/s

0.84%

baseline (io_uring)
fan_out (io_uring) — Linux, part 2 ↑ faster · dotted line = asio (callbacks) · backend = io_uring corosio asio (coroutines) -20% -10% +10% 0% fork_join/1 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/1 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. -12% fork_join/1 fork_join/16 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/16 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/16 fork_join/4 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/4 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/4 fork_join/64 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/64 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/64 fork_join_lockless/1 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. +2% fork_join_lockless/1 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/1 fork_join_lockless/16 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/16 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/16 fork_join_lockless/4 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/4 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/4 fork_join_lockless/64 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/64 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. -2% fork_join_lockless/64 nested/16 — Two-level fan-out: the parent starts N groups of 4 sub-requests each, each group awaited via its own latch; N is the number of groups. -16% nested/16 — Two-level fan-out: the parent starts N groups of 4 sub-requests each, each group awaited via its own latch; N is the number of groups. nested/16 nested/4 — Two-level fan-out: the parent starts N groups of 4 sub-requests each, each group awaited via its own latch; N is the number of groups. nested/4 — Two-level fan-out: the parent starts N groups of 4 sub-requests each, each group awaited via its own latch; N is the number of groups. nested/4 nested_lockless/16 — Same as nested, with the context in single-threaded lockless mode; N is the number of groups. nested_lockless/16 — Same as nested, with the context in single-threaded lockless mode; N is the number of groups. nested_lockless/16 nested_lockless/4 — Same as nested, with the context in single-threaded lockless mode; N is the number of groups. nested_lockless/4 — Same as nested, with the context in single-threaded lockless mode; N is the number of groups. nested_lockless/4
Detailed results
Benchmark Implementation Median CV vs asio

fork_join/1

corosio (io_uring)

53.35K ops/s

1.17%

-0.9%

fork_join/1

asio (coroutines, io_uring)

47.61K ops/s

1.20%

-11.6%

fork_join/1

asio (callbacks, io_uring)

53.83K ops/s

1.02%

baseline (io_uring)

fork_join/16

corosio (io_uring)

3.44K ops/s

0.97%

-12.3%

fork_join/16

asio (coroutines, io_uring)

3.79K ops/s

0.84%

-3.4%

fork_join/16

asio (callbacks, io_uring)

3.92K ops/s

1.20%

baseline (io_uring)

fork_join/4

corosio (io_uring)

13.74K ops/s

1.19%

-9.7%

fork_join/4

asio (coroutines, io_uring)

14.39K ops/s

0.91%

-5.3%

fork_join/4

asio (callbacks, io_uring)

15.21K ops/s

1.19%

baseline (io_uring)

fork_join/64

corosio (io_uring)

813.2 ops/s

1.01%

-15.2%

fork_join/64

asio (coroutines, io_uring)

933.8 ops/s

0.85%

-2.6%

fork_join/64

asio (callbacks, io_uring)

958.5 ops/s

0.88%

baseline (io_uring)

fork_join_lockless/1

corosio (io_uring)

55.85K ops/s

0.69%

+1.7%

fork_join_lockless/1

asio (coroutines, io_uring)

48.68K ops/s

1.11%

-11.3%

fork_join_lockless/1

asio (callbacks, io_uring)

54.89K ops/s

1.12%

baseline (io_uring)

fork_join_lockless/16

corosio (io_uring)

3.66K ops/s

0.83%

-7.9%

fork_join_lockless/16

asio (coroutines, io_uring)

3.85K ops/s

0.98%

-3.1%

fork_join_lockless/16

asio (callbacks, io_uring)

3.97K ops/s

0.92%

baseline (io_uring)

fork_join_lockless/4

corosio (io_uring)

14.55K ops/s

0.19%

-5.7%

fork_join_lockless/4

asio (coroutines, io_uring)

14.57K ops/s

1.27%

-5.6%

fork_join_lockless/4

asio (callbacks, io_uring)

15.43K ops/s

1.08%

baseline (io_uring)

fork_join_lockless/64

corosio (io_uring)

867.2 ops/s

0.73%

-10.5%

fork_join_lockless/64

asio (coroutines, io_uring)

946.7 ops/s

1.21%

-2.3%

fork_join_lockless/64

asio (callbacks, io_uring)

969.4 ops/s

0.83%

baseline (io_uring)

nested/16

corosio (io_uring)

804.1 ops/s

1.20%

-16.1%

nested/16

asio (coroutines, io_uring)

924.9 ops/s

0.94%

-3.5%

nested/16

asio (callbacks, io_uring)

958.0 ops/s

0.59%

baseline (io_uring)

nested/4

corosio (io_uring)

3.40K ops/s

1.16%

-13.3%

nested/4

asio (coroutines, io_uring)

3.77K ops/s

1.08%

-3.7%

nested/4

asio (callbacks, io_uring)

3.92K ops/s

0.90%

baseline (io_uring)

nested_lockless/16

corosio (io_uring)

856.8 ops/s

0.83%

-11.7%

nested_lockless/16

asio (coroutines, io_uring)

935.0 ops/s

1.11%

-3.7%

nested_lockless/16

asio (callbacks, io_uring)

970.5 ops/s

0.62%

baseline (io_uring)

nested_lockless/4

corosio (io_uring)

3.63K ops/s

1.08%

-8.7%

nested_lockless/4

asio (coroutines, io_uring)

3.81K ops/s

1.04%

-4.1%

nested_lockless/4

asio (callbacks, io_uring)

3.97K ops/s

0.65%

baseline (io_uring)

epoll

fan_out (epoll) — Linux, part 1 ↑ faster · dotted line = asio (callbacks) · backend = epoll corosio asio (coroutines) -15% -10% -5% +5% 0% concurrent_parents/1 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -8% concurrent_parents/1 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -5% concurrent_parents/1 concurrent_parents/16 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -9% concurrent_parents/16 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -5% concurrent_parents/16 concurrent_parents/4 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -9% concurrent_parents/4 — N independent parents each fan out to 16 sub-requests at once; N is the number of parents. -5% concurrent_parents/4 concurrent_parents_lockless/1 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -8% concurrent_parents_lockless/1 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -4% concurrent_parents_lockless/1 concurrent_parents_lockless/16 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -9% concurrent_parents_lockless/16 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -5% concurrent_parents_lockless/16 concurrent_parents_lockless/4 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -9% concurrent_parents_lockless/4 — Same as concurrent_parents, with the context in single-threaded lockless mode; N is the number of parents. -5% concurrent_parents_lockless/4
Detailed results
Benchmark Implementation Median CV vs asio

concurrent_parents/1

corosio (epoll)

3.69K ops/s

0.10%

-8.3%

concurrent_parents/1

asio (coroutines, epoll)

3.83K ops/s

1.00%

-4.8%

concurrent_parents/1

asio (callbacks, epoll)

4.02K ops/s

0.45%

baseline (epoll)

concurrent_parents/16

corosio (epoll)

3.59K ops/s

0.10%

-8.8%

concurrent_parents/16

asio (coroutines, epoll)

3.74K ops/s

0.81%

-4.9%

concurrent_parents/16

asio (callbacks, epoll)

3.93K ops/s

0.77%

baseline (epoll)

concurrent_parents/4

corosio (epoll)

3.60K ops/s

0.12%

-8.6%

concurrent_parents/4

asio (coroutines, epoll)

3.74K ops/s

0.97%

-4.8%

concurrent_parents/4

asio (callbacks, epoll)

3.93K ops/s

0.83%

baseline (epoll)

concurrent_parents_lockless/1

corosio (epoll)

3.73K ops/s

0.12%

-8.2%

concurrent_parents_lockless/1

asio (coroutines, epoll)

3.89K ops/s

0.28%

-4.5%

concurrent_parents_lockless/1

asio (callbacks, epoll)

4.07K ops/s

0.85%

baseline (epoll)

concurrent_parents_lockless/16

corosio (epoll)

3.62K ops/s

0.81%

-9.2%

concurrent_parents_lockless/16

asio (coroutines, epoll)

3.80K ops/s

0.19%

-4.8%

concurrent_parents_lockless/16

asio (callbacks, epoll)

3.99K ops/s

0.80%

baseline (epoll)

concurrent_parents_lockless/4

corosio (epoll)

3.63K ops/s

0.59%

-8.8%

concurrent_parents_lockless/4

asio (coroutines, epoll)

3.80K ops/s

0.32%

-4.6%

concurrent_parents_lockless/4

asio (callbacks, epoll)

3.98K ops/s

0.87%

baseline (epoll)
fan_out (epoll) — Linux, part 2 ↑ faster · dotted line = asio (callbacks) · backend = epoll corosio asio (coroutines) -15% -10% -5% +5% 0% fork_join/1 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/1 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. -12% fork_join/1 fork_join/16 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/16 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/16 fork_join/4 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. -8% fork_join/4 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/4 fork_join/64 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/64 — One parent starts N sub-requests to echo servers and waits for all to finish before repeating; N is the fan-out width. fork_join/64 fork_join_lockless/1 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/1 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/1 fork_join_lockless/16 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/16 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/16 fork_join_lockless/4 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/4 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/4 fork_join_lockless/64 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. fork_join_lockless/64 — Same as fork_join, with the context in single-threaded lockless mode; N is the fan-out width. -4% fork_join_lockless/64 nested/16 — Two-level fan-out: the parent starts N groups of 4 sub-requests each, each group awaited via its own latch; N is the number of groups. nested/16 — Two-level fan-out: the parent starts N groups of 4 sub-requests each, each group awaited via its own latch; N is the number of groups. nested/16 nested/4 — Two-level fan-out: the parent starts N groups of 4 sub-requests each, each group awaited via its own latch; N is the number of groups. nested/4 — Two-level fan-out: the parent starts N groups of 4 sub-requests each, each group awaited via its own latch; N is the number of groups. nested/4 nested_lockless/16 — Same as nested, with the context in single-threaded lockless mode; N is the number of groups. -9% nested_lockless/16 — Same as nested, with the context in single-threaded lockless mode; N is the number of groups. nested_lockless/16 nested_lockless/4 — Same as nested, with the context in single-threaded lockless mode; N is the number of groups. nested_lockless/4 — Same as nested, with the context in single-threaded lockless mode; N is the number of groups. nested_lockless/4
Detailed results
Benchmark Implementation Median CV vs asio

fork_join/1

corosio (epoll)

54.81K ops/s

0.95%

-8.6%

fork_join/1

asio (coroutines, epoll)

52.61K ops/s

0.98%

-12.3%

fork_join/1

asio (callbacks, epoll)

59.96K ops/s

1.01%

baseline (epoll)

fork_join/16

corosio (epoll)

3.69K ops/s

0.21%

-8.2%

fork_join/16

asio (coroutines, epoll)

3.84K ops/s

0.96%

-4.5%

fork_join/16

asio (callbacks, epoll)

4.02K ops/s

0.99%

baseline (epoll)

fork_join/4

corosio (epoll)

14.65K ops/s

0.13%

-7.6%

fork_join/4

asio (coroutines, epoll)

14.84K ops/s

0.98%

-6.4%

fork_join/4

asio (callbacks, epoll)

15.85K ops/s

0.60%

baseline (epoll)

fork_join/64

corosio (epoll)

900.2 ops/s

0.12%

-8.0%

fork_join/64

asio (coroutines, epoll)

939.3 ops/s

0.87%

-4.0%

fork_join/64

asio (callbacks, epoll)

978.8 ops/s

0.92%

baseline (epoll)

fork_join_lockless/1

corosio (epoll)

55.61K ops/s

0.21%

-8.6%

fork_join_lockless/1

asio (coroutines, epoll)

53.42K ops/s

0.69%

-12.2%

fork_join_lockless/1

asio (callbacks, epoll)

60.85K ops/s

0.89%

baseline (epoll)

fork_join_lockless/16

corosio (epoll)

3.73K ops/s

0.82%

-8.5%

fork_join_lockless/16

asio (coroutines, epoll)

3.88K ops/s

0.26%

-4.7%

fork_join_lockless/16

asio (callbacks, epoll)

4.08K ops/s

0.90%

baseline (epoll)

fork_join_lockless/4

corosio (epoll)

14.84K ops/s

0.19%

-7.7%

fork_join_lockless/4

asio (coroutines, epoll)

15.05K ops/s

0.32%

-6.5%

fork_join_lockless/4

asio (callbacks, epoll)

16.09K ops/s

0.86%

baseline (epoll)

fork_join_lockless/64

corosio (epoll)

910.3 ops/s

0.88%

-8.2%

fork_join_lockless/64

asio (coroutines, epoll)

952.6 ops/s

0.18%

-3.9%

fork_join_lockless/64

asio (callbacks, epoll)

991.7 ops/s

0.77%

baseline (epoll)

nested/16

corosio (epoll)

893.1 ops/s

0.88%

-8.4%

nested/16

asio (coroutines, epoll)

929.2 ops/s

0.23%

-4.7%

nested/16

asio (callbacks, epoll)

975.4 ops/s

0.97%

baseline (epoll)

nested/4

corosio (epoll)

3.66K ops/s

0.91%

-8.6%

nested/4

asio (coroutines, epoll)

3.78K ops/s

0.27%

-5.7%

nested/4

asio (callbacks, epoll)

4.01K ops/s

1.03%

baseline (epoll)

nested_lockless/16

corosio (epoll)

902.4 ops/s

0.10%

-9.0%

nested_lockless/16

asio (coroutines, epoll)

943.3 ops/s

0.86%

-4.9%

nested_lockless/16

asio (callbacks, epoll)

991.4 ops/s

0.87%

baseline (epoll)

nested_lockless/4

corosio (epoll)

3.71K ops/s

0.89%

-8.9%

nested_lockless/4

asio (coroutines, epoll)

3.84K ops/s

0.86%

-5.9%

nested_lockless/4

asio (callbacks, epoll)

4.08K ops/s

0.85%

baseline (epoll)

http_server

Request/response throughput of a minimal HTTP/1.1 server exchanging a fixed small request and canned response over persistent TCP loopback connections.

io_uring

http_server (io_uring) — Linux ↑ faster · dotted line = asio (callbacks) · backend = io_uring corosio asio (coroutines) -50% +50% +100% +150% 0% concurrent/1 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/1 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/1 concurrent/16 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/16 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/16 concurrent/32 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. -14% concurrent/32 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/32 concurrent/4 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/4 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/4 multithread/1 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/1 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/1 multithread/16 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/16 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. -7% multithread/16 multithread/2 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. +91% multithread/2 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. 0% multithread/2 multithread/4 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/4 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/4 multithread/8 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/8 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/8 single_conn — One client repeatedly sends a fixed small HTTP request to one server and reads the response. single_conn — One client repeatedly sends a fixed small HTTP request to one server and reads the response. single_conn single_conn_lockless — Same as single_conn, with the context in single-threaded lockless mode. single_conn_lockless — Same as single_conn, with the context in single-threaded lockless mode. single_conn_lockless
Detailed results
Benchmark Implementation Median CV vs asio

concurrent/1

corosio (io_uring)

53.45K ops/s

1.33%

+1.3%

concurrent/1

asio (coroutines, io_uring)

51.57K ops/s

1.06%

-2.3%

concurrent/1

asio (callbacks, io_uring)

52.77K ops/s

1.16%

baseline (io_uring)

concurrent/16

corosio (io_uring)

52.87K ops/s

0.99%

-14.0%

concurrent/16

asio (coroutines, io_uring)

60.33K ops/s

0.78%

-1.9%

concurrent/16

asio (callbacks, io_uring)

61.50K ops/s

0.94%

baseline (io_uring)

concurrent/32

corosio (io_uring)

52.55K ops/s

0.83%

-14.5%

concurrent/32

asio (coroutines, io_uring)

60.23K ops/s

1.05%

-2.0%

concurrent/32

asio (callbacks, io_uring)

61.44K ops/s

0.41%

baseline (io_uring)

concurrent/4

corosio (io_uring)

53.06K ops/s

1.22%

-10.9%

concurrent/4

asio (coroutines, io_uring)

58.30K ops/s

0.88%

-2.1%

concurrent/4

asio (callbacks, io_uring)

59.57K ops/s

0.92%

baseline (io_uring)

multithread/1

corosio (io_uring)

52.53K ops/s

0.74%

-14.4%

multithread/1

asio (coroutines, io_uring)

60.16K ops/s

0.78%

-2.0%

multithread/1

asio (callbacks, io_uring)

61.39K ops/s

0.28%

baseline (io_uring)

multithread/16

corosio (io_uring)

81.86K ops/s

13.26%

+90.3%

multithread/16

asio (coroutines, io_uring)

40.11K ops/s

1.71%

-6.7%

multithread/16

asio (callbacks, io_uring)

43.01K ops/s

1.66%

baseline (io_uring)

multithread/2

corosio (io_uring)

90.67K ops/s

0.46%

+91.3%

multithread/2

asio (coroutines, io_uring)

47.35K ops/s

8.15%

-0.1%

multithread/2

asio (callbacks, io_uring)

47.41K ops/s

10.18%

baseline (io_uring)

multithread/4

corosio (io_uring)

80.23K ops/s

0.39%

+72.7%

multithread/4

asio (coroutines, io_uring)

45.90K ops/s

0.35%

-1.2%

multithread/4

asio (callbacks, io_uring)

46.46K ops/s

0.27%

baseline (io_uring)

multithread/8

corosio (io_uring)

77.71K ops/s

0.53%

+78.9%

multithread/8

asio (coroutines, io_uring)

41.91K ops/s

0.95%

-3.5%

multithread/8

asio (callbacks, io_uring)

43.43K ops/s

0.80%

baseline (io_uring)

single_conn

corosio (io_uring)

53.68K ops/s

1.21%

+1.8%

single_conn

asio (coroutines, io_uring)

51.56K ops/s

0.98%

-2.3%

single_conn

asio (callbacks, io_uring)

52.75K ops/s

0.80%

baseline (io_uring)

single_conn_lockless

corosio (io_uring)

56.14K ops/s

1.05%

+4.5%

single_conn_lockless

asio (coroutines, io_uring)

52.55K ops/s

1.11%

-2.2%

single_conn_lockless

asio (callbacks, io_uring)

53.74K ops/s

1.15%

baseline (io_uring)

epoll

http_server (epoll) — Linux ↑ faster · dotted line = asio (callbacks) · backend = epoll corosio asio (coroutines) -10% -7.5% -5% -2.5% +2.5% 0% concurrent/1 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/1 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/1 concurrent/16 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/16 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/16 concurrent/32 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/32 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/32 concurrent/4 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. concurrent/4 — N client/server pairs run the single_conn request/response loop concurrently; N is the number of connections. -1% concurrent/4 multithread/1 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/1 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/1 multithread/16 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. -1% multithread/16 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/16 multithread/2 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. -8% multithread/2 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. -5% multithread/2 multithread/4 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/4 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/4 multithread/8 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/8 — 32 client/server pairs share one context serviced by N threads running the context; N is the thread count. multithread/8 single_conn — One client repeatedly sends a fixed small HTTP request to one server and reads the response. single_conn — One client repeatedly sends a fixed small HTTP request to one server and reads the response. single_conn single_conn_lockless — Same as single_conn, with the context in single-threaded lockless mode. single_conn_lockless — Same as single_conn, with the context in single-threaded lockless mode. single_conn_lockless
Detailed results
Benchmark Implementation Median CV vs asio

concurrent/1

corosio (epoll)

55.02K ops/s

0.21%

-6.0%

concurrent/1

asio (coroutines, epoll)

57.04K ops/s

1.14%

-2.6%

concurrent/1

asio (callbacks, epoll)

58.53K ops/s

0.13%

baseline (epoll)

concurrent/16

corosio (epoll)

57.82K ops/s

0.14%

-6.8%

concurrent/16

asio (coroutines, epoll)

60.14K ops/s

1.28%

-3.0%

concurrent/16

asio (callbacks, epoll)

62.02K ops/s

0.73%

baseline (epoll)

concurrent/32

corosio (epoll)

57.45K ops/s

0.13%

-6.6%

concurrent/32

asio (coroutines, epoll)

59.51K ops/s

1.16%

-3.2%

concurrent/32

asio (callbacks, epoll)

61.48K ops/s

0.83%

baseline (epoll)

concurrent/4

corosio (epoll)

57.08K ops/s

0.14%

-7.0%

concurrent/4

asio (coroutines, epoll)

60.48K ops/s

1.27%

-1.5%

concurrent/4

asio (callbacks, epoll)

61.37K ops/s

0.22%

baseline (epoll)

multithread/1

corosio (epoll)

57.35K ops/s

0.59%

-6.7%

multithread/1

asio (coroutines, epoll)

59.50K ops/s

1.08%

-3.2%

multithread/1

asio (callbacks, epoll)

61.45K ops/s

0.89%

baseline (epoll)

multithread/16

corosio (epoll)

161.9K ops/s

0.20%

-1.0%

multithread/16

asio (coroutines, epoll)

155.4K ops/s

0.31%

-4.9%

multithread/16

asio (callbacks, epoll)

163.4K ops/s

0.53%

baseline (epoll)

multithread/2

corosio (epoll)

86.93K ops/s

0.44%

-7.5%

multithread/2

asio (coroutines, epoll)

89.12K ops/s

0.51%

-5.2%

multithread/2

asio (callbacks, epoll)

94.00K ops/s

0.67%

baseline (epoll)

multithread/4

corosio (epoll)

160.1K ops/s

0.47%

-3.4%

multithread/4

asio (coroutines, epoll)

157.3K ops/s

0.61%

-5.1%

multithread/4

asio (callbacks, epoll)

165.8K ops/s

0.65%

baseline (epoll)

multithread/8

corosio (epoll)

161.1K ops/s

0.21%

-1.4%

multithread/8

asio (coroutines, epoll)

155.7K ops/s

0.29%

-4.7%

multithread/8

asio (callbacks, epoll)

163.3K ops/s

0.52%

baseline (epoll)

single_conn

corosio (epoll)

55.09K ops/s

1.04%

-5.9%

single_conn

asio (coroutines, epoll)

56.90K ops/s

0.92%

-2.8%

single_conn

asio (callbacks, epoll)

58.56K ops/s

0.85%

baseline (epoll)

single_conn_lockless

corosio (epoll)

55.80K ops/s

0.46%

-5.9%

single_conn_lockless

asio (coroutines, epoll)

57.70K ops/s

1.10%

-2.7%

single_conn_lockless

asio (callbacks, epoll)

59.30K ops/s

0.98%

baseline (epoll)

local_socket_latency

Round-trip latency of a single Unix domain stream socket write/read exchange across message sizes and concurrent pair counts.

io_uring

local_socket_latency (io_uring) — Linux ↑ faster (lower mean latency) · dotted line = asio (callbacks) · backend = io_uring corosio asio (coroutines) -20% +20% +40% 0% concurrent/1 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/1 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/1 concurrent/16 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/16 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/16 concurrent/4 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/4 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/4 concurrent_lockless/1 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/1 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/1 concurrent_lockless/16 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. -11% concurrent_lockless/16 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/16 concurrent_lockless/4 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/4 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/4 pingpong/1 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. +26% pingpong/1 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. pingpong/1 pingpong/1024 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. pingpong/1024 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. -8% pingpong/1024 pingpong/64 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. pingpong/64 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. pingpong/64 pingpong_lockless/1 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1 pingpong_lockless/1024 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1024 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. +2% pingpong_lockless/1024 pingpong_lockless/64 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/64 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/64
Detailed results
Benchmark Implementation Median CV vs asio

concurrent/1

corosio (io_uring)

6.56 µs

4.09%

+25.3%

concurrent/1

asio (coroutines, io_uring)

9.22 µs

2.67%

-5.0%

concurrent/1

asio (callbacks, io_uring)

8.78 µs

0.16%

baseline (io_uring)

concurrent/16

corosio (io_uring)

108.34 µs

1.05%

-6.5%

concurrent/16

asio (coroutines, io_uring)

105.53 µs

0.75%

-3.7%

concurrent/16

asio (callbacks, io_uring)

101.74 µs

0.75%

baseline (io_uring)

concurrent/4

corosio (io_uring)

26.94 µs

1.58%

+1.8%

concurrent/4

asio (coroutines, io_uring)

28.20 µs

1.47%

-2.8%

concurrent/4

asio (callbacks, io_uring)

27.43 µs

1.11%

baseline (io_uring)

concurrent_lockless/1

corosio (io_uring)

6.55 µs

4.18%

+23.7%

concurrent_lockless/1

asio (coroutines, io_uring)

8.77 µs

3.11%

-2.2%

concurrent_lockless/1

asio (callbacks, io_uring)

8.58 µs

3.48%

baseline (io_uring)

concurrent_lockless/16

corosio (io_uring)

108.13 µs

0.82%

-10.7%

concurrent_lockless/16

asio (coroutines, io_uring)

102.49 µs

1.04%

-5.0%

concurrent_lockless/16

asio (callbacks, io_uring)

97.65 µs

0.84%

baseline (io_uring)

concurrent_lockless/4

corosio (io_uring)

27.08 µs

0.88%

-2.6%

concurrent_lockless/4

asio (coroutines, io_uring)

27.52 µs

1.41%

-4.3%

concurrent_lockless/4

asio (callbacks, io_uring)

26.39 µs

0.74%

baseline (io_uring)

pingpong/1

corosio (io_uring)

6.53 µs

3.23%

+25.5%

pingpong/1

asio (coroutines, io_uring)

9.08 µs

2.60%

-3.6%

pingpong/1

asio (callbacks, io_uring)

8.76 µs

3.62%

baseline (io_uring)

pingpong/1024

corosio (io_uring)

7.43 µs

5.34%

+19.5%

pingpong/1024

asio (coroutines, io_uring)

9.99 µs

3.17%

-8.2%

pingpong/1024

asio (callbacks, io_uring)

9.23 µs

3.64%

baseline (io_uring)

pingpong/64

corosio (io_uring)

6.59 µs

4.50%

+24.9%

pingpong/64

asio (coroutines, io_uring)

9.18 µs

1.08%

-4.6%

pingpong/64

asio (callbacks, io_uring)

8.78 µs

3.19%

baseline (io_uring)

pingpong_lockless/1

corosio (io_uring)

6.53 µs

4.41%

+21.0%

pingpong_lockless/1

asio (coroutines, io_uring)

8.73 µs

3.80%

-5.6%

pingpong_lockless/1

asio (callbacks, io_uring)

8.27 µs

2.59%

baseline (io_uring)

pingpong_lockless/1024

corosio (io_uring)

7.45 µs

4.84%

+19.0%

pingpong_lockless/1024

asio (coroutines, io_uring)

9.04 µs

4.01%

+1.7%

pingpong_lockless/1024

asio (callbacks, io_uring)

9.20 µs

2.66%

baseline (io_uring)

pingpong_lockless/64

corosio (io_uring)

6.56 µs

3.76%

+21.0%

pingpong_lockless/64

asio (coroutines, io_uring)

8.77 µs

2.99%

-5.5%

pingpong_lockless/64

asio (callbacks, io_uring)

8.31 µs

0.28%

baseline (io_uring)

epoll

local_socket_latency (epoll) — Linux ↑ faster (lower mean latency) · dotted line = asio (callbacks) · backend = epoll corosio asio (coroutines) -20% -10% +10% +20% +30% 0% concurrent/1 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/1 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/1 concurrent/16 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/16 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/16 concurrent/4 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/4 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/4 concurrent_lockless/1 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/1 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/1 concurrent_lockless/16 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. -8% concurrent_lockless/16 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/16 concurrent_lockless/4 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/4 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. -8% concurrent_lockless/4 pingpong/1 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. pingpong/1 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. pingpong/1 pingpong/1024 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. +18% pingpong/1024 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. pingpong/1024 pingpong/64 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. pingpong/64 — One connected Unix domain socket pair ping-pongs a message client->server->client; N is the message size in bytes. pingpong/64 pingpong_lockless/1 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. -4% pingpong_lockless/1 pingpong_lockless/1024 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1024 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1024 pingpong_lockless/64 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/64 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/64
Detailed results
Benchmark Implementation Median CV vs asio

concurrent/1

corosio (epoll)

7.11 µs

3.77%

+13.4%

concurrent/1

asio (coroutines, epoll)

8.60 µs

3.20%

-4.6%

concurrent/1

asio (callbacks, epoll)

8.22 µs

3.11%

baseline (epoll)

concurrent/16

corosio (epoll)

110.13 µs

0.73%

-4.5%

concurrent/16

asio (coroutines, epoll)

112.84 µs

0.71%

-7.1%

concurrent/16

asio (callbacks, epoll)

105.39 µs

0.75%

baseline (epoll)

concurrent/4

corosio (epoll)

27.52 µs

1.65%

+1.0%

concurrent/4

asio (coroutines, epoll)

29.61 µs

1.67%

-6.5%

concurrent/4

asio (callbacks, epoll)

27.79 µs

1.31%

baseline (epoll)

concurrent_lockless/1

corosio (epoll)

7.00 µs

4.74%

+11.4%

concurrent_lockless/1

asio (coroutines, epoll)

8.29 µs

3.82%

-5.0%

concurrent_lockless/1

asio (callbacks, epoll)

7.90 µs

1.09%

baseline (epoll)

concurrent_lockless/16

corosio (epoll)

108.75 µs

0.34%

-7.5%

concurrent_lockless/16

asio (coroutines, epoll)

107.97 µs

0.77%

-6.7%

concurrent_lockless/16

asio (callbacks, epoll)

101.16 µs

1.07%

baseline (epoll)

concurrent_lockless/4

corosio (epoll)

27.60 µs

1.36%

-4.9%

concurrent_lockless/4

asio (coroutines, epoll)

28.47 µs

0.49%

-8.2%

concurrent_lockless/4

asio (callbacks, epoll)

26.31 µs

1.67%

baseline (epoll)

pingpong/1

corosio (epoll)

7.07 µs

3.51%

+13.7%

pingpong/1

asio (coroutines, epoll)

8.56 µs

3.08%

-4.5%

pingpong/1

asio (callbacks, epoll)

8.19 µs

3.62%

baseline (epoll)

pingpong/1024

corosio (epoll)

7.36 µs

4.39%

+17.7%

pingpong/1024

asio (coroutines, epoll)

9.50 µs

3.30%

-6.1%

pingpong/1024

asio (callbacks, epoll)

8.95 µs

3.55%

baseline (epoll)

pingpong/64

corosio (epoll)

7.11 µs

0.18%

+13.5%

pingpong/64

asio (coroutines, epoll)

8.61 µs

3.70%

-4.8%

pingpong/64

asio (callbacks, epoll)

8.22 µs

0.84%

baseline (epoll)

pingpong_lockless/1

corosio (epoll)

6.96 µs

5.13%

+11.7%

pingpong_lockless/1

asio (coroutines, epoll)

8.22 µs

3.88%

-4.3%

pingpong_lockless/1

asio (callbacks, epoll)

7.88 µs

4.43%

baseline (epoll)

pingpong_lockless/1024

corosio (epoll)

7.85 µs

2.88%

+10.9%

pingpong_lockless/1024

asio (coroutines, epoll)

9.19 µs

2.99%

-4.3%

pingpong_lockless/1024

asio (callbacks, epoll)

8.81 µs

2.58%

baseline (epoll)

pingpong_lockless/64

corosio (epoll)

7.51 µs

4.31%

+5.0%

pingpong_lockless/64

asio (coroutines, epoll)

8.28 µs

4.12%

-4.7%

pingpong_lockless/64

asio (callbacks, epoll)

7.90 µs

4.42%

baseline (epoll)

local_socket_throughput

Sustained byte throughput of a Unix domain stream socket pair under continuous streaming across chunk sizes and directions.

io_uring

local_socket_throughput (io_uring) — Linux, part 1 ↑ faster · dotted line = asio (callbacks) · backend = io_uring corosio asio (coroutines) -40% -20% +20% 0% bidirectional/1024 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/1024 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. -5% bidirectional/1024 bidirectional/1048576 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/1048576 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/1048576 bidirectional/16384 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/16384 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/16384 bidirectional/262144 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/262144 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/262144 bidirectional/4096 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/4096 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/4096 bidirectional/65536 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. -20% bidirectional/65536 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/65536 bidirectional_lockless/1024 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1024 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1024 bidirectional_lockless/1048576 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1048576 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1048576 bidirectional_lockless/16384 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/16384 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/16384 bidirectional_lockless/262144 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/262144 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/262144 bidirectional_lockless/4096 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. +12% bidirectional_lockless/4096 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. +3% bidirectional_lockless/4096 bidirectional_lockless/65536 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/65536 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/65536
Detailed results
Benchmark Implementation Median CV vs asio

bidirectional/1024

corosio (io_uring)

302.6 MB/s

1.03%

+10.3%

bidirectional/1024

asio (coroutines, io_uring)

260.4 MB/s

5.00%

-5.1%

bidirectional/1024

asio (callbacks, io_uring)

274.3 MB/s

3.34%

baseline (io_uring)

bidirectional/1048576

corosio (io_uring)

3.31 GB/s

1.57%

-4.8%

bidirectional/1048576

asio (coroutines, io_uring)

3.53 GB/s

2.27%

+1.5%

bidirectional/1048576

asio (callbacks, io_uring)

3.47 GB/s

2.38%

baseline (io_uring)

bidirectional/16384

corosio (io_uring)

2.42 GB/s

1.24%

-16.1%

bidirectional/16384

asio (coroutines, io_uring)

2.88 GB/s

2.69%

-0.2%

bidirectional/16384

asio (callbacks, io_uring)

2.88 GB/s

1.71%

baseline (io_uring)

bidirectional/262144

corosio (io_uring)

3.30 GB/s

1.11%

-5.6%

bidirectional/262144

asio (coroutines, io_uring)

3.52 GB/s

1.94%

+0.5%

bidirectional/262144

asio (callbacks, io_uring)

3.50 GB/s

2.59%

baseline (io_uring)

bidirectional/4096

corosio (io_uring)

996.0 MB/s

1.15%

+7.0%

bidirectional/4096

asio (coroutines, io_uring)

900.3 MB/s

2.84%

-3.3%

bidirectional/4096

asio (callbacks, io_uring)

930.6 MB/s

2.89%

baseline (io_uring)

bidirectional/65536

corosio (io_uring)

3.02 GB/s

2.29%

-20.0%

bidirectional/65536

asio (coroutines, io_uring)

3.72 GB/s

4.19%

-1.6%

bidirectional/65536

asio (callbacks, io_uring)

3.77 GB/s

2.09%

baseline (io_uring)

bidirectional_lockless/1024

corosio (io_uring)

303.0 MB/s

0.32%

+4.9%

bidirectional_lockless/1024

asio (coroutines, io_uring)

295.7 MB/s

3.42%

+2.4%

bidirectional_lockless/1024

asio (callbacks, io_uring)

288.8 MB/s

4.79%

baseline (io_uring)

bidirectional_lockless/1048576

corosio (io_uring)

3.41 GB/s

2.52%

-4.2%

bidirectional_lockless/1048576

asio (coroutines, io_uring)

3.53 GB/s

1.21%

-0.9%

bidirectional_lockless/1048576

asio (callbacks, io_uring)

3.56 GB/s

2.31%

baseline (io_uring)

bidirectional_lockless/16384

corosio (io_uring)

2.64 GB/s

1.44%

-9.9%

bidirectional_lockless/16384

asio (coroutines, io_uring)

2.93 GB/s

1.86%

+0.2%

bidirectional_lockless/16384

asio (callbacks, io_uring)

2.93 GB/s

2.88%

baseline (io_uring)

bidirectional_lockless/262144

corosio (io_uring)

3.38 GB/s

1.45%

-5.5%

bidirectional_lockless/262144

asio (coroutines, io_uring)

3.57 GB/s

2.53%

-0.3%

bidirectional_lockless/262144

asio (callbacks, io_uring)

3.58 GB/s

2.74%

baseline (io_uring)

bidirectional_lockless/4096

corosio (io_uring)

999.9 MB/s

0.56%

+11.6%

bidirectional_lockless/4096

asio (coroutines, io_uring)

925.8 MB/s

2.49%

+3.3%

bidirectional_lockless/4096

asio (callbacks, io_uring)

895.8 MB/s

4.44%

baseline (io_uring)

bidirectional_lockless/65536

corosio (io_uring)

3.21 GB/s

1.77%

-14.8%

bidirectional_lockless/65536

asio (coroutines, io_uring)

3.76 GB/s

4.44%

-0.2%

bidirectional_lockless/65536

asio (callbacks, io_uring)

3.77 GB/s

2.17%

baseline (io_uring)
local_socket_throughput (io_uring) — Linux, part 2 ↑ faster · dotted line = asio (callbacks) · backend = io_uring corosio asio (coroutines) -40% -20% +20% +40% 0% unidirectional/1024 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/1024 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. -9% unidirectional/1024 unidirectional/1048576 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/1048576 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/1048576 unidirectional/16384 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/16384 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/16384 unidirectional/262144 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/262144 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/262144 unidirectional/4096 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/4096 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/4096 unidirectional/65536 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/65536 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/65536 unidirectional_lockless/1024 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. +17% unidirectional_lockless/1024 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. +3% unidirectional_lockless/1024 unidirectional_lockless/1048576 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/1048576 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/1048576 unidirectional_lockless/16384 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/16384 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/16384 unidirectional_lockless/262144 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/262144 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/262144 unidirectional_lockless/4096 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/4096 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/4096 unidirectional_lockless/65536 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. -18% unidirectional_lockless/65536 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/65536
Detailed results
Benchmark Implementation Median CV vs asio

unidirectional/1024

corosio (io_uring)

303.5 MB/s

0.82%

+15.0%

unidirectional/1024

asio (coroutines, io_uring)

240.0 MB/s

3.83%

-9.0%

unidirectional/1024

asio (callbacks, io_uring)

263.8 MB/s

3.17%

baseline (io_uring)

unidirectional/1048576

corosio (io_uring)

3.42 GB/s

3.01%

-3.5%

unidirectional/1048576

asio (coroutines, io_uring)

3.54 GB/s

2.17%

-0.1%

unidirectional/1048576

asio (callbacks, io_uring)

3.54 GB/s

2.60%

baseline (io_uring)

unidirectional/16384

corosio (io_uring)

2.67 GB/s

2.34%

-4.0%

unidirectional/16384

asio (coroutines, io_uring)

2.73 GB/s

2.13%

-1.7%

unidirectional/16384

asio (callbacks, io_uring)

2.78 GB/s

3.80%

baseline (io_uring)

unidirectional/262144

corosio (io_uring)

3.46 GB/s

1.32%

-1.9%

unidirectional/262144

asio (coroutines, io_uring)

3.55 GB/s

1.01%

+0.7%

unidirectional/262144

asio (callbacks, io_uring)

3.53 GB/s

2.29%

baseline (io_uring)

unidirectional/4096

corosio (io_uring)

1.00 GB/s

1.12%

+14.9%

unidirectional/4096

asio (coroutines, io_uring)

845.2 MB/s

3.36%

-3.0%

unidirectional/4096

asio (callbacks, io_uring)

870.9 MB/s

0.72%

baseline (io_uring)

unidirectional/65536

corosio (io_uring)

3.32 GB/s

3.03%

-15.9%

unidirectional/65536

asio (coroutines, io_uring)

3.95 GB/s

5.85%

-0.1%

unidirectional/65536

asio (callbacks, io_uring)

3.95 GB/s

3.95%

baseline (io_uring)

unidirectional_lockless/1024

corosio (io_uring)

304.2 MB/s

0.88%

+16.8%

unidirectional_lockless/1024

asio (coroutines, io_uring)

269.5 MB/s

4.38%

+3.4%

unidirectional_lockless/1024

asio (callbacks, io_uring)

260.6 MB/s

4.69%

baseline (io_uring)

unidirectional_lockless/1048576

corosio (io_uring)

3.41 GB/s

2.81%

-4.2%

unidirectional_lockless/1048576

asio (coroutines, io_uring)

3.56 GB/s

1.41%

+0.1%

unidirectional_lockless/1048576

asio (callbacks, io_uring)

3.56 GB/s

2.67%

baseline (io_uring)

unidirectional_lockless/16384

corosio (io_uring)

2.65 GB/s

1.33%

-5.1%

unidirectional_lockless/16384

asio (coroutines, io_uring)

2.83 GB/s

0.69%

+1.4%

unidirectional_lockless/16384

asio (callbacks, io_uring)

2.79 GB/s

2.48%

baseline (io_uring)

unidirectional_lockless/262144

corosio (io_uring)

3.45 GB/s

1.24%

-3.2%

unidirectional_lockless/262144

asio (coroutines, io_uring)

3.56 GB/s

1.25%

+0.0%

unidirectional_lockless/262144

asio (callbacks, io_uring)

3.56 GB/s

2.67%

baseline (io_uring)

unidirectional_lockless/4096

corosio (io_uring)

1.00 GB/s

0.67%

+11.3%

unidirectional_lockless/4096

asio (coroutines, io_uring)

872.1 MB/s

2.14%

-3.0%

unidirectional_lockless/4096

asio (callbacks, io_uring)

899.2 MB/s

3.15%

baseline (io_uring)

unidirectional_lockless/65536

corosio (io_uring)

3.28 GB/s

1.64%

-17.7%

unidirectional_lockless/65536

asio (coroutines, io_uring)

3.99 GB/s

5.86%

+0.2%

unidirectional_lockless/65536

asio (callbacks, io_uring)

3.98 GB/s

4.49%

baseline (io_uring)

epoll

local_socket_throughput (epoll) — Linux, part 1 ↑ faster · dotted line = asio (callbacks) · backend = epoll corosio asio (coroutines) -30% -20% -10% +10% +20% 0% bidirectional/1024 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/1024 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. -11% bidirectional/1024 bidirectional/1048576 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/1048576 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/1048576 bidirectional/16384 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/16384 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/16384 bidirectional/262144 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/262144 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/262144 bidirectional/4096 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. +8% bidirectional/4096 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/4096 bidirectional/65536 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/65536 — Both ends of the socket pair write and read simultaneously; N is the chunk size in bytes. bidirectional/65536 bidirectional_lockless/1024 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1024 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1024 bidirectional_lockless/1048576 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1048576 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1048576 bidirectional_lockless/16384 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/16384 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/16384 bidirectional_lockless/262144 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/262144 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. +0% bidirectional_lockless/262144 bidirectional_lockless/4096 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/4096 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/4096 bidirectional_lockless/65536 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. -19% bidirectional_lockless/65536 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/65536
Detailed results
Benchmark Implementation Median CV vs asio

bidirectional/1024

corosio (epoll)

304.3 MB/s

0.78%

+3.2%

bidirectional/1024

asio (coroutines, epoll)

263.6 MB/s

4.34%

-10.6%

bidirectional/1024

asio (callbacks, epoll)

294.8 MB/s

2.68%

baseline (epoll)

bidirectional/1048576

corosio (epoll)

3.13 GB/s

1.70%

-5.0%

bidirectional/1048576

asio (coroutines, epoll)

3.20 GB/s

2.51%

-2.8%

bidirectional/1048576

asio (callbacks, epoll)

3.30 GB/s

3.22%

baseline (epoll)

bidirectional/16384

corosio (epoll)

3.03 GB/s

1.53%

+1.7%

bidirectional/16384

asio (coroutines, epoll)

2.83 GB/s

2.51%

-4.9%

bidirectional/16384

asio (callbacks, epoll)

2.98 GB/s

2.72%

baseline (epoll)

bidirectional/262144

corosio (epoll)

3.13 GB/s

1.23%

-4.5%

bidirectional/262144

asio (coroutines, epoll)

3.27 GB/s

2.17%

-0.3%

bidirectional/262144

asio (callbacks, epoll)

3.28 GB/s

4.57%

baseline (epoll)

bidirectional/4096

corosio (epoll)

1.00 GB/s

0.46%

+7.6%

bidirectional/4096

asio (coroutines, epoll)

888.2 MB/s

2.87%

-4.7%

bidirectional/4096

asio (callbacks, epoll)

932.4 MB/s

3.54%

baseline (epoll)

bidirectional/65536

corosio (epoll)

2.98 GB/s

1.58%

-18.0%

bidirectional/65536

asio (coroutines, epoll)

3.60 GB/s

1.68%

-1.0%

bidirectional/65536

asio (callbacks, epoll)

3.64 GB/s

2.98%

baseline (epoll)

bidirectional_lockless/1024

corosio (epoll)

305.2 MB/s

1.08%

+7.3%

bidirectional_lockless/1024

asio (coroutines, epoll)

265.0 MB/s

5.06%

-6.8%

bidirectional_lockless/1024

asio (callbacks, epoll)

284.4 MB/s

4.56%

baseline (epoll)

bidirectional_lockless/1048576

corosio (epoll)

3.16 GB/s

0.89%

-3.4%

bidirectional_lockless/1048576

asio (coroutines, epoll)

3.22 GB/s

2.67%

-1.5%

bidirectional_lockless/1048576

asio (callbacks, epoll)

3.27 GB/s

4.06%

baseline (epoll)

bidirectional_lockless/16384

corosio (epoll)

3.11 GB/s

2.10%

+1.9%

bidirectional_lockless/16384

asio (coroutines, epoll)

2.90 GB/s

2.33%

-5.2%

bidirectional_lockless/16384

asio (callbacks, epoll)

3.06 GB/s

1.54%

baseline (epoll)

bidirectional_lockless/262144

corosio (epoll)

3.12 GB/s

2.30%

-4.5%

bidirectional_lockless/262144

asio (coroutines, epoll)

3.28 GB/s

2.37%

+0.4%

bidirectional_lockless/262144

asio (callbacks, epoll)

3.27 GB/s

4.17%

baseline (epoll)

bidirectional_lockless/4096

corosio (epoll)

1.01 GB/s

0.84%

+6.6%

bidirectional_lockless/4096

asio (coroutines, epoll)

910.5 MB/s

3.20%

-4.1%

bidirectional_lockless/4096

asio (callbacks, epoll)

949.8 MB/s

3.41%

baseline (epoll)

bidirectional_lockless/65536

corosio (epoll)

3.00 GB/s

1.86%

-19.2%

bidirectional_lockless/65536

asio (coroutines, epoll)

3.61 GB/s

2.91%

-2.8%

bidirectional_lockless/65536

asio (callbacks, epoll)

3.72 GB/s

3.56%

baseline (epoll)
local_socket_throughput (epoll) — Linux, part 2 ↑ faster · dotted line = asio (callbacks) · backend = epoll corosio asio (coroutines) -20% +20% +40% 0% unidirectional/1024 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. +16% unidirectional/1024 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. -6% unidirectional/1024 unidirectional/1048576 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/1048576 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/1048576 unidirectional/16384 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/16384 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/16384 unidirectional/262144 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/262144 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/262144 unidirectional/4096 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/4096 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/4096 unidirectional/65536 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/65536 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/65536 unidirectional_lockless/1024 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/1024 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/1024 unidirectional_lockless/1048576 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/1048576 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/1048576 unidirectional_lockless/16384 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/16384 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/16384 unidirectional_lockless/262144 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/262144 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/262144 unidirectional_lockless/4096 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/4096 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/4096 unidirectional_lockless/65536 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. -15% unidirectional_lockless/65536 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. +1% unidirectional_lockless/65536
Detailed results
Benchmark Implementation Median CV vs asio

unidirectional/1024

corosio (epoll)

298.4 MB/s

0.96%

+16.1%

unidirectional/1024

asio (coroutines, epoll)

240.9 MB/s

3.83%

-6.3%

unidirectional/1024

asio (callbacks, epoll)

257.0 MB/s

4.28%

baseline (epoll)

unidirectional/1048576

corosio (epoll)

3.25 GB/s

1.31%

-6.3%

unidirectional/1048576

asio (coroutines, epoll)

3.39 GB/s

2.40%

-2.3%

unidirectional/1048576

asio (callbacks, epoll)

3.47 GB/s

3.83%

baseline (epoll)

unidirectional/16384

corosio (epoll)

3.08 GB/s

1.51%

+7.9%

unidirectional/16384

asio (coroutines, epoll)

2.76 GB/s

2.62%

-3.4%

unidirectional/16384

asio (callbacks, epoll)

2.86 GB/s

1.81%

baseline (epoll)

unidirectional/262144

corosio (epoll)

3.27 GB/s

1.66%

-5.6%

unidirectional/262144

asio (coroutines, epoll)

3.37 GB/s

1.67%

-2.7%

unidirectional/262144

asio (callbacks, epoll)

3.46 GB/s

3.93%

baseline (epoll)

unidirectional/4096

corosio (epoll)

978.7 MB/s

1.06%

+11.0%

unidirectional/4096

asio (coroutines, epoll)

842.7 MB/s

2.71%

-4.4%

unidirectional/4096

asio (callbacks, epoll)

881.8 MB/s

0.49%

baseline (epoll)

unidirectional/65536

corosio (epoll)

3.29 GB/s

2.35%

-13.9%

unidirectional/65536

asio (coroutines, epoll)

3.83 GB/s

2.77%

+0.2%

unidirectional/65536

asio (callbacks, epoll)

3.82 GB/s

2.69%

baseline (epoll)

unidirectional_lockless/1024

corosio (epoll)

299.0 MB/s

0.71%

+12.9%

unidirectional_lockless/1024

asio (coroutines, epoll)

250.0 MB/s

4.40%

-5.6%

unidirectional_lockless/1024

asio (callbacks, epoll)

264.9 MB/s

4.61%

baseline (epoll)

unidirectional_lockless/1048576

corosio (epoll)

3.28 GB/s

1.05%

-4.1%

unidirectional_lockless/1048576

asio (coroutines, epoll)

3.36 GB/s

1.57%

-1.6%

unidirectional_lockless/1048576

asio (callbacks, epoll)

3.42 GB/s

3.36%

baseline (epoll)

unidirectional_lockless/16384

corosio (epoll)

3.09 GB/s

3.18%

+5.8%

unidirectional_lockless/16384

asio (coroutines, epoll)

2.81 GB/s

1.94%

-3.8%

unidirectional_lockless/16384

asio (callbacks, epoll)

2.92 GB/s

2.13%

baseline (epoll)

unidirectional_lockless/262144

corosio (epoll)

3.28 GB/s

1.37%

-5.9%

unidirectional_lockless/262144

asio (coroutines, epoll)

3.34 GB/s

3.68%

-4.2%

unidirectional_lockless/262144

asio (callbacks, epoll)

3.48 GB/s

4.09%

baseline (epoll)

unidirectional_lockless/4096

corosio (epoll)

988.0 MB/s

0.93%

+13.2%

unidirectional_lockless/4096

asio (coroutines, epoll)

869.2 MB/s

2.52%

-0.4%

unidirectional_lockless/4096

asio (callbacks, epoll)

873.0 MB/s

2.70%

baseline (epoll)

unidirectional_lockless/65536

corosio (epoll)

3.29 GB/s

4.33%

-14.8%

unidirectional_lockless/65536

asio (coroutines, epoll)

3.92 GB/s

1.70%

+1.4%

unidirectional_lockless/65536

asio (callbacks, epoll)

3.86 GB/s

3.97%

baseline (epoll)

socket_latency

Round-trip latency of a TCP loopback connection. Each sample is one full round trip (request out, reply back) across message sizes and concurrent pair counts.

io_uring

socket_latency (io_uring) — Linux ↑ faster (lower mean latency) · dotted line = asio (callbacks) · backend = io_uring corosio asio (coroutines) -10% +10% +20% 0% concurrent/1 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/1 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/1 concurrent/16 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/16 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/16 concurrent/4 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/4 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/4 concurrent_lockless/1 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/1 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/1 concurrent_lockless/16 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. -5% concurrent_lockless/16 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/16 concurrent_lockless/4 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/4 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/4 pingpong/1 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. +13% pingpong/1 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. -2% pingpong/1 pingpong/1024 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. pingpong/1024 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. pingpong/1024 pingpong/64 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. pingpong/64 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. pingpong/64 pingpong_lockless/1 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1 pingpong_lockless/1024 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1024 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1024 pingpong_lockless/64 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/64 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. -3% pingpong_lockless/64
Detailed results
Benchmark Implementation Median CV vs asio

concurrent/1

corosio (io_uring)

14.57 µs

0.98%

+13.3%

concurrent/1

asio (coroutines, io_uring)

17.13 µs

0.50%

-2.0%

concurrent/1

asio (callbacks, io_uring)

16.81 µs

1.22%

baseline (io_uring)

concurrent/16

corosio (io_uring)

236.38 µs

0.79%

-3.8%

concurrent/16

asio (coroutines, io_uring)

232.68 µs

0.92%

-2.2%

concurrent/16

asio (callbacks, io_uring)

227.64 µs

0.96%

baseline (io_uring)

concurrent/4

corosio (io_uring)

58.57 µs

0.71%

+0.5%

concurrent/4

asio (coroutines, io_uring)

60.22 µs

0.94%

-2.3%

concurrent/4

asio (callbacks, io_uring)

58.84 µs

0.87%

baseline (io_uring)

concurrent_lockless/1

corosio (io_uring)

14.58 µs

1.06%

+10.7%

concurrent_lockless/1

asio (coroutines, io_uring)

16.70 µs

1.06%

-2.3%

concurrent_lockless/1

asio (callbacks, io_uring)

16.33 µs

0.86%

baseline (io_uring)

concurrent_lockless/16

corosio (io_uring)

235.72 µs

1.04%

-5.0%

concurrent_lockless/16

asio (coroutines, io_uring)

228.57 µs

1.04%

-1.8%

concurrent_lockless/16

asio (callbacks, io_uring)

224.43 µs

1.01%

baseline (io_uring)

concurrent_lockless/4

corosio (io_uring)

58.50 µs

1.02%

-1.2%

concurrent_lockless/4

asio (coroutines, io_uring)

59.27 µs

1.28%

-2.5%

concurrent_lockless/4

asio (callbacks, io_uring)

57.81 µs

1.11%

baseline (io_uring)

pingpong/1

corosio (io_uring)

14.52 µs

1.27%

+13.3%

pingpong/1

asio (coroutines, io_uring)

17.04 µs

1.15%

-1.7%

pingpong/1

asio (callbacks, io_uring)

16.75 µs

0.11%

baseline (io_uring)

pingpong/1024

corosio (io_uring)

14.81 µs

1.36%

+13.3%

pingpong/1024

asio (coroutines, io_uring)

17.46 µs

0.82%

-2.2%

pingpong/1024

asio (callbacks, io_uring)

17.09 µs

0.65%

baseline (io_uring)

pingpong/64

corosio (io_uring)

14.59 µs

1.34%

+13.2%

pingpong/64

asio (coroutines, io_uring)

17.14 µs

0.86%

-1.9%

pingpong/64

asio (callbacks, io_uring)

16.81 µs

0.26%

baseline (io_uring)

pingpong_lockless/1

corosio (io_uring)

14.49 µs

1.33%

+10.8%

pingpong_lockless/1

asio (coroutines, io_uring)

16.64 µs

1.06%

-2.5%

pingpong_lockless/1

asio (callbacks, io_uring)

16.24 µs

1.05%

baseline (io_uring)

pingpong_lockless/1024

corosio (io_uring)

14.81 µs

1.19%

+10.8%

pingpong_lockless/1024

asio (coroutines, io_uring)

17.01 µs

0.27%

-2.5%

pingpong_lockless/1024

asio (callbacks, io_uring)

16.59 µs

1.12%

baseline (io_uring)

pingpong_lockless/64

corosio (io_uring)

14.56 µs

1.35%

+10.6%

pingpong_lockless/64

asio (coroutines, io_uring)

16.73 µs

0.21%

-2.7%

pingpong_lockless/64

asio (callbacks, io_uring)

16.29 µs

0.97%

baseline (io_uring)

epoll

socket_latency (epoll) — Linux ↑ faster (lower mean latency) · dotted line = asio (callbacks) · backend = epoll corosio asio (coroutines) -5% +5% +10% 0% concurrent/1 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/1 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/1 concurrent/16 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/16 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/16 concurrent/4 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/4 — N independent 64-byte pingpong pairs run concurrently on one context; N is the number of connection pairs. concurrent/4 concurrent_lockless/1 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/1 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/1 concurrent_lockless/16 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. -3% concurrent_lockless/16 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/16 concurrent_lockless/4 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. concurrent_lockless/4 — Same as concurrent, with the context in single-threaded lockless mode; N is the number of connection pairs. -3% concurrent_lockless/4 pingpong/1 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. pingpong/1 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. pingpong/1 pingpong/1024 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. +7% pingpong/1024 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. -2% pingpong/1024 pingpong/64 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. pingpong/64 — One TCP connection ping-pongs a message client->server->client; N is the message size in bytes. pingpong/64 pingpong_lockless/1 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1 pingpong_lockless/1024 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1024 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/1024 pingpong_lockless/64 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/64 — Same as pingpong, with the context in single-threaded lockless mode; N is the message size in bytes. pingpong_lockless/64
Detailed results
Benchmark Implementation Median CV vs asio

concurrent/1

corosio (epoll)

14.80 µs

0.88%

+7.1%

concurrent/1

asio (coroutines, epoll)

16.32 µs

0.92%

-2.4%

concurrent/1

asio (callbacks, epoll)

15.94 µs

0.61%

baseline (epoll)

concurrent/16

corosio (epoll)

232.79 µs

0.98%

-1.4%

concurrent/16

asio (coroutines, epoll)

236.30 µs

0.99%

-2.9%

concurrent/16

asio (callbacks, epoll)

229.53 µs

0.71%

baseline (epoll)

concurrent/4

corosio (epoll)

58.22 µs

0.88%

+0.4%

concurrent/4

asio (coroutines, epoll)

60.27 µs

1.02%

-3.1%

concurrent/4

asio (callbacks, epoll)

58.48 µs

0.79%

baseline (epoll)

concurrent_lockless/1

corosio (epoll)

14.69 µs

1.26%

+6.0%

concurrent_lockless/1

asio (coroutines, epoll)

16.00 µs

0.81%

-2.4%

concurrent_lockless/1

asio (callbacks, epoll)

15.63 µs

0.97%

baseline (epoll)

concurrent_lockless/16

corosio (epoll)

231.82 µs

1.01%

-3.3%

concurrent_lockless/16

asio (coroutines, epoll)

231.30 µs

1.07%

-3.1%

concurrent_lockless/16

asio (callbacks, epoll)

224.45 µs

0.91%

baseline (epoll)

concurrent_lockless/4

corosio (epoll)

57.92 µs

1.30%

-1.3%

concurrent_lockless/4

asio (coroutines, epoll)

59.00 µs

0.95%

-3.1%

concurrent_lockless/4

asio (callbacks, epoll)

57.20 µs

1.03%

baseline (epoll)

pingpong/1

corosio (epoll)

14.74 µs

1.03%

+7.2%

pingpong/1

asio (coroutines, epoll)

16.29 µs

1.32%

-2.5%

pingpong/1

asio (callbacks, epoll)

15.88 µs

1.06%

baseline (epoll)

pingpong/1024

corosio (epoll)

15.04 µs

1.18%

+7.3%

pingpong/1024

asio (coroutines, epoll)

16.56 µs

0.99%

-2.1%

pingpong/1024

asio (callbacks, epoll)

16.22 µs

0.98%

baseline (epoll)

pingpong/64

corosio (epoll)

14.78 µs

1.17%

+7.3%

pingpong/64

asio (coroutines, epoll)

16.35 µs

1.26%

-2.6%

pingpong/64

asio (callbacks, epoll)

15.94 µs

0.90%

baseline (epoll)

pingpong_lockless/1

corosio (epoll)

14.62 µs

0.99%

+5.8%

pingpong_lockless/1

asio (coroutines, epoll)

15.92 µs

0.79%

-2.5%

pingpong_lockless/1

asio (callbacks, epoll)

15.52 µs

0.86%

baseline (epoll)

pingpong_lockless/1024

corosio (epoll)

14.96 µs

0.95%

+5.8%

pingpong_lockless/1024

asio (coroutines, epoll)

16.26 µs

0.90%

-2.4%

pingpong_lockless/1024

asio (callbacks, epoll)

15.89 µs

0.12%

baseline (epoll)

pingpong_lockless/64

corosio (epoll)

14.71 µs

0.98%

+5.8%

pingpong_lockless/64

asio (coroutines, epoll)

16.01 µs

0.66%

-2.5%

pingpong_lockless/64

asio (callbacks, epoll)

15.61 µs

0.18%

baseline (epoll)

socket_throughput

Sustained byte throughput of a TCP loopback connection under continuous streaming, varying chunk size, direction, and concurrency.

io_uring

socket_throughput (io_uring) — Linux, part 1 ↑ faster · dotted line = asio (callbacks) · backend = io_uring corosio asio (coroutines) -20% -10% +10% +20% 0% bidirectional/1024 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/1024 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/1024 bidirectional/1048576 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/1048576 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/1048576 bidirectional/16384 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/16384 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. -5% bidirectional/16384 bidirectional/262144 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/262144 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/262144 bidirectional/4096 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/4096 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/4096 bidirectional/65536 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/65536 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/65536 bidirectional_lockless/1024 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1024 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1024 bidirectional_lockless/1048576 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1048576 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1048576 bidirectional_lockless/16384 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. -8% bidirectional_lockless/16384 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/16384 bidirectional_lockless/262144 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. +13% bidirectional_lockless/262144 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. +1% bidirectional_lockless/262144 bidirectional_lockless/4096 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/4096 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/4096 bidirectional_lockless/65536 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/65536 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/65536
Detailed results
Benchmark Implementation Median CV vs asio

bidirectional/1024

corosio (io_uring)

140.6 MB/s

1.04%

+2.9%

bidirectional/1024

asio (coroutines, io_uring)

134.2 MB/s

1.24%

-1.8%

bidirectional/1024

asio (callbacks, io_uring)

136.6 MB/s

0.45%

baseline (io_uring)

bidirectional/1048576

corosio (io_uring)

3.26 GB/s

1.47%

-0.7%

bidirectional/1048576

asio (coroutines, io_uring)

3.25 GB/s

0.88%

-0.8%

bidirectional/1048576

asio (callbacks, io_uring)

3.28 GB/s

0.86%

baseline (io_uring)

bidirectional/16384

corosio (io_uring)

1.54 GB/s

0.62%

-6.2%

bidirectional/16384

asio (coroutines, io_uring)

1.56 GB/s

0.47%

-5.1%

bidirectional/16384

asio (callbacks, io_uring)

1.64 GB/s

1.87%

baseline (io_uring)

bidirectional/262144

corosio (io_uring)

3.21 GB/s

1.11%

+10.8%

bidirectional/262144

asio (coroutines, io_uring)

2.86 GB/s

0.72%

-1.1%

bidirectional/262144

asio (callbacks, io_uring)

2.89 GB/s

1.61%

baseline (io_uring)

bidirectional/4096

corosio (io_uring)

494.8 MB/s

1.09%

+0.4%

bidirectional/4096

asio (coroutines, io_uring)

485.4 MB/s

1.25%

-1.5%

bidirectional/4096

asio (callbacks, io_uring)

492.8 MB/s

1.39%

baseline (io_uring)

bidirectional/65536

corosio (io_uring)

2.76 GB/s

2.00%

+5.0%

bidirectional/65536

asio (coroutines, io_uring)

2.60 GB/s

1.95%

-1.0%

bidirectional/65536

asio (callbacks, io_uring)

2.63 GB/s

2.11%

baseline (io_uring)

bidirectional_lockless/1024

corosio (io_uring)

140.8 MB/s

0.18%

+0.9%

bidirectional_lockless/1024

asio (coroutines, io_uring)

136.4 MB/s

0.13%

-2.2%

bidirectional_lockless/1024

asio (callbacks, io_uring)

139.5 MB/s

1.34%

baseline (io_uring)

bidirectional_lockless/1048576

corosio (io_uring)

3.37 GB/s

1.25%

+3.0%

bidirectional_lockless/1048576

asio (coroutines, io_uring)

3.28 GB/s

1.49%

+0.2%

bidirectional_lockless/1048576

asio (callbacks, io_uring)

3.27 GB/s

1.11%

baseline (io_uring)

bidirectional_lockless/16384

corosio (io_uring)

1.54 GB/s

1.96%

-8.0%

bidirectional_lockless/16384

asio (coroutines, io_uring)

1.66 GB/s

1.10%

-0.5%

bidirectional_lockless/16384

asio (callbacks, io_uring)

1.67 GB/s

2.44%

baseline (io_uring)

bidirectional_lockless/262144

corosio (io_uring)

3.25 GB/s

1.02%

+13.0%

bidirectional_lockless/262144

asio (coroutines, io_uring)

2.90 GB/s

0.99%

+0.8%

bidirectional_lockless/262144

asio (callbacks, io_uring)

2.88 GB/s

0.46%

baseline (io_uring)

bidirectional_lockless/4096

corosio (io_uring)

494.3 MB/s

0.58%

-1.9%

bidirectional_lockless/4096

asio (coroutines, io_uring)

494.6 MB/s

0.28%

-1.8%

bidirectional_lockless/4096

asio (callbacks, io_uring)

503.6 MB/s

1.55%

baseline (io_uring)

bidirectional_lockless/65536

corosio (io_uring)

2.78 GB/s

0.90%

+5.1%

bidirectional_lockless/65536

asio (coroutines, io_uring)

2.61 GB/s

1.69%

-1.1%

bidirectional_lockless/65536

asio (callbacks, io_uring)

2.64 GB/s

2.18%

baseline (io_uring)
socket_throughput (io_uring) — Linux, part 2 ↑ faster · dotted line = asio (callbacks) · backend = io_uring corosio asio (coroutines) -100% +100% +200% +300% 0% multithread/2 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. +123% multithread/2 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. +7% multithread/2 multithread/4 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. +206% multithread/4 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. -1% multithread/4 multithread/8 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. +215% multithread/8 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. -1% multithread/8
Detailed results
Benchmark Implementation Median CV vs asio

multithread/2

corosio (io_uring)

3.67 GB/s

1.05%

+123.3%

multithread/2

asio (coroutines, io_uring)

1.75 GB/s

7.63%

+6.8%

multithread/2

asio (callbacks, io_uring)

1.64 GB/s

7.69%

baseline (io_uring)

multithread/4

corosio (io_uring)

5.08 GB/s

0.33%

+206.1%

multithread/4

asio (coroutines, io_uring)

1.64 GB/s

0.65%

-1.2%

multithread/4

asio (callbacks, io_uring)

1.66 GB/s

0.65%

baseline (io_uring)

multithread/8

corosio (io_uring)

5.13 GB/s

0.17%

+214.9%

multithread/8

asio (coroutines, io_uring)

1.61 GB/s

0.94%

-1.0%

multithread/8

asio (callbacks, io_uring)

1.63 GB/s

0.60%

baseline (io_uring)
socket_throughput (io_uring) — Linux, part 3 ↑ faster · dotted line = asio (callbacks) · backend = io_uring corosio asio (coroutines) -20% +20% +40% +60% 0% unidirectional/1024 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. +51% unidirectional/1024 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/1024 unidirectional/1048576 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/1048576 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/1048576 unidirectional/16384 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/16384 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/16384 unidirectional/262144 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/262144 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/262144 unidirectional/4096 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/4096 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/4096 unidirectional/65536 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/65536 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/65536 unidirectional_lockless/1024 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/1024 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. -2% unidirectional_lockless/1024 unidirectional_lockless/1048576 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. -5% unidirectional_lockless/1048576 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/1048576 unidirectional_lockless/16384 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/16384 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/16384 unidirectional_lockless/262144 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/262144 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. +2% unidirectional_lockless/262144 unidirectional_lockless/4096 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/4096 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/4096 unidirectional_lockless/65536 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/65536 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/65536
Detailed results
Benchmark Implementation Median CV vs asio

unidirectional/1024

corosio (io_uring)

139.2 MB/s

0.31%

+51.1%

unidirectional/1024

asio (coroutines, io_uring)

90.70 MB/s

0.50%

-1.6%

unidirectional/1024

asio (callbacks, io_uring)

92.14 MB/s

1.18%

baseline (io_uring)

unidirectional/1048576

corosio (io_uring)

3.41 GB/s

1.66%

+3.5%

unidirectional/1048576

asio (coroutines, io_uring)

3.31 GB/s

2.10%

+0.3%

unidirectional/1048576

asio (callbacks, io_uring)

3.30 GB/s

0.48%

baseline (io_uring)

unidirectional/16384

corosio (io_uring)

1.54 GB/s

3.75%

+16.9%

unidirectional/16384

asio (coroutines, io_uring)

1.32 GB/s

1.75%

-0.2%

unidirectional/16384

asio (callbacks, io_uring)

1.32 GB/s

1.06%

baseline (io_uring)

unidirectional/262144

corosio (io_uring)

3.23 GB/s

1.54%

+6.1%

unidirectional/262144

asio (coroutines, io_uring)

3.07 GB/s

1.19%

+1.0%

unidirectional/262144

asio (callbacks, io_uring)

3.04 GB/s

0.67%

baseline (io_uring)

unidirectional/4096

corosio (io_uring)

491.5 MB/s

1.27%

+36.5%

unidirectional/4096

asio (coroutines, io_uring)

354.4 MB/s

0.83%

-1.6%

unidirectional/4096

asio (callbacks, io_uring)

360.0 MB/s

0.52%

baseline (io_uring)

unidirectional/65536

corosio (io_uring)

2.72 GB/s

0.87%

+3.1%

unidirectional/65536

asio (coroutines, io_uring)

2.64 GB/s

2.73%

-0.2%

unidirectional/65536

asio (callbacks, io_uring)

2.64 GB/s

2.12%

baseline (io_uring)

unidirectional_lockless/1024

corosio (io_uring)

139.7 MB/s

0.91%

+49.0%

unidirectional_lockless/1024

asio (coroutines, io_uring)

92.09 MB/s

0.45%

-1.8%

unidirectional_lockless/1024

asio (callbacks, io_uring)

93.74 MB/s

1.32%

baseline (io_uring)

unidirectional_lockless/1048576

corosio (io_uring)

3.34 GB/s

1.53%

-4.7%

unidirectional_lockless/1048576

asio (coroutines, io_uring)

3.52 GB/s

0.43%

+0.5%

unidirectional_lockless/1048576

asio (callbacks, io_uring)

3.51 GB/s

0.76%

baseline (io_uring)

unidirectional_lockless/16384

corosio (io_uring)

1.62 GB/s

2.27%

+21.0%

unidirectional_lockless/16384

asio (coroutines, io_uring)

1.33 GB/s

1.46%

-0.6%

unidirectional_lockless/16384

asio (callbacks, io_uring)

1.34 GB/s

2.07%

baseline (io_uring)

unidirectional_lockless/262144

corosio (io_uring)

3.23 GB/s

1.21%

+6.8%

unidirectional_lockless/262144

asio (coroutines, io_uring)

3.10 GB/s

1.13%

+2.2%

unidirectional_lockless/262144

asio (callbacks, io_uring)

3.03 GB/s

1.03%

baseline (io_uring)

unidirectional_lockless/4096

corosio (io_uring)

494.4 MB/s

0.99%

+35.3%

unidirectional_lockless/4096

asio (coroutines, io_uring)

361.0 MB/s

0.29%

-1.2%

unidirectional_lockless/4096

asio (callbacks, io_uring)

365.4 MB/s

1.55%

baseline (io_uring)

unidirectional_lockless/65536

corosio (io_uring)

2.73 GB/s

2.34%

+1.8%

unidirectional_lockless/65536

asio (coroutines, io_uring)

2.64 GB/s

2.13%

-1.4%

unidirectional_lockless/65536

asio (callbacks, io_uring)

2.68 GB/s

2.22%

baseline (io_uring)

epoll

socket_throughput (epoll) — Linux, part 1 ↑ faster · dotted line = asio (callbacks) · backend = epoll corosio asio (coroutines) -10% +10% +20% 0% bidirectional/1024 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/1024 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. -3% bidirectional/1024 bidirectional/1048576 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/1048576 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/1048576 bidirectional/16384 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/16384 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/16384 bidirectional/262144 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. +13% bidirectional/262144 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. +2% bidirectional/262144 bidirectional/4096 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/4096 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/4096 bidirectional/65536 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/65536 — Both ends of the connection write and read simultaneously; N is the chunk size in bytes. bidirectional/65536 bidirectional_lockless/1024 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1024 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1024 bidirectional_lockless/1048576 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1048576 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/1048576 bidirectional_lockless/16384 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/16384 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/16384 bidirectional_lockless/262144 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/262144 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/262144 bidirectional_lockless/4096 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/4096 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/4096 bidirectional_lockless/65536 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. -5% bidirectional_lockless/65536 — Same as bidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. bidirectional_lockless/65536
Detailed results
Benchmark Implementation Median CV vs asio

bidirectional/1024

corosio (epoll)

134.2 MB/s

0.37%

-3.7%

bidirectional/1024

asio (coroutines, epoll)

134.9 MB/s

0.79%

-3.2%

bidirectional/1024

asio (callbacks, epoll)

139.4 MB/s

0.93%

baseline (epoll)

bidirectional/1048576

corosio (epoll)

3.24 GB/s

2.03%

-1.1%

bidirectional/1048576

asio (coroutines, epoll)

3.26 GB/s

1.14%

-0.7%

bidirectional/1048576

asio (callbacks, epoll)

3.28 GB/s

0.90%

baseline (epoll)

bidirectional/16384

corosio (epoll)

1.64 GB/s

0.44%

-1.8%

bidirectional/16384

asio (coroutines, epoll)

1.63 GB/s

1.71%

-2.4%

bidirectional/16384

asio (callbacks, epoll)

1.67 GB/s

1.45%

baseline (epoll)

bidirectional/262144

corosio (epoll)

3.23 GB/s

0.81%

+12.5%

bidirectional/262144

asio (coroutines, epoll)

2.92 GB/s

2.11%

+1.7%

bidirectional/262144

asio (callbacks, epoll)

2.87 GB/s

0.66%

baseline (epoll)

bidirectional/4096

corosio (epoll)

506.4 MB/s

0.56%

+1.1%

bidirectional/4096

asio (coroutines, epoll)

491.0 MB/s

1.43%

-2.0%

bidirectional/4096

asio (callbacks, epoll)

500.9 MB/s

1.38%

baseline (epoll)

bidirectional/65536

corosio (epoll)

2.57 GB/s

2.19%

-2.4%

bidirectional/65536

asio (coroutines, epoll)

2.59 GB/s

1.89%

-1.4%

bidirectional/65536

asio (callbacks, epoll)

2.63 GB/s

1.51%

baseline (epoll)

bidirectional_lockless/1024

corosio (epoll)

135.0 MB/s

1.20%

-4.5%

bidirectional_lockless/1024

asio (coroutines, epoll)

137.6 MB/s

1.28%

-2.7%

bidirectional_lockless/1024

asio (callbacks, epoll)

141.4 MB/s

0.91%

baseline (epoll)

bidirectional_lockless/1048576

corosio (epoll)

3.25 GB/s

2.77%

-2.2%

bidirectional_lockless/1048576

asio (coroutines, epoll)

3.27 GB/s

0.87%

-1.6%

bidirectional_lockless/1048576

asio (callbacks, epoll)

3.33 GB/s

1.88%

baseline (epoll)

bidirectional_lockless/16384

corosio (epoll)

1.65 GB/s

0.29%

-2.4%

bidirectional_lockless/16384

asio (coroutines, epoll)

1.67 GB/s

0.77%

-1.4%

bidirectional_lockless/16384

asio (callbacks, epoll)

1.69 GB/s

1.75%

baseline (epoll)

bidirectional_lockless/262144

corosio (epoll)

3.23 GB/s

1.02%

+10.0%

bidirectional_lockless/262144

asio (coroutines, epoll)

2.86 GB/s

1.38%

-2.5%

bidirectional_lockless/262144

asio (callbacks, epoll)

2.93 GB/s

2.33%

baseline (epoll)

bidirectional_lockless/4096

corosio (epoll)

510.0 MB/s

1.42%

-0.2%

bidirectional_lockless/4096

asio (coroutines, epoll)

496.9 MB/s

1.06%

-2.8%

bidirectional_lockless/4096

asio (callbacks, epoll)

511.1 MB/s

1.13%

baseline (epoll)

bidirectional_lockless/65536

corosio (epoll)

2.52 GB/s

3.26%

-5.2%

bidirectional_lockless/65536

asio (coroutines, epoll)

2.60 GB/s

2.95%

-2.4%

bidirectional_lockless/65536

asio (callbacks, epoll)

2.66 GB/s

2.19%

baseline (epoll)
socket_throughput (epoll) — Linux, part 2 ↑ faster · dotted line = asio (callbacks) · backend = epoll corosio asio (coroutines) -20% +20% +40% +60% 0% multithread/2 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. +45% multithread/2 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. -3% multithread/2 multithread/4 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. +38% multithread/4 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. +0% multithread/4 multithread/8 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. +25% multithread/8 — 32 bidirectional connection pairs share one context serviced by N threads running the context, with 64 KiB chunks; N is the thread count. 0% multithread/8
Detailed results
Benchmark Implementation Median CV vs asio

multithread/2

corosio (epoll)

3.40 GB/s

0.78%

+44.6%

multithread/2

asio (coroutines, epoll)

2.28 GB/s

2.00%

-2.9%

multithread/2

asio (callbacks, epoll)

2.35 GB/s

1.70%

baseline (epoll)

multithread/4

corosio (epoll)

4.87 GB/s

0.44%

+38.0%

multithread/4

asio (coroutines, epoll)

3.53 GB/s

0.56%

+0.1%

multithread/4

asio (callbacks, epoll)

3.53 GB/s

0.57%

baseline (epoll)

multithread/8

corosio (epoll)

4.95 GB/s

0.22%

+25.4%

multithread/8

asio (coroutines, epoll)

3.94 GB/s

0.19%

-0.3%

multithread/8

asio (callbacks, epoll)

3.95 GB/s

0.35%

baseline (epoll)
socket_throughput (epoll) — Linux, part 3 ↑ faster · dotted line = asio (callbacks) · backend = epoll corosio asio (coroutines) -20% +20% +40% +60% 0% unidirectional/1024 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. +35% unidirectional/1024 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/1024 unidirectional/1048576 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/1048576 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/1048576 unidirectional/16384 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/16384 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. -3% unidirectional/16384 unidirectional/262144 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/262144 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/262144 unidirectional/4096 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/4096 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/4096 unidirectional/65536 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/65536 — One writer streams to one reader as fast as possible; N is the write/read chunk size in bytes. unidirectional/65536 unidirectional_lockless/1024 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/1024 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/1024 unidirectional_lockless/1048576 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. -6% unidirectional_lockless/1048576 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/1048576 unidirectional_lockless/16384 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/16384 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/16384 unidirectional_lockless/262144 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/262144 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/262144 unidirectional_lockless/4096 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/4096 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/4096 unidirectional_lockless/65536 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. unidirectional_lockless/65536 — Same as unidirectional, with the context in single-threaded lockless mode; N is the chunk size in bytes. +0% unidirectional_lockless/65536
Detailed results
Benchmark Implementation Median CV vs asio

unidirectional/1024

corosio (epoll)

126.1 MB/s

1.10%

+34.6%

unidirectional/1024

asio (coroutines, epoll)

91.85 MB/s

0.45%

-1.9%

unidirectional/1024

asio (callbacks, epoll)

93.66 MB/s

0.88%

baseline (epoll)

unidirectional/1048576

corosio (epoll)

3.38 GB/s

0.96%

-0.7%

unidirectional/1048576

asio (coroutines, epoll)

3.36 GB/s

2.04%

-1.3%

unidirectional/1048576

asio (callbacks, epoll)

3.40 GB/s

0.40%

baseline (epoll)

unidirectional/16384

corosio (epoll)

1.58 GB/s

1.29%

+15.6%

unidirectional/16384

asio (coroutines, epoll)

1.33 GB/s

1.39%

-3.0%

unidirectional/16384

asio (callbacks, epoll)

1.37 GB/s

1.28%

baseline (epoll)

unidirectional/262144

corosio (epoll)

3.35 GB/s

1.21%

+10.0%

unidirectional/262144

asio (coroutines, epoll)

3.04 GB/s

1.49%

-0.2%

unidirectional/262144

asio (callbacks, epoll)

3.04 GB/s

2.22%

baseline (epoll)

unidirectional/4096

corosio (epoll)

477.0 MB/s

1.07%

+30.0%

unidirectional/4096

asio (coroutines, epoll)

358.2 MB/s

0.83%

-2.4%

unidirectional/4096

asio (callbacks, epoll)

366.8 MB/s

0.70%

baseline (epoll)

unidirectional/65536

corosio (epoll)

2.76 GB/s

3.74%

+2.4%

unidirectional/65536

asio (coroutines, epoll)

2.69 GB/s

1.14%

-0.3%

unidirectional/65536

asio (callbacks, epoll)

2.70 GB/s

2.02%

baseline (epoll)

unidirectional_lockless/1024

corosio (epoll)

126.1 MB/s

0.80%

+33.3%

unidirectional_lockless/1024

asio (coroutines, epoll)

93.22 MB/s

1.39%

-1.5%

unidirectional_lockless/1024

asio (callbacks, epoll)

94.59 MB/s

0.33%

baseline (epoll)

unidirectional_lockless/1048576

corosio (epoll)

3.35 GB/s

1.22%

-6.3%

unidirectional_lockless/1048576

asio (coroutines, epoll)

3.55 GB/s

2.34%

-0.8%

unidirectional_lockless/1048576

asio (callbacks, epoll)

3.57 GB/s

1.29%

baseline (epoll)

unidirectional_lockless/16384

corosio (epoll)

1.56 GB/s

2.38%

+13.6%

unidirectional_lockless/16384

asio (coroutines, epoll)

1.35 GB/s

1.63%

-2.0%

unidirectional_lockless/16384

asio (callbacks, epoll)

1.37 GB/s

1.51%

baseline (epoll)

unidirectional_lockless/262144

corosio (epoll)

3.35 GB/s

1.59%

+7.5%

unidirectional_lockless/262144

asio (coroutines, epoll)

3.07 GB/s

1.21%

-1.4%

unidirectional_lockless/262144

asio (callbacks, epoll)

3.11 GB/s

1.77%

baseline (epoll)

unidirectional_lockless/4096

corosio (epoll)

476.5 MB/s

1.12%

+28.4%

unidirectional_lockless/4096

asio (coroutines, epoll)

364.4 MB/s

1.20%

-1.8%

unidirectional_lockless/4096

asio (callbacks, epoll)

371.0 MB/s

0.39%

baseline (epoll)

unidirectional_lockless/65536

corosio (epoll)

2.74 GB/s

4.50%

+0.3%

unidirectional_lockless/65536

asio (coroutines, epoll)

2.73 GB/s

2.28%

+0.1%

unidirectional_lockless/65536

asio (callbacks, epoll)

2.73 GB/s

2.83%

baseline (epoll)
Edit this Page