[sgerbino]

Regenerates the published benchmark pages from a full run of benchmark-report on develop at 0b3830d, which includes the io_uring work from #409. The last refresh was on 2026-10-02. Only the generated pages and charts change.

Source data: run 42119198783, 7 iterations of 2 s on the Linux, Windows and macOS machines.

What moved

corosio on io_uring against asio on io_uring, across all 113 Linux rows:

	vs asio coroutines	vs asio callbacks
Faster / within noise / slower	82 / 27 / 4	62 / 41 / 10
Geometric mean	+17.7%	+15.3%
http_server	+37%	+36%
fan_out	+4%	+1%
accept_churn	0%	−1%

These counts come from a separate analysis of the same data, which calls a row a tie within 2% or its noise. The generated pages use their own threshold, so their summaries can differ slightly.

Worth checking

macOS local_socket_latency shows asio at about 48.6 µs and corosio at about 1.8 µs. Both asio flavors land on nearly the same value, and it is stable across iterations, so this may reflect how the benchmark behaves on macOS rather than a real 26× difference. The cause is not known. The pages show the measured values unchanged.

Testing

The pages were generated with report_page.py as described in bench/README.md. The Antora site was not built locally; the Documentation workflow on this PR builds it.

[cppalliance-bot]

An automated preview of the documentation is available at https://413.corosio.prtest3.cppalliance.org/index.html

If more commits are pushed to the pull request, the docs will rebuild at the same URL.

2026-11-06 03:38:53 UTC

[cppalliance-bot]

GCOVR code coverage report https://413.corosio.prtest3.cppalliance.org/gcovr/index.html
LCOV code coverage report https://413.corosio.prtest3.cppalliance.org/genhtml/index.html
Coverage Diff Report https://413.corosio.prtest3.cppalliance.org/diff-report/index.html

Build time: 2026-11-06 03:55:05 UTC