Re: Performance improvement with 81f730d4

[Kevin Wolf <kwolf@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1342 bytes --]

Hi Lukáš,

Am 19.04.2023 um 08:02 hat Lukáš Doktor geschrieben:
> Hello Paolo,
> 
> the perf-ci detected and bisected the 81f730d4 - block, block-backend:
> write some hot coroutine wrappers by hand - as a performance
> improvement with 4K read/writes across various profiles. Without
> pinning it was about 10%, with pinned qemu about 15%.
> 
>     https://ldoktor.github.io/tmp/RedHat-virtlab722/v8.0.0/130-improvement.html

I must admit that I don't really understand how to read this page, and
which build corresponds to which QEMU revision (or even which one is
older and which one is newer).

All the build names are links, but they only result in a 404.

> Based on the commit message I guess it's expected, so my question is
> whether I should report such obvious improvements or only report
> improvements/regressions that are not that obvious?

I think it's useful to know that we did indeed fix the performance
problem, so as far as I am concerned, please do report them.

When we're addressing a regression, the most interesting part would of
course be comparing to the version before the regression was introduced
(pre-d8fbf9aa85a in this case) to see if the regression was fully fixed.
Maybe the page you linked above does already explain this and I just
don't know where to look? :-)

Kevin

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]