#techempower-fwbm

[FrameworkBenchmarks] nateberkopec opened pull request #2849: All Ruby frameworks use Puma 3.9.x (master...pumaver) https://git.io/vH6JR

[FrameworkBenchmarks] nateberkopec opened pull request #2850: Rails: Auto-tune unicorn and puma (master...rails-auto-tune) https://git.io/vH6kG

dantti: are you around?

yup

ive got some cutelyst data for you

:)

however, we're still seeing some weird mysql performance degredation

so keep that in mind for your mysql tests

(and if you have any input for that there's an open issue on github right now)

right, I'm more of a pg guy :P

i can get you a couple output logs for some tests if you'd like to see them. let me know. got some other things going on at the moment

interesting it seem that the half tests were mostly worse

on prefork mode they improved performance :P

well like i said, ignore the mysql stuff. otherwise, the rest fall within our acceptable range, mostly +/- 3%

half = cpu_cores/2 and each thread pinned to 2 cpus trying to render HT unusable

I think I'll make a PR now to update Qt, it's likely to improve a bit due some code that is atomic instead of using mutexes

thanks

woot

did you finally get a benchmark run done?

RX14: "finally"? ouch

well it's a hard thing to do

wranging so amny different things

that all have to work

wrangling*

haha it is it is, we have a full run in progress, but there is serious mysql issues

hmm :(

posted open issue on github

yeah i saw

what was the framework you were working on?

crystal mostly

none of the tests use mysql

oh good stuff with crystal

oh nice

i believe they were largely affected by the ulimit increase

yeah i saw some earlier results

again?

i could imagine that

there was a huge bump by my patch

but i didn't reslise ulimit would effect it too

i dont know if this is related to what you already saw:

crystal-rawjson24,231 580,743 +2,296.7% crystal-rawplaintext 47,969 2,379,579 +4,860.7%

yeah i saw similar before ulimit i think

you posted a partial run's json before

which I had a peek at

happened to have crystal in because it's early in the alphabet

ah ok cool

not sure if we'll post this either, unless it's to get community attention for the mysql issues but im tracking that now

yeah the previous results were 2 380 585 and 519 321

so the JSON seems to bave done better

by 12%

on the latest results

it would be nice to see the json/plaintext ranking though to see where crystal comes in

as that's not affected by the mysql issues

it's currently 5th in json and 4th in plaintext but only at 'h' tests in the run

oh it's not complete?

no this one is still running after yesterdays toolset updates

takes about 3 days

i didn't realise it dumped the JSON before the run completed

each framework creates it's own json file, and we have an inhouse tool that merges them and keeps us updated

ah cool

[FrameworkBenchmarks] nbrady-techempower opened pull request #2851: MySQL Update to 5.7 (master...mysql-5.7) https://git.io/vHikl