<antocuni>
LarstiQ: yes, I'll do a blog post but it's too early now, it's not even committed
jacob22 has quit [Quit: Konversation terminated!]
jacob22 has joined #pypy
jcea has joined #pypy
<LarstiQ>
antocuni: I wasn't being serious, but I do look forwards to it
otisolsen70 has quit [Read error: Connection reset by peer]
otisolsen70 has joined #pypy
mjacob has quit [Ping timeout: 240 seconds]
mjacob has joined #pypy
otisolsen70 has quit [Quit: Leaving]
jcea has quit [Ping timeout: 264 seconds]
Atque has quit [Remote host closed the connection]
Atque has joined #pypy
jcea has joined #pypy
Dejan has quit [Ping timeout: 256 seconds]
greedom has quit [Remote host closed the connection]
<antocuni>
what is the standard/best way nowadays to check that a branch doesn't impact performances?
<antocuni>
I guess I need to start some buildbot and then look at speed.pypy.org, but was wondering whether anyone has recent experience with it
<cfbolz>
antocuni: if you break something badly, it will show it
<antocuni>
yes but I wanted to check whether gc-custom-trace-memo affects performance BEFORE merging it
<antocuni>
so I guess the best way is to rebase the branch on top of "pypy3.9-jit-latest" (whatever revision it is) and then compare
<antocuni>
but e.g. it seems that the only way to compare a branch in speed.pypy.org is the "comparison" tab
<cfbolz>
antocuni: yes, as opposed to which tab?
<antocuni>
e.g. "changes"
<antocuni>
with changes, you can see a nice table with green and reds to see whether it's faster or slower
<antocuni>
with "comparison", you get a graph and have to look closely to spot regressions
<antocuni>
anyway, I merged the current py3.9 into gc-custom-trace-memo and started the jit-benchmark-linux-x86-64 builder on it
<antocuni>
I suppose that once it finishes, the result will be automatically available on "comparison" tab?
<cfbolz>
antocuni: yes
Dejan has joined #pypy
mgornyz has joined #pypy
mgorny has quit [Ping timeout: 246 seconds]
jcea has quit [Ping timeout: 272 seconds]
mgorny has joined #pypy
mgornyz has quit [Ping timeout: 246 seconds]
mgornyz has joined #pypy
mgorny has quit [Ping timeout: 246 seconds]
Techcable has quit [Remote host closed the connection]
Techcable has joined #pypy
<antocuni>
so, I have benchmark results. I cannot put a link to show them because the "permalink" feature seems to be broken
<antocuni>
but here is a screenshot:http://antocuni.eu/misc/0xdgLr66.png
<antocuni>
most benchmarks seems unchanged; a tiny minority seem to have some very small speedup, and a not-so-tiny minority seem to have some moderate slow down
<antocuni>
I don't know how to interpret the results: is it a real slowdown or the classical random variations?