<fijal>
but this is a random old checkout in benchmarks
<cfbolz>
I've never seen this error, doesn't ring a bell. I am not aware of any changes to this stuff in the last few years
<fijal>
there are differences between checkout and what we have in benchmarks
<fijal>
commits don't explain much
<fijal>
*but* I had explosions while translating in benchmarks, but not on checkout
<fijal>
arigato: ^^^
<kenaan>
arigo py3.6 7b256d2286ec /pypy/objspace/std/: Generate a file that hopefully covers all corner cases of startswith() and endswith() on bytes, bytearrays and unicod...
xcm has quit [Ping timeout: 246 seconds]
xcm has joined #pypy
inhahe has quit [Ping timeout: 244 seconds]
inhahe has joined #pypy
<krono>
hi all. I found a `signature.finishsigs` in my codebase. Does that do anything when nobody called `signature.signature` on any of the methods in the first place?
<arigato>
fijal: maybe it was fixed, but already some years ago
<arigato>
there were certainly very-rare random translation crashes
<fijal>
arigato: so what should I do?
<fijal>
assume this one no longer exists or assume this is a JIT bug?
adamholmberg has joined #pypy
<arigato>
no, it's unlikely to be a JIT bug
<arigato>
I'd guess it's more likely to be that you've got a new machine where the randomness is more different from what we tested so far
<arigato>
in fact I remember that one of these bugs was found only on ppc
<fijal>
"ah"
<fijal>
ok, so I think arm64 works
<arigato>
yay
<fijal>
I'm unable to finish benchmark run because of this
adamholmberg has quit [Ping timeout: 246 seconds]
<fijal>
but otherwise it works
<fijal>
I just need to make a docker image, at one stage
<arigato>
(I think the condition should be "being able to translate with everything current", if it works repeatedly then we can definitely say it works)
<fijal>
yeah
<arigato>
how do the other benchmarks look like?
<fijal>
hard to say since after hours of running it crashes
<fijal>
right?
<arigato>
ok, so let me ask about just a couple of benchmarks, maybe in --fast mode
<fijal>
so which one do you want?
<fijal>
richard looks good
<fijal>
I
<arigato>
most interesting would be the compare the speed-ups with the one we get on x86-64
<arigato>
richards is fine
<fijal>
m *also* a bit skeptical because we are on a VM
<arigato>
ok yes
<fijal>
so maybe it's reliable maybe it's not
xcm has quit [Remote host closed the connection]
dansan has quit [Ping timeout: 272 seconds]
xcm has joined #pypy
<fijal>
richards it's 38 times faster on my mac and 27 on ARM64
<fijal>
say chameleon is 7x faster (vs 3x faster on my mac)
<fijal>
django is roughly the same
<Alex_Gaynor>
the same as it is on x86-64, or the same as cpython?
<fijal>
the speedups pypy vs cpython are (with a large variation) similar on arm and x86
<fijal>
but the variance is huge, we don't have a reliable hardware machine
<fijal>
and possibly other reasons
<fijal>
(you can end up with any number you want if you pick your benchmarks correctly)
adamholmberg has joined #pypy
Rhy0lite has joined #pypy
speeder39_ has quit [Quit: Connection closed for inactivity]
xcm is now known as Guest13854
Guest13854 has quit [Read error: Connection reset by peer]
xcm has joined #pypy
dansan has joined #pypy
dansan has quit [Ping timeout: 246 seconds]
forgottenone has quit [Read error: Connection reset by peer]
forgottenone has joined #pypy
dansan has joined #pypy
dansan_ has joined #pypy
dansan has quit [Ping timeout: 268 seconds]
marky1991 has joined #pypy
marky1991 has quit [Remote host closed the connection]
<cfbolz>
fijal: sounds cool, small blog post?
<fijal>
cfbolz: or a big one!
<fijal>
;-)
<cfbolz>
Even better
adamholmberg has quit [Remote host closed the connection]
adamholmberg has joined #pypy
adamholmberg has quit [Remote host closed the connection]
adamholmberg has joined #pypy
marky1991 has joined #pypy
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
xcm has quit [Ping timeout: 248 seconds]
catern has quit [Ping timeout: 245 seconds]
xcm has joined #pypy
<kenaan>
fijal arm64 4bddf4462063 /rpython/jit/backend/aarch64/TODO: close to be merged branch
<kenaan>
fijal default f3b7650ebfc0 /: merge arm64 support
<Alex_Gaynor>
fijal: how many lines of code in the diff were outside rpython/jit/backend/aarch64
<fijal>
Alex_Gaynor: nearly 0 - there is a hack for aarch64 in translationoption which adds 3 lines and a comment
<fijal>
why?
<Alex_Gaynor>
fijal: Hoping for evidence that we designed the JIT backend abstraction well :-)
<fijal>
I can give you some evidence that we didn't ;-)
<fijal>
at least a few things could have been done a level before
<fijal>
and in fact I did one branch in the meantime that did that
<fijal>
moved a few things to rewrite.py, from backend
<Alex_Gaynor>
hehe
<fijal>
so the total amount of changes outside of the backend is >0
<fijal>
*but* you can call it a cleanup
<Alex_Gaynor>
the question is "is there cleanup every single time you do a new backend"
<fijal>
you mean mandatory or "what the fuck I cant handle it anymore" cleanup?
<Alex_Gaynor>
mandatory :-)
<fijal>
no
<fijal>
the thing I did should have been done a long time ago and have been incredibly isolated - it was about clearing memory
<fijal>
so instead of hardcoding a call to memzero in assembler, we now create a CALL_N to memset
<fijal>
or no sorry, about copying contents of unicode
<fijal>
and string
dansan_ has quit [Ping timeout: 248 seconds]
<arigato>
the memzeroing thing is still duplicated in all backends and will be cleanup up during the next one :-)
<fijal>
arigato: so yay, thanks!
<fijal>
seems to be working
<arigato>
great
<fijal>
I left the TODO, all those items are to-be-done-in-the-fullness of time
<arigato>
OK
lritter has joined #pypy
marky1991 has quit [Ping timeout: 245 seconds]
gaze__ has quit [Ping timeout: 248 seconds]
<meredydd>
Hi all! I don't want to interrupt the champagne re aarch64, but I wanted to enquire about the possibility of a bounty/paid work to update the pypy sandbox to Python 3. We use (and love) the v2 sandbox, but EOL is looming and our customers are understandably antsy. Would anyone have any interest in getting the sandbox working on a Python 3 version,
<meredydd>
and if so how would I find such a person?
<meredydd>
(I would volunteer, but I'm busy dragging a different Python compiler kicking and screaming into Python3-land :) )
abrown_ has joined #pypy
<arigato>
hi!
abrown_ has quit [Client Quit]
<meredydd>
Howdy
<arigato>
the pypy sandbox requires some updates, independently of python2/python3; then there might be some issues specific to python3 doing too much stuff to sanely control with the sandbox. All in all I'd think it's not much work to get it back again, but would probably be more work in the long run to continue to have it running
abrown has joined #pypy
<meredydd>
We might be OK with that, depending on the amount of work
<arigato>
cool. maybe you should drop a mail to pypy-z@python.org
<meredydd>
which list is that?
<arigato>
that's our internal list
<meredydd>
I don't see it listed publicly on mail.python.org
<meredydd>
Does it accept external email?
<arigato>
yes, it should. it's where we discuss money-related matters, where we could find if one of the core devs (maybe me) would be willing to help
<meredydd>
Great, thanks. I'll package up my question in an email
marky1991 has joined #pypy
marky1991 has quit [Remote host closed the connection]