cfbolz changed the topic of #pypy to: #pypy PyPy, the flexible snake | IRC logs: and | the pypy angle is to shrug and copy the implementation of CPython as closely as possible, and staying out of design decisions
jcea has joined #pypy
derpydoo has quit [Ping timeout: 256 seconds]
jinsun has quit [Ping timeout: 252 seconds]
jinsun has joined #pypy
jinsun__ has joined #pypy
jinsun has quit [Killed ( (Nickname regained by services))]
jinsun__ is now known as jinsun
jcea has quit [Remote host closed the connection]
jcea has joined #pypy
jcea has quit [Ping timeout: 248 seconds]
<mattip> pypy recieved $50 from "github sponsors", it seems they are distributing money to lots of different orgs
<mattip> I wonder what the criteria were for deciding the amounts
<mgorny> mattip: now you'll have to release pypy3.10 x)
<mattip> mgorny: :)
otisolsen70 has joined #pypy
otisolsen70 has quit [Remote host closed the connection]
otisolsen70 has joined #pypy
<mattip> tumbleweed: I tried the ponyorm tests.
<mattip> We are missing CALL_METHOD_KW, and tests run 3x slower with PyPy
<LarstiQ> huh, a project not compatible with pytest
<tumbleweed> ah, right
jcea has joined #pypy
<mattip> we have CALL_METHOD_KW, but something in the way they use it doesn't work
<mattip> for NumPy, I have spent a few days optimizing a tight loop in C
<mattip> basically doing what the JIT does: eliding constants, creating specialized paths for different types
<mattip> too bad it is C and not Python :)
<tumbleweed> mattip: yeah, their bytecode thing is a bit suspect
<tumbleweed> they compile python to bytecode and decompile it to understand what it is doing and turn it into sql queries
<tumbleweed> but their "evaluation" loop makes some *big* assumptions about what's going on, and doesn't track stack usage very accurately
<tumbleweed> I helped port it to Python 3.11 (which is why it was on my mind) and there was a lot of confusion and hair-pulling in the process
<mattip> well, that means we are somehow not quite compatible with CPython when they handle CALL_METHOD_KW for python3.8
<tumbleweed> they may be making assumptions about the order of bytecode
<tumbleweed> i.e. some optimizations cpython / pypy does break their assumptions
<mattip> we pass most of the tests, I think only one failure is a success
<tumbleweed> yeah
<tumbleweed> my experience was thousands of them failing at a time
<Hodgestar> Are there lots more bytcode changes coming in Python 3.12?
<mattip> so far no
<mattip> here are the faster-cpython ideas though, so there may be more changes
<Hodgestar> Thanks!
<Hodgestar> I can't tell from that page what will actually be in 3.12. It's a mixture of very big projects (e.g. tracing JIT for CPython that can stitch traces together) and very small ones. The better calling convention seems nice. Maybe HPy could collaborate on the better calling convention somehow? Or at least give feedback. I know faster-python is likely mostly thinking about calling out to C code in CPython itself, but it would be good if extensions and HPy
<Hodgestar> could benefit too.
ellemoeiselleva has joined #pypy
ellemoeiselleva has quit [Read error: Connection reset by peer]
jinsun has quit [Ping timeout: 252 seconds]
jcea has quit [Ping timeout: 252 seconds]
epony has quit [Read error: Connection reset by peer]
epony has joined #pypy
<mattip> almost ready to release. Any comments to the release note or blog post are welcome
derpydoo has joined #pypy
epony has quit [Ping timeout: 268 seconds]
srs has joined #pypy
srs has quit [Remote host closed the connection]
epony has joined #pypy
catern has quit [Ping timeout: 260 seconds]
jinsun has joined #pypy
catern has joined #pypy
epony has quit [Ping timeout: 268 seconds]
otisolsen70 has quit [Quit: Leaving]