cfbolz changed the topic of #pypy to: #pypy PyPy, the flexible snake https://pypy.org | IRC logs: https://quodlibet.duckdns.org/irc/pypy/latest.log.html#irc-end and https://libera.irclog.whitequark.org/pypy | the pypy angle is to shrug and copy the implementation of CPython as closely as possible, and staying out of design decisions
itamarst has joined #pypy
itamarst has quit [Quit: Connection closed for inactivity]
BarrensZeppelin has joined #pypy
<BarrensZeppelin> cfbolz The blog post on the known bits abstract domain was a nice read. 🙌
<cfbolz> BarrensZeppelin: thank you! I'm glad someone's reading it 😅
<BarrensZeppelin> 🙂
<BarrensZeppelin> Do you have an intuition about how the new optimizations will affect PyPy performance?
<BarrensZeppelin> Of course it's application specific, but if the interpreter code itself uses lots of bitwise operations (for some reason / maybe for GC?), it could yield nice speedups for most applications.
BarrensZeppelin has quit [Quit: Client closed]
<xorAxAx> yeah, nobody except RSqueak uses that seriously, but dont tell anybody :)
<cfbolz> BarrensZeppelin: it doesn't really help most python applications, unfortunately
<cfbolz> I've mostly done it for https://docs.pydrofoil.org/en/latest/
<mjacob> cfbolz: i didn't know about pydrofoil. nice!
<mjacob> cfbolz: do you know who uses it?
<cfbolz> mjacob: probably no one, to a first order of approximation 😂
<cfbolz> mjacob: it's my pet project, I have some more features planned that could make it more useful (like python scriptability)
<cfbolz> it's fairly far though, can boot linux on both arm64 and risc-v64
<mjacob> more useful for those who are disappointed by the 18x speedup?
<cfbolz> mjacob: it's 18x compared to the extremely slow sail baseline interpreter. qemu is way faster
<nikolar> cfbolz: sorry, i wasn't here earlier, what exactly is your project
<mjacob> do you know how it compares to the approach chosen by pydgin?
<cfbolz> mjacob: pydgin is probably faster, but it's a user mode only emulator
<cfbolz> mjacob: pydrofoil supports the various exception levels, and all the (in arm64 extremely complex!) details of address translation
<cfbolz> I have a todo to also add support and tests for the hypervisor extension of risc-v
<nikolar> cool project cfbolz
<cfbolz> htx
<cfbolz> thx
Dejan has joined #pypy
<larstiq_> ooh, I have two blogposts to read
<cfbolz> hehe, that's a very indirect way to add to people's todo list
<larstiq_> well, I do enjoy your posts so it's not a todo in that sense :)
<larstiq_> it's just that I'm currently working so will take half a day before I can
Diggsey has joined #pypy
jcea has joined #pypy
Dejan has quit [Quit: Leaving]
<korvo> cfbolz: I haven't dug into the source; do you do knownbits for bigints too, or just for int32/int64?
<korvo> knownbits for bigints could possibly speed up algos which use bigints as bitsets.
<cfbolz> korvo: just word-sized ints
<korvo> cfbolz: Maybe it's time to revive tagged pointers?
<korvo> Also I should add pydrofoil to rpypkgs.
<cfbolz> tagged pointers, unlikely. it's a mess
<cfbolz> korvo: just as a warning, pydrofoil-arm is a real beast to build
<korvo> cfbolz: Understandable. I don't think I have ARM machines for that.
<cfbolz> no, pydrofoil-arm is version where arm is the guest CPU
<cfbolz> it's also a beast to build on an x86 host
<korvo> What's the constraint? I have a chonky second-hand workstation with 150 GiB of RAM, and Nix should make dependency management easy.
<cfbolz> if you have that much RAM it's ok. will just take a while (about 4x longer than a pypy build)
itamarst has joined #pypy
<korvo> Noted, thanks. I appreciate the warning.
jcea has quit [Ping timeout: 252 seconds]
otisolsen70 has joined #pypy
Diggsey has quit [Quit: Connection closed for inactivity]
otisolsen70 has quit [Read error: Connection reset by peer]
Dejan has joined #pypy
jcea has joined #pypy