06:43 <mattip> any comments/thoughts on the M1 blog post?

06:43 <mattip> fijal: ^^^

06:45 <fijal> let me have a look

06:45 <fijal> I mean, it's very minimal :)

06:46 <fijal> maybe it's worth noting that arm64 support has been around for a while (sponsored by RPi foundation?), but there are a few interesting OS X specifics that needed adressing

06:46 <fijal> (but I'm also ok if not)

06:58 <mattip> thanks, I will add that

07:30 <cfbolz> fijal: the arm64 support was sponsored by crossbar and arm ltd, I think

07:30 <cfbolz> https://www.pypy.org/posts/2019/07/pypy-jit-for-aarch64-7161523403247118006.html

08:07 <mattip> I added a mention, without the sponsor. How does it look now?

08:13 <fijal> ok

08:13 <fijal> mattip: seems to work, more or less?

08:34 <mattip> fijal: what seems to work, the blog post or to the M1 port?

08:34 <mattip> if you are OK with the post I will publish it soon

08:34 <mattip> agreed it is more of an extended tweet than a real blog post, but there doesn't seem to be more to say

08:56 <fijal> yes I'm ok with the post

08:56 <fijal> arm64 seems to more or less work

08:56 <fijal> mattip: I agree

09:20 <mattip> ok, publishing

09:22 <cfbolz> awesome

09:22 <cfbolz> will tweet

10:50 <j4at> mattip: the opencollective url is incorrect

10:55 <cfbolz> Ah, an extra ')'

11:09 <cfbolz> j4at: thanks a lot, I pushed a fix

11:10 <j4at> ;)

12:24 <j4at> cgbolz: does pypy have logic to disable inlining locally or does it only disable it globally ?

12:31 <j4at> cfbolz *

13:14 <cfbolz> j4at: so by default there is some heuristic for inlining decisions

13:15 <cfbolz> I'm not sure what the hooks expose

13:15 <j4at> the hooks doesn't expose anything other than stoping inlining globally

13:15 <j4at> Can't do anything with that

13:21 <j4at> I will do retracing seeems easier with what is exposed rn. I learned how to read the operations kinda I will try to use that. Needs per loop/bridge `get_stats_snapshot()` tho.

13:27 <j4at> Without retracing If I use `max()` builtin function with a list in warmup. max with tuples becomes 2x slower.

13:27 <j4at> there is a huge performance lose there

17:28 <krono> Nice blogpost, everyone. Kudos for the work!

21:52 <antocuni> I'm reading a lot about WebAssembly nowadays, and I stumbled upon this page: https://github.com/WebAssembly/design/blob/main/Tooling.md

21:52 <antocuni> it's nice to see that pypy is explicitly mentioned as one of the examples of JITs