cfbolz changed the topic of #pypy to: #pypy PyPy, the flexible snake | IRC logs: and | so many corner cases, so little time
Aksam has joined #pypy
<Aksam> cfbolz good day, As I was reading through the Cross-layer workload characterization of meta-tracing JIT VMs ,I see your were part of the research. Are the artifacts (source code) on the research still in available ,if yes kindly share with me.
<cfbolz> Aksam: I don't have it, no. Probably Berkin Ilbeyi, you should send him a mail
<cfbolz> Aksam: as for your PYPYLOG question, it would help me to know what precisely you are trying to do, to be able to answer it
<Aksam> cfbolz I want to get the deeper understanding how PYPYLOG gets the execution time for the different phases in the JIT
Aksam has quit [Ping timeout: 256 seconds]
Aksam has joined #pypy
<cfbolz> Aksam: the JIT? Or the GC?
<Aksam> GC
<Aksam> cfbolz GC phase
<cfbolz> That's different though
<cfbolz> There is no complete infrastructure for that
<cfbolz> Aksam: also i would be interesting in understanding *why* you are looking into GC time
<Aksam> I want to undestand the GC overhead during tracing. I know that during tracing we have a GC phase according to that cross-layer characterization paper
<cfbolz> Ok
<cfbolz> Aksam: I think the logs contain the proper nesting of the two, but I'm not 100% sure
<Aksam> cfbolz Where can I find the full documentation for logs .The one pypy has few details where I believe can't help me to find exactly what I need .Thank you again
<cfbolz> Aksam: there is no proper documentation. It's very much an internal feature. You can read the script that is mentioned in the blog post that I linked
<fijal> Aksam: I think we are also happy to answer questions
<fijal> I *think* the GC would show as the lowermost layer, but the GC is incremental, so you can't quite ascribe the "gc caused by the JIT" in a foolproof way
<fijal> (there are a few other complicated reasons I think why you can't)
<cfbolz> Indeed, but starting with 'this GC phase was done during the jit' is a starting point
<cfbolz> (we should rescue the pictures of that blog post somehow)
Aksam has quit [Quit: Client closed]
Aksam has joined #pypy
Aksam has quit [Ping timeout: 256 seconds]
<cfbolz> mattip: I am tempted to just merge speedup-dict-copy. the code is fine and makes dict.copy faster (it just didn't solve the other 3.9 issues, but I am doing them in other ways)
Aksam has joined #pypy
Aksam has quit [Quit: Client closed]
nimaje has quit [Remote host closed the connection]
nimaje has joined #pypy
<mattip> cfbolz: sounds good to me
<cfbolz> mattip: ok, will wait till tomorrow to get another nightly in
<mattip> I am updating stdlibs to 3.7.13, 3.8.13, 3.9.12. Let's see what new failures there are
<cfbolz> mattip: right, good
<cfbolz> so so I'll wait for a run of that, so we don't have too many possible sources of new problems
<mattip> cool, thanks