github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
Geth MoarVM: 34fac5f4ef | (Stefan Seifert)++ | src/spesh/optimize.c
Enable optimization of speshresolve with ustable guard indexes

Even if there's no clearly winning guard index, we can at least replace speshresolve by sp_speshresolve and get rid of the logging and enable JIT compilation for that frame. The comments suggest that it has always been intended this way and that the return was just an oversight happening due to the 2 halfs of the function being written a month apart.
08:06
nine A new data point: disabling the JIT hurts Inline::Perl5's master branch a lot (~33 %) while it hurts my optimized branch just a little (~13 %). 09:47
And another one: according to perf enabling the profiler reduces time spent in MVM_interp_run by ~ 90 %. And that time consists mostly of entering JITed code. 09:49
lucasb dunno if relevant to you, but I stumble upon a paper in HN: Predicting Variable Types in Dynamically Typed Programming Languages arxiv.org/abs/1901.05138 12:49
dogbert11 M#1035 13:38
synopsebot M#1035 [open]: github.com/MoarVM/MoarVM/issues/1035 Mysterious error when running t/spec/S02-types/sethash.t with FSA_SIZE_DEBUG=1
jnthn lucasb: Using machine learning, eh... :) 15:03
timotimo nine: i can suggest activating that env var that lets the jit create a pid.map file in itmp 19:22
yoleaux 10:55Z <Xliff> timotimo: Do we really want people to be able to set an attribute using the $. form?
timotimo for use with perf
in /tmp*
nine timotimo: oh, how do I then use that with perf? 19:56
Ah, perf report --symfs=/tmp 20:09