cfbolz changed the topic of #pypy to: PyPy, the flexible snake (IRC logs: https://quodlibet.duckdns.org/irc/pypy/latest.log.html#irc-end ) | use cffi for calling C | if a pep adds a mere 25-30 [C-API] functions or so, it's a drop in the ocean (cough) - Armin
marky1991_2 has joined #pypy
marky1991 has quit [Ping timeout: 256 seconds]
jcea has quit [Ping timeout: 260 seconds]
marky1991_2 has quit [Ping timeout: 240 seconds]
lobo_ has quit [Ping timeout: 256 seconds]
lobo_ has joined #pypy
tsaka__ has joined #pypy
Dejan has quit [Quit: Leaving]
oberstet has joined #pypy
jcea has joined #pypy
<mattip>
cfbolz: is map-improvements ready for merging?
<cfbolz>
not
<cfbolz>
I haven't done a benchmark run
<cfbolz>
and I need to hunt a bit to see whether the theoretical bad cases occur in practice
<mattip>
under what conditions would you expect them to occur?
<cfbolz>
mattip: it's a bit like with list strategies
<cfbolz>
If you make a huge amount of instances, where all attributes 'a' are ints
<cfbolz>
And then you change one to a float, stuff becomes slower
<cfbolz>
mattip: also, should we merge it if it helps none of our benchmarks (but does help pierre's stuff?)
<antocuni>
it might be worth making pierr's stuff an official benchmark
<antocuni>
it's a reasonable use case
<cfbolz>
ok, but it's also a bit cheating, because we do extremely well on it ;-)
<antocuni>
if we include only the benchmarks in which we are bad, we will give the wrong impression in the other way
<cfbolz>
it's a problem that the benchmarks really have two purposes, marketing tool and a way for us to check our progress
<antocuni>
yes
<simpson>
cfbolz: I think you nailed it. Like, is a benchmark a way to connect with users and test their desires, or is it a way to compare Python implementations?
<antocuni>
but also from the non-marketing point of view, it's good to have "good benchmarks", because they tell us if we introduce any regression
<antocuni>
else it's too easy to disable some optimization and say it doesn't matter because benchmarks are not affected
<cfbolz>
that's fair
<cfbolz>
(but I suggest we wait until pierre has a paper version, so we can assume the code is final)
toad_polo has quit [Quit: Bridge terminating on SIGTERM]
astrojl_matrix has quit [Quit: Bridge terminating on SIGTERM]
panosl[m] has quit [Quit: Bridge terminating on SIGTERM]