These are chat archives for dropbox/pyston

3rd
Feb 2016
An Long
@aisk
Feb 03 2016 02:42
@kmod got it thx
Kevin Modzelewski
@kmod
Feb 03 2016 03:44
This might be silly, but I'm thinking maybe we shouldn't call the rewriter a "mini tracing jit"
and maybe just call it a "per-bytecode JIT"
I'm not really sure it makes sense to talk about method-jit-vs-tracing-jit at this level of granularity
I think it might reduce confusion ("how does this compare to PyPy's JIT") and also make it more clear what it is
Kevin Modzelewski
@kmod
Feb 03 2016 22:50
nice, I'm seeing some performance wins from refcounting :)
right now in the 5-10% range on some microbenchmarks
but the wins seem to get quite large when the heap is large
Rudi Chen
@rudi-c
Feb 03 2016 22:53
Nice!!!
I think that group did 2 or 3 papers in a series on that project
Kevin Modzelewski
@kmod
Feb 03 2016 23:41
oh cool, I'll have to give that a read