I claim at least a part of the jitted code change is a fluke. A part is due to disabling of retracing, which never worked properly as far as I know anyway, we'll reenable that somehow at some point. I would not put it in the release announcement.
On Fri, Oct 16, 2015 at 2:33 PM, Matti Picus <matti.pi...@gmail.com> wrote: > On 16/10/15 15:23, Phyo Arkar wrote: >> >> warmup time improvement of 20% or so at the cost of a minor regression in >> jitted code speed. >> >> 20% Warmup is huge! That will skyrocket in bencmarks. >> >> But Minor regression? how minor ? >> > > speed.pypy.org (which measures only jit performance without warmup) went > from 7.10 to 6.90, but of course that is just for those benchmarks > > > Matti > _______________________________________________ > pypy-dev mailing list > pypy-dev@python.org > https://mail.python.org/mailman/listinfo/pypy-dev _______________________________________________ pypy-dev mailing list pypy-dev@python.org https://mail.python.org/mailman/listinfo/pypy-dev
