getting slower on expansion

0 views

Skip to first unread message

Marco Maggi

unread,

Jun 13, 2011, 7:41:00 AM6/13/11

to mosh-develo...@googlegroups.com

I did the hard-disk-shaking fetch of "master" branch and I
get the feeling that expanding libraries has become a bit
slower with "mosh". I this supposed to happen?

I will have to do some preparation to perform an actual
measurement.

TIA
--
Marco Maggi

higepon

unread,

Jun 13, 2011, 9:50:47 AM6/13/11

to mosh-develo...@googlegroups.com

Hi.

> get the feeling that expanding libraries has become a bit
> slower with "mosh". I this supposed to happen?

Thank you for your letting us know about it.
The last drastic change we've done is "annotated pair optimization",
but I don't think it made only "mosh" slower.

> I will have to do some preparation to perform an actual
> measurement.

I really appreciate your work.
Cheers.

Taro.

okuoku

unread,

Jun 13, 2011, 2:56:07 PM6/13/11

to mosh-develo...@googlegroups.com

2011/6/13 Marco Maggi <mrc...@gmail.com>:

Several (possible) reason:
- We had bumped up Boehm GC(Garbage collecter) version in this June.
The last commit which has old GC is
8ed45dd5345cb1950d63be4667fde0dfbd4203b4
Current GC operation is now a bit more precise and it could take more
time to work.
(Especially if you see "finalization cycle" message from GC -- these
objects are just leaked before.)

- You cannot measure expansion time with "mosh --disable-acc" because
this option is unreliable.
Please use "nmosh --disable-acc" instead or remove ~/.mosh every time
of your measurement.