epsilon-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[epsilon-devel] eAM: how big is garbage collector impact?


From: Luca Saiu
Subject: [epsilon-devel] eAM: how big is garbage collector impact?
Date: Fri, 21 Nov 2003 01:18:05 +0100
User-agent: Mozilla/5.0 (X11; U; Linux i586; en-US; rv:1.3a) Gecko/20021212

Hi Matteo. I was trying a variant of the 'same old' :-) screen hack on a slow machine that I normally use as a terminal; it's a Pentium/120MHz. Your laptop is very fast, but however the same effect I saw might be noticeable. Could you please try modifying eam/gc/gc.c replacing the two #undef's near the beginning with #define's? Then you should compile, install, recompile the screen hack, run it and look at the garbage collector output. Which is usually reported as larger, "marking time" or "sweeping time"? Is there any noticeable delay associated to collections? On your machine probably not, and maybe not even on my laptop, which I haven't now (it's still being repaired); but there is some (the application stops for about 250 milliseconds per collection) on this slow machine. At least on this machine the sweep time strongly dominates. If, as I think, the program behaves in the same way also on your machine then I should implement "mark and don't sweep"; that should dramatically reduce sweeping time (even if not marking time).

  Thanks in advance.

--
Luca Saiu, maintainer of GNU epsilon
http://www.gnu.org/software/epsilon





reply via email to

[Prev in Thread] Current Thread [Next in Thread]