develooper Front page | perl.perl6.internals | Postings from February 2002

Re: GC performance

Thread Previous | Thread Next
From:
Bryan C. Warnock
Date:
February 27, 2002 20:36
Subject:
Re: GC performance
Message ID:
200202280433.g1S4XqQ27295@smtp.capita.com
On Wednesday 27 February 2002 20:57, Dan Sugalski wrote:
> I presume the ### lines were the two that took longest, at 21 seconds
> each, more or less?

Yeah, as flagged by WorkShop.

>
> I think some sort of "X  full memory allocations per collection"
> scheme would be a good thing, and tuning the DOD side of things would
> definitely be in order.

Haven't found the problem yet.  I did some walking through the debugger, and 
it looked like you were consistently reclaiming 90 entry slots on DOD, and 
29K on GC.  So then I went back to a call count, and here's what I found.

life.pasm

Generations DOD GC
------------------------
   5      133      10
 100    8,644     279
 500   91,310   4,214
1000  265,294  20,615

I did a graphical mapping of the DOD and GC calls, and the GC pattern was 
interesting.  (Indicative of a leak.  I'm going to patch the output to show 
a generation loop, and then post and interpret.)

-- 
Bryan C. Warnock
bwarnock@capita.com

Thread Previous | Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About