[Gcl-devel] Re: Windows random test results

[Mike Thomas]

Hi again.

During the cleanup last night I copped trouble from my last minute 
modifications, so I've reverted CVS to a known stable -O3 level.

The random tester running on that stable CVS version of GCL between 
about 00.40 and 07.20 this morning did 7000 iterations on an 867MHz 
512KB PIII under Windows NT2000 Professional:
=====================================================================
CL-TEST>(loop-random-int-forms 1000 8)
     1 | 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 
1500 1600
1700 1800 1900 2000
     2 | 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 
1500 1600
1700 1800 1900 2000
     3 | 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 
1500 1600
1700 1800 1900 2000
     4 | 100 200 300 400 500 600 700 800 900 1000
=====================================================================

I suspect from reports on the mailing list that this is much slower than 
the Linux version and would be interested to see results for the same 
command, version etc on other OS/CPU combinations.  The command is 
precisely that sent to me by Camm a few days ago.

Memory usage is reported as nearly 84,000 KBytes and VM size is 156,736 
KBytes on the Windows task manager so there has been some growth through 
the night, but I don't know whether this is to be expected or whether it 
signals memory leak problems.  Similar OS space monitor data from Linux 
etc would likewise be appreciated if someone can find the time.

Thinking some more I was running a Maxima build and a number of other 
tasks when I measured last night so I suppose it's not suprising that 
the ratio of memory usage to VM size is substantially different. 
Never-the-less the VM size is about 50% larger all up.

I'm hoping to let this run all day to get a really solid feel for 
stability on Windows.

Despite the late night last minute problems I am still feeling very good 
about this.

Mike Thomas


Mike Thomas wrote:
> Hi all.
>
> I've checked in the changes to fix the object initialisation bug on 
> Windows in stable branch.  While doing the code cleanup I had the 
> preliminary stable branch -O3 GCL running the HEAD random tester and 
> here are the results so far:
>
> CL-TEST>(loop-random-int-forms 1000 8)
>      1 | 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 
> 1500 1600 1700
>
> Memory usage constantly hovers between 32,000 and 33,000 KBytes while 
> the VM size is stable on 106,168 KBytes.
>
> A concurrent CVS Maxima build at the object file load stage was around 
> 38,000 and 91,220 KBytes respectively.
>
> CVS Maxima builds and passes all tests.
>
> I'll aim to leave a random tester going tonight once I've rebuilt the 
> new version but it all looks good from here.
>
> Camm, is the console output the only place I need to look for errors 
> from the random tester?
>
> Cheers
>
> MIke Thomas.