On Mon, 23 Dec 2002 20:30:52 +1100, Thorsten Froehlich wrote
(in message <3e06d7cf@news.povray.org>):

> In article <0001HW.BA2D0F4B0002F2930F175220@news.povray.org> , 
> Martin Crisp <Spa### [at] tesseractcomau>  wrote: 
> 
>> 4 machines, 4 identical slowdowns.
> 
> No, that just does not make sense.  And as you have noticed, 

Unfortunately, sense or not, it is what I'm seeing.

BUT: Not with the preemptive threading turned on! [Only just tried 
that.]

The differences it made were:
1) the render progress dialog shows details!
2) I can type this while it is rendering in the background [set on 
lowest throttle, of course]
3) no slowdown after ~30 seconds, updates to the preview are smooth 
and continuous, what I would 'expect', rather than a line exactly 
every 60 seconds.
4) if I wait another minute or few (currently line 216 of 480) I 
predict I'll see a render finish 'cleanly'. It did.
5) a fatal error(-10000) when I quit partway through a render.
6) quitting after a render finished was substantially quicker than 
previous (a few seconds).

> nobody else is having such a problem.  Maybe you have a 

[with preemptive threading off:]
Nobody else is reporting it the same as me, perhaps. If someone's 
menubar clock shows the time in seconds, and it updates for the 
first little while of a render, then stops dead - would they report 
that as a slowdown, or a freeze? Several others have reported 
freezes. Others have reported that renders never start on big 
scenes. Which is what I get if I try to render (2^n)-1 differences 
of spheres nested inside each other using a while loop (and a chunk 
then taken out of all of them by subtracting a box), when n>11. 
(n=11 starts to render after about 20 seconds or so, n=12 doesn't 
after 10 minutes). Spheres inside spheres, radii halving.



> corrupted download or something like that.  Or you got a virus 

I've used 2 separate downloads, one on my local machines, and one 
on the graphic designer's machine (the G4 500 with CarbonLib 1.5). 
I downloaded this latter several days ago, but only unpacked it 
today. The .smi is mounted without reporting a checksum error. 
Corrupted download is very unlikely. I haven't transferred any data 
to that machine from home, so viral contamination is unlikely too. 
[And I've no other 'symptoms' of a virus and the only other 
slowdown I have is my newsreader when it's trying to deal with 
subscriptions with more than 50,000 articles. Or old versions of 
applescript that have memory leaks when they open and close files]


> that attached itself to POV-Ray and thus you copied it each 
> time.  What you report does not even make sense :-( 

If it made sense I (probably - I majored in Comp Sci, graphics and 
digital image processing my best subjects) wouldn't be needing to 
ask.
[But I don't read C very well and haven't kept up with what Carbon 
did to the Mac API, so looking at the source wouldn't help. :-(]

I'm just reporting a problem I have (and from what others have said 
that they may have too, but are describing differently) in the hope 
of finding a solution.

Does knowing it works fine with preemptive threading turned on help 
any?

Have Fun
Martin [much happier :-)]
-- 
Kinky:
What I do that you wouldn't
Perverted:
What you do that I wouldn't

Almost always SMASHed

Post a reply to this message