| From: | Oliver Roberts |
| Date: | 16 Aug 2000 at 13:09:33 |
| Subject: | Re: Multitask lock bug (was Re: Questions) |
On 10-Aug-00 06:59:56 BST, Mark Olsen wrote:
> On 09-Aug-00, you wrote:
>>> I've tried a 240 mhz board on my computer instead of my normal 160 mhz
>>> 603e(overclocked to 200 mhz), and guess what. The infamous multitask
>>> bug occurs quite often. And i even have a ventilator standing on my
>>> motherboard blowing down on the PPC cpu.
>> So WHAT is the conclusion? Is there a simmilar bug on both processors?
>> Someone!? I'm going crazy :)
> The conclusiong is that it seems not to be a software problem. As the one
> board works where there other fail, in the same software environment, it
> must be something else. And it is /NOT/ cooling.
I'm not so sure that it's not a software problem - it could be a bug in
WarpOS, and it's just that the bug does not cause problems on all systems,
but if the bug was fixed, WarpOS would then work on all systems. Just a
theory :)
I have a 240MHz 603e board, and I generally don't have any problems.
However, under heavy WarpOS multitasking, WarpOS just seems to lock up.
There are no crashes, hits, exceptions or anything - WarpOS just seems to
get stuck and doesn't respond. If you try to run any PPC program after
this has happened (such as stat), it just sits there and does nothing.
One way which is almost guaranteed to reproduce this problem, is to
create a script with about 100 "run multiview test.jpg" lines in it - i.e.
tries to a display an image 100 times, with each process running in
parallel. With my WarpJPEG and/or WarpPNG datatypes, this will usually
result in WarpOS coming to a standstill, as above. I notice that the
ak* datatypes appear to workaround this problem, by only allowing one
decoder to run concurrently (whereas my datatypes may well have 100
decoders/tasks running concurrently, in the this case).
Has Sam Jordan made any comments about this problem?