HTT/SMP

[Stephen Montgomery-Smith]

Kris Kennaway wrote:
> On Mon, Feb 28, 2005 at 06:39:21PM -0600, Stephen Montgomery-Smith wrote:
> 
>>I have a dual Xeon system, for which I had hyperthreading enabled.  I 
>>was running a CPU intensive floating point program (specifically using 
>>the fftw3 port with threads).  I found that using 4 (hyper) processors 
>>really did give me a speed advantage, even though I do find web sites 
>>that suggested this advantage was minimal.  (So for example, 4 processes 
>>on this system seemed to run at the equivalent of 2GHz on these 3GHz 
>>computers, so the advantage was about 1.3 times.)
>>
>>But then after about a day the computer had a panic.  I found that 4 
>>CPU's under FreeBSD 5.3 is apparently unstable, and there were 
>>statements to the effect that this would be fixed in FreeBSD 5.4.  Is 
>>this going to happen?  I am currently running a rather recent version of 
>>FreeBSD 5.3-Stable, maybe a week or two old, so I am concerned that 
>>maybe this won't get committed.
> 
> 
> Depends on the panic.  What was it?
> 
> Kris

I'm sorry, but I didn't write it down.  Something about a page being 
messed up or something.  Since I had already seen messages to the effect 
that HTT on dual processors had problems, I assumed that it was just 
that.  I can tell you that I now have HTT disabled, and it is not 
panicing at all.

Was there some fix committed (which would suggest that maybe this is 
some other problem)?  If so, I could reenable HTT and wait for another 
panic to occur, and then write it down.

Thanks, Stephen