On Fri, 17 Sep 1999, you wrote:
> Hello all!
>
> I've taken the plunge and picked up a pair of Celeron 400's and an Abit
> BP6 board. This is my first foray into the wild yonders of SMP, so
> forgive my naive questions.
>
> I was under the impression that since the kernel and libc libraries
> natively support SMP, applications would automatically make use of both
> processors without intervention. I'm wondering whether I was incorrect.
>
> As a test, I ripped an audio track from a CD and started 'notlame' (also
> tested 'bladeenc' with similar results) to encode the wav file. Using
> top, the encoding process never used more than 50% of CPU time. This
> makes me think that it's only making use of one processor. Even using
> 'nice -19 <blah>' never yielded much more than 50% CPU.
>
> Was I naive to think that applications "auto-magically" benefit from
> SMP? Anyone know of an MP3 encoder that will make use of the second
> processor?
Steve,
Were you using Top to see application activity? I think you probably were.
When Top shows 50% CPU being used in a dual CPU system, that means that the
ripper/encoder were using all of 1 CPU. KPM shows things a little differently.
Anyways.. To answer your question. If you want a single application to use
both CPUs, it has to be capable of multi-threading. Without that, the program
can never use more than 1 CPU at a time. I'm not aware of too many
applications that can do that. I know Photoshop when used with NT can use more
than 1 CPU.
But, there is an advantage.. Take the CD Ripping and Encoding for example. I
use Krabber with CD Paranoia and Bladenc. Krabber allows you to start more
than 1 encoder at a time. So, I have it start 2. That way, the CD is done
ripping in 1/2 the time. Or, I can run Seti@HOME reniced at 20 and still have
plenty of horsepower left to do other things so that it gets full use of the
other processor. I can also compile apps and still do other things while the
app compiles.
Darin -
--
Cthulhu for President in 2000 - Why settle for the lesser evil????
