in reply to use dual-core or quad-core

Basically that's a stupid thing to try to achieve in user space, because scheduling is the kernel's task.

What you can do is start the desired number threads or processes to keep your cores busy, and you can probably the set the priority of your processes. But you shouldn't try to decided which core executes which program. That's your operating system's job.

We could better help you if you told us why your are trying to achieve things like that, perhaps there are better solutions.

Replies are listed 'Best First'.
Re^2: use dual-core or quad-core
by baxy77bax (Deacon) on May 20, 2008 at 09:02 UTC
    well,

    i'm trying to replicate the idea behind the the BLAST program(it is an alignment program - it realy doesn't matter what kind a algorithm it is). the program is written i c and it has an option to use one, tow ... or number of cores. the program calculates the similarity of the alignment. i'm also calculating something similarly to the things BLAST calculates, and in some segments my program is even faster than blast, but i cannot compare myself to BLAST algorithm when multiple core system is turned on. so i'm looking for a way to speed up my performance.

    PS

    i now that perl is not intended for hard core calculations and lot of people told me not to do this in perl, but everything is in the approach.

[reply]
      blast is optimized to hell. If you say you are faster in perl under some circumstances, you are doing something that is not comparable ;)

      Honestly, it makes not much sense to use perl for huge scientific calculations and even less trying it on multiple CPUs. In C (even without OpenMP), you are magnitudes (much more than 2,4,8 times) faster.

      Here are a lot of bioinformatic guys, what do you want to do?

[reply]
        i'm not aligning, i'm not using the heuristics behind finding matches between two sequences. only the calculation after scoring is interesting to me. i gave the example , because this was the only program i could use to see differences between parallel and linear computing

[reply]
      I think the "number of cores" is just the number of parallel threads or processes used, which shouldn't be too hard to in perl.

      Beware that threads can have nasty pitfalls in perl, so if you don't have a lot of data to share between the parallel calculations you could try to start separate processes instead.

[reply]
[reply]

In Section Seekers of Perl Wisdom