Is mprime significantly slower than Prime95?
 

Hi,

I installed Ubuntu 7.04 just couple of weeks ago. Until then I used Windows XP pro edition.

I have noticed that in windows, Prime95 took 0.071 ms per Iteration, while mprime takes about 0.120 ms per Iteration on Ubuntu 7.04, on the same computer. No other modifications has been made in it. The times refer to when the computer is idle.

Does anybody know what's the cause for this much slower calculation time?

Is there any way I could improve that?

Thanks in advanced.

Was the same Mersenne number being tested? (Don't post it.)

What was the version number of mprime and of Prime95?

On typing "top" is mprime using 99-100% of the processor? ("q" to quit.)

Is your computer a multi-core one or one that has hyperthreading?

1) The mersenne numbers are not the same. The one I test with mprime is bigger than the numbers I tested with Prime95.

2) Both are version 2414.

3) No, it uses about 35-49% of the CPU.

4) My CPU is Pentuim 4, one core.

Thanks.

[QUOTE=abcd;113148]1) The mersenne numbers are not the same. The one I test with mprime is bigger than the numbers I tested with Prime95.

2) Both are version 2414.

3) No, it uses about 35-49% of the CPU.

4) My CPU is Pentuim 4, one core.

Thanks.[/QUOTE]

A bigger number will take longer. Please post the two numbers in the form 332xxxxx or whatever, to disguise your numbers but to let us know the FFT sizes used.

Running mprime under linux or prime95 under windows should not make much difference.

49% usage indicates your machine is a hyperthreading one. This can be switched off in BIOS. Doing this is up to you and others here might have more to say on this. Your computer is, however, now being used at "100%". This all assuming there is nothing else processor intensive under "top".

ps. Beware that you will not be able to boot a SMP linux kernel when there is no hyperthreading (nor multi-core.)

[quote]ps. Beware that you will not be able to boot a SMP linux kernel when there is no hyperthreading (nor multi-core.)[/quote]We do it all the time. Most distros ship a SMP kernel by default, to cover everyone.

Single core (obviously!) Prescott P4:

[code]$ uname -a
Linux p4 2.6.18-5-686 #1 SMP Sun Aug 12 21:57:02 UTC 2007 i686 GNU/Linux[/code]

[QUOTE=Xyzzy;113159]We do it all the time. Most distros ship a SMP kernel by default, to cover everyone.
[/QUOTE]

:blush: I made a mistake above; What really went wrong was that I tried to run a non-HT chip with the motherboard in HT mode and the SMP kernel panicked.

I notice that Ubuntu now have a default "generic kernel".

On windows, most of the numbers were between 32**** and 35****.
Now, the number I'm testing is 399****.
Please note that I am not talking about the overall time to test a number, but the time it takes between Iterations.

My knowledge of linux is poor, but i'll try it.

Thanks.

That looks about right.

Go to [URL="http://www.mersenne.org/bench.htm"]http://www.mersenne.org/bench.htm[/URL]
And scroll down.

You will notice that the numbers that you had been testing are in the purple column, while the current number is in the green. The bigger the number the more time per iteration (based upon the FFT size).

Thanks.

Note that the "iteration" in an LLtest of 2^n-1
consists primarily of squaring a n bit number.