Dual core anyone?

[Prime95]

Do any of you mersennaires use a dual-core machine runniing Windows? If so, I'd like you to test drive a new version. This version would run two threads. If you give prime95 900MB to use and both threads are in stage 2, they each get 450MB. If only one thread is in stage 2 it gets the whole 900MB.

[James Heinrich]

I can pull my Opteron165 off LMH duty and test your new version for a week (unfortunately I'm selling it next Wednesday).

[Prime95]

No need. I've done some trivial testing here. If someone happened to be using a dual core machine, then some real-world feedback might have been useful.

[James Heinrich]

Well, if you want some real-world feedback over the next 6 days or so, you know how to reach me :)

[drew]

Originally Posted by Prime95

No need. I've done some trivial testing here. If someone happened to be using a dual core machine, then some real-world feedback might have been useful.

I just noticed this thread. I've got a 2.16 GHz Core Duo laptop if you're still looking for someone to test your code.

[Prime95]

Grab ftp://mersenne.org/gimps/p95tst.zip This is a pre-pre-pre-alpha v25 version. Its ability to read v24 save files and v24 worktodo files is untested.
It is probably safest to run it in a brand new directory. Create a worktodo.txt file. Tell it how much memory to use and threads to run (Test/Worker threads dialog). Let it rip.

It should be good about reducing stage 2 memory use when both threads are running stage 2 at the same time. It is less good about switching back to max memory use when one ends stage 2. Let me know when you find any problems or have suggested improvements.

[drew]

Originally Posted by Prime95

Grab ftp://mersenne.org/gimps/p95tst.zip This is a pre-pre-pre-alpha v25 version. Its ability to read v24 save files and v24 worktodo files is untested.
It is probably safest to run it in a brand new directory. Create a worktodo.txt file. Tell it how much memory to use and threads to run (Test/Worker threads dialog). Let it rip.

It should be good about reducing stage 2 memory use when both threads are running stage 2 at the same time. It is less good about switching back to max memory use when one ends stage 2. Let me know when you find any problems or have suggested improvements.

Will do.

Just to add my two cents...for normal usage, would a better strategy simply be to try to avoid having concurrent P-1 factoring? If Prime95 predicts that two tests will finish at the same time, it can take a break from one, perform a P-1 on the next test in the queue, then go back to the first test before the second test finishes.

Drew

[drew]

Originally Posted by Prime95

Grab ftp://mersenne.org/gimps/p95tst.zip This is a pre-pre-pre-alpha v25 version. Its ability to read v24 save files and v24 worktodo files is untested.
It is probably safest to run it in a brand new directory. Create a worktodo.txt file. Tell it how much memory to use and threads to run (Test/Worker threads dialog). Let it rip.

It should be good about reducing stage 2 memory use when both threads are running stage 2 at the same time. It is less good about switching back to max memory use when one ends stage 2. Let me know when you find any problems or have suggested improvements.

Ok, first bug.

It tried to contact Primenet to get new work, and Got the following message:

PrimeNet error 7: Invalid parameter
parameter c: Invalid character: '@' in 'Genuine Intel(R) CPU T2600 @ 2.16GHz'

If I can't get the Primenet interface working, is there a pair of P-1 factoring test someone can recommend for me? Also, George, how do I configure the worktodo.txt to perform two P-1 tests on separate threads?

Thanks,
Drew

[Prime95]

Originally Posted by drew

Ok, first bug.

It tried to contact Primenet to get new work, and Got the following message:

PrimeNet error 7: Invalid parameter
parameter c: Invalid character: '@' in 'Genuine Intel(R) CPU T2600 @ 2.16GHz'

Yeah, the server access isn't ready for testing. I posted here because the Mersenne-aires run manual tests.

If I can't get the Primenet interface working, is there a pair of P-1 factoring test someone can recommend for me? Also, George, how do I configure the worktodo.txt to perform two P-1 tests on separate threads?

This ought to work:

Pfactor=exp,how_far_factored,double_check_flag
[thread #2]
Pfactor=...

P.S. Thanks for the bug report. I'll fix that one.

[drew]

Originally Posted by Prime95

Yeah, the server access isn't ready for testing. I posted here because the Mersenne-aires run manual tests.



This ought to work:

Pfactor=exp,how_far_factored,double_check_flag
[thread #2]
Pfactor=...

P.S. Thanks for the bug report. I'll fix that one.

Ok, that worked, but now I'm stuck in a loop (both threads behave the same way):


Optimal P-1 factoring of M28515913 using up to 1024 MB of memory.
Assuming no factors below 2^68 and 1 primality tests saved if a factor is found.
Optimal bounds are B1=150000, B2=2925000
Chance of finding a factor is an estimated 3.59%
Using FFT length 1536K
Available memory is 512MB
using 478 MB of memory.
Available memory is 1016MB.
Using 957MB of memory.
Restarting thread with new memory settings.
Optimal P-1 factoring of M28515913 using up to 1024 MB of memory.
Assuming no factors below 2^68 and 1 primality tests saved if a factor is found.
Optimal bounds are B1=150000, B2=2925000
Chance of finding a factor is an estimated 3.59%
Using FFT length 1536K
Available memory is 512MB
using 478 MB of memory.
Available memory is 1016MB.
Using 957MB of memory.
Restarting thread with new memory settings.
.
.
.
ad infinitum

[Prime95]

Are they really stuck or are they making progress toward completing stage 2?

I'm changing the "Using nMB of memory." output to "Using nMB of memory. Processing x relative primes (y of z already processed)."
This should make it clear that forward progress is being made.