Prime95 version 25.7

[WraithX]

Originally Posted by Prime95

That all looked good. Workers always use 1 thread. Since you selected multithreading using 5 threads you get the 1 worker thread and 4 helper threads.

Ah, ok, thank you. Since this was my first time working on an exponent, I thought it strange that I was given an LL test and it started doing P-1.

BTW, is there a way to tell it to stop working (not get additional assignments) once it has finished working on the current assignment?

[Mini-Geek]

Originally Posted by WraithX

Ah, ok, thank you. Since this was my first time working on an exponent, I thought it strange that I was given an LL test and it started doing P-1.

BTW, is there a way to tell it to stop working (not get additional assignments) once it has finished working on the current assignment?

Advanced > Quit GIMPS
then select the right option to finish your current work and then stop.

[WraithX]

Originally Posted by Mini-Geek

Advanced > Quit GIMPS
then select the right option to finish your current work and then stop.

What happens if I set, under Preferences..., "Days of work to queue up" equal to zero? Will that still try to queue up work after the current job is done? I'm not looking to "Quit GIMPS"/have my computer removed from the server. I'd just like to decide what I want to spend my resources on once this test is done.

[James Heinrich]

Originally Posted by James Heinrich

If one (or more) cores are idle, and another core is working on something that can reasonably exploit more multithreading, respawn the other work with the appropriate number of threads to fully utilize the CPU.

To illustrate, right now only 3 of 4 cores are being used on my machine:
#1 = ECM, now idle
#2,3 = LL, 2 threads
#4 = P-1

The P-1 worker is using all available RAM, so ECM eventually ran out of stage-1 work, so it's sitting idle until #4 finishes stage2. It would be more productive to either:
a) respawn #2,3 as 3-threaded or respawn #4 as 2-threaded
b) let #1 start working on some other low-memory task from another worker (stage1 of the next P-1, or start on the next LL test, etc).

I suspect that option (b) is a more efficient use of the CPU, multi-threaded workers tend to have far less than 100% efficient use of additional threads, especially past 2-threaded.

[adminc]

Im running Ubuntu Hardy x64 bit.

not sure if this is what is supposed to happen but when first running the program, it seems to ask twice for the memory limits. It first asks for the limits, and then says 8mb of memory is not enough, asks me for the mem limits again and then gets them right.

Secondly, if I close the prime95 process (by typing 17 i think) and then try to run it again, I dont get the console output I get when I run it the first time.

Also when asking for the priority at which I want t run, it gives bounds from 1 - 10. On Linux though I believe its actually called the "nice" value and functions differently than in windows. Perhaps you could look into this.

[db597]

Is there any migration plan for those on 24.x? I have 4 separate copies of 24.x running on my quad core. When I move to 25.7, is there any easy way to transfer this work over? I wouldn't want to end up downloading 4 new primes to test by mistake.

[ckdo]

Originally Posted by adminc

Secondly, if I close the prime95 process (by typing 17 i think) and then try to run it again, I dont get the console output I get when I run it the first time.

mprime produces no screen output unless told to do so, once it is configured.

"./mprime -d" gives just the output with no menus (you'll have to Ctrl-C it);
"./mprime -m" gives the menus as well.

STH
Carsten

[James Heinrich]

feature:
It nicely shows e.g. " (using 1234MB)" at the end of P-1 worker window titles; can this be replicated for ECM workers, please? (I think that's the only 2 that use large emounts of RAM, right?)

bug:
The whole anything-more-than-1800MB allocated problem (runs out of memory, even though memory is available, when set to >1800MB) from the v25.6 thread seems to still be present, please let me know how I can provide more useful feedback.

[adminc]

Not sure if this is supposed to happen, but it seems as though one thread goes lightning fast while the other is super slow.

Specs: Ubuntu Hardy x64
Core 2 duo@ 2.33Ghz

One thread takes about 1,000 seconds for each P-1 output, while the other takes approx 13,000 seconds.

Also, when setting up initially, p95 asks for the memory bounds twice:

It asks for day memory, then night memory, and says 8MB is not enough, and then asks for more memory. The second time I set the memory. it takes.

If I can provide any more info please tell me.

[James Heinrich]

Originally Posted by James Heinrich

The whole anything-more-than-1800MB allocated problem

It's having problems now even when set to 1800MB:

[Oct 8 06:33] Restarting worker with new memory settings.
[Oct 8 06:33] Optimal P-1 factoring of M46061153 using up to 1800MB of memory.
[Oct 8 06:33] Assuming no factors below 2^68 and 2 primality tests saved if a factor is found.
[Oct 8 06:33] Optimal bounds are B1=560000, B2=16660000
[Oct 8 06:33] Chance of finding a factor is an estimated 6.54%
[Oct 8 06:33] Using FFT length 2560K
[Oct 8 06:33] Available memory is 1776MB.
[Oct 8 06:33] Using 1765MB of memory.  Processing 79 relative primes (316 of 480 already processed).
[Oct 8 06:34] Memory allocation error.  Trying again using less memory.
[Oct 8 06:34] Using FFT length 2560K
[Oct 8 06:34] Available memory is 1420MB.
[Oct 8 06:34] Using 1417MB of memory.  Processing 62 relative primes (316 of 480 already processed).
[Oct 8 06:34] Memory allocation error.  Trying again using less memory.
[Oct 8 06:34] Using FFT length 2560K
[Oct 8 06:34] Available memory is 1136MB.
[Oct 8 06:34] Using 1130MB of memory.  Processing 48 relative primes (316 of 480 already processed).
[Oct 8 06:35] Memory allocation error.  Trying again using less memory.
[Oct 8 06:35] Using FFT length 2560K
[Oct 8 06:35] Available memory is 909MB.
[Oct 8 06:35] Using 905MB of memory.  Processing 37 relative primes (316 of 480 already processed).
[Oct 8 06:36] Memory allocation error.  Trying again using less memory.
[Oct 8 06:36] Using FFT length 2560K
[Oct 8 06:36] Available memory is 727MB.
[Oct 8 06:36] Using 721MB of memory.  Processing 28 relative primes (316 of 480 already processed).
[Oct 8 06:36] Memory allocation error.  Trying again using less memory.
[Oct 8 06:36] Using FFT length 2560K
[Oct 8 06:36] Available memory is 581MB.
[Oct 8 06:36] Using 578MB of memory.  Processing 21 relative primes (316 of 480 already processed).
[Oct 8 06:38] M46061153 stage 2 is 71.71% complete.
[Oct 8 07:33] Available memory is 581MB.
[Oct 8 07:33] Using 578MB of memory.  Processing 21 relative primes (337 of 480 already processed).
[Oct 8 08:32] Available memory is 581MB.
[Oct 8 08:32] Using 578MB of memory.  Processing 21 relative primes (358 of 480 already processed).

[Prime95]

Originally Posted by db597

Is there any migration plan for those on 24.x?

Not yet. Most people should continue using 24.x.