P-1 factoring wasting cpu cycle?

[engracio]

A couple days ago I downloaded a wu and my ws did what it was suppose to do. Ran P-1 factoring and found no factor and continued to prp the wu.

[Mon Feb 21 21:57:09 2011]
UID: engracio, 55459*2^19355890+1 completed P-1, B1=55000, B2=426250, We1: F4C8BB45, AID: 00000000000000000000000000100F9D

Last night I guess we had a power outage and it resulted an error on Prime95. I decided to cancel the prp and grabbed another wu due to the error.

While checking my other ws in the herd I found the same wu now being P-1 by another ws. Knowing full well it has been P-1/factored already by another machine. Is this acceptable duplicate of work??? It takes about an hour now to P-1 a wu on a Q6600 with 1 worker and 2 threads.

[Feb 23 11:59] Worker starting
[Feb 23 11:59] Setting affinity to run worker on any logical CPU.
[Feb 23 11:59] Optimal P-1 factoring of 55459*2^19355890+1 using up to 1024MB of memory.
[Feb 23 11:59] Assuming no factors below 2^56 and 2.1 primality tests saved if a factor is found.
[Feb 23 11:59] Optimal bounds are B1=55000, B2=426250
[Feb 23 11:59] Chance of finding a factor is an estimated 0.346%
[Feb 23 11:59] Using all-complex AMD K10 type-3 FFT length 1728K, Pass1=384, Pass2=4608, 2 threads
[Feb 23 11:59] Setting affinity to run helper thread 1 on any logical CPU.
[Feb 23 11:59] 55459*2^19355890+1 stage 1 is 47.67% complete.

More importantly what happens if a factor was found by the ws, would it continue to prp the wu?? Just asking.

e

[Frodo42]

hmmm I think you are right there is a problem with P-1 factoring being done more than once.

I just saw it when I got back home to my computer where there had been a power down while I was away, when I started mprime it started doing P-1 tests on 3 of the 4 tests that were running even though I can see in results.txt that these P-1 tests have been done before.

I haven't checked if the problem also occurs if the program gets a clean shutdown, don't want to waste cpu-cycles if it does, but will check next time I need to reboot the computer (might be weeks though).

I'm using linux 64-bit mprime version 26.5 on the computer where I have seen this behavior.

[enderak]

This has reportedly been fixed in the newer builds of 26.5... See here: http://www.mersenneforum.org/showthread.php?t=15224

[engracio]

This has reportedly been fixed in the newer builds of 26.5... See here: http://www.mersenneforum.org/showthread.php?t=15224

I also reported that issue and it was fixed with Prime95 26.5b5. The original issue was that after P1'ing, Prime95 would not update the wu and delete the 56, 2.1 and whenever the box is rebooted, Prime 95 would start P1'ing all over again. If the box is set up to reboot every day, everyday it would P1 the wu.

My issue here is that when P1'ing is completed and gets prp all the way up to 98% but if for whatever reason it had to be deleted/cancel, the new machine would start P1 all over again. Somehow the P1 factoring result stays locally. My other question is if P1 finds a factor does it get reported and drop/get another wu from the server???

If that is by design due to the network/server traffic, I am fine with that. Just an fyi to those that cares.

[engracio]

hmmm I think you are right there is a problem with P-1 factoring being done more than once.

I just saw it when I got back home to my computer where there had been a power down while I was away, when I started mprime it started doing P-1 tests on 3 of the 4 tests that were running even though I can see in results.txt that these P-1 tests have been done before.

I haven't checked if the problem also occurs if the program gets a clean shutdown, don't want to waste cpu-cycles if it does, but will check next time I need to reboot the computer (might be weeks though).

I'm using linux 64-bit mprime version 26.5 on the computer where I have seen this behavior.

Update to 26.5build 5 and you should be okay.

[Frodo42]

Well just tried upgrading to build 5 and got the same thing again ... wasting a few more hours on P-1 then
It's only the three test that are somewhere around 26% that it retests, the one that's something like 88% done it just goes on PRP'ing ... makes sense if it hadn't already been doing those P-1 tests once.

[enderak]

Well the problem is that the earlier builds didn't update the worktodo properly after P-1 was completed. So the first time you run build 5, it will do the P-1 again because that's what the worktodo says to do. After that it should be updated properly and it shouldn't happen again.

[Frodo42]

Thank you for the explanation, sounds like it is going to work out fine then