Low n ECM Factoring Reservation - Page 3

ltd · 2006-04-25, 16:01

@Greenbank:
Composite removed from ecm server list.
Thanks for your effort.
Your ecm results are now visible on (The msieve factor will follow later)

www.beta.psp-project.de/stats.html

Transfer is not yet automatic generated as the handling of rejected results from the llrnet server was more important. Now rejected results count as double checks again if the result comes from a different user then the first result.

Lars

Greenbank · 2006-04-25, 16:03

98749*2^242+1 is:

(msieve 1.06 again).

Tue Apr 25 16:56:48 2006 prp36 factor: 506326223436922423044617328637914119
Tue Apr 25 16:56:48 2006 prp43 factor: 1378355476952984680473416674360244746420663
Tue Apr 25 16:56:48 2006 elapsed time 00:15:15

Forgot to say, this was run on a 3.0GHz Xeon. May see how long it will take for the next largest...SIQS/MPQS has its limits...[EDIT] Nope, next largest is 640 bits which is too big for msieve.

Greenbank · 2006-04-25, 16:23

Oh and I'm having some problem with ecmclient.

Whenever it gets a block of B1=11e6 (45 digit level) it runs for 20 minutes and then just fails. I have to watch ecmclient and ^C every time it gets some of these blocks of work. The 35 and 40 digit levels (B1=1e6 and B1=3e6) are fine.

Not entirely sure what it is, I am running this on a PowerMac G5 with GMP-ECM 6.0.1 with GMP 4.1.4 (haven't upgraded to 4.2 yet!). Will try and find out why later.

On another note, I should be 50% of the way through that 8T sieving range tomorrow and so I'll email you the fact.txt then.

rogue · 2006-04-25, 17:01

Quote:

Originally Posted by Greenbank

Oh and I'm having some problem with ecmclient.

Whenever it gets a block of B1=11e6 (45 digit level) it runs for 20 minutes and then just fails. I have to watch ecmclient and ^C every time it gets some of these blocks of work. The 35 and 40 digit levels (B1=1e6 and B1=3e6) are fine.

Not entirely sure what it is, I am running this on a PowerMac G5 with GMP-ECM 6.0.1 with GMP 4.1.4 (haven't upgraded to 4.2 yet!). Will try and find out why later.

Alex, turn on debugging via -debug and send me the log.

Greenbank · 2006-04-25, 17:14

Will have a look at it tomorrow. For the moment I've made a wrapper script for ecm that just exits with "ERROR" if it gets a B1 over 3e6 otherwise it execs the real ecm binary. It seems to be working fine like this.

ltd · 2006-04-25, 17:23

Is it possible that the problem comes from the runtime of the stage one of 1400 seconds compared to the frequency setting of the ecmnet client which looks to me as if it is set much smaller then that. (Only guessing from the server logs so i could be completly wrong.)
So maybe there is a problem when the client wants to switch between stage 1 and 2?

Lars

Greenbank · 2006-04-25, 17:43

Aha, good point. I only run this when I'm at the machine so I've been using a low frequency setting. Will try that out tomorrow.

rogue · 2006-04-25, 18:12

Quote:

Originally Posted by ltd

Is it possible that the problem comes from the runtime of the stage one of 1400 seconds compared to the frequency setting of the ecmnet client which looks to me as if it is set much smaller then that. (Only guessing from the server logs so i could be completly wrong.)
So maybe there is a problem when the client wants to switch between stage 1 and 2?

Lars

Scenario 1 is not likely. The frequency is not on a timer. Scenario 2 would not be controlled by the client, but by GMP-ECM.

In debug mode, it will generate the actual command sent to the system to process a curve. What happens if you execute that directly from the command prompt?

Alex, when you switch to GMP 4.2, let me know. I have a patch for the addmul_1.asm and submul_1.asm on PowerPC that didn't make it into that release.

Greenbank · 2006-04-26, 09:25

Quote:

Originally Posted by rogue

Scenario 1 is not likely. The frequency is not on a timer. Scenario 2 would not be controlled by the client, but by GMP-ECM.

In debug mode, it will generate the actual command sent to the system to process a curve. What happens if you execute that directly from the command prompt?

Alex, when you switch to GMP 4.2, let me know. I have a patch for the addmul_1.asm and submul_1.asm on PowerPC that didn't make it into that release.

[2006-04-25 19:00:00 GMT] a: Received 79817_1191 (length 14) : ECM with B1=11000000 * 4589 curves
[2006-04-25 19:00:00 GMT] 79817_1191: ./ecm 11000000 < work_a.in > work_a.out
This is the PSP Project beta ECMNET Server running on ECMNET version 2.7.1
[2006-04-25 19:08:48 GMT] a: Quitting due to detected interruption

[2006-04-25 19:08:48 GMT] Total Time: 1:08:59 Curves: 145 Total Work: 8 Total Factors Found: 0

$ cat work_a.out
GMP-ECM 6.0.1 [powered by GMP 4.1.4] [ECM]
Input number is 79817*2^1191+1 (364 digits)
Using B1=11000000, B2=25577181640, polynomial Dickson(12), sigma=1937413883
Step 1 took 384213ms

So it looks like it didn't complete Stage 2 for the first curve and so no curves had been completed and it shut down.

The reason it is so slow is that I'm still running 4 copies of proth_sieve in the background and so the machine is pegged at 100% CPU usage anyway.

Can you send the two asm files to my usual email please Mark. Ta very much.

rogue · 2006-04-26, 12:12

Quote:

Originally Posted by Greenbank

GMP-ECM 6.0.1 [powered by GMP 4.1.4] [ECM]
Input number is 79817*2^1191+1 (364 digits)
Using B1=11000000, B2=25577181640, polynomial Dickson(12), sigma=1937413883
Step 1 took 384213ms

So it looks like it didn't complete Stage 2 for the first curve and so no curves had been completed and it shut down.

The reason it is so slow is that I'm still running 4 copies of proth_sieve in the background and so the machine is pegged at 100% CPU usage anyway.

Can you send the two asm files to my usual email please Mark. Ta very much.

If Stage 2 does not complete, then the client assumes that the process was cancelled. Could you run GMP-ECM (outside of ECMNet) on that number? I believe that there is a "verbose" flag for GMP-ECM. I would be curious to see why it fails.

Greenbank · 2006-04-26, 13:14

Trying ecm manually although it will take a while. I'm doing this ECM work at the same time as running 4 instances of proth_sieve. I want to load the machine so that my current sieve ranges finish after 9am on Monday 14th (they are currently due to finish late Sunday or very early Monday) when I won't be at the machine.

2006-04-25, 16:01	#23
ltd Apr 2003 2²×193 Posts	@Greenbank: Composite removed from ecm server list. Thanks for your effort. Your ecm results are now visible on (The msieve factor will follow later) www.beta.psp-project.de/stats.html Transfer is not yet automatic generated as the handling of rejected results from the llrnet server was more important. Now rejected results count as double checks again if the result comes from a different user then the first result. Lars Last fiddled with by ltd on 2006-04-25 at 16:01

2006-04-25, 16:03	#24
Greenbank Jul 2005 2×193 Posts	987492^242+1 is: (msieve 1.06 again). Tue Apr 25 16:56:48 2006 prp36 factor: 506326223436922423044617328637914119 Tue Apr 25 16:56:48 2006 prp43 factor: 1378355476952984680473416674360244746420663 Tue Apr 25 16:56:48 2006 elapsed time 00:15:15 Forgot to say, this was run on a 3.0GHz Xeon. May see how long it will take for the next largest...SIQS/MPQS has its limits...[EDIT] Nope, next largest is 640 bits which is too big for msieve. Last fiddled with by Greenbank on 2006-04-25 at 16:07*

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
5*2^n-1 Reservation Thread	VBCurtis	Riesel Prime Search	661	2021-05-08 02:21
LLR reservation	hhh	Prime Cullen Prime	6	2007-09-17 04:27
Sieve reservation	hhh	Prime Cullen Prime	5	2007-08-03 04:37
P-1 reservation	hhh	Prime Cullen Prime	1	2007-05-16 20:12
15k reservation	gribozavr	15k Search	3	2005-03-11 14:04

2006-04-25, 16:23	#25
Greenbank Jul 2005 602₈ Posts	Oh and I'm having some problem with ecmclient. Whenever it gets a block of B1=11e6 (45 digit level) it runs for 20 minutes and then just fails. I have to watch ecmclient and ^C every time it gets some of these blocks of work. The 35 and 40 digit levels (B1=1e6 and B1=3e6) are fine. Not entirely sure what it is, I am running this on a PowerMac G5 with GMP-ECM 6.0.1 with GMP 4.1.4 (haven't upgraded to 4.2 yet!). Will try and find out why later. On another note, I should be 50% of the way through that 8T sieving range tomorrow and so I'll email you the fact.txt then.

2006-04-25, 17:14	#27
Greenbank Jul 2005 2·193 Posts	Will have a look at it tomorrow. For the moment I've made a wrapper script for ecm that just exits with "ERROR" if it gets a B1 over 3e6 otherwise it execs the real ecm binary. It seems to be working fine like this.

2006-04-25, 17:23	#28
ltd Apr 2003 772₁₀ Posts	Is it possible that the problem comes from the runtime of the stage one of 1400 seconds compared to the frequency setting of the ecmnet client which looks to me as if it is set much smaller then that. (Only guessing from the server logs so i could be completly wrong.) So maybe there is a problem when the client wants to switch between stage 1 and 2? Lars

2006-04-25, 17:43	#29
Greenbank Jul 2005 110000010₂ Posts	Aha, good point. I only run this when I'm at the machine so I've been using a low frequency setting. Will try that out tomorrow.

2006-04-26, 13:14	#33
Greenbank Jul 2005 2×193 Posts	Trying ecm manually although it will take a while. I'm doing this ECM work at the same time as running 4 instances of proth_sieve. I want to load the machine so that my current sieve ranges finish after 9am on Monday 14th (they are currently due to finish late Sunday or very early Monday) when I won't be at the machine.