I´ve never used GMP-ECM on exponents that large. What is the max amount of mem it uses during Stage 2, for the bounds you mentioned on your post?

Never mind, I had missed one previous post of yours, where you refer the amount of mem used...

M85027 falls...
 

another sub 100k exponent succumbs to ecm - ran 425 curves in total

Resuming ECM residue saved with Prime95
Input number is 4758184975...7139017727 (25596 digits)
Using B1=1000000, B2=1000000000, polynomial Dickson(6), sigma=0:3673943552503015
Step 1 took 3847484ms
Step 2 took 1217010ms
********** Factor found in step 2: 113574028377227867558212550573836752813871
Found prime factor of 42 digits: 113574028377227867558212550573836752813871

136.3 bits.

Using the instructions given at [url]http://www.mersenneforum.org/showpost.php?p=207360&postcount=69[/url], I found that the group order of the lucky elliptic curve is:

2[SUP]2[/SUP] * 3[SUP]3[/SUP] * 5 * 19 * 3457 * 17027 * 33769 * 107273 * 311407 * 685859 * 243064937

So you would not have been able to find that 42-digit prime factor with the bound B2 = 100 B1 generally used on Prime95. You can also see that the group order is multiple of 12, as expected.

[QUOTE=Gordon;412763]another sub 100k exponent succumbs to ecm - ran 425 curves in total

Resuming ECM residue saved with Prime95
Input number is 4758184975...7139017727 (25596 digits)
Using B1=1000000, B2=1000000000, polynomial Dickson(6), sigma=0:3673943552503015
[B]Step 1 took 3847484ms[/B]
Step 2 took 1217010ms
********** Factor found in step 2: 113574028377227867558212550573836752813871
Found prime factor of 42 digits: 113574028377227867558212550573836752813871

136.3 bits.[/QUOTE]

How come Step 1 took the time listed above? If you´ve used Prime95 for Stage 1, GMP-ECM should have taken virtually no time on Stage 1. It seems you are duplicating work running Stage 1 twice, and you can avoid it:

Use the following command (adapting it to your particular Setup): ecm -v -resume xxxxxxx.txt 1e6-1e6 1e9. The key point is the 1e6-1e6 syntax, that instructs GMP-ECM not to run Stage 1.

And by the way, congrats on that very nice find... you´re making me envious :))

3847484 ms is little more than 1 hour, So almost 19 CPU days were lost in this way while computing the 425 curves.

[QUOTE=lycorn;412801]

Use the following command (adapting it to your particular Setup): ecm -v -resume xxxxxxx.txt 1e6-1e6 1e9. [B]The key point is[/B] the 1e6-1e6 syntax, that instructs GMP-ECM not to run Stage 1.

[/QUOTE]

..and that's the bit I missed. :redface:

I just stopped and restarted one instance of gmp-ecm and restarted with the correct parameters...

Step 1 took 0ms

Will cut the run time of each curve by 75%, you live and learn :tu:

[QUOTE=alpertron;412809]3847484 ms is little more than 1 hour, So almost 19 CPU days were lost in this way while computing the 425 curves.[/QUOTE]

It's not as bad as it could have been though, I am running 4 instances of gmp-ecm concurrently...so the waste of wall-time wasn't TOO bad :no:

[QUOTE=Gordon;411964]Haswell, 3.4GHz, no overclock. Does stage 1 in about 5 minutes.

I run a batch of 300 curves (stage 1 only) then copy that output file into a sub-folder. Start GMP-ECM running on that for the stage 2

Run another batch of 300 stage 1 curves, repeat process, and again.

Even running 3 copies of GMP-ECM at the same time, cpu utilisation stays below 40%. Memory usage is only about 3gb per instance and with 32 I have plenty to play with.[/QUOTE]

Just to update.

Takes 201 seconds to run stage 1 on m85121.

I run 4 instances of prime95 at once to generate batches of results.

Then 4 instances of gmp-ecm to do stage 2. They only take just under 4gb each.

So the Stage 1 timing is much better than quoted from your previous post. 201 s is well under 5 minutes.
I must get a Haswell... definitely. Or a Skylake.

On another note, I found funny that the CPU utilization with 3 instances running stayed below 40%. If you have a quad core, each instance of GMP-ECM should take ~ 25% (1 core).

You might want to experiment with the B2 values a bit, you are now spending 6x longer in stage 2 than in stage 1 (stg1 201 sec vs. stg2 1271 sec). The 'rule of thumb' is spending the same time in stage 2 as in stage 1.

[QUOTE=lycorn;412884]So the Stage 1 timing is much better than quoted from your previous post. 201 s is well under 5 minutes.
I must get a Haswell... definitely. Or a Skylake.

On another note, I found funny that the CPU utilization with 3 instances running stayed below 40%. If you have a quad core, each instance of GMP-ECM should take ~ 25% (1 core).[/QUOTE]

4 instances is currently holding at ~50%