About M1277

[storm5510]

Originally Posted by VBCurtis

It's not worth your time to increase B1 by 3%. Add a zero, or don't bother. You'll spend 98% of your time repeating an exact calculation that was already done, with essentially zero chance to find a factor.

There's a reason others have said "more P-1 is not useful". It's useful to run P-1 at bounds 10x those already done, but not useful at a smaller ratio (e.g. doubling B1 isn't worth the effort).

Essentially, you're running a P-1 double check, which simply isn't useful.

I agree, and won't go on with it myself. All that can be done for now is continuing the parade of ECM's. Prime95 does not appear to provide a way of not repeating B1's from zero on up. This leaves GMP-ECM. I have learned in recent days that trying to skip B1's is dubious. e.g. 1e12-1e12. A much easier-to-read document does not address this. It is attached below. What I wanted to do was throw as many Sigma values at it as I could. Apparently, this will not work either. I was using M4363 as a test subject. It still has many thousands of curves to be ran with B1 at 110e6 according to mersenne.org.

[BigNumberGuy]

First off, Yea i see the moved now
Second, I dont think its throttling, it only has like 60C to 70C
and yes i have 8 cores so apparently maximum is 800% (mprime uses 100%, usually it uses like about 600%)

[retina]

Originally Posted by S485122

When thermal throttling or power usage limitations kic- in the processor will still be showing the same percentage of CPU usage AFAIK.

Correct.

Originally Posted by S485122

The processor might be limited by the memory bandwidth.
Some ranges of processors (from both Intel and ARM) are based on a processor having two types of cores : high-performance and high-efficiency.

These things also won't show any change in percentage usage.

The percentage usage is only the OS task switching. You can't see things like throttling, memory starvation, and all the other CPU related bottlenecks from the percentage usage value the OS states.

Crappy unoptimised code will still show 100% usage even if has cache misses everywhere and massive delays waiting for data from the memory. The same for any spinners waiting for some interlock to become free. It is all 100% usage as far as the OS knows.

[BigNumberGuy]

Currently Running ECM on M1277 with B1,B2=7.6 billion 100 curves,

Current Progress: Curve 7, 52.29% (prime 3974756953)

[VBCurtis]

I hope you mean you set B1 to 7.6e9 and let B2 be whatever GMP-ECM chooses. If you've set B2 = B1, that means you're not running stage 2 at all and that's not a good plan (to put it mildly).

[BigNumberGuy]

unfortunately whenever i set b2 to anything more than about 8 billion my process ends with "segmentation fault". wtf does that mean?
method of fixing would be welcome
i have 8 gb ram

[ryanp]

Originally Posted by BigNumberGuy

Currently Running ECM on M1277 with B1,B2=7.6 billion 100 curves,

Current Progress: Curve 7, 52.29% (prime 3974756953)

Are you running these with GMP-ECM or with mprime / Prime95?

[BigNumberGuy]

mprime.
idk how GMP-ECM really work.

[storm5510]

Originally Posted by BigNumberGuy

Currently Running ECM on M1277 with B1,B2=7.6 billion 100 curves,

Current Progress: Curve 7, 52.29% (prime 3974756953)

I really don't see what you're doing with this. What is your B1?

Originally Posted by VBCurtis

I hope you mean you set B1 to 7.6e9 and let B2 be whatever GMP-ECM chooses. If you've set B2 = B1, that means you're not running stage 2 at all and that's not a good plan (to put it mildly).

He's running far too low regardless of how. I think it was you who told me several years ago that ECM's on M1277 needed a B2 of 10e14, or something similar. Following the convention of B2 = B1 * 100 would mean his B1 should be 10e12. I would not want to start something my twice-removed descendants would have to complete, if the tech remained unchanged.

[kruoli]

Originally Posted by storm5510

… the convention of B2 = B1 * 100 …

This does not apply if one is using GMP-ECM for stage 2 (which is basically a must for such a small exponent). It will also not apply to Prime95/mprime if/when George implements a GMP-ECM-like ECM stage 2. You (general "you") can do stage 1 in Prime95 and export it to GMP-ECM or build GMP-ECM with gwnum or run GMP-ECM directly. Use the suggested B1 for ECM in this thread, select a maximum memory and GMP-ECM will find a good B2 for this settings.

[VBCurtis]

Originally Posted by BigNumberGuy

mprime.
idk how GMP-ECM really work.

You're wasting your time & electricity. mprime is the wrong tool for the task, and your curves aren't going to find a factor.
Run GMP-ECM, or wait for an upcoming P95/mprime with enhanced ECM.