mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Conjectures 'R Us (https://www.mersenneforum.org/forumdisplay.php?f=81)
-   -   Software/instructions/questions (https://www.mersenneforum.org/showthread.php?t=9742)

rebirther 2021-08-24 06:04

[QUOTE=gd_barnes;586374]Confirmed #3 in my last post: The file provided to me for R546 was already sieved to P=1G.[/QUOTE]


There could be 2 issues why its not working. The amount of k's or the multicore option -W, in the past I ran without -W and it was working but its a mess with 30 folders and merging every file. I always running 16cores + 14HT

gd_barnes 2021-08-24 06:38

[QUOTE=rebirther;586377]There could be 2 issues why its not working. The amount of k's or the multicore option -W, in the past I ran without -W and it was working but its a mess with 30 folders and merging every file. I always running 16cores + 14HT[/QUOTE]


That would be a mess!

I would like to do a comparison between srsieve and srsieve2: How many CPU hours did it take you to run that full sieve?

Here is what I did: I sieved P=715M-1G in ~8.5 hours on 4 cores. So that's ~34 CPU hours for a P=285M range. Extrapolating: If I sieved the entire range we'd have: 1000M total-range / 285M range that I sieved * 34 CPU hours = ~120 CPU hours. To be fair, add about ~10% since the higher P-ranges sieve faster. Therefore:

I estimate using srsieve it would have taken me ~135 CPU hours to sieve S548 for n=2.5K-10K to P=1G.

How does that compare to how much CPU time that it took you running srsieve2?

rebirther 2021-08-24 07:01

[QUOTE=gd_barnes;586380]That would be a mess!

I would like to do a comparison between srsieve and srsieve2: How many CPU hours did it take you to run that full sieve?

Here is what I did: I sieved P=715M-1G in ~8.5 hours on 4 cores. So that's ~34 CPU hours for a P=285M range. Extrapolating: If I sieved the entire range we'd have: 1000M total-range / 285M range that I sieved * 34 CPU hours = ~120 CPU hours. To be fair, add about ~10% since the higher P-ranges sieve faster. Therefore:

I estimate using srsieve it would have taken me ~135 CPU hours to sieve S548 for n=2.5K-10K to P=1G.

How does that compare to how much CPU time that it took you running srsieve2?[/QUOTE]

With -W16 around 20h if its working.

gd_barnes 2021-08-24 07:37

[QUOTE=rebirther;586382]With -W16 around 20h if its working.[/QUOTE]

So...16 threads/cores * 20 hours = 320 CPU hours.

I estimated 135 CPU hours running srsieve with multiple instances. That's with an old (but very reliable) version of srsieve.

Consider the following when sieving in the future:

Split up the k's remaining into 8, 10, 12, or 16 pieces depending on the # of cores/threads you want to use. That's more efficient than splitting up P-ranges while running multiple cores on all k's. Sieve them all at the same time with separate instances of srsieve (or try it with srsieve2). Then use srfile to combine all of the files at the end. I'm pretty sure you'll save CPU time that way.

Of course that defeats the purpose of having a multi-threaded program, which is its ease of use. But if its using that much more CPU time then you might consider the extra personal time worth it.

My sieving machine is not real fast and not over-clocked. It's an AMD Ryzen 16 core/32 thread running at 3.2 Ghz.

rebirther 2021-08-24 08:46

[QUOTE=gd_barnes;586386]So...16 threads/cores * 20 hours = 320 CPU hours.

I estimated 135 CPU hours running srsieve with multiple instances. That's with an old (but very reliable) version of srsieve.

Consider the following when sieving in the future:

Split up the k's remaining into 8, 10, 12, or 16 pieces depending on the # of cores/threads you want to use. That's more efficient than splitting up P-ranges while running multiple cores on all k's. Sieve them all at the same time with separate instances of srsieve (or try it with srsieve2). Then use srfile to combine all of the files at the end. I'm pretty sure you'll save CPU time that way.

Of course that defeats the purpose of having a multi-threaded program, which is its ease of use. But if its using that much more CPU time then you might consider the extra personal time worth it.

My sieving machine is not real fast and not over-clocked. It's an AMD Ryzen 16 core/32 thread running at 3.2 Ghz.[/QUOTE]

Mine is a 3950x but cant run 16core llr, too hot, reaching 95°C, with 16+14 the clockspeed is lower with 80°C and faster for the smaller tests

gd_barnes 2021-08-24 09:19

[QUOTE=rebirther;586391]Mine is a 3950x but cant run 16core llr, too hot, reaching 95°C, with 16+14 the clock speed is lower with 80°C and faster for the smaller tests[/QUOTE]

That's too bad about the heat on your machine. Mine is also a 3950X. It isn't great for small and medium-length LLR testing but is an excellent siever and seems almost equal to my Intels on longer LLR tests.

The benchmark that I gave you for the 4 cores running srsieve was with 20 instances of LLR running at the same time. So 24 threads were in operation at that time. That is generally the most that I will run on it at one time.


All times are UTC. The time now is 17:51.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.