PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
1) Message boards : Number crunching : Yuri's Night Challenge (Message 150053)
Posted 4 days ago by Profile JeppeSNProject donor
Should I be running this on CPU too? What's the optimal number of threads?

It is incredibly slow on CPU. Try using all your threads on one task. /JeppeSN
2) Message boards : Number crunching : WW & 3080 ranking? (Message 150030)
Posted 6 days ago by Profile JeppeSNProject donor
I noticed in the "computer types" link in the menu at bottom left, that about 2/3 of systems running PrimeGrid are Linux. They outnumber Windows systems by about 2:1 ratio.

Even before The Science Cloud and its many (sometimes short-lived) hosts, Linux had the lead by numbers of machines: Page from 2019-Apr-29

Not sure how much you must upgrade or modify your OS before BOINC sees you as a new computer.

However, in terms of recent credit, Windows is the leading OS.

/JeppeSN
3) Message boards : Number crunching : Which wingman tasks get allocated to (Message 149959)
Posted 7 days ago by Profile JeppeSNProject donor
It is completely random. Consider yourself lucky you got the task first. Sometimes you will be the one who gets the task "late". /JeppeSN
4) Message boards : Number crunching : Yuri's Night Challenge (Message 149941)
Posted 9 days ago by Profile JeppeSNProject donor
2021 is the 100th anniversary of the birth of D. D. Wall, so a nice year to find some Wall-Sun-Sun near-results (like Wall did at IBM at the same time as the Soviet space program was planning to launch a man into space). /JeppeSN
5) Message boards : Wieferich and Wall-Sun-Sun Prime Search : PRPnet near finds to PG account? (Message 149879)
Posted 11 days ago by Profile JeppeSNProject donor
As far as I have heard, all the near-Wieferich and near-WSS that were found on PRPNet were correct (that was probably already known), and also no near-Wieferich or near-WSS were missed!

They can still be seen at:
https://u-g-f.de/PRPNet/findlist_adv.php?proj=WFS
https://u-g-f.de/PRPNet/findlist_adv.php?proj=WSS

Of course this is not an answer to your question.

/JeppeSN
6) Message boards : General discussion : A new class of primes? (Message 149865)
Posted 12 days ago by Profile JeppeSNProject donor
That article made me think of another kind of counting problem:
How many ways do you get a prime number by changing one zero bit in a particular prime number into a one?

For example:
17 in binary (most significant bit first) is 10001; when you you change that to 10011 you have 19, but that's it, any other zero bit changed to one is results in a composite number: 10101 is 21, 11001 is 25; so the answer for 17 is 1 (unless you keep going with higher powers of 2; is there a limit? 17+8192=8209 another prime). NB amazingly adding powers of 2 to 17 produces a few unexpected perfect squares: 17+8=25, 17+32=49, 17+64=81, 17+512=529

The basic problem is finding how many primes you can generate from another prime by adding powers of 2, skipping those powers of 2 that already occur in the additive decomposition into powers of 2 of the prime.


I take it to mean that you can replace one existing zero (excluding leading zeroes, only internal zeros are allowed (Mersenne primes have no zeroes to try)) by a one. The first prime for which you can produce 2 other primes in this way, is 43. The first where you can produce 3 other primes, is 149. Continuing like that, I get:
[1, 2] [2, 43] [3, 149] [4, 4421] [5, 5441] [6, 49169] [7, 542021] [8, 2376131] [9, 19154321]
/JeppeSN
7) Message boards : General discussion : A new class of primes? (Message 149863)
Posted 12 days ago by Profile JeppeSNProject donor
Mathematicians Find a New Class of Digitally Delicate Primes

https://www.quantamagazine.org/mathematicians-find-a-new-class-of-digitally-delicate-primes-20210330

I am so glad we have Quanta Magazine. /JeppeSN
8) Message boards : Number crunching : Too much work when joining (Message 149860)
Posted 13 days ago by Profile JeppeSNProject donor
JeppeSN wrote:
... or ensure that PrimeGrid uses only 50% of the "processors" (meaning 50% of the logical processors) ...

JeppeSN, please correct me if I'm wrong here, I have the impression that "use at most 50% of the CPUs" setting only affects BOINC's calculation of the quantity of tasks it will ask for download from the server.
The actual distribution of work load among the cores is managed by the OS (with the optional help of third party software, like Process Lasso for Windows).

It should affect the amount of tasks BOINC will process at the same time (no matter if it has a buffer of additional, waiting tasks or not). You are absolutely correct the OS (operating system, like Windows or Linux) will distribute the work load. Hopefully, most OSs will know to put the different tasks onto separate physical cores (as long as the number of processes (where the OS must be able to see which ones are relevant) is small enough to allow it). I heard even Windows has become better at this.

But as you say, you can control/overwrite this, or just observe it, with a tool.

/JeppeSN
9) Message boards : Number crunching : Too much work when joining (Message 149850)
Posted 13 days ago by Profile JeppeSNProject donor
Peter Hucker, when all participants unanimously say that hyperthreading (HT) is bad for you when you crunch PrimeGrid, and that you must either turn off HT in the BIOS, or ensure that PrimeGrid uses only 50% of the "processors" (meaning 50% of the logical processors), then you should believe them!

If you run on all logical threads (HT), the speed of each process will be less than half the normal value, and as a result, the throughput will be smaller than when you run it the way everybody recommends.

Hyperthreading works well for most typical software, programs and apps. That is because the software most people use on their computers, does very complex tasks, and is programmed on a high level, and as a result, does not put heavy pressure on the CPU and memory. But PrimeGrid's apps, on the other hand, do the same type of "simple" (mathematical) calculation (i.e. the same CPU instructions) over and over again, and the apps are very highly optimized to squeezing the most out of the hardware. There is a difference between having two independent, "real" CPU cores, versus having two hyperthreads running on one physical core. For "normal" software, that does not resemble PrimeGrid's, it is not so big a problem to have to share some ressource with another thread. But for PrimeGrid's apps (perhaps except for some sieving apps), two hyperthreads on the same physical core will just be fighting each other to get access to a ressource of which there is only one, and the overhead of shifting from one to the other, and invalidating cache from the opposite thread, etc. (someone can teach me the precise details), will make things slower than without HT.

But if you do not believe it, it is easy to measure yourself.

/JeppeSN
10) Message boards : Generalized Fermat Prime Search : GFN-14/13/... MEGA Prime Search on GFN Server (Message 149839)
Posted 14 days ago by Profile JeppeSNProject donor
Congratulations to mmonnin (@PrimeGrid) with the GFN-13 Mega. I hope you answer Stream's messages soon so we can add you and the prime on The PrimePages. /JeppeSN


Next 10 posts
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2021 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 3.45, 3.94, 4.02
Generated 18 Apr 2021 | 11:54:49 UTC