mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   PrimeNet (https://www.mersenneforum.org/forumdisplay.php?f=11)
-   -   Continue on someone's expired first time LL assignment (https://www.mersenneforum.org/showthread.php?t=26267)

rikfo 2020-12-04 11:01

Continue on someone's expired first time LL assignment
 
Let's say someone does not complete their LL test and the assignment expires. Does the prime95 software repeatedly upload the remainder after some interations allowing someone else to continue from that iteration or do you have to restart from the beginning?

retina 2020-12-04 11:10

[QUOTE=rikfo;565220]Does the prime95 software repeatedly upload the remainder after some interations allowing someone else to continue from that iteration or do you have to restart from the beginning?[/QUOTE]You have to run the test from the beginning.

The server doesn't store intermediate residues, The client doesn't upload intermediate residues.

kruoli 2020-12-04 11:15

Doesn't Prime95 upload 1024 bit residues regularly? In any case, that data is not sufficient for continuing a computation.

retina 2020-12-04 11:30

[QUOTE=kruoli;565222]Doesn't Prime95 upload 1024 bit residues regularly?[/QUOTE]What it does upload is only a tiny part of the residue. The full residue is kept only locally.

kriesel 2020-12-04 17:45

The interim residues logged in prime.log as uploaded at 500K iterations from start and every multiple of 5M from there on are IIRC only least-significant-64-bits. These get logged in results.txt:
[CODE][Sun Apr 19 00:51:13 2020]
M79335979 interim PRP residue F884B6F970C2E554 at iteration 10000000
[/CODE]Final residues in results.json.txt are 2048bit. Both are far shorter than the full exponent-bits-long residue, currently ~100,000,000 bits, which is what's needed to continue a test, and can not be derived from the truncated values.

Periodically someone proposes periodic saving of full-length interim residues from all clients to the server, to potentially save lost time from abandoned work.
It gets shot down on the basis of bandwidth limitations, storage requirements, and cost.

rikfo 2020-12-04 20:57

Hmm okay. The maximum size of the residue is equal to the exponent number of bits, right? So as you say, right now it would be in the order of ~100-400 Mbits or ~13-50 Mbytes? That seems reasonable to store "somewhere". If everyone devotes some of their space all residues of the current assignments could easily be stored. But I guess it gets very complicated to keep track of and to synchronize the progress.

Maybe it could also provide the option for those who work in a team or for someone who works on different computers to spread the work of the same exponent on different computers?

In the end maybe its not worth the effort at all...

VBCurtis 2020-12-04 21:14

This has been discussed many times, and the effort to share residues is much more than the work saved by aborted tests.

The people serious enough to do more than 24 hr of work on an exponent are also often resourceful enough to copy a Prime95 folder to another computer if something fails to finish a test. So, the use case for managing residue files is pretty small. It also adds a whole bunch of bandwidth, which is not free for every user.

Uncwilly 2020-12-04 21:16

[QUOTE=rikfo;565283]Maybe it could also provide the option for those who work in a team or for someone who works on different computers to spread the work of the same exponent on different computers?[/QUOTE]If you only had access to computers part time, say like at a library or internet cafe, you could put Prime95 on a thumb drive and move those files around. But you can't practically have 2 computers work on parts of the same PRP at the same time.


All times are UTC. The time now is 23:00.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.