Posts by yoyo_rkn

21) Message boards : Number crunching : Long running 8T task - nearly 5 days !! (Message 1436)
Posted 18 Aug 2019 by Profile yoyo_rkn
Post:
In the last phase 8 gnfs jobs are started for roughly 45 minutes. Afterwards the results of the 8 gnfs jobs are put into nfs.dat.
Then yafu.exe runs to check if it is enough gnfs. If not again 8 gnfs jobs are started and so on.
If it is enough, yafu.exe computes the result. This is the single run at the end which should not take longer than 2 hours.

But I see that gnfs runs single threaded over many days. So not 8 gnfs jobs are running in parallel, only 1. This is also the explanation for running many days.
As you restarted BM the gnfs phase was already over and just the combining was done.

yoyo
22) Message boards : Number crunching : Long running 8T task - nearly 5 days !! (Message 1432)
Posted 17 Aug 2019 by Profile yoyo_rkn
Post:
That is a behaviour which I didn't saw before.
Is there only one gnfs process running? It should be 8 of them.
23) Message boards : Number crunching : Long running 8T task - nearly 5 days !! (Message 1430)
Posted 17 Aug 2019 by Profile yoyo_rkn
Post:
Files in the slot directory are still changing?
24) Message boards : Number crunching : Long running 8T task - nearly 5 days !! (Message 1428)
Posted 16 Aug 2019 by Profile yoyo_rkn
Post:
I would let it run, since nfs.dat is only 3 hours old and ggnfs was recently updated.
I assume that nfs is in it's last phase combining everything, which is single threaded.
25) Message boards : Number crunching : Long running 8T task - nearly 5 days !! (Message 1426)
Posted 16 Aug 2019 by Profile yoyo_rkn
Post:
The percentage is just a artifical estimation by BOINC. It doesn't has anything todo with real completion. Important is if nfs.dat or factor.log are still changing.

Single threaded run for 24 hours is very very strange. It shouldn't be much more than 2 hours.
26) Message boards : Number crunching : Long running 8T task - nearly 5 days !! (Message 1424)
Posted 15 Aug 2019 by Profile yoyo_rkn
Post:
If you hav not changed it, the 8t uses 8 cores. But, at the end (and this is valid for all yafu apps) it runs single threaded. This can take (I would estimate for 8t) 1 hour. But the good thing is that the task is near it's end.
27) Message boards : Number crunching : Slower and slower, then sticks at 100% (Message 1421)
Posted 9 Aug 2019 by Profile yoyo_rkn
Post:
Have fun with other projects.
I talked a lot with other people. Nobody claims that his computer is freezed by yafu or yoyo or that a 4t task uses 8 cores.
It it the oposite, most user blame that not all cores are used by yafu and even if the are used the cpu doesn't go hot.
So my assumption is you have a prblem with your PC (memory faults, software,...)

I asked you to provide symptoms and not just blaming for
- usage of 8 cores by 4t task and
- where you got this wrong estimated flops from.
But you provided nothing. You just posted a long aggressive text.

I'll be happy that you leave. Have fun with other projects and good luck in your life.
28) Message boards : Number crunching : YAFU 4T freeze computer (exept task manager !!!! (Message 1419)
Posted 8 Aug 2019 by Profile yoyo_rkn
Post:
Please screenshots or something similar to see that 4t uses 8 cores!!!!!
Where and how do you see it? You just claim it, but didn't provide any symptoms.

The server configuration for 4t is:
    <plan_class>
        <name>                  4t          </name>
        <min_ncpus>             4           </min_ncpus>
        <max_threads>           4          </max_threads>
        <nthreads_cmdline/>
    </plan_class>

It works perfect on every host where no app_config was done and the standards from the project are used. It uses exactly 4 cores.
29) Message boards : Number crunching : Slower and slower, then sticks at 100% (Message 1417)
Posted 8 Aug 2019 by Profile yoyo_rkn
Post:
You are blaming and crying without reading my answers. I find your postings very aggressive, but I try to answer again:

I think I explained above why some wus runs only minutes and others hours.
But again.
Factorization is a very computing intensive task. The method to do it is NFS (number field sieve) . The wu estimation is based on this and the wu will run for hours.
BUT
There are also some trial mechanisms which have a good chance to find factors. So yafu runs them first to find factors without going the long way. Those methods are:
- trying known primes up to 10000
- rho method
- different ecm methods
- p-1, p+1 method
If with those trial factorings wu will run only minutes if a factor is found. If not yafu goes the long way.
This is independent of fast or slow host. It just depends on the wu and if trial factoring finds factors or not. It is also not predictable if trial factoring finds factors or not.
The 4t wus are NOT estimated with 10.000Gflops as you wrote. Don't know where you got this from.
E.g. 4t wu 4048662 is estimated with 537,830.42 GFLOPS
wu 4065784 with 477,447.87 GFLOPS
The estimated flops depending on the length of the composite we try to factor.

Regarding nfs.dat.
On all of my hosts nfs.dat changes at least every 45 minutes.

We run more than 2500 workunits successfully every day. I'm not saying every workunit doen't have any problems. I'm not God or Moses. I just run the project in my spare time beside my full day job. But it is definitely NOT the case that every workunit has problems.
All the 3 projects, yafu, yoyo and RNAWorld runs with 0 full time people attached to it and with a budget of nearly 0 EUR.

So please stop blaming, we run more than 2500 successful workunits per day.
30) Message boards : News : Aliquot sequence 2272032 has terminated!!! (Message 1415)
Posted 7 Aug 2019 by Profile yoyo_rkn
Post:
Aliquot sequence 2272032 has terminated!!!

31) Message boards : Number crunching : Slower and slower, then sticks at 100% (Message 1414)
Posted 6 Aug 2019 by Profile yoyo_rkn
Post:
Regarding runtime prediction an percentage which is displayed in BOINC.

At the end the runtime is not predictable. Only a max runtime is predictable. Therefore I give the workunit this max runtime estimation.
But very often the workunit is much faster, only seconds or minutes. This is because we are doing kind of trial factorisation and very often those trials leads to finding a factor and the workunit finishes much faster.
BOINC "learns" from such computations and "thinks" that a workunit needs only a fraction of my estimation. Based on this BOINC shows the percentage of completion. This percentage is NOT done by the yafu application.
If a workunit doesn't find early factors during this trial factorization it has to go the full NFS (number field sieve) way. When BOINC sees that the workunit takes much longer, the percentage increase is slower and slower and might stay longer time at 100%.
During this last phase NFS writes a checkpoint file nfs.dat. This is used for restarting the app and to continue from the checkpoint.
The nfs.dat is roughly update every 45 minutes.
32) Message boards : Number crunching : Slower and slower, then sticks at 100% (Message 1413)
Posted 6 Aug 2019 by Profile yoyo_rkn
Post:
The nfs.dat file is the checkpoint file. If this is found on restart it is used to continue from it.

The statement that there is no checkpoint is wrong.
33) Message boards : News : Aliquot sequence 2297910 has terminated!!! (Message 1408)
Posted 4 Aug 2019 by Profile yoyo_rkn
Post:
Aliquot sequence 2297910 has terminated!!!

34) Message boards : News : Aliquot sequence 1280190 has terminated!!! (Message 1407)
Posted 29 Jul 2019 by Profile yoyo_rkn
Post:
Aliquot sequence 1280190 has terminated!!!

35) Message boards : News : Aliquot sequence 1794744 and 2311524 have terminated!!! (Message 1406)
Posted 22 Jul 2019 by Profile yoyo_rkn
Post:
Aliquot sequence 1794744 and 2311524 have terminated!!!


36) Message boards : News : Aliquot sequence 2155242 has terminated!!! (Message 1403)
Posted 11 Jul 2019 by Profile yoyo_rkn
Post:
Aliquot sequence 2155242 has terminated!!!

37) Message boards : News : Aliquot sequence 1480050 has terminated!!! (Message 1399)
Posted 1 Jul 2019 by Profile yoyo_rkn
Post:
Aliquot sequence 1480050 has terminated!!!

38) Message boards : News : Aliquot sequence 2133740 has terminated!!! (Message 1398)
Posted 23 Jun 2019 by Profile yoyo_rkn
Post:
Aliquot sequence 2133740 has terminated!!!

39) Message boards : News : Aliquot sequence 2080320 has terminated!!! (Message 1397)
Posted 17 Jun 2019 by Profile yoyo_rkn
Post:
Aliquot sequence 2080320 has terminated!!!

40) Message boards : News : Aliquot sequence 2199210 has terminated!!! (Message 1395)
Posted 27 May 2019 by Profile yoyo_rkn
Post:
Aliquot sequence 2199210 has terminated!!!



Previous 20 · Next 20




Datenschutz / Privacy Copyright © 2011-2020 Rechenkraft.net e.V. & yoyo