Welcome to MilkyWay@home

Posts by Jake Bauer

1) Message boards : News : New Nbody version (Message 62638)
Posted 27 Oct 2014 by Jake Bauer
Post:
2 days is absurd and you shouldn't need this long to compute any of these simulations running on 6 cores. I would recommend aborting if it looks like it is going to take this long.
2) Message boards : News : New Nbody version (Message 62623)
Posted 24 Oct 2014 by Jake Bauer
Post:
I almost have a solution. I'm just testing now. Abort these if you wish, but they shouldn't actually take that long.
3) Message boards : News : New Nbody version (Message 62602)
Posted 20 Oct 2014 by Jake Bauer
Post:
Yes. This is absolutely true. If I recall, the algorithm that did this was changed recently, and we had to significantly revise this calculation. The issue with time estimates in n-body simulations is the the time estimate is potentially dependent on any parameters you use to describe your system, the number of bodies (resolution), and whatever n-body solver you are using. By time, I of course am referring to CPU and not clock time. Actually, this issue has piqued my interest in this topic. I might take another stab at it to see if we can get a better time estimate.

Jake
4) Message boards : News : New Nbody version (Message 62598)
Posted 19 Oct 2014 by Jake Bauer
Post:
This is normal. N-body simulations are proving very difficult to estimate completion times for. Apologies. It should not actually take that long (probably no more than 15 hours).

Jake
5) Message boards : News : New Nbody version (Message 62527)
Posted 9 Oct 2014 by Jake Bauer
Post:
This succeeded. I see some checkpointing issues, but nothing critical. Am I missing something?

Jake
6) Message boards : News : New Nbody version (Message 62503)
Posted 7 Oct 2014 by Jake Bauer
Post:
This appears to be an issue with that specific work unit. RHO_MAX = 3000 is absurd. If it's just an isolated instance, I wouldn't worry.


Jake
7) Message boards : News : New Nbody version (Message 62414)
Posted 28 Sep 2014 by Jake Bauer
Post:
Is this an isolated instance or does this consistently happen?
8) Message boards : News : N-Body Release 1.42 (Message 62203)
Posted 19 Aug 2014 by Jake Bauer
Post:
Thanks! In response to your earlier point, we have confirmed that this is a library issue. This should be fixed upon the next release (I want to say a couple weeks from now, but don't quote me).
9) Message boards : News : N-Body Release 1.42 (Message 62193)
Posted 17 Aug 2014 by Jake Bauer
Post:
I will look into this. It looks like a library issue. We tried something different this time around. I will suggest that revert back to the previous way of doing things to avoid this error.

Jake
10) Message boards : News : N-Body Release 1.42 (Message 62149)
Posted 10 Aug 2014 by Jake Bauer
Post:
I'm a slave to this project. It's not a problem to work Sundays at all!

Please post to the MilkyWay main page. Truth-be-told, I am the worst person to ask about server operations. Travis is probably your best bet. The runs in question have been stopped so this is only a temporary issue.

You mentioned the time estimation issues:

It turns out that for this application, time estimation is extremely difficult. N-body simulations are notorious for this. I believe the last person to work on this in our group was Jeff. There are no actual 42 hour workunits to my knowledge, and if there are, appropriate credit should be awarded. Like I said, this should be resolved as the queue is cleared of these old workunits in the coming week.

I apologize for this. But maybe someone else will be able to make something of your suggestion. Definitely try posting on the main page.

P.S. I'll think I'll have that cold beer now.
11) Message boards : News : N-Body Release 1.42 (Message 62147)
Posted 10 Aug 2014 by Jake Bauer
Post:
This is the error I had mentioned. It would be in the workunit name. For instance:

de_nbody_08_05_orphan_sim

will be in the name of the workunit.

This simply refers to the date that the run was started. You should stop seeing these errors very soon.
12) Message boards : News : N-Body Release 1.42 (Message 62135)
Posted 7 Aug 2014 by Jake Bauer
Post:
The histogram files were updated. Anything dated before 08_06 will definitely error. This is what is being seen. Abort these workunits if you see them. Only ones with 08_05 in them should be valid.

Apologies for the inconvenience, but the development team saw no other way.
13) Message boards : Number crunching : New N-Body Release 1.42 (Message 62132)
Posted 7 Aug 2014 by Jake Bauer
Post:
The histogram format was changed. If you are crunching units posted earlier than 08_06, you will see this error. These should stop as soon as the queue clears.

Jake
14) Message boards : News : N-Body Release 1.42 (Message 62131)
Posted 7 Aug 2014 by Jake Bauer
Post:
Thank you for your reply. I have noticed a slow validation process so far. However, keep in mind that on 1 core, these simulations take half a day to run. It is conceivable that there has not been enough time for validation to occur. I will be monitoring this closely and provide more information later.

Jake
15) Message boards : News : N-Body Release 1.42 (Message 62111)
Posted 5 Aug 2014 by Jake Bauer
Post:
Dear users,

N-Body will be released at 4 pm EST today. Features of this release include an improved dwarf galaxy model and new data representation features. Post here with issues.

As this is likely my last release on this project (due to my move to England for graduate school), I would like to thank all of you for helping me do great science during my time affiliated with Dr. Newberg's group. I will still be active in the community, but I am proud to hand my role in the project off to Sidd. I look forward to a continued relationship with the BOINC community in the future.

Cheers,
Jake Bauer
16) Message boards : News : New N-Body Runs (6/28) (Message 61965)
Posted 28 Jun 2014 by Jake Bauer
Post:
These are the new runs:

de_nbody_06_28_orphan_sim_0
de_nbody_06_28_orphan_sim_1
de_nbody_06_28_orphan_sim_2
de_nbody_06_28_orphan_sim_3

ps_nbody_06_28_orphan_sim_0
ps_nbody_06_28_orphan_sim_1
ps_nbody_06_28_orphan_sim_2
ps_nbody_06_28_orphan_sim_3


Most people's issues with the last set should be fixed. Post errors here please. Also, if you are having issues still with the ones from 6/10 and 6/11, expect to stop seeing those work units as the queue vacates in the next few weeks.

Thanks,

Jake
17) Message boards : News : New N-Body Runs (Message 61947)
Posted 25 Jun 2014 by Jake Bauer
Post:
The workunits take vastly different amounts of time to complete. This is a problem that we at MW@Home have been working on to assign appropriate credit to crunchers. Our goal is to ultimately perfect this art and prevent the assignment of non-useful simulations that are time-expensive. You are right to say that for the same simulation, the wall clock time on your 4 core machine should be half that of your dual AMD cores. I can go into more detail. If you send me a private message, I would be glad to explain the science of how the workunits are very difficult to assess computationally. I hope that I can answer any questions you may have.

Jake
18) Message boards : News : New N-Body Runs (Message 61946)
Posted 25 Jun 2014 by Jake Bauer
Post:
I have fixed the underlying issue. It was necessary that these runs be up for this period of time for the science. This issue will not occur in the future. Expect to stop getting problematic workunits in the next week.

Apologies,

Jake
19) Message boards : News : New N-Body Runs (Message 61832)
Posted 3 Jun 2014 by Jake Bauer
Post:
I have posted:

de_nbody_06_03_orphan_sim_0
de_nbody_06_03_orphan_sim_1
de_nbody_06_03_orphan_sim_2
de_nbody_06_03_orphan_sim_3

ps_nbody_06_03_orphan_sim_0
ps_nbody_06_03_orphan_sim_1
ps_nbody_06_03_orphan_sim_2
ps_nbody_06_03_orphan_sim_3

Please alert me to any problems.

Best wishes and good crunching,

Jake Bauer
20) Message boards : News : More New N-Body Runs (Message 61292)
Posted 1 Mar 2014 by Jake Bauer
Post:
MilkyWay@Home runs 2 applications:

Separation tries to pull out structures called "stellar streams" from a background (see astrophysics section of the website) by running a complicated probability calculation. Separation is an older project and has since acquired a GPU version to supplement the CPU calculations.

N-body is tasked with the problem of replicating the creation of stellar streams from collisions in an arbitrary Galactic gravitational potential (basically a measure of how strong the gravity of the Milky Way is at any point and space, and in principle, time as well). What you guys are doing is running a single n-body simulation and telling us what the likelihood of your simulation fitting the observed data. In principle, this will eventually tell us where dark matter is in the Milky Way system. This is a newer application and only runs on CPUs at present, although this has potential to change in the near future.

I hope this helps!

Jake


Next 20

©2021 Astroinformatics Group