Welcome to MilkyWay@home

Posts by Matt Arsenault

1) Message boards : Number crunching : N-Body CPU: does it checkpoint? (Message 57470)
Posted 10 Mar 2013 by Matt Arsenault
Post:
It does checkpoint
2) Message boards : News : N-body screensaver available for testing (Message 54458)
Posted 18 May 2012 by Matt Arsenault
Post:
Dear Creator,

couldyou pls give me a hint how to run the screensaveron Ubuntu 11.10

I have errors:
<stderr.txt>
01:24:22 (15538): Can't open init data file - running in standalone mode


<stderrgfx.txt>
Failed to connect to shared scene
Failed to connect to shared scene
Failed to connect to shared scene
Failed to connect to shared scene
Failed to connect to shared scene
Could not attach to simulation after 5 attempts

TIA!
It should work. The running standalone part might be the problem
3) Message boards : News : N-body screensaver available for testing (Message 54457)
Posted 18 May 2012 by Matt Arsenault
Post:
^ ok then...


Nice screensaver, but I do have some suggestions for future builds:
[quote]1. Needs more continuity. Since the WU's that are available finish for my computer in a mere 30 seconds, the graphics can only run for... 30 seconds. After that, the screensaver just stops (or rather, just stops progressing and hangs). It still functions, but it does not continue over to the next WU. In fact, when you quit the screensaver to look at the graphics for the next one that is already running, the screensaver fails to load (with the error being that it cannot attach to a running process, being the application itself) until the next WU arrives and runs.

Could there be a way for the screensaver to identify when the WU finishes, stay on, and then reattach to the next WU that runs?
I wasn't really sure what it should do when the workunit finishes

2. Needs more... fluidity. When I watched your demo, the particle motion seemed rather fluid and smooth. In contrast, when I viewed the same from my end, it was quite jumpy.. or rather, it just updated whenever it felt like. IN fact, i dont think it updated per .001 Gyr (as it did in the demo). It just updated all over the place, and it was slow at updating. I did not notice a regular interval. is this something that can be fixed? it would look much better if it acted the way your demo had it.
This is one of the problems I mentioned. The way it runs now it just updates the display whenever it can and lets' the actual simulation go ahead on its own. The demo video made the simulation stop so that everything is smooth. Another option which I might make the default just updates the displayed scene every 10 seconds or so jumping to the current view.

3. a little thing, but the fact that you couldn't pause the screen from rotating annoyed me. could you add an option to enable and disable screen rotation? I would really like to see the particles move on their own instead of seeing it happen while flying around ;)
There already is, it (and all the other on/off preferences) aren't enabled in the web preferences yet.
4) Message boards : Number crunching : Computation Error (Message 54410)
Posted 14 May 2012 by Matt Arsenault
Post:
This is an Nvidia driver bug. With the most recent set of drivers after the display goes to sleep or something like that tasks can't run correctly
5) Message boards : Number crunching : Linux x86_64 1.02 (opencl_amd_ati) - MultiGPU - fglrx asic hang happened (fglrx driver crash) (Message 54409)
Posted 14 May 2012 by Matt Arsenault
Post:
There are a lot of crashes with Catalyst and the OpenCL driver so probably. There is a horrible defect where the Linux Catalyst driver can't reset the GPU after some error, so even if there is an application error the driver problem makes it much worse since you have to reboot after.
6) Message boards : Number crunching : No Credit! (Message 54393)
Posted 12 May 2012 by Matt Arsenault
Post:
My machine is beavering away at this project, with, usually, two tasks running simultaneously, and it is reporting results, but the work done credit has not changed for months now!
You're using a very ancient version which won't work. You can't use the old applications that are floating around the internet
7) Message boards : Number crunching : N-Body programs, "Error in Computation" (Message 54392)
Posted 12 May 2012 by Matt Arsenault
Post:
I don't see any tasks with errors on your system, but if you're running with the BOINC data dir on a FAT32 drive or something under XP there could be an error when writing checkpoints
8) Message boards : Number crunching : Settings for the project (Message 54391)
Posted 12 May 2012 by Matt Arsenault
Post:
See here http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=2891
9) Message boards : Number crunching : Can't update preferences. (Message 54390)
Posted 12 May 2012 by Matt Arsenault
Post:
I temporarily disabled some of the new graphics preferences so it should work now
10) Message boards : News : N-body screensaver available for testing (Message 54378)
Posted 12 May 2012 by Matt Arsenault
Post:
I wrote a screensaver for the nbody application. You can manually install it and try it from here: http://milkyway.cs.rpi.edu/milkyway/download/nbody_0.92/ (available for Linux, Windows and OS X 64 bit. OS X requires 10.7 Lion). It requires OpenGL 3.2 so it needs a DirectX 10 class GPU (Something newer than Radeon HD 2xxx or GeForce 8000 series)

The screensaver has various mouse and keyboard controls for viewing the simulations. A list of all controls can be accessed by pressing the h or ? keys. You can manipulate the camera by dragging with the mouse and zooming with the scroll wheel.

There are a couple of known problems which you may run into:
    - For workunits with larger timesteps, it seems very jumpy. There's an option to update at some interval instead of at every opportunity. There is another option to stop the simulation to wait for the graphics, so that every timestep is drawn as a frame and the animation is smooth.
    - The orbit trace view will not do what you expect except when using the option to stop the simulation for the graphics
    - The optional axes are ugly
    - The preference descriptions aren't very good

I added a bunch of preferences to the web project preferences, but they're not quite finished yet (e.g. the default values don't show up in the boxes when you don't have them set).

I made a small demo video a while ago demonstrating it: http://milkyway.cs.rpi.edu/milkyway/download/random_stuff/nbody_graphics_sample.mp4

Try it out, and see if you like it! If you run into any problems or have any recommendations for future versions, please leave a comment below.

11) Message boards : Number crunching : freebsd wine and milkyway (Message 54338)
Posted 6 May 2012 by Matt Arsenault
Post:
Some of my computers run freebsd/amd64, and I found out that the computers which running XP is faster to get milkyway credit than freebsd/amd64.
The funniest thing is, running boinc under wine in freebsd is much faster than running in freebsd itself. The different is so big (like in milkyway under wine to complete the task usually take 2-3 hrs, but under freebsd it take 17-20hrs:( in the same computers even if I tried to speed them up using nice).

Anybody knows why running milkyway under freebsd wine is much faster than under freebsd itself in freebsd/amd64?

Edit:
You can see what I mean is by checking computer ID 368456 which is boinc under freebsd/amd64 wine, and compuer ID 436052 is boinc without wine (both are the same computer)

That's slow enough to be consistent with me screwing up and releasing a debug build. I'll try to check later but I don't think I did that
12) Message boards : Number crunching : getting errors with new v1.02 separation application? (Message 54284)
Posted 2 May 2012 by Matt Arsenault
Post:
but saying that, with boinc logging the detected gpu's and their capabilities - someone in the know should be able to make a "check box" option to choose the gpus - and perhaps their capabilities like we can kinda do with the cpu options or post a good thorough "sticky" somewhere with ALL the options and examples of use as hunting for cc_config stuff took quite some time to compile.
Or it could actually be fixed instead of adding such an embarassingly awful workaround.
13) Message boards : Number crunching : getting errors with new v1.02 separation application? (Message 54271)
Posted 1 May 2012 by Matt Arsenault
Post:
So it appears that when Boinc does the hardware query, it's only taking the first card it reads for it's project configurations. I'm tempted to swap cards, but I don't have time for that shenanigans.

If it's app specific, then perhaps the milkyway@home guys need to add 20 lines of code to check the capability of the additional gpu's so that people who are total n00bs like me don't have to muck about in cc_config.

thank you and good night.
BOINC's multiGPU support is completely nonexistent. It only uses the GPU it decides the "most capable." This is why you need to enable all GPUs in the first place with cc_config.xml
14) Message boards : Number crunching : Nbody tasks ending within 1-2 minutes after starting (Message 54178)
Posted 26 Apr 2012 by Matt Arsenault
Post:
It's working normally
15) Message boards : Number crunching : New BOINC Manager = no MWAH WUs (Message 54082)
Posted 20 Apr 2012 by Matt Arsenault
Post:
Should be fixed
16) Message boards : Number crunching : No Work for "new" 4850 (Message 54081)
Posted 20 Apr 2012 by Matt Arsenault
Post:
It was a scheduler problem. It was requiring OpenCL 1.1 for no reason, but the 4850 only supports 1.0 (and barely that).
17) Message boards : Number crunching : Done WUs disappearing? (Message 54080)
Posted 20 Apr 2012 by Matt Arsenault
Post:
The ops pages will show you failure rate per OS and per error code over the last 24 hours or 7 days (although the 7days probably only actually works over the last n-days results were saved from). They aren't that great but I occasionally look through the most popular errors to see if there's a problem. Nearly all the errors are driver problems of one form or another at any given time.
18) Message boards : Application Code Discussion : Errors: Unhanded Exception Task 174285468 (Message 54079)
Posted 20 Apr 2012 by Matt Arsenault
Post:
If the driver crashes and you're getting BSOD's, that makes is a driver (except there are a bunch of potential problems that can cause a driver reset from an application, but the fact that it causes a driver reset is an AMD defect but that's not really related) or possibly hardware problem.
19) Message boards : Number crunching : New BOINC Manager = no MWAH WUs (Message 53996)
Posted 12 Apr 2012 by Matt Arsenault
Post:
It works for me. The other complaint I saw about this was also using an R700. Is it working for anybody else with an R700/Radeon 4xxx?
20) Message boards : Number crunching : Nbody error (Message 53995)
Posted 12 Apr 2012 by Matt Arsenault
Post:
Plum_EMD_100K_9783_0

<core_client_version>6.12.34</core_client_version>
<![CDATA[
<message>
Incorrect function. (0x1) - exit code 1 (0x1)
</message>
<stderr_txt>
Argument parsing error: -0.3255476365281918500000000: unknown option
<search_application> milkyway_nbody_0.80_windows_x86_64__mt.exe 0.80 Windows x86 double OpenMP, Crlibm </search_application>
22:37:51 (4768): called boinc_finish

</stderr_txt>
]]>

There was a batch of bad units at the start of the run (the nbody_100K_Plum_EMD ones) which should disappear and be replaced with Plum_EMD_100k ones which should work.


Next 20

©2020 Astroinformatics Group