I'm trying to run BRP built from the source code on a computer with under 256 MB of RAM.
I keep getting the error:
Binary Radio Pulsar Search needs 247.96 MB of RAM but only 185.57 is avaliable for use.
Copyright © 2024 Einstein@Home. All rights reserved.
Minimum RAM
)
I did a little more research and it looks like this project only requires around 90 MB of ram. Does anyone know why this error is coming up?
RE: I did a little more
)
There are 2 versions of the BRP apps, one that runs on GPU and another that runs on CPU, the one that needs less than 90Mb is the GPU one, that uses the GPU memory instead of the main memory...
Hmmm....is this on the
)
Hmmm....is this on the Raspberry Pi, perhaps?? Android ??
I would be very interested in hearing about BOINC apps on ARM architectures. Feel free to contact me via private message.
Cheers
HBE
RE: Android ?? I would
)
There is one calculating app for android, it crunches for collatz
https://play.google.com/store/search?q=collatz&c=apps
other BOINC apps are for PC-control and statistc purposes
https://play.google.com/store/search?q=boinc&c=apps
Alexander
@milkyway an interesting post
)
@milkyway an interesting post can be found her:
http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=3026&sort=5
regarding crunching on raspberri pi.
RE: @milkyway an
)
Hi!
Thanks for the pointer. But I don't quite agree with the poster that now the GPU on the RaspPi is open for general purpose computing. My understanding is that the firmware for the GPU is still closed and will likely remain so. The graphics driver is now open source, but from what I have read, this will not be enough to (say) compute an FFT on the GPU.
Cheers
HB
Hi! A little update on the
)
Hi!
A little update on the Raspberry Pi situation, after some research and tests:
1) There are now several Operating System images available for the RaspPi, the one that would probably best work with BOINC & number crunching is the "Raspbian" Linux distribution, which is based on Debian "Wheezy". BOINC is included in the distribution, so it can be downloaded and installed with a single command line
2) Compiling the app: With some minor changes, the app can be built on the Raspi and seems to give valid results, more testing would have to be done, tho. Tests were done on the new 512 MB version of model B.
3) Performance: At the level of overclocking that most Raspi users have reported as save, a full BRP4 task would take ca 1 Mio seconds, or 12 days. However, the current BRP4 tasks are really a bundle of 8 independent sub-tasks. One option would be to have a special job queue for ARM based small devices such as RaspPis and Androids using "unbundled" tasks. That would take some time to set up tho. I don't see us do this specifically for the Raspi, but once we get this working on Android, we should consider it (just my personal opinion, tho).
Does it make sense to crunch on the Raspi? I think there are actually BOINC projects that are better suited for the Raspi, like "Radioactive@Home" and "QuakeCatcherNetwork" (QCN). In those two projects, which are both experimenting with Raspberry Pi apps, the Pi would not be used so much for number crunching but mainly as a controller for environmental sensors (seismic and radioactivity, respectively). The low price and low power consumption would make the Pi ideal for outdoor installations, for example. Still E@H on Raspi might be a nice way to "bake-in" your Pi and test the stability at a given overclocking setting.
As Android phones and tablets are often using a newer generation of ARM CPUs (ARMv7 as opposed to the Raspi's ARMv6), the situation looks brighter on that platform. There are also a lot more around than Raspis (however the former would only crunch when connected to a charger, while Raspis are typically connected permanently to the mains and would be able to crunch 24/7).
Cheers
HB
Another link to the MW
)
Another link to the MW forum:
http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=2826#53769
I published new project on GitHub https://github.com/matszpk/native-boinc-for-android. This is fully fledged native BOINC client for Android and Boinc manager based on AndroBOINC.
I prepared applications for MilkyWay@Home, Primegrid and Enigma@Home (Polish project).