Posts by MarkJ
|
1)
Posted 19 Sep 2020 by MarkJ Post: Gianfranco has also pushed the 7.16.11 into buster-backports for those of you running Debian. If you are running Ubuntu he has a ppa here |
|
2)
(Message 7067)
Posted 19 Sep 2020 by MarkJ Post: Upgrade to BOINC 7.16.11 |
|
3)
Posted 8 Aug 2020 by MarkJ Post: Since that last post the Raspberry Pi foundation have released Raspberry Pi OS (64 bit) which is really the aarch64 version of Debian with a few tweaks for the Pi4 (it works on Pi3 but they only have 1GB memory). To sum up the current state of Rpi apps on Asteriods at home: 1. The ARMv6 Pi’s all fail (that’s Pi zero and the original Pi). 2. The current app works on Pi2 and up but doesn’t use neon which makes it extremely slow. 3. There isn’t any support for aarch64. |
|
4)
Posted 12 May 2020 by MarkJ Post: In case Kyong reads this, here are the CPU features of the different Raspberry Pi's that I have at hand. I don't have a Pi Zero but expect they would be the same as a Pi B+. I think -march=armv7-a should cover everything from a Pi2 upwards. Pi model B+ (ARMv6) half thumb fastmult vfp edsp java tls Pi2 model B (ARMv7) half thumb fastmult vfp edsp neon vfpv3 tls vfpv4 idiva idivt vfpd32 lpae evtstrm Pi3 model B & Pi3 model B+ (ARMv7) half thumb fastmult vfp edsp neon vfpv3 tls vfpv4 idiva idivt vfpd32 lpae evtstrm crc32 Pi4 model B (ARMv8 armhf mode) half thumb fastmult vfp edsp neon vfpv3 tls vfpv4 idiva idivt vfpd32 lpae evtstrm crc32 Pi4 model B (ARMv8 aarch64 mode) fp asimd evtstrm crc32 cpuid |
|
5)
Posted 11 May 2020 by MarkJ Post: I have converted some of the 3B+ (1 hour labor for each one) to the Buster os and they seem to be working now. Although units have a predicted completion of 2 days. It might be right. My Pi4’s came in at 35 hours and they’re faster. Usually with OS upgrades I find it easier to reimage the SD card from the iso file you download of Raspbian lite.and then reinstall BOINC on them. I have the config stuff on an NFS server which is just a Pi with an external hard drive. |
|
6)
Posted 11 May 2020 by MarkJ Post: I expect the project has compiled the app using the current compiler. It doesn’t look like they’ve used the -march=armv7-a compile option either because tasks take so long on the Pi4. As for the ARMv6 failing that suggests an app problem, probably an incorrect compile option again. We would need Kyong to recompile them. |
|
7)
Posted 9 May 2020 by MarkJ Post: Upgrading to Buster didn't get me going, the GLIBC error is now replaced with this error instead: process got signal 4 I can’t test point 2, but are you sure you upgraded to buster? The kernel should be 4.19.97 and the version of BOINC should be 7.14.2. Yours appear to be older. I tried a Pi4 on here after the app upgrade and the tasks took 35 hours so I don’t think they are using neon which the Pi2 and later have. I hate to think how long a 1st generation Pi would take. |
|
8)
Posted 9 May 2020 by MarkJ Post: They updated their ARM application a couple of days ago (version 102.13) and it now needs GLIBC 2.28 to run, which I don't have since I use Raspbian Stretch lite. Is there some reason why you aren’t upgrading to buster-lite? The Glibc comes with the operating system, so upgrading it and probably BOINC should get you going again. |
|
9)
Posted 15 Jun 2019 by MarkJ Post: I think the issue is Kyong is too busy running the project on his own. He doesn't have time to read the message boards or post about things. He really needs some other people to keep an eye on the message boards and tell him when the server has issues and maybe help with other maintenance tasks. |
|
10)
(Message 6177)
Posted 24 Feb 2019 by MarkJ Post: The GTX 1660 Ti would be in more people’s reach. l ask that we get the app updated to support the Turing based cards please. Hopefully it would just be a matter of recompiling it with the latest CUDA compiler but I bet it’s not that simple. |
|
11)
Posted 1 Dec 2018 by MarkJ Post: Update Sent a private message to Kyong and pointed him to this message thread, hopefully he will look into it. |
|
12)
Posted 1 Dec 2018 by MarkJ Post: The server software is supposed to try different apps to see which is faster and more reliable. However Asteroids doesn't have it configured properly so its not trying all the apps. BOINC project servers have a configuration parameter <version_select_random_factor> which isn't set high enough. They can turn on <debug_version_select> to debug it. It needs at least 10 successfully validated results for each app before it can get an APR (Average Processing Rate). You can see the APR figures for each app when you look at your computer details under your account (on here, after you sign in). Its supposed to pick the fastest app and occasionally send one of the others to see how they go. |
|
13)
(Message 5955)
Posted 15 Sep 2018 by MarkJ Post: The current app is rather dated and does not utilise the neon extension that the ARM v7 and later have. In theory all it would take to use them would be a recompile with an up to date gcc with the appropriate options. Unfortunately Kyong is rather busy just keeping the project running. As for using the Videocore IV that would require coding changes and there is no guarantee that it would be faster. You also have to remember there is only one Videocore but there are 4 CPU cores so it would probably be more beneficial to optimise the CPU part. |
|
14)
(Message 5890)
Posted 8 Jul 2018 by MarkJ Post: How many of the users are ready to donate some money to pay for the additional disk space and any other changes needed to make it useful? Does the project offer a way to send such money? Kyong has mentioned they're getting another server (hopefully a storage one). Not sure what time frame they are looking at though. From the server status page I understand they have a couple of general purpose servers (IBM x3550 M4's) with 8 x 2.5 inch drive bays. They use expensive SAS disks that only go up to 600GB each but have higher spindle speed (10k RPM). Great for a database or compute server but not so good for storage. |
|
15)
(Message 5733)
Posted 10 Mar 2018 by MarkJ Post: According to the Server Status page there are 230,000 work units in progress, so we got through 40,000 since my last post. Its still going to take 3 to 4 days before we get it down to the 100,000 needed to generate more work. Hoarding work leads to the database getting bloated and those hosts are probably not able to complete the work in time. Like I said before the project needs some limits on how many can be cached, which is why this project goes from “feast to famine” regularly. Seti have limits of 100 CPU and 100 per GPU, which is a bit low but if Asteroids were to adopt something like that there would be a regular stream of work all the time and we wouldn’t have to wait two weeks to get tasks validated. |
|
16)
(Message 5725)
Posted 10 Mar 2018 by MarkJ Post: I read elsewhere that there needs to be no work to send and less than 100,000 tasks in progress before it will generate more. So those of you with the 270,000+ tasks stop hoarding them and finish them. Perhaps the project needs to implement restrictions on how many tasks in progress each machine can have. |
|
17)
(Message 5700)
Posted 20 Feb 2018 by MarkJ Post: The 384.111 driver seems to work, so I think we can safely say it’s something in the older driver. Due to a bug in the BOINC client it truncates the CUDA minor version number to two digits so it appears as 384.11. I reported it on the BOINC mailing lists when the driver came out. |
|
18)
(Message 5594)
Posted 17 Dec 2017 by MarkJ Post: Why wouldn’t you use Raspbian? Its got better support. In the astronomy field Asteroids, Einstein and Seti all have apps available for it. Unfortunately the Asteroids app isn’t optimised for it so is rather slow. |
|
19)
(Message 5593)
Posted 17 Dec 2017 by MarkJ Post: On the Applications page it shows for the Raspberry Pi plan class arm doing 16 GFlops and armv6l doing zero so I would suggest dropping support for arm v6 completely. |
|
20)
(Message 5555)
Posted 17 Nov 2017 by MarkJ Post: Resurrecting an old thread. It would seem the server selection requires 1000+ tasks before trying a different app. I have PM’ed Kyong the relevant config option but have yet to receive a reply. |
Next 20