Posts by ahorek's team

21) (Message 8368)
Posted 12 May 2024 by ahorek's team
Post:
it's indeed a bug in the AVX implementation, that mostly affects hosts with legacy Intel CPUs like Sandy Bridge on Linux.

we already have a fix, but releasing it would take some time (about a week)

could you test this version? it would be helpful if you could double-check it works as expected on these systems. You can also run regular tasks using the anonymous platform with it.
https://fileport.io/wWq7w3weQMBq

thanks and sorry for the inconvenience!
22) (Message 8363)
Posted 11 May 2024 by ahorek's team
Post:
the new arm64 native application now utilizes asimd instruction set, so it should be more close and likely faster than the emulated x86_64 version.
23) (Message 8362)
Posted 11 May 2024 by ahorek's team
Post:
version 102.20 has been released. It includes new hand-written optimizations for asimd instruction set, that should give a substantial performance boost on supported 64-bit ARM v8 processors like Apple Mx and RPis.
24) (Message 8361)
Posted 10 May 2024 by ahorek's team
Post:
the new app is finally out. Give it a try!
25) (Message 8353)
Posted 23 Apr 2024 by ahorek's team
Post:
we have decided to detect available features by the app instead of having multiple apps and relying on what the boinc client reports. Once (hopefully soon) we release a new version, this boinc feature won't be relevant to us anymore.

btw Primegrid has used the same technique for a long time. http://www.primegrid.com/forum_thread.php?id=10534

until then, there's still an option to run optimized versions on Windows via anonymous platform (if your hardware supports it). Compared to FMA it's up to 20% faster on Intel and slightly faster on AMD (Zen 4). There are rumors that Zen 5 will improve the AVX512 performance, which might be interesting.
26) (Message 8331)
Posted 27 Mar 2024 by ahorek's team
Post:
I've tested the alpha 8.0 version and have no issues getting new tasks. Could you double-check if CPU work is enabled in your account manager preferences?
Also, the client may not request new work, if your host is busy running other projects.
27) (Message 8323)
Posted 27 Mar 2024 by ahorek's team
Post:
you can download avx512 apps from
https://asteroidsathome.net/boinc/download/period_search_10215_windows_x86_64__avx512_win10.exe
https://asteroidsathome.net/boinc/download/period_search_10215_x86_64-pc-linux-gnu__avx512_linux

ensure your CPU has avx512 support (like AMD Zen 4, Intel Ice Lake, Intel Sapphire Rapids etc.).
28) (Message 8316)
Posted 25 Mar 2024 by ahorek's team
Post:
is there anything in the event log?
29) (Message 8312)
Posted 19 Mar 2024 by ahorek's team
Post:
Keith's issue was unrelated to the android's app.

The new android app still hasn't been released, so the "Number of lc points is greater than POINTS_MAX = 1000" problem is still present. The code is ready, we just need to build it and upload a new release to the server. Unfortunately, I don't have access to the server, so it has to wait for Georgi & Radim once they have time to do it.
30) (Message 8308)
Posted 18 Mar 2024 by ahorek's team
Post:
Cause I always get SSE3 app and never get AVX/FMA ones. I think my CPU 5700X support these.

you're right, the fact that BOINC is stuck with SSE3 and never tries new, more optimized apps (on some hosts) is a bug either with the server configuration or the BOINC server code.
however, there are more related issues with this approach. BOINC doesn't reliably report all necessary features on all platforms. Windows is the most important one, but the same issue is on BSD and Macs all of them have their specific way of detecting features, so we have decided to develop a new app that selects the best implementation at runtime instead of relying on what the BOINC client says. It's already in the testing phase.

if you have this issue and don't want to wait for the next release, using an anonymous platform as Keith suggested is a workaround to get the best performance. I admit it's inconvenient for users who aren't familiar with it, but please be patient, we're working on a solution :)
31) (Message 8295)
Posted 9 Mar 2024 by ahorek's team
Post:
I have some news from admins, there was a DOS attack from China's network on the website 1000 to 8000 connections at a time. They took some preventive measures. It's hard to say if there's any connection with the Challenge though, it could be just a coincidence.
32) (Message 8292)
Posted 9 Mar 2024 by ahorek's team
Post:
sorry for the trouble, the server should easily handle +200 (/4500) more users during the challenge. Perhaps the webserver ran out of free connections or more likely someone's deliberately hitting it too hard (by downloading an insane amount of work, crawling the website etc.). Unfortunately, only the university has access to the firewall & logs and it may take some time for them to investigate what exactly has happened especially during a weekend. They're aware of the issue, but I don't know more yet.
33) (Message 8262)
Posted 7 Feb 2024 by ahorek's team
Post:
sometimes the queue gets dry and it takes a day until admins fill it up again.

you can monitor available work here
https://grafana.kiska.pw/d/boinc/boinc?orgId=1&var-project=Asteroids@Home&from=now-30d&to=now
34) (Message 8248)
Posted 29 Jan 2024 by ahorek's team
Post:
yes, it's most likely 2 issues:
1. scheduling problem - we have no control over task scheduling. It doesn't seem to be related to this project, rather than to Boinc (client) itself. Please discuss the issue on their github https://github.com/BOINC/boinc/issues/5402
2. POINTS_MAX - we're aware of this issue. Android's app is outdated and it can't handle some types of tasks anymore. We are in the process of recompiling old apps with the recent code and the android app seems to be the last one... we'll fix it eventually, but it may take some time, please be patient :)
35) (Message 8245)
Posted 28 Jan 2024 by ahorek's team
Post:
Bitdefender no longer detects it as malicious software, so I guess, they've fixed it. Note that our binary is still the same :)
36) (Message 8236)
Posted 23 Jan 2024 by ahorek's team
Post:
Hello gentleman74,
the current rpi version is compiled with march=armv8-a

+simd option - is already enabled by default, see https://gcc.gnu.org/onlinedocs/gcc/AArch64-Options.html
+crc - these instructions aren't used by the app, so enabling this option will have no effect

unlike x86 apps, we don't have handwritten optimizations for ARMs, so NEON optimizations depend on the compiler's auto vectorization capabilities.

unfortunately playing with these compiler flags without code changes and manual optimizations won't make a difference. It's not that easy, but if you're interested in development I can give you a link to the recent source code (PM me).
37) (Message 8224)
Posted 19 Jan 2024 by ahorek's team
Post:
Hi, the app has some driver issues on slow cards, especially on the Radeon R5 family + Windows

we will try to fix it in the code, but for now, disabling TDR protection might work
https://learn.microsoft.com/en-us/windows-hardware/drivers/display/tdr-registry-keys

anyway, don't expect a good performance. I would estimate ~6 hours for a single work unit. Your CPU doesn't even support AVX, but it's most likely a more effective option for this project than utilizing a GPU.
38) (Message 8197)
Posted 13 Jan 2024 by ahorek's team
Post:
are you running 20/30 gpu tasks in parallel? that's not a good idea and it won't be any better.

perhaps try with recent drivers? 516.94 is a bit older
https://www.nvidia.com/Download/driverResults.aspx/216860/en-us/
39) (Message 8152)
Posted 27 Nov 2023 by ahorek's team
Post:
you can test each application individually
https://asteroidsathome.net/boinc/download/period_search_10214_windows_x86_64__sse3_win10.exe
https://asteroidsathome.net/boinc/download/period_search_10213_windows_x86_64__avx_win10.exe
https://asteroidsathome.net/boinc/download/period_search_10213_windows_x86_64__fma_win10.exe

download a sample workunit, rename it to "period_search_in" and place it in the same folder. Then you can reliably compare runtimes for each version.

results may vary based on the CPU you have, there are also other things to consider - power consumption, running multiple WUs, operating system etc.

usually, AVX/FMA apps have similar performance. Some server Xeons have better FMA performance so they may run faster with FMA. On the other hand, even if Bulldozer/Piledriver support superior AVX, it's actually slower than SSE3.

to give you a rough idea about general performance, here are (single thread) results on some architectures I have
40) (Message 8116)
Posted 18 Oct 2023 by ahorek's team
Post:
https://asteroidsathome.net/boinc/download/period_search_10214_windows_x86_64__sse2_win10.exe

https://www.virustotal.com/gui/file/41b0472160ebb829dc70a3513bb541daa4a08fb0535bcf943f438adbe98c4855?nocache=1
11/72 AV suites flag it as potentially suspicious, but BitDefender is known for high false positives.

Boinc apps are CPU-intensive (on purpose) which could be one reason why they can look suspicious. If you want to be 100% sure, try reporting it https://www.bitdefender.com/consumer/support/answer/29358/


Previous 20 · Next 20