Posts by Sergei Chernykh

1) Message boards : Number crunching : Drop in WU size? (Message 711)
Posted 1 day ago by Sergei Chernykh
Post:
We've had a bunch of very GPU-friendly (highly parallelizable) workunits recently, but as far as I can tell, we're back to regular workunits again. All workunits were calibrated to have approximately the same amount of computations for CPU version.

Run times on GPU are usually stable too, but sometimes they can vary a lot.
2) Message boards : News : The search up to 10^20 (Message 707)
Posted 2 days ago by Sergei Chernykh
Post:
What would the RAM requirement be for the search to 10^21? I would guess that it wouldn't quite be a 10x jump to 15GB, but it could still be quite substantial.

The memory required is O(sqrt(N)), so it will not be a 10x jump:
CPU version would require ~1400 MB of memory.
GPU version would require ~3700 MB of GPU memory.
3) Message boards : Random stuff : Amicable Pair Reporting Format (Message 702)
Posted 23 days ago by Sergei Chernykh
Post:
Thanks - I presume X means exotic? ( in relation to type)

Yes.
4) Message boards : News : Credit badges (Message 700)
Posted 23 days ago by Sergei Chernykh
Post:
"Pool" user is opaque, i.e. the project doesn't know who exactly runs tasks behind pool's user account. It can only be done on corresponding pool's side.
5) Message boards : Random stuff : Amicable Pair Reporting Format (Message 697)
Posted 23 days ago by Sergei Chernykh
Post:
The first line has format "Type Discoverer Year". You can read how the type of amicable pair is defined here: https://en.wikipedia.org/wiki/Amicable_numbers#Regular_pairs
There are two discoverers for each amicable pair in this BOINC project because each work unit is processed by two different users and then the results from both users are compared. These two users are then listed in the order they completed the corresponding work unit.
6) Message boards : Bug tracker : Computation error at 6 seconds (Message 695)
Posted 26 days ago by Sergei Chernykh
Post:
It's a 560 Ti so no it has 1024 of memory.

Also I added another project and allocated this card towards it to see if same error. Did not fail, so successfully crunching with this card on another project.

If that 2GB memory minimum is a project specific thing then I guess all is working as intended, must have missed the line that mentioned it.

You need to have 2 GB of GPU memory to run GPU version of Amicable Numbers, so it's normal that 560 Ti errors out.
7) Message boards : Bug tracker : Computation error at 6 seconds (Message 693)
Posted 28 days ago by Sergei Chernykh
Post:
OpenCL.cpp, line 440: clEnqueueWriteBuffer returned error -4

Error -4 is CL_MEM_OBJECT_ALLOCATION_FAILURE. What are your second GPU's specs? Does it have at least 2 GB of memory?
8) Message boards : Number crunching : Auto-ban faulty hosts (Message 691)
Posted 29 days ago by Sergei Chernykh
Post:
It can still run CPU tasks. But if you consider only GPUs for Amicable Numbers, this host will be auto-removed after 3 months of inactivity.
9) Message boards : Number crunching : Very long WUs? (Message 686)
Posted 8 Dec 2017 by Sergei Chernykh
Post:
more than 1 hr to finish a single wu ? with 100% cpu and gpu usage ?
used to be only 15 mins, all of my cards are showing this.
something wrong ???????

Actually, there is something wrong with your settings. You've set kernel size for NVIDIA GPUs to 18 which is too small for GTX 1070/1080. Set it to at least 21 here: https://sech.me/boinc/Amicable/prefs.php?subset=project

P.S. I've set it to 21 for you, but you can always change it to whatever value you want. Let's hope it improves WU run times.
10) Message boards : Number crunching : Very long WUs? (Message 685)
Posted 8 Dec 2017 by Sergei Chernykh
Post:
Are you talking about GPU version? It varies depending on WU - some ranges get processed fast (compared to CPU), some don't.

P.S. The project started the new subrange "2^2*11*..." a few days ago, so it's normal that WU run times changed.
11) Message boards : Number crunching : WU size or difficulty change? (Message 680)
Posted 4 Dec 2017 by Sergei Chernykh
Post:
Hi guys!

Yes I've noticed it's getting harder atm as we moved up to "2^2*..." and discoveries of new number has almost stopped.

Sergei, when are you expecting us to to hit that "range" when it would be possible to find new ones and could you please explain why?

...just planning to put some more in to find them, but electricity ain't cheap over in Finland and I want to hit that 100 000 found ones as my next goal =)

Mankka*

Just as I answered before, the project has now reached "2^2*11*..." subrange which is relatively rich with new amicable pairs.
12) Message boards : Number crunching : BOINC Client 7.8.4 recommended for MAC OS (Message 679)
Posted 3 Dec 2017 by Sergei Chernykh
Post:
jamezz
Mac version is compiled from the same source code, so there shouldn't be more than 10-20% difference in run times compared to Windows and Linux versions.
13) Message boards : Getting started : Errors on Geforce GTX 770 (Message 675)
Posted 29 Nov 2017 by Sergei Chernykh
Post:
it needs 2GB memory and the card has only 2GB so it's stupid to try to run 2 WUs parallel on that card, right?

Yes, each WU requires 2 GB of memory, so you need 4 GB to run 2 WUs in parallel.
14) Message boards : Getting started : Errors on Geforce GTX 770 (Message 673)
Posted 29 Nov 2017 by Sergei Chernykh
Post:
I would say to try reducing kernel size here: https://sech.me/boinc/Amicable/prefs.php?subset=project
But you'll need to use a separate user account for it.
15) Message boards : Getting started : Errors on Geforce GTX 770 (Message 672)
Posted 29 Nov 2017 by Sergei Chernykh
Post:
Try to reboot and run BOINC only, don't start any programs that use GPU (including web-browser). If errors go away, it means there is still not enough GPU memory to run both Amicable and other software at the same time.
16) Message boards : Number crunching : Ways too many credits via Amicable...? (Message 670)
Posted 27 Nov 2017 by Sergei Chernykh
Post:
You can also limit number of CPUs used by Amicable Numbers here: https://sech.me/boinc/Amicable/prefs.php?subset=project - set "Max # CPUs" to whatever value suits you. BOINC should automatically schedule all other CPUs to PrimeGrid tasks (in your case). It should work better than split settings in BOINC client.
17) Message boards : Number crunching : no task (Message 668)
Posted 26 Nov 2017 by Sergei Chernykh
Post:
Check BOINC client's log to see what went wrong. But it already looks like this new machine has started processing tasks.
18) Message boards : Number crunching : Ways too many credits via Amicable...? (Message 665)
Posted 22 Nov 2017 by Sergei Chernykh
Post:
1) Credit points from different projects can't be compared directly.
2) Are you sure BOINC actually splitted run time between two projects the way you intended?
19) Message boards : Number crunching : Long running work units now, but sitting at 100% and not uploading? (Message 662)
Posted 9 Nov 2017 by Sergei Chernykh
Post:
Yes, something was definitely wrong with these two WUs. Let's hope it doesn't happen again.
20) Message boards : Number crunching : Long running work units now, but sitting at 100% and not uploading? (Message 660)
Posted 9 Nov 2017 by Sergei Chernykh
Post:
You should also check your PC - one of the tasks finished with message "Disk usage limit exceeded".[/url]
21) Message boards : Number crunching : Long running work units now, but sitting at 100% and not uploading? (Message 659)
Posted 9 Nov 2017 by Sergei Chernykh
Post:
Current work units are actually shorter than average. Just cancel WUs that can't finish.
22) Message boards : Number crunching : Interesting distribution of numbers - my discoveries stopped. (Message 657)
Posted 7 Nov 2017 by Sergei Chernykh
Post:
Read the last two messages here: https://sech.me/boinc/Amicable/forum_thread.php?id=95&postid=639#639
23) Message boards : Bug tracker : Using too much memory... (Message 655)
Posted 7 Nov 2017 by Sergei Chernykh
Post:
Try to reduce kernel size for AMD GPUs in project preferences: https://sech.me/boinc/Amicable/prefs.php?subset=project
It should reduce memory size used by GPU tasks.
24) Message boards : Bug tracker : Using too much memory... (Message 652)
Posted 7 Nov 2017 by Sergei Chernykh
Post:
Current applications use ~440 MB per one CPU task and ~2 GB per one GPU task. Don't run many tasks at the same time: one CPU + one GPU task is enough.
25) Message boards : Bug tracker : "Show names" in Task list - only shows task numbers. (Message 650)
Posted 27 Oct 2017 by Sergei Chernykh
Post:
It was turned off because of performance issues.
26) Message boards : Bug tracker : AMD mesa OpenCL kernel compilation error, fix suggestion (Message 645)
Posted 20 Oct 2017 by Sergei Chernykh
Post:
It looks like it just crashes on your PC now when it tries to compile OpenCL code:

SIGSEGV: segmentation violation
Stack trace (19 frames):
../../projects/sech.me_boinc_Amicable/amicable_OpenCL_v_2_06(boinc_catch_signal+0x65)[0x4373a5]
/usr/lib/libpthread.so.0(+0x117e0)[0x7f967051a7e0]
/usr/lib/libLLVM-4.0.so(_ZN4llvm11Instruction15eraseFromParentEv+0x2c)[0x7f961c4b4b9c]
/usr/lib/libLLVM-4.0.so(+0x1938224)[0x7f961d60b224]
/usr/lib/libLLVM-4.0.so(+0x1937c3b)[0x7f961d60ac3b]
/usr/lib/libLLVM-4.0.so(+0x1939890)[0x7f961d60c890]
/usr/lib/libLLVM-4.0.so(_ZN4llvm13FPPassManager13runOnFunctionERNS_8FunctionE+0x311)[0x7f961c4df791]
/usr/lib/libLLVM-4.0.so(_ZN4llvm13FPPassManager11runOnModuleERNS_6ModuleE+0x32)[0x7f961c4df7d2]
/usr/lib/libLLVM-4.0.so(_ZN4llvm6legacy15PassManagerImpl3runERNS_6ModuleE+0x312)[0x7f961c4def62]
/usr/lib/libMesaOpenCL.so.1(+0x1ddb4e)[0x7f961faf8b4e]
/usr/lib/libMesaOpenCL.so.1(+0x1de160)[0x7f961faf9160]
/usr/lib/libMesaOpenCL.so.1(+0x1d9862)[0x7f961faf4862]
/usr/lib/libMesaOpenCL.so.1(+0x1cac04)[0x7f961fae5c04]
/usr/lib/libMesaOpenCL.so.1(+0x1aa16f)[0x7f961fac516f]
/usr/lib/libOpenCL.so.1(clBuildProgram+0x4b)[0x7f96700ed65b]
../../projects/sech.me_boinc_Amicable/amicable_OpenCL_v_2_06[0x411629]
../../projects/sech.me_boinc_Amicable/amicable_OpenCL_v_2_06[0x406a21]
/usr/lib/libc.so.6(__libc_start_main+0xea)[0x7f966fa4a4ca]
../../projects/sech.me_boinc_Amicable/amicable_OpenCL_v_2_06[0x407855]
27) Message boards : Bug tracker : AMD mesa OpenCL kernel compilation error, fix suggestion (Message 644)
Posted 20 Oct 2017 by Sergei Chernykh
Post:
Fixed it in version 2.06: https://sech.me/boinc/Amicable/apps.php
https://github.com/SChernykh/Amicable/commit/b7cbf824d13774f85e0b53c00bbacc6757bb9293
28) Message boards : Bug tracker : AMD mesa OpenCL kernel compilation error, fix suggestion (Message 643)
Posted 19 Oct 2017 by Sergei Chernykh
Post:
You don't need to edit the binary, source code is here: https://github.com/SChernykh/Amicable/tree/boinc-opencl-version-128-bit
You can edit and compile it yourself or wait for me. I'll update the GPU version tomorrow.
29) Message boards : Number crunching : WU size or difficulty change? (Message 640)
Posted 14 Oct 2017 by Sergei Chernykh
Post:
Judging by the distribution of known pairs in "2^2*..." range, I don't expect many new pairs until we start "2^2*11*..." subrange which we'll reach somewhere in the beginning of December at current speed. But I plan to request adding Amicable Numbers to main BOINC projects list (both on BOINC web-site and in BOINC client), so it'll hopefully attract some more users.

P.S. Don't worry, there are still ~1,388,000 amicable pairs below 1020 left undiscovered, so they will add ~2,776,000 to user counters of found pairs (each pair is found by two users).
30) Message boards : Bug tracker : Amicable pairs discovered not updating... (Message 635)
Posted 11 Oct 2017 by Sergei Chernykh
Post:
Yes, it's related to the change from "2*..." to "2^2*...". The distribution of amicable pairs is not uniform, they tend to come in waves - each range has a lot of small waves (distributed seemingly randomly) and one big wave in the end. There are still new pairs being found currently - just not that many, you can see today's new pairs here.

P.S. It's definitely not a bug because the search didn't miss any known pair in the range "2*..."
31) Message boards : Number crunching : WU size or difficulty change? (Message 629)
Posted 3 Oct 2017 by Sergei Chernykh
Post:
Yes, WUs changed a few days ago. The project was running through "2*..." range, but it's running through "2^2*..." range now and run times per WU can be 30-40% higher than before.
32) Message boards : News : MacOS GPU version released (Message 625)
Posted 16 Sep 2017 by Sergei Chernykh
Post:
GPU requirements are the same as for Windows and Linux versions: AMD (HD 5xxx or newer) or NVIDIA (GTX 4xx or newer) GPU with OpenCL support and at least 2 GB of video memory.

The MacOS GPU version has been confirmed to work on OS X El Capitan and MacOS Sierra. Older MacOS versions have poor OpenCL support and may fail to run the GPU version.
33) Message boards : Bug tracker : Total number of tasks falls (Message 624)
Posted 13 Sep 2017 by Sergei Chernykh
Post:
Work units that were completed and validated more than one week ago are automatically deleted.
34) Message boards : Number crunching : GPU version requirements for the search up to 10^20 (Message 622)
Posted 13 Sep 2017 by Sergei Chernykh
Post:
I know this is a low priority - I just got my (modified) MacPro up on Amic, and CPU tasks are running now. It appears there is no Mac version of the GPU applications... :-/

(Yeah, I know, I may be the ONLY one on this project with a Mac with an AMD card in it...)

I finally figured out how to compile OpenCL version for MacOS and added AMD and NVIDIA OpenCL application versions for MacOS 10.7 or newer: https://sech.me/boinc/Amicable/apps.php
35) Message boards : Number crunching : How are the APs sorted? (Message 621)
Posted 11 Sep 2017 by Sergei Chernykh
Post:
The project discovered a first example of type (10,4) amicable pair:
10,4 "BOINC: Cautilus, Hooker63" 2017
60839375973463962290=2*5*13*17*41*59*73*83*109*127*241*563
75707483892585778510=2*5*503*77549*261071*743423

Finding new types of amicable pairs is a rare event.

It was also close to beat the lowest m/n ratio for known amicable pairs, but didn't break the record yet.
36) Message boards : Number crunching : How are the APs sorted? (Message 619)
Posted 11 Sep 2017 by Sergei Chernykh
Post:
They're sorted by the smaller member in a pair. The number "65" is type of amicable pair. In short, first digit is how many prime factors are there in the smaller member that don't divide the larger member, second digit is how many prime factors are there in the larger member that don't divide the smaller member. You can find strict definition here: http://mathworld.wolfram.com/AmicablePair.html
37) Message boards : Number crunching : GPU version requirements for the search up to 10^20 (Message 617)
Posted 1 Sep 2017 by Sergei Chernykh
Post:
I'll look into making Mac GPU version next week.
38) Message boards : News : The search up to 10^20 (Message 614)
Posted 16 Aug 2017 by Sergei Chernykh
Post:
I haven't decided yet. The search up to 10^20 will take more than 4 years at current speed, so 10^21 will take more than 40 years which is just too much. So it all depends on how many volunteers will be running the project in 4 years from now.
39) Message boards : News : The search up to 2^64 is complete! (Message 612)
Posted 16 Aug 2017 by Sergei Chernykh
Post:
Congratulations to everyone who took part in this search!

- There are 2,390,655 amicable pairs with smaller member below 2^64 in total
- BOINC volunteers found 552,874 new amicable pairs below 2^64

A more detailed analysis of amicable pairs distribution will follow later this month. And don't forget to edit your preferences and check that you're all set up to run the current search up to 10^20.
40) Message boards : News : Credits for the old application (search up to 2^64) doubled (Message 609)
Posted 13 Aug 2017 by Sergei Chernykh
Post:
There are no new WUs, I've just spawned more tasks for existing WUs to speed up things. Check https://sech.me/boinc/Amicable/workunit.php?wuid=2591144 for example. It has 3 tasks "In progress" now, so it should finish as soon as any of these 3 tasks is finished.
41) Message boards : News : Credits for the old application (search up to 2^64) doubled (Message 607)
Posted 12 Aug 2017 by Sergei Chernykh
Post:
Are we all done with 2^64?

Yes, all WUs are out, we're now waiting for current tasks in progress to complete.
42) Message boards : Number crunching : Extend deadline fpr 10^20 WU? (Message 604)
Posted 9 Aug 2017 by Sergei Chernykh
Post:
Would it lead to issues on the server side when extending the deadline a few days?

The problem is that the project uses replication: each workunit is sent to two volunteers, and if one of them holds it, the other one will have to wait until he gets credit for it.

Even if you miss deadline for some workunits, they'll just be sent to someone else. Just make sure that if you're already running a workunit, wait until it completes if you're going to switch to other projects for a few days.
43) Message boards : Bug tracker : RAM problem (Message 601)
Posted 7 Aug 2017 by Sergei Chernykh
Post:
Nothing is wrong, the last batches for the 2^64 application need a lot of RAM: https://sech.me/boinc/Amicable/forum_thread.php?id=88
44) Message boards : Number crunching : Extend deadline fpr 10^20 WU? (Message 600)
Posted 7 Aug 2017 by Sergei Chernykh
Post:
The deadline is 3 days, current workunits finish in only ~3 hours on your PC. I don't see a big issue here.
45) Message boards : Getting started : A note to the Project Dev/Lead (Message 596)
Posted 3 Aug 2017 by Sergei Chernykh
Post:
1) 2^64 app needs a lot of memory per thread now, it can't be avoided at this stage of the search.
2) 10^20 is not supposed to run on GPUs with less than 2 GB memory. It's probably a bug that BOINC assigned it to your GT 440 OEM 1.5 GB.
3) 10^20 app uses only integer arithmetic, you can look at performance comparison here: https://sech.me/boinc/Amicable/gpu_list.php

P.S. 10^20 app does use 32-bit floats in a few places outside of main loops, but they take a negligible amount of total execution time. It can even run on GPUs that don't support double precision at all.
46) Message boards : Bug tracker : Daily quota error in event log? (Message 593)
Posted 26 Jul 2017 by Sergei Chernykh
Post:
looks like host #2209 decided to take its turn at getting blacklisted. lol
i replaced the RAM today, and set up dual channel properly (not colour-coded on this board). hopefully this clears things up for the future. :)

You set kernel size = 23 for NVIDIA cards. Host #2209 has GTX 750 ti and it just doesn't have enough memory to run kernerl size = 23. Try to decrease it.
47) Message boards : Number crunching : GPU version requirements for the search up to 10^20 (Message 590)
Posted 21 Jul 2017 by Sergei Chernykh
Post:
I've checked BOINC client source code, and found this in "client/gpu_nvidia.cpp":
// return 1/-1/0 if device 1 is more/less/same capable than device 2.
// factors (decreasing priority):
// - compute capability
// - software version
// - available memory
// - speed
//
// If "loose", ignore FLOPS and tolerate small memory diff
//
int nvidia_compare(COPROC_NVIDIA& c1, COPROC_NVIDIA& c2, bool loose) {
    if (c1.prop.major > c2.prop.major) return 1;
    if (c1.prop.major < c2.prop.major) return -1;
    if (c1.prop.minor > c2.prop.minor) return 1;
    if (c1.prop.minor < c2.prop.minor) return -1;
    if (c1.cuda_version > c2.cuda_version) return 1;
    if (c1.cuda_version < c2.cuda_version) return -1;
    if (loose) {
        if (c1.available_ram> 1.4*c2.available_ram) return 1;
        if (c1.available_ram < .7* c2.available_ram) return -1;
        return 0;
    }
    if (c1.available_ram > c2.available_ram) return 1;
    if (c1.available_ram < c2.available_ram) return -1;
    double s1 = c1.peak_flops;
    double s2 = c2.peak_flops;
    if (s1 > s2) return 1;
    if (s1 < s2) return -1;
    return 0;
}


You can try to move this into the beginning of "nvidia_compare":
    if (c1.available_ram > c2.available_ram) return 1;
    if (c1.available_ram < c2.available_ram) return -1;

and then build BOINC client from source.
48) Message boards : Number crunching : GPU version requirements for the search up to 10^20 (Message 587)
Posted 20 Jul 2017 by Sergei Chernykh
Post:
BOINC ver 7.6, however, wrongfully detects it as GPU0 and reports ONLY this card to the Amicable Numbers Server. The Monitor is NOT connected to this card.

I think the problem is that it doesn't detect other cards. GPU detection is 100% on client side, so you should try to wipe BOINC configs and then check client logs (Tools -> Event log in BOINC Manager) to see why it ignores some GPUs.
49) Message boards : News : Finish line of the current search and next steps (Message 583)
Posted 19 Jul 2017 by Sergei Chernykh
Post:
I am curious to know if anyone knows how many more tasks there are for 2_64? Before it is finished

Approx. 83,000 work units left, or 4 weeks more work at current speed.

P.S. Server status page
50) Message boards : Bug tracker : Formal credit for new APs - how is it handled? (Message 581)
Posted 17 Jul 2017 by Sergei Chernykh
Post:
Everything is fixed now. There were others with the same bug.


Next 50


©2018 Sergei Chernykh