Posts by Sergei Chernykh

151) Message boards : Number crunching : Amicable Numbers has slowed down (Message 964)
Posted 1 Nov 2018 by Sergei Chernykh
Post:
My only guess is as numbers grow larger, internal buffers don't fit in CPU cache anymore. Hence the slowdown.
152) Message boards : Number crunching : Problem with valid wus count on All Tasks page? (Message 958)
Posted 30 Oct 2018 by Sergei Chernykh
Post:
Old WUs get deleted 1 week after they are verified.
153) Message boards : Number crunching : Probability (Message 955)
Posted 29 Oct 2018 by Sergei Chernykh
Post:
This page shows it: https://sech.me/boinc/Amicable/server_status.php
ETA at current speed (search up to 1020) Mar 2019
154) Message boards : Number crunching : Probability (Message 953)
Posted 28 Oct 2018 by Sergei Chernykh
Post:
There are 4,197,265 known amicable pairs below 1020 or approximately 1 pair per 23.825*1012 numbers. So the probability is low, very low if you just check numbers one by one. If you search them in more clever order, it can be much higher for some ranges. This project was founding thousands pairs per day at some stage, but now it's chewing through last ranges which are unlikely to have new pairs.
155) Message boards : Number crunching : Work unit errors (Message 943)
Posted 4 Oct 2018 by Sergei Chernykh
Post:
It can happen if:
- this is an old work unit which expired for someone else
- then it was sent to you
- then this "someone else" finally sends it to the server
- the server validates it and cancels all remaining "in progress" tasks (including yours)

P.S. But your 3 errors are just computing errors, they didn't happen immediately after start.
156) Message boards : Number crunching : Work unit errors (Message 941)
Posted 4 Oct 2018 by Sergei Chernykh
Post:
I see only 3 errors out of 925 work units for your PC. Time limit for work unit is 3 days from the time it was sent to client.
157) Message boards : Number crunching : GPU version: kernel size tuning and less UI lags (Message 934)
Posted 29 Sep 2018 by Sergei Chernykh
Post:
From what I heard, RTX series has separate integer and FP units, and this project is integer only. This is probably why RTX gets such boost.
158) Message boards : Number crunching : Calculations made (Message 928)
Posted 24 Sep 2018 by Sergei Chernykh
Post:
Calculations here are integer only, so no flops. One work unit runs in ~40 minutes on Ryzen 7 1700 @ 3.6 GHz as far as I remember, so it translates to ~7*1013 integer operations. You can take it as 1 task = 70,000 GFLOP in your terms.
159) Message boards : Number crunching : Calculations made (Message 925)
Posted 24 Sep 2018 by Sergei Chernykh
Post:
Divide your total credit by 6836.19: 1968824 / 6834.19 = 288 tasks completed so far.
160) Message boards : Number crunching : All work is ending in Error (Message 919)
Posted 17 Sep 2018 by Sergei Chernykh
Post:
Try to reduce "Kernel size for NVIDIA GPU", 23 might be too high.
161) Message boards : Number crunching : All work is ending in Error (Message 916)
Posted 13 Sep 2018 by Sergei Chernykh
Post:
It looks like it's fixed now. New GPU versions didn't give any errors so far, and almost 100 WUs are already finished.
162) Message boards : Number crunching : All work is ending in Error (Message 915)
Posted 13 Sep 2018 by Sergei Chernykh
Post:
I've updated Windows & Linux OpenCL versions, can you check that they run fine? MacOS version will follow soon.

P.S. MacOS OpenCL version is now updated.
163) Message boards : Number crunching : All work is ending in Error (Message 913)
Posted 13 Sep 2018 by Sergei Chernykh
Post:
Sorry, will fix it ASAP.
164) Message boards : Number crunching : All work is ending in Error (Message 908)
Posted 12 Sep 2018 by Sergei Chernykh
Post:
It's fixed now.
165) Message boards : Number crunching : All work is ending in Error (Message 907)
Posted 12 Sep 2018 by Sergei Chernykh
Post:
I've found the bug, will update CPU version later today.
166) Message boards : Number crunching : All work is ending in Error (Message 906)
Posted 12 Sep 2018 by Sergei Chernykh
Post:
I can confirm it started crashing on latest work units, but I don't know why yet. I'll try to fix it today.
167) Message boards : Number crunching : GPU - Double Precision? (Message 903)
Posted 10 Sep 2018 by Sergei Chernykh
Post:
It's more than 3-4x because you need to:
1) Do 4 multiplications
2) Add 4 results together, doing 3 64-bit additions
In total, it translates to ~12 low level 32-bit operations. I tried it before, it doesn't improve speed.
168) Message boards : Number crunching : GPU - Double Precision? (Message 901)
Posted 10 Sep 2018 by Sergei Chernykh
Post:
It doesn't make sense to rewrite 64-bit multiplication as 4 32-bit multiplications because this is what OpenCL compiler already does for GPUs with weak FP double performance.
169) Message boards : Number crunching : GPU - Double Precision? (Message 899)
Posted 10 Sep 2018 by Sergei Chernykh
Post:
They're not vectorizable - the algorithm requires 64x64 bit multiplications. But if you can optimize this GPU code, do a pull request.
170) Message boards : Number crunching : GPU - Double Precision? (Message 897)
Posted 10 Sep 2018 by Sergei Chernykh
Post:
You can look at the GPU code here: https://github.com/SChernykh/Amicable/blob/boinc-opencl-version-128-bit/Amicable/kernel.cl

It uses 64-bit integer arithmetic which is kind of equivalent to FP double arithmetic when it comes to performance. Yes, GPUs with a lot of double GFlops capability will be much faster.
171) Message boards : Bug tracker : Validation not working? (Message 893)
Posted 8 Sep 2018 by Sergei Chernykh
Post:
"Workunits waiting for validation 1385"
The server is clogged now because amicable pairs cache backend crashed yesterday and it's reloading all data from disk now (> 700 GB of data). The issue should resolve itself as soon as cache is fully loaded. It's still about 12 hours to go.
172) Message boards : Random stuff : Cross-Project stats page broken. :( (Message 890)
Posted 29 Aug 2018 by Sergei Chernykh
Post:
Well, the page is broken because http://boinc.netsoft-online.com returns error 403 when the server asks it for data.
173) Message boards : Random stuff : Cross-Project stats page broken. :( (Message 888)
Posted 26 Aug 2018 by Sergei Chernykh
Post:
There are no http links on that page, background picture is also https. I'll look into the bug itself tomorrow.
174) Message boards : Bug tracker : Amicable doesn't respect usage limits (Message 884)
Posted 3 Aug 2018 by Sergei Chernykh
Post:
Don't mix Tmax and Tjmax. Tjmax for i7 2600k is 100C. It starts throttling at 90C. 75C is totally safe. If your processor reports 75C, it's a temperature of its internal sensor, temperature at the heat spreader (which must not be higher than Tmax) will be much less than 75C.
175) Message boards : Bug tracker : Amicable doesn't respect usage limits (Message 882)
Posted 31 Jul 2018 by Sergei Chernykh
Post:
75°C is normal temperature for CPU, nothing bad will happen to it. Amicable runs at the lowest possible priority, so even if it uses all cores, the system should be responsive.
176) Message boards : Bug tracker : Amicable doesn't respect usage limits (Message 880)
Posted 28 Jul 2018 by Sergei Chernykh
Post:
This was already answered in the second post of this thread. CPU time limitation done by BOINC doesn't work well with Amicable, there was a lot of invalid results.
177) Message boards : Bug tracker : Amicable doesn't respect usage limits (Message 878)
Posted 26 Jul 2018 by Sergei Chernykh
Post:
You have 4 cores per WU, but you also need to set "Max # jobs" to 1. Otherwise BOINC will start 2 WUs that will take all 8 cores.

P.S. Your CPU shouldn't overheat when running Amicable even at 100% on all cores. It doesn't even use SSE, it's only integer math on Amicable. Check your CPU cooler.
178) Message boards : Bug tracker : Amicable doesn't respect usage limits (Message 876)
Posted 26 Jul 2018 by Sergei Chernykh
Post:
If I reduce the available core, it crash.

This crash is fixed already, try again.
179) Message boards : Bug tracker : All CPU WU finished in error !!! (Message 874)
Posted 24 Jul 2018 by Sergei Chernykh
Post:
All CPU versions have been updated.
180) Message boards : Bug tracker : All CPU WU finished in error !!! (Message 873)
Posted 24 Jul 2018 by Sergei Chernykh
Post:
I've fixed it: https://github.com/SChernykh/Amicable/commit/3b984aaf6c7aed4791ec457217fa65131ffb3cc3.
The bug was only in the CPU version. I'll update binaries later today. It's been a long time since I updated them, so first I need to remember how to do it, find all needed credentials and test the new binaries thoroughly.
181) Message boards : Bug tracker : All CPU WU finished in error !!! (Message 872)
Posted 24 Jul 2018 by Sergei Chernykh
Post:
I'll look into it ASAP.
182) Message boards : Getting started : not getting work units (Message 869)
Posted 4 Jul 2018 by Sergei Chernykh
Post:
It won't reset your credits and RAC, just make sure you keep your credentials (email and password) for each project in a safe place before you reinstall BOINC.
183) Message boards : Getting started : not getting work units (Message 867)
Posted 4 Jul 2018 by Sergei Chernykh
Post:
The last relevant line I saw is "NVIDIA GPU: job cache fu...". I don't know how it can happen, you probably need to manually delete all BOINC files and reinstall BOINC.
184) Message boards : Getting started : not getting work units (Message 865)
Posted 4 Jul 2018 by Sergei Chernykh
Post:
You can look at logs in Tools -> Event log (Ctrl+Shift+E).
185) Message boards : Getting started : not getting work units (Message 863)
Posted 4 Jul 2018 by Sergei Chernykh
Post:
Check your BOINC client logs. It usually shows a detailed reason for not getting work units.
186) Message boards : Getting started : How do the points get turned into Gridcoins? (Message 860)
Posted 20 Jun 2018 by Sergei Chernykh
Post:
I didn't get into the details, all I know is that you need to run Gridcoin software on your PC that runs BOINC: https://gridcoin.us/Guides/boinc-install.htm
187) Message boards : Number crunching : Massive Improvement to Runtimes With New Nvidia Driver (Message 854)
Posted 1 Jun 2018 by Sergei Chernykh
Post:
He runs 397.93 on Windows 10.
188) Message boards : News : Power outage (Message 851)
Posted 24 May 2018 by Sergei Chernykh
Post:
The server was down today between 11:04 and 14:20 (CEST) because of a major power outage in Hetzner datacenter. Everything should be up and running again now.
189) Message boards : Number crunching : Is source code available? (Message 848)
Posted 17 May 2018 by Sergei Chernykh
Post:
So the app_name must include info on 10^20. What app_name is set on the server, Sergei?

app_name is amicable_10_20
190) Message boards : Number crunching : Is source code available? (Message 846)
Posted 17 May 2018 by Sergei Chernykh
Post:
It says "syntax error" because you have mismatched tags:
 <app_name>amicable</name>
191) Message boards : Number crunching : Is source code available? (Message 842)
Posted 17 May 2018 by Sergei Chernykh
Post:
Thank you for quick response!!! It'll be help me a lot!

Do you have any issues with running the existing Linux binary on your system? Don't forget to make a pull request if your modifications could be useful for other users.
192) Message boards : Number crunching : Is source code available? (Message 840)
Posted 17 May 2018 by Sergei Chernykh
Post:
https://github.com/SChernykh/Amicable/tree/boinc-version-128-bit
193) Message boards : Number crunching : Newbie questions about the project (Message 838)
Posted 14 May 2018 by Sergei Chernykh
Post:
Yes, Feb 2019, and yes, it's still an open question.
194) Message boards : Getting started : Raspberry Pi support? (Message 835)
Posted 12 May 2018 by Sergei Chernykh
Post:
It works on my Raspberry Pi 3 Model B. I use it to compile 64-bit ARM application. But you'll need to install 64-bit Linux on it to run Amicable Numbers.

P.S. https://sech.me/boinc/Amicable/forum_thread.php?id=37#300
195) Message boards : Number crunching : Newbie questions about the project (Message 832)
Posted 10 May 2018 by Sergei Chernykh
Post:
When are WUs rich in amicable pairs going to start?

They're already finished. Next time will be in 2019 when the project starts the search up to 1021.
196) Message boards : Number crunching : Newbie questions about the project (Message 830)
Posted 8 May 2018 by Sergei Chernykh
Post:
what it means to have a factorization between a certain factorization and another, could you define it?

Factorization order is defined as follows: take two factorizations A and B in normalized form (all factors ordered from smallest to largest), compare respective factors one by one until there is a difference. If n-th factor from A is less than n-th factor from B then factorization of A precedes factorization of B. If n-th factor matches, but its power is less than in B (for example, 172 and 173) then again factorization of A precedes factorization of B.

Example: 3^7*5^2*13*311*32189 precedes 3^7*5^2*13*487*1399 because 311 < 487

Also, how can you spot all the numbers with largest prime factor between two certain numbers when you perform the second type of search?

The second part of the search goes through all numbers of the form N*p < 1020 where p is a prime number between numbers in the command line and N goes through all amicable candidates which are < 1020/p. All these candidates are calculated when a work unit starts to execute, it takes a few seconds to do.
197) Message boards : News : GPU version fix (Message 827)
Posted 8 May 2018 by Sergei Chernykh
Post:
We caught up more than a month ago, it's all new work now.
198) Message boards : Number crunching : Newbie questions about the project (Message 824)
Posted 4 May 2018 by Sergei Chernykh
Post:
There is no short answer to these questions, it's better to read the article describing the algorithm: https://sech.me/ap/articles.html#a1 - especially part 3, "Reducing search space".
199) Message boards : Number crunching : Newbie questions about the project (Message 822)
Posted 3 May 2018 by Sergei Chernykh
Post:
/task_size parameter is how many numbers to check in this work unit. It's used to calculate WU progress.
200) Message boards : Random stuff : Project Status Page (Message 819)
Posted 1 May 2018 by Sergei Chernykh
Post:
I7-5960X is one of the reference CPUs here: https://sech.me/boinc/Amicable/gpu_list.php. Those numbers might be a bit outdated because they were measured for the search up to 264. I'll update them in a few days.


Previous 50 · Next 50


©2024 Sergei Chernykh