mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware > GPU Computing

Reply
 
Thread Tools
Old 2020-05-06, 01:47   #1607
ixfd64
Bemusing Prompter
 
ixfd64's Avatar
 
"Danny"
Dec 2002
California

29×79 Posts
Default

Quote:
Originally Posted by James Heinrich View Post
Could you perchance provide a link?
https://github.com/Bdot42/mfakto

It's currently only available as source code.

Last fiddled with by ixfd64 on 2020-05-06 at 01:48
ixfd64 is offline   Reply With Quote
Old 2020-05-06, 03:08   #1608
kracker
ἀβουλία
 
kracker's Avatar
 
"Mr. Meeseeks"
Jan 2012
California, USA

17·127 Posts
Default

Quote:
Originally Posted by ixfd64 View Post
Hey everyone, mfakto 0.15pre7 is now available on GitHub. Here are the changes:
...
Thanks!

Also, binaries for windows.
Attached Files
File Type: zip mfakto-0.15pre7-x64.zip (798.8 KB, 35 views)
kracker is online now   Reply With Quote
Old 2020-05-06, 07:31   #1609
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

22·5·211 Posts
Default

Quote:
Originally Posted by ixfd64 View Post
Hey everyone, mfakto 0.15pre7 is now available on GitHub. Here are the changes:
  1. kracker has added preliminary support for newer architectures up to RDNA
  2. Some compiler warnings have been fixed
  3. mfakto should compile out of the box in Microsoft Visual Studio
  4. Of course, more documentation updates

Please note that I wasn't able to test the latest code on the Mac Pro at my office due to the COVID-19 pandemic. Feel free to slap me with a trout if I broke anything.
How about increasing max gpusievesize from 128 to 2047 (megabits), as has been done in mfaktc, and tested there to provide several percent increase in throughput, especially on faster gpus? see also
https://mersenneforum.org/showpost.p...postcount=3197
https://mersenneforum.org/showpost.p...postcount=3199
https://mersenneforum.org/showpost.p...postcount=3201

Thanks to you both for the update, and to kracker for the Windows build. Mfakto has been languishing without support for a while.

Last fiddled with by kriesel on 2020-05-06 at 08:00
kriesel is online now   Reply With Quote
Old 2020-05-06, 11:54   #1610
James Heinrich
 
James Heinrich's Avatar
 
"James Heinrich"
May 2004
ex-Northern Ontario

1011011100112 Posts
Default

Quote:
Originally Posted by ixfd64 View Post
https://github.com/Bdot42/mfakto
It's currently only available as source code.
Perhaps a release could be tagged on GitHub? The latest tagged release (v0.14) is just over 6 years old.
James Heinrich is offline   Reply With Quote
Old 2020-05-06, 17:55   #1611
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

37×227 Posts
Default

Is it worth replace 0.15pre6 on an integrated graphics card with 0.15pre7?
Uncwilly is online now   Reply With Quote
Old 2020-05-07, 02:38   #1612
kracker
ἀβουλία
 
kracker's Avatar
 
"Mr. Meeseeks"
Jan 2012
California, USA

86F16 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
Is it worth replace 0.15pre6 on an integrated graphics card with 0.15pre7?
I doubt it - assuming it's already working.
kracker is online now   Reply With Quote
Old 2020-05-09, 05:45   #1613
ixfd64
Bemusing Prompter
 
ixfd64's Avatar
 
"Danny"
Dec 2002
California

29×79 Posts
Default

I've been investigating the issue where the macOS version would fail to find factors on Intel HD Graphics. From what I got from various sources, the issue is due to buggy drivers shipped with the OS. So unless Apple fixes the drivers, then mfakto likely won't support Intel HD Graphics on macOS in the foreseeable future. But this probably isn't a huge deal anyway because integrated GPUs don't produce many GHz-days.

Now for another problem: on one of my computers, the MinGW build would give an "entry point not found" error unless libstdc++-6.dll is in the same folder. Even after it runs, mfakto would terminate immediately after the "Select device" message. Compiling with static linking enabled didn't make a difference. This does not happen when I build mfakto using Visual Studio.

Another likely related issue: building with MinGW using the -m32 flag causes a compiler error:

Code:
c:\winbuilds\include\c++\4.8.3\cstdlib:178:10: error: expected unqualified-id before '__int128'
   inline __int128
          ^
Again, Visual Studio builds the 32-bit version just fine. I checked out an earlier commit to make sure I didn't break anything in my development branch, and the results were the same. Anyone got any ideas?

Last fiddled with by ixfd64 on 2020-05-12 at 17:26
ixfd64 is offline   Reply With Quote
Old 2020-05-09, 14:07   #1614
James Heinrich
 
James Heinrich's Avatar
 
"James Heinrich"
May 2004
ex-Northern Ontario

3·977 Posts
Default

I believe I know the answer, can somebody with knowledge of the source please confirm for me: How does mfakto handle discovering composite factors?
(I've been using M110393069 for testing which has a thousand composite factors)
I believe mfakto simply returns the composite factor as-discovered with no attempt to determine if said factor is composite or prime, correct?

Assuming thus, a two-part question:
a) how practical would it be for mfakto to detect that a discovered factor is composite?
b) if factor is composite, how practical would it be for mfakto to split the factor into primes?
James Heinrich is offline   Reply With Quote
Old 2020-05-12, 17:33   #1615
ixfd64
Bemusing Prompter
 
ixfd64's Avatar
 
"Danny"
Dec 2002
California

29×79 Posts
Default

It turns out the MinGW build that comes with Win-builds doesn't correctly compile OpenCL programs. I was able to compile a working version of mfakto using the package from SourceForge: https://sourceforge.net/projects/mingw-w64

Also, most MinGW installations are single-target and can only compile for the same architecture. In other words, the 64-bit version of MinGW can generally only build the 64-bit version of mfakto.
ixfd64 is offline   Reply With Quote
Old 2020-05-16, 23:55   #1616
ixfd64
Bemusing Prompter
 
ixfd64's Avatar
 
"Danny"
Dec 2002
California

43638 Posts
Default

I've pushed a few more changes to mfakto. They're pretty minor, but here are the highlights:
  • fixed several compiler warnings. macOS users should no longer see them.
  • fixed two potential compiler errors on macOS
  • updated the output text to be more consistent and use correct terminology (example: "inifile" to "INI file")
  • enabled universal x86_64 builds for macOS
  • refactored the makefiles to resolve redundancies and use more standard variable names

On the documentation side, I've added troubleshooting information to the README file.
ixfd64 is offline   Reply With Quote
Old 2020-05-31, 16:46   #1617
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

22×5×211 Posts
Default UHD920?

The integrated graphics processor (IGP) on an avx512-capable 1035G1 cpu is a UHD920. Has anyone gotten mfakto to run on a UHD920 reliably? My attempts on Win10 on a refurbished system have yielded only about half the short self tests succeeding.
kriesel is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
gpuOwL: an OpenCL program for Mersenne primality testing preda GpuOwl 2393 2020-08-07 17:09
mfaktc: a CUDA program for Mersenne prefactoring TheJudger GPU Computing 3292 2020-07-29 18:12
LL with OpenCL msft GPU Computing 433 2019-06-23 21:11
OpenCL for FPGAs TObject GPU Computing 2 2013-10-12 21:09
Program to TF Mersenne numbers with more than 1 sextillion digits? Stargate38 Factoring 24 2011-11-03 00:34

All times are UTC. The time now is 21:31.

Sat Aug 8 21:31:33 UTC 2020 up 22 days, 17:18, 1 user, load averages: 1.35, 1.58, 1.58

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.