mfaktc: a CUDA program for Mersenne prefactoring

[lalera]

Originally Posted by James Heinrich

I assumed the versions posted in #3593 and/or #3595 supersede that. What does #3572 offer that the later two don't?

hi,
the difference is that it is no "blackbox" compilation
you have the complete code

[James Heinrich]

Originally Posted by lalera

you have the complete code

Oh, ok. I've included it now.

[lalera]

Originally Posted by James Heinrich

Oh, ok. I've included it now.

hi,
thank you!

[lalera]

Originally Posted by James Heinrich

Oh, ok. I've included it now.

hi,
the download link does not work

[James Heinrich]

Originally Posted by lalera

the download link does not work

Wasn't playing nice with + special character in filename.

[lalera]

Originally Posted by James Heinrich

Wasn't playing nice with + special character in filename.

hi,
now it works
thank you !

[Rodrigo]

Originally Posted by Rodrigo

Thank you. I've never compiled a program before, have only used the ready-made executables available for download. Not unwilling to give it a shot, but it's hard to avoid the sense that this would be way above my pay grade.

Is there a comprehensive guide here to compiling mfaktc for Kubuntu (Ubuntu)?

Thankfully, recompiling turned out to be unnecessary. I had read up on the procedure for compiling a program and downloaded the elements required to do it, but after installing NVIDIA driver 510 and rebooting as Mark Rose instructed, out of curiosity I then launched the existing version of mfaktc.exe, and lo and behold it started running as if nothing had happened.

Bullet dodged.

[Luminescence]

Originally Posted by rebirther

Can you pls also include cc6.1-cc9.0 and a short guide how did you compile it?

I've attached the big executable for CC6.0 up to CC9.0

It should run on any Pascal GPU onwards, even some rare ones like Jetson/Tegra devices.

To compile it:

• Install the CUDA toolkit version you want to use and remember the directory where you put it.
• Install Visual Studio. I use the the community version of Visual Studio 2022.
• Install make via Chocolatey using the command choco install make in an administrative command promt or Powershell.
• Edit the Makefile.win in the ...\mfaktc\src folder:
  

  - CUDA_DIR should be set to the root directory of the toolkit folder
  - Remove /GL from the CFLAGS and CUFLAGS
  - Remove -ccbin="C:\..." from CUFLAGS
  - Remove /LTCG #/ltcg:pgo from the LFLAGS
  - Replace LIBS = $(CUDA_DIR)\lib\x64\cudart.lib with LIBS = $(CUDA_DIR)\lib\x64\cudart_static.lib
  - Edit the --generate-code arch=compute_xx,code=sm_xx flags to whatever you want to compile mfaktc for and remove lines that are unnecessary

• Finally, search for "Developer Command Prompt for VS 2022" or "Developer Powershell for VS 2022" in your start menu and use the command cd to navigate into your mfaktc source folder.
• Enter make -f Makefile.win all and it should (hopefully) generate the executable and two additional files in the ...\mfaktc directory

The zip file also includes the source directory and you can use my Makefile.win as a template.

[kriesel]

Outstanding! Both here and in the MMFF thread.

[chalsall]

Originally Posted by kriesel

Outstanding! Both here and in the MMFF thread.

+1.

[rebirther]

As from the readme mfaktc/o cant run multiple GPUs on a single instance. It would be very helpful for BOINC if this can be changed in the future.