Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

troubles while compiling for cuda less than 11.6 #8

Open
pixel-wipe opened this issue Mar 20, 2022 · 6 comments
Open

troubles while compiling for cuda less than 11.6 #8

pixel-wipe opened this issue Mar 20, 2022 · 6 comments

Comments

@pixel-wipe
Copy link

pixel-wipe commented Mar 20, 2022

Build
Windows:

Program was prepared using CUDA 11.6 - for any other version manual change in VS project config files is needed.
 Exe under /Releases/ was build using compute_cap=86, for cards 30xx. If you have older card, you must rebuild 
program using older CUDA/lower CCAP.

Linux:

Go to WifSolverCuda/ subfolder and execute make all. If your device does not support compute capability=86 (error 
"No kernel image is available for execution on the device"), do the change in Makefile (for example 1080Ti requires 
COMPUTE_CAP=61).

i was try to fix <Import Project="$(VCTargetsPath)\BuildCustomizations\CUDA 11.6.targets" />
in WifSolverCuda.vcxproj

but compiling failed yet after it

explain please where and how i may fix source code to compile for cuda less than 11.6 ?
thx a lot )

@PawelGorny
Copy link
Owner

Hello,
Which version do you want to use? For which card?
If compilation fails, please include the error output.
Usually it is enough to replace both "11.6" in project file with your version, but maybe you must also change computation capabilities, by default it is high, 86.
Or you may always try to launch precompiled version from /releases/ on github, there is one file for cards 30xx and other compiled used cuda 10.2 and comp.cap. 5.4

@pixel-wipe
Copy link
Author

Usually it is enough to replace both "11.6" in project file with your version, but maybe you must also change computation capabilities, by default it is high, 86. Or you may always try to launch precompiled version from /releases/ on github, there is one file for cards 30xx and other compiled used cuda 10.2 and comp.cap. 5.4

i has found only one "11.6" param... will try to check it one more time
ive try to recompile for linux(compcap 37 and cuda 10), but precompiled that present at ur git is only for windows

@PawelGorny
Copy link
Owner

There is already a Makefile for linux, you may launch it "make all". Only change param COMPUTE_CAP=86, it will use the existing cuda. Maybe you will need to change path CUDA_HOME=/usr/local/cuda, but normally it works.

@pixel-wipe
Copy link
Author

make all failed wtih any c_c, send u screens with errors

@PawelGorny
Copy link
Owner

I see that you are using vast.ai, right?

  1. Do you use correct image? I recommend using "devel-ubuntu". Recently I used [nvidia/cuda:11.3.1-devel-ubuntu18.04] and did not have any problems.
  2. Maybe you should use older cuda, like 10.2? Try "10.2-devel-ubuntu18.04"

@Totulik
Copy link

Totulik commented Mar 30, 2022

WifSolverCuda.vcxproj
33 to
77 compute_86,sm_86 to compute_61,sm_61
112 to

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants