Closed eneas77 closed 2 years ago
I'd use -DCOMPUTE_CAPABILITY=80
-DCOMPUTE_CAPABILITY (compute_capability) = 60,70 GPU code only. CUDA compute capability (multiplied by 10) of GPU. Valid values (noninclusive) are 35, 50, 60, 70, 75 etc., and can be comma-separated, e.g. 35,60. Multiple compute capabilites will increase the size of executables. The full list of compute capabilities can be found on the NVIDIA website.
https://github.com/TinkerTools/tinker9/blob/master/doc/manual/m/install/buildwithcmake.rst
it worked nicely! Thank you very much
Hi,
After compiling tinker9, the next message apperas when running all.tests:
Primary GPU package : CUDA
GPU Device : Setting Device ID to 0 from GPU utilization Current file: /apps/tinker/tinker9/src/acc/mdpq.cpp function: _ZN6tinker18mdCopyPosToXyz_accEv line: 11 This file was compiled: -acc=gpu -gpu=cc60 -gpu=cc70 Rebuild the application with -gpu=cc86 to use NVIDIA Tesla GPU 0
My GPU actual card is a GeFOrce RTX 3060 Ti, which is a Ampere architecture. Is that message related with some rebuilt requirements?
Thank you kindly