I have code which, based on the availability of GPUs, either runs a GPU keel or the same code on the CPU. It does this at runtime and the program is always linked to CUDA.
However, I may want to run on computers that don't have any GPUs. We tried installing the CUDA Toolkit on the computer, but libcudart is missing. How do I get this?
Both the desktop machines (which have GPUs) and the grid machines (which don't) are ruing RHEL 6.7 and we used [literally] the same CUDA installer for both, but didn't select the driver on the grid (since there's no GPU and it would probably yell at us).
برچسب:
نویسنده: استخدام کار