This is the source code of Vocale. Vocale is a software project written as part of a master thesis in GPU virtualization at the University of Oslo.
Vocale is an extension to the Qemu hypervisor that emulates the CUDA library in Qemu's virtual machines. Applications in the virtual machines can thus use any GPU in the physical machine to accelerate their operation. Function calls and kernel launches are forwarded to the host and executed using the real CUDA library.
We reference developers to the readme.pdf file in the source code for installation instructions. This file also tells where one may download the thesis. Vocale is not a complete solution, and there are issues regarding performance and security that needs to be solved.
Features
- CUDA support in virtual machines