I assume this is possible but I have not been able to find any explicit confirmation of this.
I am just starting with GPGPU and my PC only has one PCI-e X16 slot. Currently I have an Nvidia card and my original intention was to use that with CUDA for my programs. The problem is that with Nvidia, if you share the card between graphics and GPGPU you only get 5 seconds to run your program before it times out. Their suggestion is to get a second card. To run two cards I would need a new motherboard, new memory etc, so it is no longer a cost-effective solution.
So I am looking at an ATI card and Opencl. But before I spend the money I would like to know if I can share the card between graphics and GPGPU and if so, what are the limitations if I do this.