So finally AMD's Kaveri APU is coming out, and it has unified memory access
between CPU and GPU. Unified memory is likely going to occur with Intel's
processors too. The problem with GPGPU programming with current programming
models is the need to copy from main memory to GPU visible memory. With Kaveri,
the GPU can now access the same memory that the CPU does. Does this in any way
make an OpenCL port of Povray easier/feasible? Or would it still be a complete
rewrite?
Post a reply to this message
|
