| Modern graphics hardware is capable of performing many floating point calculations in parallel. Current graphics hardware is able to perform ten times faster than current CPUs for certain types of calculations. This performance gap is expected to increase in the future.; This thesis presents methods for allowing compilers to take advantage of graphics hardware for general purpose computations. Methods are discussed for performing general purpose computations on graphics hardware, as well as methods to apply standard parallel compiling techniques to graphics hardware. A framework is also discussed that allows such a compiler to be rapidly developed by using existing compiler tools. |
