Cilus
laborare est orare
Asigh is right abt the GPU processing thing. But till date Stream processors of a GPU do not support complex instruction sets. For example finding square root of a number can be performed by a single instruction in CPU, but not possible in GPU. What is the most advantage of GPU processing is it can perform a simple operation, say a color invert operation in very very large number of elements simultaneously. It mainly works in vectorized way.
For example you are performing an operation X=X+2 over 1 million values of X. Here CPU will execute it sequentially, although use of SIMD instruction set will increase the performance, whereas a GPU will perform the same operation over the 1 million element as a vectorized unit. So a single GPU cycle will be required to perform this operation over all X.
For example you are performing an operation X=X+2 over 1 million values of X. Here CPU will execute it sequentially, although use of SIMD instruction set will increase the performance, whereas a GPU will perform the same operation over the 1 million element as a vectorized unit. So a single GPU cycle will be required to perform this operation over all X.