Programming The Vertex Shader, Continued
There is also a possibility to mask the registers or constants, to access less than all the four values of a data entry.
You can see that despite the small number of different instructions and despite the missing jumps and conditional branches the vertex programs can be pretty powerful. The most important restriction however is clearly the program limitation to only 128 instructions per program and the execution time of at least 1 clock per instruction.
More about gpus
