Abstract: The Very Long Instruction Word (VLIW) architecture is widely used for DSP. In the hybrid coding process, since the precedures of inverse quantization and inverse scan are frequently called and the methods of optimizing the loops that frequently appears in programs are studied. With TM1300 as an example, the character of the VLIW architecture is introduced. Optimizing these two proceduces would improve the speed of wole decoding process. These methods are applied to optimize the inverse scan and inverse quantization in the MPEG4. The instruction cycles required for the optimized algorithm is about 60% of the original one.