We need dotp4 instruction in TCI6482 and TCI6416 processors to to do signed multiplication between two pair of four 8 bit values. Right now this instruction is not available in these processors. We are using dotp2 to do the operation of dotp4. But, dotp2 requires 16 bit format. For this, 8 bit value is converted into 16 bit, then dot product performed by dotp2 instruction.If we have dotp4, conversion of 8 bit to 16 can be avoided and at one time, eight 8 bit values can be taken for dot product in dotp4 instead of four in dopt2. This will results in speeding up the processes time.
So, We need dotp4 instruction. Is it possible for you to create dotp4 instruction or can you help us to write an assembly code as a function for dotp4? So, that we can use this dotp4 as a function for our coding.
