Dot Product on TigerSHARC

loop:

// load 8 coeffs, 8 samples, do 8 MACs

YR7:4 = Q[j0+=4]; YR11:8 = Q[k0+=4]; MR3:0 += R7:6 * R11:10;;

// load 8 coeffs, 8 samples, do 8 MACs

XR7:4 = Q[j0+=4]; XR11:8 = Q[k0+=4]; MR3:0 += R13:12 * R17:16;;

// load 8 coeffs, 8 samples, do 8 MACs

YR15:12 = Q[j0+=4]; YR19:16 = Q[k0+=4]; MR3:0 += R15:14 * R19:18;;

// branch, load 8 coeffs, 8 samples, do 8 MACs

if NLC0E,jump loop;

XR15:12 = Q[j0+=4]; XR19:16 = Q[k0+=4]; MR3:0 += R5:4 * R9:8;;

With unrolling,

SW pipelining:

0.125 Cycles/Tap