VLIW is a PARALLEL computer architecture .It requires a lot of simulation IN THE AREA where you will apply it ..You need to know the ALGORITHMS that you want to speed up ,..Is not like an EMBEDDED chips with VLIW sounds COOL !so i will code WHATEVER! .DATA SEARCH by BRUTE FORCE is a good AREA .. but NEVER EMBEDDED .. unless is some NEURONAL embedded . ..Which is yet to be DEFINED!
If you decide to do something in here ..is better to TACKLE algorithms in simulation and locate loop rolling and CONVERSIONS to loop UNFOLDING to see where is the parallizing POSIBILITIES .. is a field that has been extensively studied
in VECTOR MACHINES .. and now applied to DSP with multiple execution UNITS ..
I ALMOST FORGOT the most important ATTRIBUTE for those machines ..
Is the DATA MODEL .. the data has to be INDEPENDANT or loose .. if to calculate a result r you need r(n-1) one cyle earlier ,well it won't work