I'm trying to get a routine written to implement a fixed point 64-bit Biquad ( DF2T form ). I have a very specific question regarding the Multi-Precision multiply instructions described on P3-17 of the SHARC Processor Programming Reference.
"The multiplier supports the following data operations for 64-bit data.
MRF = Rx * Ry (SSF);
My specific question is whether the Rx and Ry are themselves 64-bit quantities ( i.e. a pair of universal registers ) and if so whether there are any guidelines on which pairs of registers to use or if there are any constraints thereon.
More generally, could anyone point me at any example code illustrating the use of these instructions ( ideally in SIMD mode )