Is it possible to make floating point operations go faster on Blackfin processors?
There are two variants of the floating point library available, depending on the compiler option specified: -ieee-fp This uses slower IEEE-compliant floating point routines. -fast-fp This uses faster floating point routines, with some deviations from IEEE. The -fast-fp library deviates from the IEEE standard because it doesn’t perform all the necessary checking for NaNs on input.
