Anything mathemactical, which doesn't use functions provided by libm, which is compiled to use soft-float like the rest of the system (these are things like trig functions, exponents, sqrts, etc.) will benefit from using hard float (vfp). The degree to which they will benefit depends on how much they do floating point calculations. If you have something like a spreadsheet (or other mathematical program) or something that draws lots of cirlces (needs trig for the coords) then you ought to also compile libc using hardfloat and link in the resultant libm to realise the maximum speed improvement.

Perhaps, but you may need to resort to assembly to actually use them. By all means try adding the -mcpu flag and then if it fails to build remove it. The answer is I don't know what the default mcpu flag is for the toolchain, and what flags that implies, sorry.


Si