DSPIC 64-bit Floating Point

[oxo]

Hi all,

Does anyone have timing comparisons for 32 vs 64 bit floating point on DSPIC?

(yeah, I know I could try it myself, but this may be quicker)

[Ttelmah]

They generally take about 2.5* the time.
A * is slightly less than double (about 190 versus 110 instructions to load two variables, multiply and save the result). Division was about 530 to 190. sin about 8500 versus 3200.
So the multiply is the fastest (<2*) and the others get up to about 2.8*.

These are 'rough figures' I noted in a file when I looked into this....

[oxo]

Thanks very muchly.