8-bit MicrocontrollersのトピックRe: converting 32-bit Q1.31 type to double

converting 32-bit Q1.31 type to double

fixed-term_apar — Thu, 02 May 2019 12:24:48 GMT

Is there any possible way to convert 32-bit Q1.31 type (tFrac32) variable to a double variable in C language?

Re: converting 32-bit Q1.31 type to double

kef2 — Fri, 03 May 2019 06:28:54 GMT

Sure. Fraction type is real value times power of two (31 in your case), then convert result from FP to integer. So to convert it back you need to convert your fraction back to float and divide by power of two.

float f = (float)frac / 2147483648.0;

float f = (float)frac * 0.4656612873077392578125E-9; // use multiply by reciprocal of 2^31

or even more faster using ldexp() function from math.h

float f = ldexp(frac, -31);

ldexp multiplies first argument by 2^N, where N is second ldexp argument. ldexp() should be lot faster than multiply on MCUs without FP coprocessor.

Before using it please check ldexp() works well with 0.0 as first argument. I saw few not cheap compilers having implemented it wrong, unexpected results doing something like ldexp(0.0, -30). When first argument is 0.0 should return 0.0 no matter what is 2nd argument.

Re: converting 32-bit Q1.31 type to double

fixed-term_apar — Fri, 03 May 2019 07:18:44 GMT

Thank you for your reply!

But can we do a direct type casting like

tFrac32 y;

double x ;

x = (double)( y);

Re: converting 32-bit Q1.31 type to double

kef2 — Fri, 03 May 2019 07:59:50 GMT

No, unless compiler supports nonstandard extension and nonstandard tFrac32 type. In C++ you could probably create tFrac32 class and define typecast operator for it.