pwr(x,y) error

[JerryR]

Anyone notice that pwr(x,y) evaluates inaccurately at higher values of y?
For instance: (int32)pwr(10,7)= 9999997 instead of 10000000.

Using PIC18F25K20, with compiler version 5.093.

Any workaround or solution?

Thanks!

[Ttelmah]

Remember this is being done using logarithms to allow support for things
like fractional powers. Using fp numbers there are only 6 actual significant
digits in an fp number. Result it's always going to give errors like this.
Solutions. Well for a positive power, using integers, just multiply using
integers. 10*10*10*10*10*10*10 done using int32 arithmetic, will take
less time that pow(10,7), and give an accurate result.
Beyond this, a PIC24/30/33 supports 'double' fp maths, and this should give
a better result.

[JerryR]

Ttelmah:

Hi Thanks for the quick reply. Yes, I though that's the case. I'll use your "work-around, thanks. The pwr(x,y) method would have made my code a lot slimmer.

All the best

[Ttelmah]

If your power is always of 10, then why not just have a lookup table?.

[JerryR]

Ahhh, clever. Thanks!

[Ttelmah]

Hundreds of times faster and smaller.

[JerryR]

One little change, however. 10^0= 1.

const unsigned int32 powers10[] = {1, 10, 100, 1000, 10000, 100000, 1000000, 10000000, 100000000, 1000000000, 10000000000};

Thanks again, good solution!

[Ttelmah]

I'm sure I typed '1'!....

Silly billy.