Paul A. Bristow said: (by the date of Wed, 30 Nov 2011 11:08:53 -0000)
- find some arbitrary precision library, that will work with Units and VERY big factorials, like 100! or even better 1000!
There are indeed at least two related efforts on arbitrary precision library going on.
A multiprecision library by Christopher Kormanyos based on his e_float library which is in the sandbox .
and, based on this, an extension to Boost.Math (already user-defined type ready and able to use NTL and GMP) that John Maddock is currently busy working on.
These should allow you to calculate factorial 100! to a few hundred decimal digits.
For your amusement, 100! is (to 50 decimal digits)
100! = 9.332621544394415268169923885626670049071596826438162146859e+157
I'm using apcalc, a very nice software ;-) 933262154439441526816992388562667004907159682643816214685929638952175999932\ 29915608941463976156518286253697920827223758251185210916864000000000000000000000000
But I'm not at all sure that this is the way forward for your purpose. For example, in calculating gamma functions, it is customary to use logs to avoid overflow.
And we need to make sure we can walk before we can run - combining arbitrary precision with Boost.Units (where double precision is usually enough to represent the accuracy of physical things).
hmm, you are right. I probably will need to go into logs. Boost.Units has no effect on runtime, it checks units during compilation. So using them shouldn't affect at all. I'm glad that I installed 1.48 with Steven's help. And now I can convert my code to use units :) first thing will be to figure out how to declare a type meter^-0.5 maybe my previous try that didn't work with boost 1.42 will start to work with 1.48: typedef derived_dimension<length_base_dimension,root<2> >::type quantum_wavefunction_1D; best regards -- Janek Kozicki http://janek.kozicki.pl/