7) I hope to find time to test this code further with
a tiny fixed-point class.
This would be interesting to test.
OK. I will try very hard to make the time for it next weekend or sooner.
Matthieu, I have successfully tested your <complex.hpp> code with a fixed-point template class. I used my own Q15.16 fixed-point representation. Everything works fine, assuming we modify your header file according to the suggestions we are currently discussing. It was a very fun experience. I actually have complex calculations with 50 decimal digits of precision and 4 decimal digits of precision in the same file. Wow! There is absolutely fascinating potential for generic numeric programming here! Your complex<T> code also helped me find weaknesses in my fixed-point code. My test code is included this e-mail. I encourage you to try to compile and run this test code. If you have any problems, don't hesitate to contact me. You need to simply call do_complex() and do_complex_fixed() from main(). You need to go to the boost sandbox and get John's <boost/multiprecision/...> stuff from the "big_number" directory. My tests run OK with boost 1.49 patched with our proposed Boost.Multiprecision. My fixed_point.h, two util*.h header files and a modified version of your <complex.hpp> are included. My fixed-point stuff is not boost stuff. Best regards, Chris.