Jonathan Wakely wrote:
It needs three of them because they do different things.
GCC requires MPFR in order to do floating point calculations are
compile-time, so that e.g. pow(0.5, 4) can be replaced with a constant
value instead of calling a function at run-time. GCC uses MFPR to get
accurate, platform-independent results that don't depend on the host
math library, which might be inaccurate or buggy. MPC is used to do
similar optimisations for complex numbers. MPFR requires GMP.
Just curious...
Is there a big difference when using these libraries from system available math library ? Is this
difference more noticeable in some platforms than in other ones ?
I guess (and I hope) that not all math libraries are inaccurate or buggy, as if this is the case and
if the compiled program does math calculations using these inaccurate libraries... no need to
optimize a code which will give inaccurate results.
