Bitwise logic gmpxx.h problems
marc.glisse at normalesup.org
Thu Jul 31 14:37:38 CEST 2008
On Thu, 31 Jul 2008, Torbjorn Granlund wrote:
> > OK, I'll use the terse C++'y std::numeric_limits<double>::max_exponent.
> I am not sure it is as portable as DBL_MAX_EXP from <cfloat>, but we'll
> see. It's easy to switch anyway if one causes trouble.
> I indeed got some failures. I'll try DBL_MAX_EXP next.
Could you tell which compilers failed, for future reference?
> Programming for C++ is real science, not plain computer science; you
> need to take an experimental approach as not even the C++ experts can
> tell what works and what does not work, or whether certain constructs
> are valid or invalid. :-)
Yes, the experimental approach is nice. Looking at gmpxx as a compiler
from c++ to gmp+c, I use ltrace to examine the generated gmp calls.
And things are going to change a lot with the new standard. Using rvalue
references (a feature that basically gives temporary objects a different
type) it should be possible to achieve results similar to the current
gmpxx in a much simpler way. But that is not for right now.
More information about the gmp-bugs