On Tuesday, 18 March 2014 at 08:04:21 UTC, Iain Buclaw wrote:
Really? That is a problem? I'm not familiar with LDC or LLVM but I would at least expect the backend be able to provide that information on command. Or is this the fact that they still use native reals for CTFE? If so, then you can look at longdouble implementation in either DMD MSVC or GDC for that.
It is indeed only a question of polishing Kai's implementation of 'real' operations on the basis of llvm::APFloat, which then also allows us to drop Rainer's longdouble code for MSVC builds.
David
