On Mon, Jan 07, 2002 at 11:29:56AM +0000, Peter Naulls wrote: > > > > As shown below, the issues seems to be the non-standard treatment of NaN > > values. Has think come up from other packages? Does anybody here see a > > workaround, or have other suggestions? > > > > Thanks, Dirk > > The only thing unusual about IEEE values on ARM is that the word > ordering of doubles is opposite to that on x86. I suspect that it's > this you're running into. Many compilers have special cases for this. > > Otherwise, to the best of my experience (apart from some minor quirks of > which x86 also has its fair share) ARM behaves totally as expected for > IEEE operations.
I was also under the impression that the FP emulation is pretty close to IEEE standard. I could be way off base, but I seem to recall that gdb has issues with FP on ARM, so i'm not sure your tests using it are valid. Can you reproduce the same problem with a simple test program (eg. not under the debugger)? -R (yes, I'm a newcomer to this list, but hopefully most know me from netwinder.org... and yes, I do intend to install Debian on one of my 'winders one of these days!)
