(Wonder if there should be a code check in the tree to look for double/quad
	    references, and flag 'em so that they'll all be built with "-msoft-quad-float"
	    when being built under NetBSD/SPARC64?)

	tv asked me if we should change the sparc64 compiler to use this flag by
	default for now.  what do other people think?

No, that causes other problems.  It prevents any code using `long double'
from building properly.  That means when you try to generate certain
files in libc they bomb out because gcc does not generate the right
stubs.

What we need to do is fix the compiler not to gratuitously up-cast
values to 128-bit floating point for intermediates.  Unless there
really is a `long double' in the source, no 128-bit floating point
should be used.  (BTW does any other platform even pretend to support
128-bit floating point?)

Eduardo