Nick Ing-Simmons wrote:
Having been through the process of validating FPUs a couple of times I have no confidence that 0*Inf will reliably yield a NaN.
I bow to your superior knowledge... :)
My point was that (should we even go down this road) we should follow the rules (such as they are) for generating +-Inf and NaN under the IEEE standards (where supported) with the existing system library calls and avoid flipping bits because we are too clever for our own good (does this sentence contain enough parentheticals to parse as LISP yet?). If a given system library/FPU doesn't support the IEEE representation, we fall back to something else (I don't know what).
Even generating an Inf isn't as easy as it should be.
And of course VAX and S390 at least use other formats which may have neither.
Those are both ripe for Configure overrides; I know that VMS can use either a compliant or non-compliant library (the latter being VAX compatible, IIRC).
Since we went to the trouble of home brewing our own atoi, we should probably make /some/ effort to handle these special cases...
John
-- John Peacock Director of Information Research and Technology Rowman & Littlefield Publishing Group 4501 Forbes Boulevard Suite H Lanham, MD 20706 301-459-3366 x.5010 fax 301-429-5748
