Hope this isn't a stupid question.  I am new to internals of floating point
format.  Writing a utility and need to clarify my understanding...

I note from Hitech C manual.pdf, page 81, that the bias used by HiTech in 24
and 32 bit floating point numbers is 127.  This is equal to 2^(m-1)-1 where
m is the number of bits in the exponent.  The subtraction of 1 from the
exponent bias is to allow a slot for Inf and NaN representations in the
IEEE754 standard.

So, does this mean that HiTech supports Inf and NaN arithmetic?  I have
never used Inf or NaN, probably never will, but I assume it is important
because it means that division/multiplication/addition/subtraction routines
must be slightly different?

It is also possible they just use this bias to be consistent with IEEE 754,
but don't actually support Inf or NaN values.  For example, Inf is a special
case where the biased exponent field is all ones (biased exponent = 255, :.
exponent = 128), and the mantissa field is all zeros.  NaN is an Inf except
that mantissa is not all zeros.  Rather than interpret such a number as Inf
or NaN, you could just say it was the biggest exponent, rather than the
special case of Infinity.

So is the 127 bias used to support Inf and NaN?

p.s. Is it called 'NaN' because multiplying a mantissa by Infinity is silly,
indicating a bug?

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics