IEEE vs LCAS

[David Hough]

Fred says:

>    This means IEEE-754 math will conform to LCAS provided that trapping
>    for "invalid", "overflow", and "zero-divide" exceptions are enabled.
>    This conflicts with the NCEG IEEE proposal that all traps be disabled
>    by default.

The way I'd put it is:

	This means LCAS will conform to IEEE 754 provided that "acted on"
	is expanded to encompass
	recording the exception in a flag and continuing with the
	754-defined default.

In my view LCAS is a misdirected effort.  For those who don't know, it's
an attempt to provide a meta-specification on how arithmetic works
in a language independent way, mostly independent of how the arithmetic
works, except for some curious restrictions that rule out Cray division
and IEEE exception handling as we currently know them.  So an implementation
like the original UCSD Pascal for Apple-][, that gave you the option of
pressing the space bar before rebooting after an overflow, might well be
LCAS-compliant - the behavior was documented well enough -
without being any good for numerical work.

IEEE 754 was too meta for its own good.  The current NCEG IEEE subeffort to 
nail down a specific language and a (fairly) specific type of arithmetic
is more in the right direction, provided it doesn't get crippled by 
attempts to preserve unwarranted compatibility with all the mistakes in
ANSI-C.