[comp.sys.hp] HP floating point '040 problem

smiller@wet.UUCP (Gregory Shane Miller) (06/16/91)

Earlier a gentlmen posted that not all the FPU stuff goes on the '040
and that the rest is handled by exception handlers (I'm sorry I've don't
have his name or email address).  Well this is absolutely correct.  The
stuff that did not go to the CPU was primarily the trig and some (as I
recall) log functions.  When you execute a CPU instruction for sin for 
example it causes an exception since of course the instruction does not
exist.  A number of things to bear in mind:

	[a] where are the exception handlers?  These would be somewhat
	    different from software function of sin()
	[b] When the exception occurs, at least on IBMs, no disk writes
	    are likely to occur since the exception have more priority
	    than disk writes.
	[c] When you re-compilied the source code, what libraries did
	    you use?  Did you include the proper commands to link in 
	    exception handlers?

I do not have UNIX stuff (only IBM MS-DOS) and hope I have not insulted your
intelligence or over stepped my grounds.  Good luck.
-- 
G. Shane Miller [ smiller@wet.UUCP (415) 453-4926 ]