🚀 go-pugleaf

I would like to write code for gcc 2.2 that will take the absolute
value of a double in the fastest possible way (for a tight inner
loop).

I currently do this

  double num;
  if (num < 0)
	num *= -1;

but it occurs to me that rather than doing the test, possibly followed
by a multiply, it may be cheaper to just force the IEEE sign bit to be
positive somehow (just one operation).  Any suggestions on how best to
do this for a double, float or int will be most appreciated.

I'm running on a sun IPX, using gcc -O.



Thanks,

-Bob

------------------------------------------------------
R. Schmid     <uunet!cbmvax!schmid>
              <schmid@cbmvax.cbm.commodore.com>

Robert Schmid
VLSI Design
Commodore Technology
1200 Wilson Drive
West Chester, PA  19380
------------------------------------------------------