[comp.graphics] Inverse Square Root Algorithm Request

bseeg@spectra.COM (Bob Seegmiller) (02/08/90)

At the '86 SIGGRAPH Conference, in the Image Rendering class or at one
of the Forums, Rob Cook of Pixar gave passing mention of a method for
computing the inverse square root (n^-(.5)) of a number that was
faster(!?) or comparable in speed to taking just the square root
itself.  I've lost the note, and my own hack at it via Newton's method
has proved to be highly unstable.  Does anyone out there know of or
remember the reference?

Thanks in advance.

bseeg@spectra.COM (Bob Seegmiller) (02/08/90)

  SORRY to repost: inews mangled the .sig file inclusion which had my
		   return address

At the '86 SIGGRAPH Conference, in the Image Rendering class or at one
of the Forums, Rob Cook of Pixar gave passing mention of a method for
computing the inverse square root (n^-(.5)) of a number that was
faster(?!) or comparable to the square-root method itself.  I've lost
the note, and my own hack at it via Newton's method has proved to be
highly unstable.  Does anyone out there know of or remember the
reference?

Thanks in advance.
  Bob Seegmiller
  UUCP: ...!nosc!spectra!bseeg