Re: Floating point constant question

Scott Mahlke <mahlke@crhc.uiuc.edu> writes:


>I have a question regarding the following compiler optimization:
>
> x = y / 500.0; ===> x = y * 0.002;
> ...
>Is this or similar optimizations dealing with FP constants ever legal? Do
>any compilers out there do optimizations such as this? Thanks.


It's only a valid replacement if you don't affect the accuracy of the
calculation. For cases where you are dividing by a power of 2, it is
always safe. This is the rule I apply to the compilers I write.


In IEEE single precision, 500.0 is exact (43fa0000), while 0.002 is
actually represented by 0.00200000009499490261077880859375 (3b03126f). On
the other hand, both 512.0 and 0.001953125 are exact.


                                                                regards,
                                                                joe


--
Full Name: Joseph M. Orost
EMail: joe@babel.ho.att.com, attmail!orost
Organization: AT&T Bell Laboratories
SurfaceMail: 943 Holmdel Rd.; Cruz Plaza; Holmdel, NJ 07733
Phone: +1 (908) 946-1115
--

Return to the comp.compilers page.
Search the comp.compilers archives again.