Re: multiplication by constant - quick comparison of algorithms

bgb@iexist.att.com (Brian G Beuning) writes:
>Isn't it a little dangerous to use sub? Won't it overflow sometimes when
>the multiply would not.


As long as the target machine supports non-overflow faulting logical shift
left and non-faulting subtract, the shift & subtract trick works fine.
The "overflow" does not prevent the correct answer from being obtained.


        e.g. x * 7 = (x << 3) - x


        For a large input, say 0x12492492:


        (0x7ffffffe) = ((0x92492490) - 0x12492492)


On some machines, the operations in question are called "unsigned".
(e.g., subu).


If you need to fully mimic the potential faulting of the multiply
instruction (e.g. for compiling Ada), then the shift and subtract
technique cannot be used. However, multiple adds can still be used (and
even shifts if the target machine supports an overflow-checking arithmetic
left shift).


                                                                regards,
                                                                joe


--
Full Name: Joseph M. Orost
EMail: joe@babel.ho.att.com
Organization: Echo Logic
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.