Re: Multiplication by a constant - how?

Preston Briggs wrote:
>
> title="Multiplication by Integer Constants",


I will take a look, thank you.


> Bernstein's techniques won't always find optimal code. One way around
> the "problem" is to write a little program that exhaustively computes
> optimal sequences and let it crunch for a long time.


I've done it with memoizing algorithm that tries and builds a tree of
numbers/decompositions. It produces EXACTLY the same sequence as
Visual C compiler. When I ran it for all 32K numbers, the longest
sequence was 11 operations. By the way, with memoization it really
runs very fast - factoring all 32K numbers took well under half a
minute.


> Since most integer multipliers are pretty fast these days, you'll want
> to always compare against that alternative. (i.e., use the multiplier
> if it's faster)


Yes and no. Not always. On Intel you have to reserve EAX+EDX for mul/imul,
but this forces my register allocator to become non-symmetrical, and forces
it sometimes spill when it should not. Multiplication with shifts can be
done in usually one, sometimes two arbitrary registers.


Regards --


Sergey.Solyanik@bentley.com
--

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