Re: how to generate code for (a,b):=(b,a)

Dave Lloyd <Dave@occl-cam.demon.co.uk> wrote:


[snip]


>With our Algol 68 compiler, the equivalent code generates (x86):
>PROC foo (E:\tests\a68\perf\swaps.a68)[6]
> di := cx
> cx := bx
> bx := ax
> ax := di
> CALL bar
> RETURN
>
>(Yes I know it should be just JMP foo with no return, one day soon I'll
>tidy that one up).


          Shouldn't that be foo run with
                    CALL foo
and ending with
                    JMP bar
? If you then end up optimizing the tail one level up, you'd change
the CALL foo to JMP foo, too and so on up as far as you go. It seems
weird to me to have all but the bottom level call being jumpified
instead of the top being the one left.


    1) Am I missing something? (Quite possibly, but what?)
    2) Is either way better than the other or is it just a preference
issue?


Sincerely,


Gene Wirchenko
--

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