[patch] Lno branch merge part 5 -- #s of loop iterations

Richard Henderson rth@redhat.com
Thu Jul 8 00:20:00 GMT 2004


On Wed, Jul 07, 2004 at 06:49:15AM +0200, Zdenek Dvorak wrote:
> + /* Determine the number of iterations according to condition (for staying
> +    inside loop) BASE0 + STEP0 * i (CODE) BASE1 + STEP1 * i, computed in TYPE.
> +    Store the results to NITER.  */

Please expand on the relationship of these arguments.  It's not
immediately clear, for instance, why 

> +       if (zero_p (step0))
> + 	{
> + 	  if (mmax)
	....
> + 	}
> +       else
> + 	{
> + 	  if (mmin)
	....
> + 	}

zero step0 implies something wrt max and nonzero step implies
something about min.  I'm guessing that this is pre-inc vs post-inc?
Are there assumptions about one or the other being set, but not both?
Are some required to be constantants?

Otherwise it looks good.


r~



More information about the Gcc-patches mailing list