[patch] Dependence analysis fixes
Sebastian Pop
sebastian.pop@inria.fr
Wed Apr 11 17:38:00 GMT 2007
On 4/11/07, Zdenek Dvorak <rakdver@atrey.karlin.mff.cuni.cz> wrote:
> I found another problem. In analyze_miv_subscript, we claim that if the
> difference of subscripts is a constant c and one of the steps of the
> subscript does not divide c, then the subscripts are independent. This
> is incorrect, consider e.g. {{0,+,2},+,3} and {{5,+,2},+,3}; what we
> really want to test is that the greatest common divisor of the steps
> does not divide c.
>
> OK once it passes bootstrap and regtesting?
>
Yes, sure. Thank you.
Sebastian
More information about the Gcc-patches
mailing list