This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [patch,commited] Do not use chrecs for description of overlaps


On 1/27/07, Daniel Berlin <dberlin@dberlin.org> wrote:
On 1/26/07, Zdenek Dvorak <rakdver@atrey.karlin.mff.cuni.cz> wrote:
> I think the right think to do is not to complicate things unless
> it is useful.  We do not use more elements than two now, so pretending that
> we do would make sense only if someone wants to extend the code.  This
> does not seem to be the case, on the contrary, Sebastian mentioned
> completely rewriting the code, basing it on the omega solver.  So I am
> somewhat inclined to keep the things as simple as possible.

First, thanks Zdenek for the cleanup. I went through the changes, and they look okay.


Okay, then could you at least assert that we don't try to build conflict functions greater than the max limit?

a gcc_assert (n <= MAX_DIM) in conflict_fn would do the right thing.


Right now it will just overwrite random memory if you tell it to build
one of size 3.

The maximum is in the function ending in _1_2 where we write
two conflicting functions (the params of the 2 variable access function):

+       *overlaps_a = conflict_fn (2, ova1, ova2);
+       *overlaps_b = conflict_fn (1, ovb);

Sebastian


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]