This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: [Fortran] Rewrite section vs. section dependency analysis
- From: Sebastian Pop <sebastian dot pop at cri dot ensmp dot fr>
- To: Roger Sayle <roger at eyesopen dot com>
- Cc: fortran at gcc dot gnu dot org, gcc-patches at gcc dot gnu dot org
- Date: Wed, 5 Apr 2006 15:07:47 +0200
- Subject: Re: [Fortran] Rewrite section vs. section dependency analysis
- References: <Pine.LNX.4.44.0604041124490.31781-100000@www.eyesopen.com>
Roger Sayle wrote:
>
> The restructuring of the code below also simplifies the task of
> implementing more complex dependency analysis when the ranges
> overlap (GCD test, Banerjee test, delta test, omega test, etc..).
> I've no immediate plans but I suspect the way forward will be
> to convert the GMP mpz_t into INTEGER_CSTs (if they fit) then
> construct a POLYNOMIAL_CHREC tree node for each fortran array
> section and pass them to the middle-end's SCEV infrastructure
> (possibly after the omega library is integrated into the tree).
> This would allow us to handle things like a(1:5:2) = a(2:6:2).
>
The omega test is independent of the polynomial_chrec infrastructure,
so you can directly formulate an omega problem just using integers.
In general when the test is simple enough to be caught by the gcd
test, omega is performing the dependence test as fast as the simpler
test: if you opt for the omega test, you don't need the other
dependence tests on the side.
Sebastian