This is the mail archive of the 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] Add support to data-ref to figure out the dependencies between a[x+i] and a[x+i+1] (PR 31995)

Andrew Pinski/R&D/SCEA wrote on 05/21/2007 02:51:30 PM:
> OK?  Bootstrapped and tested on i686-linux-gnu with no regressions.
> Thanks,
> Andrew Pinski
> ChangeLog:
>  * tree-chrec.c (evolution_function_is_affine_multivariate_p):
>  Add loopno argument.  Use evolution_function_is_invariant_rec_p
>  instead of evolution_function_is_constant_p.
>  Update calls to evolution_function_is_affine_multivariate_p.
>  * tree-chrec.h  (evolution_function_is_affine_multivariate_p):
>  Add loopno argument. 
>  * tree-scalar-evolution.c (gather_chrec_stats): Call
>  evolution_function_is_affine_multivariate_p with a  loop
>  number of 0.
>  * tree-data-ref.c (analyze_miv_subscript): Likewise.
>  (analyze_overlapping_iterations): Likewise.
>  (access_functions_are_affine_or_constant_p):
>  Likewise.
>  (build_classic_dist_vector_1): If the access functions
>  are equal, don't do anything.
>  * gcc.dg/vect/vect-102a.c: New test.

I knew I forgot something, the patch itself.

Andrew Pinski

Attachment: fixpr31995.diff.txt
Description: Text document

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