On 11/04/15 05:26, Richard Biener wrote:
> On Tue, Nov 3, 2015 at 7:11 PM, Nathan Sidwell <nathan@acm.org> wrote:
>> Richard,

> this all seems a little bit fragile and relying on implementation details?
> Is the attribute always present?  Is the call argument always a constant
> that fits in a HOST_WIDE_INT (or even int here)?  Are there always enough
> 'dims' in the tree list?  Is the 'dim' value always an INTEGER_CST that
> fits a HOST_WIDE_INT (or even an int here)?


> If so I'd like to see helper functions to hide these implementation details
> from generic code like this.

Like this?

I've added two helper functions to omp-low.c, one to  get the internal fn arg 
number and the other to get a dimension value, given an axis number.  omp-low 
seemed the most appropriate point -- that's  where the dimension processing is, 
and the generation of these internal fn calls.

(Bernd, I'll fixup the dimension folding patch to use these calls before 
applying it.)

ok?

nathan