On 10/24/2013 11:23 AM, Robert Ellenberg wrote: > It seems like the "right" way is to pass pure inputs by reference using > constant pointers:
I almost agree! I think the right way to pass big inputs is to use pointers to constant memory. > Instead of this: > > pmCartCartDot(PmCartesian v1, PmCartesian v2, double * dot) > > It would look like this with "constant pointer to constant" inputs: > > pmCartCartDot(PmCartesian const * const v1, PmCartesian const * const v2, > double * const dot) I like the first const in the declaration of v1 and v2, but I don't think the second const (or the const in the declaration of dot) is all that useful. It says pmCartCartDot() won't change v1, v2, or dot, but those variables are all locals, so who cares? I guess the second const (and the only const on dot) would give us a compile-time warning if the called function modifies where in memory its local copy of the pointer is pointing to, but that seems like a small benefit for the added syntactic complexity. My preference would be: pmCartCartDot(const PmCartesian *v1, const PmCartesian *v2, double *dot) > I've tested this change on a few of the posemath functions already without > any issues, but I'd like to know what the community thinks before I start > making major changes. I approve (modulo the minor quibble above), and I'd welcome a change like this in master for 2.6. -- Sebastian Kuzminsky ------------------------------------------------------------------------------ October Webinars: Code for Performance Free Intel webinars can help you accelerate application performance. Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from the latest Intel processors and coprocessors. See abstracts and register > http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
