Hi Greg --

> 2. Is it possible to provide feedback about optimizations that
>    can't be done?  The idea of using (sub)domains to provably
>    escape bounds checking for arrays is powerful, for example,
>    and we would like to see the array accesses where this can't
>    be shown.

Though subdomains (and index types) were introduced in large part to 
support compile-time elimination of bounds checks, in practice this 
analysis and optimization has never been implemented.  Today, bounds 
checks are always on by default (and done in a reasonably bulletproof but 
very unoptimized way) unless the --no-bounds-checks flag is thrown (either 
directly, or through the --no-checks or --fast meta-flags).

Once such an optimization is implemented, I agree that we'd want the 
compiler to produce reports indicating places where bounds checks have or 
have not been eliminated (along with other performance optimization 
reports).

I believe that some Chapel compiler optmizations have reports today, but 
most of these are oriented more at the developer than the end-user, 
unfortunately.  I'm hoping that others on the team might call out 
instances of such reports that they're familiar with.

-Brad


------------------------------------------------------------------------------
_______________________________________________
Chapel-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/chapel-users

Reply via email to