[tree-ssa] tree-ssa vs. fold

[Jeff Sturm]

On Fri, 6 Jun 2003, Roger Sayle wrote:
> There are a number of hooks or callbacks it would be nice for
> fold to have, so that language front-ends such as java and ada
> have some form of control over the optimizations/transformations
> that it performs.  Avoiding bit-field accesses when generating
> .class files comes to mind.

Definitely.  I know little about that code, but I know Andrew Haley
has fixed several fold bugs for Java.  No doubt some bugs remain.

However I think that's a different problem...

> Does a save_expr hook seem a reasonable compromise/workable?

Is there anything language-specific about save_expr?  However rewrite_expr
is fixed, it ought to benefit all frontends.  (That said, my Java example
doesn't ICE when compiled with gcc or g++.  Hum.)

Jeff