Two problems with gimplify_init_constructor

[Richard Henderson]

On Thu, Aug 12, 2004 at 05:08:12PM -0400, Richard Kenner wrote:
> If you have a machine that references some objects via a constant pool,
> we may have a MEM for that constant.  Unless we make those in a unique
> alias set (a real possibility), an indirect store to the same type would
> appear to disallow CSE of that constant.

Oh, that.  Yes, we can work out something for read-only memory.


r~