[PATCH] Overhaul __builtin_constant_p processing (take 2)

[Roger Sayle]

On Wed, 22 Jan 2003, Mike Stump wrote:
> So, the question is, Roger, is there a way to put in a check on -O0,
> and avoid any costly work?

I believe that the call to purge_builtin_constant_p can be conditionalized
on flag_gcse.  I'll re-run a bootstrap and regression test, and post a
"(take 3)".  It won't help bootstrap times, but will avoid any overhead
at -O0 and -O1.

Roger
--