This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Problem with peephole to peephole2 conversion
On Thu, Aug 25, 2005 at 06:49:06PM +0530, Ashwin Kolhe wrote:
> I am actually trying to find out WHY and WHEN peep2_reg_dead_p
> was introduced.
When the peephole2 pass was introduced. Another feature of the
peephole2 pass is that we can allocate *new* scratch registers,
and the pattern fails to match if no such scratch is available.
Since exact lifetime information is already required in the pass,
it makes sense to expose that already computed information to
the C predicate of the patterns.
r~