This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [ping] Re: PATCH: Disable sched1 when optimizing for size
On Sat, 2009-11-21 at 13:32 +0100, Richard Guenther wrote:
> I do not see why scheduling instructions should have an effect on
> code-size. Appearantly it does for arm, so why not disable scheduling
> from code inside the arm backend instead?
>
Because it increases max-live which can cause more spilling and a larger
number of registers to save/restore in the prologue/epilogue.
Note, I'm only talking about the pre-regalloc scheduling pass.
As to why do it in the ARM backend only, the answer is obvious -- I
could, but then other ports that would also benefit from this (which is
all of those that do scheduling as far as I can see) would not benefit.
R.