This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Minor improvement to spilling
On 05/28/10 16:06, Vladimir N. Makarov wrote:
On 05/28/2010 10:57 AM, Vladimir N. Makarov wrote:
On 05/27/2010 12:09 AM, Jeff Law wrote:
Bootstrapped and regression tested on x86_64-unknown-linux-gnu (and
many months ago i686-pc-linux-gnu)
Jeff, the patch looks ok to me (with comment changes about 3 passes).
I very like the idea of the patch (I thought about this too several
months ago when worked on one round-robin issue).
I've started Spec2000 to see what the patch is doing for the
performance. I'll post the results at the end of the day to be
absolutely sure that is everything ok but I am sure the results will
be positive.
The score is practically the same (0.2-0.3% better with the patch) on
x86/x86_64. The code is a bit smaller too especially for x86
SPECInt2000 (about 0.15% smaller code).
I didn't really expect to see a measurable runtime improvement, so to
see some bias towards better runtime scores is definitely a plus. I did
expect to see a small codesize improvement. It's worth noting that this
also improves reloads predictability in a teeny tiny way.
Thanks for the benchmark sanity checking,
Jeff