On 8 August 2012 17:24, Niklas Sörensson <
nikl...@gmail.com> wrote:
> Hello Vegard,
>
> That's interesting and a bit surprising, but I may just be forgetting
> something I've done since then. If you can bisect this and pin down a
> relevant commit I'd be interested and might be able to comment on
> what's going on.
It's this change:
| commit d699be0e76bb70e7a50d77ac643af840a2de0898
| Author: Niklas Sorensson <
nikl...@gmail.com>
| Date: Mon Feb 27 12:36:43 2012 +0100
|
| Patch submitted by Niklas Een.
You can verify it easily just by running d699be0 against d699be0^ a
single time (since it doesn't change the search tree):
| restarts : 1021
| conflicts : 428111 (6472 /sec)
| decisions : 4553588 (0.00 % random) (68835 /sec)
| propagations : 44477282 (672348 /sec)
| conflict literals : 21464953 (7.74 % deleted)
| Memory used : 66.00 MB
| CPU time : 66.1521 s
vs.
| restarts : 1021
| conflicts : 428111 (7275 /sec)
| decisions : 4553588 (0.00 % random) (77384 /sec)
| propagations : 44477282 (755855 /sec)
| conflict literals : 21464953 (7.74 % deleted)
| Memory used : 66.00 MB
| CPU time : 58.8437 s
Turning var_Undef, etc. back into macros makes it fast again. I tried
changing the "const" to "static const", but it remained the same. I'm
using gcc 4.6.3.
Vegard