Messages in this thread | | | Date | Wed, 14 Apr 1999 21:59:39 +0200 (CEST) | From | Andrea Arcangeli <> | Subject | Re: [patch] page cache in per-inode RB-trees |
| |
On Wed, 14 Apr 1999, Andrea Arcangeli wrote:
>thing (that I used while benchmarking RB in userspace). All the rb-code is >not inline but it's fastcall. It's not inline because I don't want to lose >L1 caching.
I am a bit worried by rb_search... Such function is short enough to be put inline for real...
So here it is a new patch:
ftp://e-mind.com/pub/linux/kernel-patches/rbtree-search-inline-2.2.5_arca11.bz2
It's again against 2.2.5_arca11 and it replaces rbtree-2.2.5_arca11 completly. It would be also interesting to know if putting rb_search inline make real differences (it was just a fastcall..).
Well if you are going to benchmark the rb-tree code probably is better that you apply this little variation that will _only_ avoid an asm "call" for every search.
Andrea Arcangeli
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |