Messages in this thread | | | Subject | Re: [PATCH v4 12/13] fair.c: Use generic rbtree impl in fair scheduler | From | Peter Zijlstra <> | Date | Wed, 27 Jun 2012 14:36:34 +0200 |
| |
On Tue, 2012-06-26 at 16:59 -0500, Daniel Santos wrote: > Also, to address why we're not using entity_before (or a less() > function) directly, there's two main reasons (one that doesn't even > affect CFS). The first reason is that an "is equal" evaluation would > also be required for insertions in trees with unique keys,
Right, I saw that feature in your code, are there any such trees in the kernel? It would be good to have a kernel code conversion for each variant your code supports.
> as well as > all lookups. This doesn't doesn't affect CFS because it isn't doing > lookups (it only cares about leftmost) and duplicate keys are allowed. > Right.. there's that.
Too bad, partial order is slightly cheaper at times. A well, nothing for it.
| |