Re: [PATCH] anobjrmap 9 priority mjb tree

[Rajesh Venkatasubramanian]

On Mon, 12 Apr 2004, Hugh Dickins wrote:
> On Mon, 12 Apr 2004, Martin J. Bligh wrote:
> >
> > If it were just a list, maybe RCU would be appropriate. It might be
> > rather write-heavy though ? I think I played with an rwsem instead
> > of a sem in the past too (though be careful if you try this, as for
> > no good reason the return codes are inverted ;-()
>
> Yes, I think all the common paths have to write, in case the
> uncommon paths (truncation and swapout) want to read: the wrong
> way round for any kind of read-write optimization, isn't it?

In common workloads e.g., add libc mapping using  __vma_prio_tree_insert,
mostly you do not add new nodes to the tree. Instead, you just add to
a vm_set list. I am currently considering using rwsem to optimize
such cases. Similarly __vma_prio_tree_remove can also be optimized
in some common cases. I don't know whether it will help. Let us see...

Rajesh



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/