benh at kernel.crashing.org
Tue Apr 27 08:28:00 EST 2004
On Mon, 2004-04-26 at 17:16, Anton Blanchard wrote:
> The performance guys have pointed out how slow our userspace segment
> SLB replacement code is. At the moment naca->slb_size isnt very useful,
> we just hardcode it at boot. On top of that, do_slb_bolted doesnt even
> use it.
> So heres a patch to get rid of it. Looking at make_slbe before and after:
I think we should rewrite that critical code path in asm in fact... (and
simulate it). Note that for things like the SLB size, if we ever have
a different size, we can always use dynamic patching...
** Sent via the linuxppc64-dev mail list. See http://lists.linuxppc.org/
More information about the Linuxppc64-dev