disabling vgaHW slow IO path

Daniel Stone daniel at fooishbar.org
Mon Jul 30 22:52:40 PDT 2007

On Tue, Jul 31, 2007 at 11:41:04AM +1000, Dave Airlie wrote:
> While profiling my server startup yesterday I noticed the
> xf86SlowBcopy was a considerable hit on the graph,
> I really don't think this level of IO acess between bytes is necessary
> on modern hardware but I don't want to break anything by accident,
> I've attached two patches, one for the vgaHW module that adds another
> entry point to disable the behaviour and one for the intel driver
> which to disable it if it finds the vgaHW symbol..
> I was tempted to just disable the ioport access on i386 as it is 2007
> but maybe someone can point me at some hw that needs it..

Why not do the reverse, and have it explicitly _enabled_?  Given that,
up until 6.9, we weren't using this, but instead a delay loop that has
been more or less meaningful for the last ten years due to clocks being
drastically higher, I doubt there's much hardware that still relies on
the ioport.

We don't need yet another arbitrary line in ScreenInit for drivers:
we've got enough of those already ...

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.x.org/archives/xorg/attachments/20070731/b2af2471/attachment.pgp>

More information about the xorg mailing list