[Mesa-dev] RFC: Haswell resource streamer/hw-generated binding tables (v2)

[Eric Anholt]

Abdiel Janulgue <abdiel.janulgue at linux.intel.com> writes:

> On Thursday, October 10, 2013 01:04:08 PM Eric Anholt wrote:
>> 
>> My basic comment on resource streamer: We need performance data showing
>> that it is a win before we commit it.  I'm not planning on reviewing the
>> changes until we get that data.
>
> At it's current form, I don't expect that much performance win from it. I view 
> it more as a groundwork laying feature. Taking advantage of that optimization 
> is another question. But how can we benefit from it if you do not even review 
> it in the first place?

As a general rule, we don't land code whose purpose is performance
improvement if it doesn't actually improve performance.  If more work is
needed to make it actually improve performance, then we wait until then.

As I understand it, the thing that you think will make this eventually
actually improve performance is state flagging that indicates which
individual surfaces need updating.  Since that should improve
performance even in the non-resource-streamer case, it can be pursued
independently.

I write a lot of code to improve performance that seems like it should
obviously help, and it doesn't pan out in testing.  I save those
branches, and sometimes later something changes that makes them
worthwhile, but a lot of my branches linger forever.  It's just part of
how trying to improve performance works.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/mesa-dev/attachments/20131011/1714b366/attachment.pgp>