[Mesa-dev] i965: Batch emission refactoring

[Pohjolainen, Topi]

On Wed, Apr 29, 2015 at 06:54:34PM +0300, Francisco Jerez wrote:
> "Pohjolainen, Topi" <topi.pohjolainen at intel.com> writes:
> 
> > On Tue, Apr 28, 2015 at 03:07:35PM -0700, Kenneth Graunke wrote:
> >> On Wednesday, April 22, 2015 11:47:20 PM Topi Pohjolainen wrote:
> >> > Currently batch emission logic is bolted into using the current
> >> > gl-state and currently bound user shader programs as input. This
> >> > series refactors the api to allow caller to give individual bits of
> >> > information needed explicitly instead of the emission logic
> >> > deducing them from the current state.
> >> > 
> >> > This is needed to support blorp style gl-state-agnostic launching
> >> > of internal utility shaders - shaders used for 2D blitting and
> >> > buffer clearing/resolving.
> >> > 
> >> > I have a follow-up series ready that is actually leveraging this,
> >> > this series is simple set of refactors. I didn't mean it to, but
> >> > it actually fixes one pigit test on ILK due to the way formats
> >> > are set for texture surfaces: arb_copy_image.arb_copy_image-formats.
> >> > 
> >> > Patches 6-13 all address texture surface setup. They move all the
> >> > decision making of values into the hardware agnostic dispatcher
> >> > leaving the hw-specific part just to deal with formatting.
> >> > 
> >> > Topi Pohjolainen (18):
> >> >   i965: Refactor rb surface setup to allow caller to store offsets
> >> >   i965: Expose and refactor brw_update_renderbuffer_surfaces()
> >> >   i965: Refactor and expose brw_upload_binding_table()
> >> >   i965: Remove dependency to tex object in default color setup
> >> >   i965: Refactor sampler state setup
> >> >   i965: Move texture buffer dispatch into single location
> >> >   i965/gen8: Use miptree format in the surface setup
> >> >   i965: Move tex miptree and format resolving into dispatcher
> >> >   i965: Move texture swizzle resolving into dispatcher
> >> >   i965: Pass integer format flag as parameter to surface setup
> >> >   i965: Refactor effective depth calculation
> >> >   i965: Pass texture target as parameter for surface setup
> >> >   i965: Pass slice details as parameters for surface setup
> >> 
> >> I requested a small change on this patch.
> >> 
> >> >   i965/wm/gen6: Refactor program offset setup
> >> 
> >> I NAK'd this one.
> >> 
> >> The rest of this 18 patch series looks great to me and is:
> >> Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>
> >> 
> >> It looks like Curro landed different texture surface state refactoring
> >> patches in the meantime, though...so the two of you will need to decide
> >> how to sort that out :(
> >
> > It looks like I need to do more rebasing than this. There are now these
> > two in master:
> >
> > e17dc00 i965/gen8: Factor out texture surface state set-up from gen8_update_text
> > 6f26ffa i965/gen7: Factor out texture surface state set-up from gen7_update_text
> >
> > Curro, neither has any review tags, why is this?
> 
> Ahh, sorry for the collision.  I also landed "i965: Add helper functions
> to calculate the slice pitch of an array or 3D miptree." without
> apparent review at the same time.  I proposed both the texture and
> miptree refactor so long ago (in December 2013 IIRC) that I had kind of
> lost hope of it getting reviewed anytime soon, and they were a real mess
> to rebase (I actually introduced regressions accidentally during my
> regular rebases due to the constant changes in those code paths).  I
> warned several days before I pushed them and nobody seemed to care.
> 
> In fact I got some R-b tags from Paul Berry at some point, but by the
> time I decided to push the patches things had changed so much that it
> didn't seem fair anymore to include his tags...
> 
> Which of your patches do they conflict with?  If they happened to change
> the API in a different direction that you were meaning to, maybe we can
> have a chat about it tomorrow?

Well, your changes would have been easy to put on top of mine. But I have
seven patch incremental refactoring of which none apply anymore.