[Mesa-dev] [PATCH 18/40] i965/blorp: Enable blits on gen8

[Pohjolainen, Topi]

On Mon, Apr 18, 2016 at 12:44:50AM -0700, Kenneth Graunke wrote:
> On Saturday, April 16, 2016 4:42:46 PM PDT Topi Pohjolainen wrote:
> > Signed-off-by: Topi Pohjolainen <topi.pohjolainen at intel.com>
> > ---
> >  src/mesa/drivers/dri/i965/brw_blorp.cpp       | 6 +++++-
> >  src/mesa/drivers/dri/i965/brw_blorp_blit.cpp  | 8 +++++---
> >  src/mesa/drivers/dri/i965/intel_mipmap_tree.c | 3 +--
> >  3 files changed, 11 insertions(+), 6 deletions(-)
> > 
> > diff --git a/src/mesa/drivers/dri/i965/brw_blorp.cpp b/src/mesa/drivers/dri/
> i965/brw_blorp.cpp
> > index b567b42..8c69f6f 100644
> > --- a/src/mesa/drivers/dri/i965/brw_blorp.cpp
> > +++ b/src/mesa/drivers/dri/i965/brw_blorp.cpp
> > @@ -210,7 +210,7 @@ void
> >  brw_blorp_exec(struct brw_context *brw, const brw_blorp_params *params)
> >  {
> >     struct gl_context *ctx = &brw->ctx;
> > -   uint32_t estimated_max_batch_usage = 1500;
> > +   const uint32_t estimated_max_batch_usage = brw->gen >= 8 ? 1800 : 1500;
> 
> May as well just make it 1800 all the time...I think these numbers are
> pretty bogus anyway.

Fine by me, this is un upper limit making sure we don't end up batch
wrapping in the middle. Of course, on gen < 8 we may unnecessarily flush too
early in some case if there were say 1799 bytes available. Whether this makes
any real difference, I can't tell.

> 
> >     bool check_aperture_failed_once = false;
> >  
> >     /* Flush the sampler and render caches.  We definitely need to flush the
> > @@ -236,6 +236,10 @@ retry:
> >     case 7:
> >        gen7_blorp_exec(brw, params);
> >        break;
> > +   case 8:
> > +   case 9:
> > +      gen8_blorp_exec(brw, params);
> > +      break;
> >     default:
> >        /* BLORP is not supported before Gen6. */
> >        unreachable("not reached");
> > diff --git a/src/mesa/drivers/dri/i965/brw_blorp_blit.cpp b/src/mesa/
> drivers/dri/i965/brw_blorp_blit.cpp
> > index e055513..006075c 100644
> > --- a/src/mesa/drivers/dri/i965/brw_blorp_blit.cpp
> > +++ b/src/mesa/drivers/dri/i965/brw_blorp_blit.cpp
> > @@ -252,8 +252,8 @@ brw_blorp_copytexsubimage(struct brw_context *brw,
> >         src_mt->num_samples > 8 || dst_mt->num_samples > 8)
> >        return false;
> >  
> > -   /* BLORP is only supported for Gen6-7. */
> > -   if (brw->gen < 6 || brw->gen > 7)
> > +   /* BLORP is only supported from Gen6 onwards. */
> > +   if (brw->gen < 6)
> >        return false;
> >  
> >     if (_mesa_get_format_base_format(src_rb->Format) !=
> > @@ -354,7 +354,7 @@ brw_blorp_framebuffer(struct brw_context *brw,
> >                        GLbitfield mask, GLenum filter)
> >  {
> >     /* BLORP is not supported before Gen6. */
> > -   if (brw->gen < 6 || brw->gen >= 8)
> > +   if (brw->gen < 6)
> >        return mask;
> >  
> >     /* There is support only for four and eight samples. */
> > @@ -1660,6 +1660,8 @@ brw_blorp_blit_program::texel_fetch(struct brw_reg 
> dst)
> >        texture_lookup(dst, SHADER_OPCODE_TXF, gen6_args, s_is_zero ? 2 : 5);
> >        break;
> >     case 7:
> > +   case 8:
> > +   case 9:
> 
> Seems like this ought to go in a different patch?

Good catch, moved into patch 11: i965/blorp: Prepare sampling for gen9.

> 
> >        switch (key->tex_layout) {
> >        case INTEL_MSAA_LAYOUT_IMS:
> >           /* From the Ivy Bridge PRM, Vol4 Part1 p72 (Multisampled Surface 
> Storage
> > diff --git a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c b/src/mesa/
> drivers/dri/i965/intel_mipmap_tree.c
> > index 9f754e7..f69d629 100644
> > --- a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
> > +++ b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
> > @@ -2165,8 +2165,7 @@ intel_miptree_updownsample(struct brw_context *brw,
> >                             struct intel_mipmap_tree *dst)
> >  {
> >     /* There is support only for four and eight samples. */
> > -   const bool use_blorp = brw->gen < 8 &&
> > -                          src->num_samples != 2 && dst->num_samples != 2 &&
> > +   const bool use_blorp = src->num_samples != 2 && dst->num_samples != 2 &&
> >                            src->num_samples <= 8 && dst->num_samples <= 8;
> >  
> >     if (use_blorp) {
> > 
>