On Thu, May 29, 2014 at 01:53:54PM -0700, Jordan Justen wrote: > Signed-off-by: Jordan Justen <jordan.l.jus...@intel.com> > --- > src/mesa/drivers/dri/i965/brw_tex_layout.c | 16 +++++++++++++--- > 1 file changed, 13 insertions(+), 3 deletions(-) > > diff --git a/src/mesa/drivers/dri/i965/brw_tex_layout.c > b/src/mesa/drivers/dri/i965/brw_tex_layout.c > index 8a0912d..9293229 100644 > --- a/src/mesa/drivers/dri/i965/brw_tex_layout.c > +++ b/src/mesa/drivers/dri/i965/brw_tex_layout.c > @@ -318,9 +318,19 @@ void > brw_miptree_layout(struct brw_context *brw, struct intel_mipmap_tree *mt) > { > bool multisampled = mt->num_samples > 1; > - mt->align_w = intel_horizontal_texture_alignment_unit(brw, mt->format); > - mt->align_h = > - intel_vertical_texture_alignment_unit(brw, mt->format, multisampled); > + > + if (brw->gen == 6 && mt->array_spacing_lod0) { > + /* On gen6, we use array_spacing_lod0 for stencil/hiz because the > + hardware doesn't support multiple mip levels on stencil. Stencil > uses > + W tiling, so we force W tiling alignment when array_spacing_lod0 is > + requested. */
The style of the comment: /* Some... * ...and more */ > + mt->align_w = 64; > + mt->align_h = 64; I didn't really understand this. What happens if this isn't done? > + } else { > + mt->align_w = intel_horizontal_texture_alignment_unit(brw, mt->format); > + mt->align_h = > + intel_vertical_texture_alignment_unit(brw, mt->format, > multisampled); > + } > > switch (mt->target) { > case GL_TEXTURE_CUBE_MAP: > -- > 2.0.0.rc4 > > _______________________________________________ > mesa-dev mailing list > mesa-dev@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/mesa-dev _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev