Okay I misunderstood that the shader access was relative to the view.
Looks like your patch is a good bug fix.
My original test code on master branch did work on Nvidia.
Jon
On 01/14/2014 12:06 PM, Chris Forbes wrote:
Could do that, but one side or the other *must* account for the view
having a nonzero MinLayer.
Colors[] is indexed based on the layer index relative to the
underlying texture; expectedLayer is currently relative to the
underlying texture; the value used in the shader needs to be relative
to the view [so expectedLayer - view's MinLayer]
Did this actually work on nvidia on master?
On Wed, Jan 15, 2014 at 5:36 AM, Jon Ashburn <[email protected]> wrote:
see inline
On 01/11/2014 05:39 PM, Chris Forbes wrote:
We want to sample from the last layer in the view, but we were actually
trying to sample from a layer beyond that [by the texture's minimum
layer].
This might have worked on hardware where the sampler clamps the layer
index before sampling [NV?], but that relies on undefined behavior.
Signed-off-by: Chris Forbes <[email protected]>
---
tests/spec/arb_texture_view/rendering_levels.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tests/spec/arb_texture_view/rendering_levels.c
b/tests/spec/arb_texture_view/rendering_levels.c
index ef29f78..86bf055 100644
--- a/tests/spec/arb_texture_view/rendering_levels.c
+++ b/tests/spec/arb_texture_view/rendering_levels.c
@@ -205,7 +205,7 @@ test_render_layers(void)
glClear(GL_COLOR_BUFFER_BIT);
expectedLayer = l + numLayers[l] - 1;
- draw_3d_depth(-1.0, -1.0, 2.0, 2.0, expectedLayer);
+ draw_3d_depth(-1.0, -1.0, 2.0, 2.0, numLayers[l] - 1);
with this change then the draw depth is inconsistent with the indexing into
the Colors array
how about this:
expectedLayer = numLayers[l] - 1;
expected[0] = Colors[expectedLayer][0] / 255.0;
expected[1] = Colors[expectedLayer][1] / 255.0;
_______________________________________________
Piglit mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/piglit
_______________________________________________
Piglit mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/piglit
