Differnt drivers give a wide range of results. Since the test is not about the
quality of linear filtering just increase the tolerance.
---
tests/spec/arb_fragment_program/sparse-samplers.c | 21 +++++++++++++++++++++
1 file changed, 21 insertions(+)
diff --git a/tests/spec/arb_fragment_program/sparse-samplers.c
b/tests/spec/arb_fragment_program/sparse-samplers.c
index 40d2408..0400008 100644
--- a/tests/spec/arb_fragment_program/sparse-samplers.c
+++ b/tests/spec/arb_fragment_program/sparse-samplers.c
@@ -38,16 +38,26 @@ PIGLIT_GL_TEST_CONFIG_BEGIN
PIGLIT_GL_TEST_CONFIG_END
+extern float piglit_tolerance[4];
+
static int texrect_w;
static int texrect_h;
static const float r[] = {1, 0, 0, 1};
static const float g[] = {0, 1, 0, 1};
static const float b[] = {0, 0, 1, 1};
static const float w[] = {1, 1, 1, 1};
+static float linear_tolerance[4];
+static float nearest_tolerance[4];
static bool
test_nearest(int x)
{
+ int i;
+
+ /* Set default tolerance for nearest filtering. */
+ for (i = 0; i < 4; i++)
+ piglit_tolerance[i] = nearest_tolerance[i];
+
return (piglit_probe_rect_rgba(x, 0,
texrect_w, texrect_h, r) &&
piglit_probe_rect_rgba(x + texrect_w, 0,
@@ -73,6 +83,11 @@ test_linear(int x)
const float offset = 1.0f/(float)texrect_w;
const float lfm = 0.5f - offset;
const float lfp = 0.5f + offset;
+
+ /* Increase tolerance a bit for linear filtering. */
+ for (i = 0; i < 4; i++)
+ piglit_tolerance[i] = linear_tolerance[i];
+
for (i = 0; i < 4; i++)
average[i] = (r[i] *lfm*lfm +
g[i] *lfm*lfp +
@@ -158,4 +173,10 @@ piglit_display(void)
void
piglit_init(int argc, char **argv)
{
+ int i;
+
+ for (i = 0; i < 4; i++) {
+ nearest_tolerance[i] = piglit_tolerance[i];
+ linear_tolerance[i] = 0.02;
+ }
}
--
1.8.1.2
_______________________________________________
Piglit mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/piglit
