Make combine_two_pixels() handle the fact that the GPU has limited subpixel interpola...

[movit] / util.h

diff --git a/util.h b/util.h
index 432171ac015d3fd5921b5eb9fe9c0ccffd92b9b3..082d1146a7f4c86c2d935ffeaa62307f6575b884 100644 (file)
--- a/util.h
+++ b/util.h
@@ -36,19 +36,12 @@ std::string output_glsl_mat3(const std::string &name, const Eigen::Matrix3d &m);
 
 // Calculate where to sample, and with what weight, if one wants to use
 // the GPU's bilinear hardware to sample w1 * x[0] + w2 * x[1].
-static inline void combine_two_samples(float w1, float w2, float *offset, float *total_weight)
-{
-       assert(w1 * w2 >= 0.0f);  // Should not have differing signs.
-       if (fabs(w1 + w2) < 1e-6) {
-               *offset = 0.5f;
-               *total_weight = 0.0f;
-       } else {
-               *offset = w2 / (w1 + w2);
-               *total_weight = w1 + w2;
-       }
-       assert(*offset >= 0.0f);
-       assert(*offset <= 1.0f);
-}
+//
+// Note that since the GPU might have limited precision in its linear
+// interpolation, the effective weights might be different from the ones you
+// asked for. sum_sq_error, if not NULL, will contain the sum of the
+// (estimated) squared errors of the two weights.
+void combine_two_samples(float w1, float w2, float *offset, float *total_weight, float *sum_sq_error);
 
 #ifdef NDEBUG
 #define check_error()