Open
Description
glm_vec4_refract incorrectly computes k. Per GLSL:
k = 1.0 - eta * eta * (1.0 - dot(N, I) * dot(N, I))Stepping through the function:
glm_vec4 const dot0 = glm_vec4_dot(N, I); // dot(N, I)
glm_vec4 const mul0 = _mm_mul_ps(eta, eta); // eta * eta
glm_vec4 const mul1 = _mm_mul_ps(dot0, dot0); // dot(N, I) * dot(N, I)
glm_vec4 const sub0 = _mm_sub_ps(_mm_set1_ps(1.0f), mul0); // 1.0 - eta * eta
glm_vec4 const sub1 = _mm_sub_ps(_mm_set1_ps(1.0f), mul1); // 1.0 - dot(N, I) * dot(N, I)
glm_vec4 const mul2 = _mm_mul_ps(sub0, sub1); // k = (1.0 - eta * eta) * (1.0 - dot(N, I) * dot(N, I))Which is incorrect. Additionally, the _mm_movemask_ps logic is confusing (_mm_cmplt_ss only sets the lowest element while _mm_movemask_ps creates a mask from all four elements).