When creating shaders and using maths functions it is expected that Blender should match functions in other DCC applications, game engines and shading languages such as GLSL and OSL. This patch adds missing functions to the Blender maths node. Ideally, it would be nice to have these functions available to vectors too but that is not part of this patch. This patch adds the following functions trunc, snap, wrap, compare, pingpong, sign, radians, degrees, cosh, sinh, tanh, exp, smoothmin and inversesqrt. Sign function is based on GLSL and OSL functions and returns zero when x == 0. Differential Revision: https://developer.blender.org/D5957
221 lines
4.3 KiB
GLSL
221 lines
4.3 KiB
GLSL
void math_add(float a, float b, float c, out float result)
|
|
{
|
|
result = a + b;
|
|
}
|
|
|
|
void math_subtract(float a, float b, float c, out float result)
|
|
{
|
|
result = a - b;
|
|
}
|
|
|
|
void math_multiply(float a, float b, float c, out float result)
|
|
{
|
|
result = a * b;
|
|
}
|
|
|
|
void math_divide(float a, float b, float c, out float result)
|
|
{
|
|
result = safe_divide(a, b);
|
|
}
|
|
|
|
void math_power(float a, float b, float c, out float result)
|
|
{
|
|
if (a >= 0.0) {
|
|
result = compatible_pow(a, b);
|
|
}
|
|
else {
|
|
float fraction = mod(abs(b), 1.0);
|
|
if (fraction > 0.999 || fraction < 0.001) {
|
|
result = compatible_pow(a, floor(b + 0.5));
|
|
}
|
|
else {
|
|
result = 0.0;
|
|
}
|
|
}
|
|
}
|
|
|
|
void math_logarithm(float a, float b, float c, out float result)
|
|
{
|
|
result = (a > 0.0 && b > 0.0) ? log2(a) / log2(b) : 0.0;
|
|
}
|
|
|
|
void math_sqrt(float a, float b, float c, out float result)
|
|
{
|
|
result = (a > 0.0) ? sqrt(a) : 0.0;
|
|
}
|
|
|
|
void math_inversesqrt(float a, float b, float c, out float result)
|
|
{
|
|
result = inversesqrt(a);
|
|
}
|
|
|
|
void math_absolute(float a, float b, float c, out float result)
|
|
{
|
|
result = abs(a);
|
|
}
|
|
|
|
void math_radians(float a, float b, float c, out float result)
|
|
{
|
|
result = radians(a);
|
|
}
|
|
|
|
void math_degrees(float a, float b, float c, out float result)
|
|
{
|
|
result = degrees(a);
|
|
}
|
|
|
|
void math_minimum(float a, float b, float c, out float result)
|
|
{
|
|
result = min(a, b);
|
|
}
|
|
|
|
void math_maximum(float a, float b, float c, out float result)
|
|
{
|
|
result = max(a, b);
|
|
}
|
|
|
|
void math_less_than(float a, float b, float c, out float result)
|
|
{
|
|
result = (a < b) ? 1.0 : 0.0;
|
|
}
|
|
|
|
void math_greater_than(float a, float b, float c, out float result)
|
|
{
|
|
result = (a > b) ? 1.0 : 0.0;
|
|
}
|
|
|
|
void math_round(float a, float b, float c, out float result)
|
|
{
|
|
result = floor(a + 0.5);
|
|
}
|
|
|
|
void math_floor(float a, float b, float c, out float result)
|
|
{
|
|
result = floor(a);
|
|
}
|
|
|
|
void math_ceil(float a, float b, float c, out float result)
|
|
{
|
|
result = ceil(a);
|
|
}
|
|
|
|
void math_fraction(float a, float b, float c, out float result)
|
|
{
|
|
result = a - floor(a);
|
|
}
|
|
|
|
void math_modulo(float a, float b, float c, out float result)
|
|
{
|
|
result = c_mod(a, b);
|
|
}
|
|
|
|
void math_trunc(float a, float b, float c, out float result)
|
|
{
|
|
result = trunc(a);
|
|
}
|
|
|
|
void math_snap(float a, float b, float c, out float result)
|
|
{
|
|
result = floor(safe_divide(a, b)) * b;
|
|
}
|
|
|
|
void math_pingpong(float a, float b, float c, out float result)
|
|
{
|
|
result = (b != 0.0) ? abs(fract((a - b) / (b * 2.0)) * b * 2.0 - b) : 0.0;
|
|
}
|
|
|
|
/* Adapted from godotengine math_funcs.h. */
|
|
void math_wrap(float a, float b, float c, out float result)
|
|
{
|
|
float range = b - c;
|
|
result = (range != 0.0) ? a - (range * floor((a - c) / range)) : c;
|
|
}
|
|
|
|
void math_sine(float a, float b, float c, out float result)
|
|
{
|
|
result = sin(a);
|
|
}
|
|
|
|
void math_cosine(float a, float b, float c, out float result)
|
|
{
|
|
result = cos(a);
|
|
}
|
|
|
|
void math_tangent(float a, float b, float c, out float result)
|
|
{
|
|
result = tan(a);
|
|
}
|
|
|
|
void math_sinh(float a, float b, float c, out float result)
|
|
{
|
|
result = sinh(a);
|
|
}
|
|
|
|
void math_cosh(float a, float b, float c, out float result)
|
|
{
|
|
result = cosh(a);
|
|
}
|
|
|
|
void math_tanh(float a, float b, float c, out float result)
|
|
{
|
|
result = tanh(a);
|
|
}
|
|
|
|
void math_arcsine(float a, float b, float c, out float result)
|
|
{
|
|
result = (a <= 1.0 && a >= -1.0) ? asin(a) : 0.0;
|
|
}
|
|
|
|
void math_arccosine(float a, float b, float c, out float result)
|
|
{
|
|
result = (a <= 1.0 && a >= -1.0) ? acos(a) : 0.0;
|
|
}
|
|
|
|
void math_arctangent(float a, float b, float c, out float result)
|
|
{
|
|
result = atan(a);
|
|
}
|
|
|
|
void math_arctan2(float a, float b, float c, out float result)
|
|
{
|
|
result = atan(a, b);
|
|
}
|
|
|
|
void math_sign(float a, float b, float c, out float result)
|
|
{
|
|
result = sign(a);
|
|
}
|
|
|
|
void math_exponent(float a, float b, float c, out float result)
|
|
{
|
|
result = exp(a);
|
|
}
|
|
|
|
void math_compare(float a, float b, float c, out float result)
|
|
{
|
|
result = (abs(a - b) <= max(c, 1e-5)) ? 1.0 : 0.0;
|
|
}
|
|
|
|
void math_multiply_add(float a, float b, float c, out float result)
|
|
{
|
|
result = a * b + c;
|
|
}
|
|
|
|
/* See: https://www.iquilezles.org/www/articles/smin/smin.htm. */
|
|
void math_smoothmin(float a, float b, float c, out float result)
|
|
{
|
|
if (c != 0.0) {
|
|
float h = max(c - abs(a - b), 0.0) / c;
|
|
result = min(a, b) - h * h * h * c * (1.0 / 6.0);
|
|
}
|
|
else {
|
|
result = min(a, b);
|
|
}
|
|
}
|
|
|
|
void math_smoothmax(float a, float b, float c, out float result)
|
|
{
|
|
math_smoothmin(-a, -b, c, result);
|
|
result = -result;
|
|
}
|