Undesired regular Perlin noise - c++

I'm trying to implement fBm onto a sphere for a planet. To create my sphere, I convert it to such from a cube.
Unfortunately, the fBm that gets generated appears as mirrored patches. In addition, it only does it on 2 faces (wrapping the values for the other faces).
This leads to a similarly stretched look when rendered as a sphere
The noise function is the improved noise as described by Ken Perlin,
I adapted this for HLSL:
float fade(float t) { return t * t * t * (t * (t * 6 - 15) + 10); }
float lerp(float t, float a, float b) { return a + t * (b - a); }
float grad(int hash, float x, float y, float z) {
int h = hash & 15; // CONVERT LO 4 BITS OF HASH CODE
float u = h<8 ? x : y, // INTO 12 GRADIENT DIRECTIONS.
v = h<4 ? y : h==12||h==14 ? x : z;
return ((h&1) == 0 ? u : -u) + ((h&2) == 0 ? v : -v);
}
int p[512] = { 151,...180 }; //0-255 twice
float noise(float x, float y, float z) {
int X = (int)floor(x) & 255; // FIND UNIT CUBE THAT
int Y = (int)floor(y) & 255; // CONTAINS POINT.
int Z = (int)floor(z) & 255;
x -= floor(x); // FIND RELATIVE X,Y,Z
y -= floor(y); // OF POINT IN CUBE.
z -= floor(z);
float u = fade(x), // COMPUTE FADE CURVES
v = fade(y), // FOR EACH OF X,Y,Z.
w = fade(z);
int A = p[X ]+Y, AA = p[A]+Z, AB = p[A+1]+Z, // HASH COORDINATES OF
B = p[X+1]+Y, BA = p[B]+Z, BB = p[B+1]+Z; // THE 8 CUBE CORNERS,
return lerp(w, lerp(v, lerp(u, grad(p[AA ], x , y , z ), // AND ADD
grad(p[BA ], x-1, y , z )), // BLENDED
lerp(u, grad(p[AB ], x , y-1, z ), // RESULTS
grad(p[BB ], x-1, y-1, z ))),// FROM 8
lerp(v, lerp(u, grad(p[AA+1], x , y , z-1 ), // CORNERS
grad(p[BA+1], x-1, y , z-1 )), // OF CUBE
lerp(u, grad(p[AB+1], x , y-1, z-1 ),
grad(p[BB+1], x-1, y-1, z-1 ))));
}
This implementation has worked as expected in a previous project, however for this project it appears to instead create a smoothed out grid when I use the vertex position as the input.
It's a unit cube, so the values aren't integers but I can't figure out why it's not creating the typical Perlin noise texture.
Any help would be greatly appreciated, I'll provide more information if it's needed.

The array of ints p can't be accessed by the function so I'm assuming the values in it are undefined.
A quick fix is to make the array static, but this is really slow.
So now I need to pass in the array. But I'm having trouble with that.

I use the noise function below in a Dx11 planet rendering project. I've included an fBm function too. I found it (written in GLSL) on the WebGL shader programming website ShaderToy.
It was written by the godlike inigo quilez, who authored the site.
Give it a try, I hope it's of some help. All credit should go to inigo quilez for his work. Porting it to HLSL is trivial. I've only tested in in shader model 5, but I'm sure it'll work under 4 at least.
// hash based 3d value noise
// function taken from https://www.shadertoy.com/view/XslGRr
// Created by inigo quilez - iq/2013
// License Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.
// ported from GLSL to HLSL
cbuffer cbNoiseParameters
{
float _rOctaves;
float _rLacunarity;
float _rFrequency;
float _rAmplitude;
float _rGain;
float _rOffset;
};
float hash( float n )
{
return frac(sin(n)*43758.5453);
}
float noise( float3 x )
{
// The noise function returns a value in the range -1.0f -> 1.0f
float3 p = floor(x);
float3 f = frac(x);
f = f*f*(3.0-2.0*f);
float n = p.x + p.y*57.0 + 113.0*p.z;
return lerp(lerp(lerp( hash(n+0.0), hash(n+1.0),f.x),
lerp( hash(n+57.0), hash(n+58.0),f.x),f.y),
lerp(lerp( hash(n+113.0), hash(n+114.0),f.x),
lerp( hash(n+170.0), hash(n+171.0),f.x),f.y),f.z);
}
float fBm( float3 vPt )
{
float octaves = _rOctaves;
float lacunarity = _rLacunarity;
float frequency = _rFrequency;
float amplitude = _rAmplitude;
float gain = _rGain;
float offset = _rOffset;
float value = 0.f;
for( int i = 0; i < octaves; ++ i )
{
value += noise( vPt * frequency ) * amplitude;
amplitude *= gain;
frequency *= lacunarity;
}
return value;
}

Related

How to generate geometry for part of a sphere

There are few algorithms to generate the points needed to render a sphere, but I couldn't not find a good algorithm nor to adapt a full sphere algorithm to generate the points for a part of the sphere.
Let's assume I have the min/max latitude, min/max longitude and the sphere radius. How do I generate a mesh to be able to render this part of the sphere?
I made hemisphere & part of hemisphere mesh. I edit songho Ahn's source code slightly and you can refer from following link :
http://www.songho.ca/opengl/gl_sphere.html
the main method of class Sphere are buildVertices*
You can get desired result from editing this method.
I made hemisphere like this :
void HemiSphere::buildVerticesSmooth()
{
...
float sectorStep = 2 *PI / sectorCount;
float stackStep = (PI / 2) / stackCount;
float sectorAngle, stackAngle;
for(int i = 0; i <= stackCount; ++i)
{
stackAngle = -PI / 2 + i * stackStep; // starting from -pi/2 to 0
xy = radius * cosf(stackAngle); // r * cos(u)
z = radius *( 1 + sinf(stackAngle) ); // r * sin(u)
...
and I also made semi-hemisphere which cut its floor so that it have even floor on the bottom. To do this, leave x y coordinate unchanged, just edit z coordinate to 0.
void HemiSphere::buildVerticesSmooth()
{
....
float sectorStep = 2 *PI / sectorCount;
float stackStep = (PI / 2) / stackCount;
float sectorAngle, stackAngle;
float d = sqrt(radius*radius - flat_radius*flat_radius)/radius;
for(int i = 0; i <= stackCount; ++i)
{
stackAngle = -PI / 2 + i * stackStep; // starting from -pi/2 to 0
xy = radius * cosf(stackAngle); // r * cos(u)
z = radius *( d + sinf(stackAngle) ); // r * sin(u)
if (z<0.0) z = 0.0;
....
Hope this helps.

Perlin Noise algorithm does not seem to produce gradient noise

I am attempting to implement Perlin Noise in c++.
Firstly, the problem (I think) is that the output is not what I expect. Currently I simply use the generated Perlin Noise values in a greyscaled image, and this is the results I get:
However, from my understanding, it's supposed to look more along the lines of:
That is, the noise I am producing currently seems to be more along the lines of "standard" irregular noise.
This is the Perlin Noise Algorithm I have implemented so far:
float perlinNoise2D(float x, float y)
{
// Find grid cell coordinates
int x0 = (x > 0.0f ? static_cast<int>(x) : (static_cast<int>(x) - 1));
int x1 = x0 + 1;
int y0 = (y > 0.0f ? static_cast<int>(y) : (static_cast<int>(y) - 1));
int y1 = y0 + 1;
float s = calculateInfluence(x0, y0, x, y);
float t = calculateInfluence(x1, y0, x, y);
float u = calculateInfluence(x0, y1, x, y);
float v = calculateInfluence(x1, y1, x, y);
// Local position in the grid cell
float localPosX = 3 * ((x - (float)x0) * (x - (float)x0)) - 2 * ((x - (float)x0) * (x - (float)x0) * (x - (float)x0));
float localPosY = 3 * ((y - (float)y0) * (y - (float)y0)) - 2 * ((y - (float)y0) * (y - (float)y0) * (y - (float)y0));
float a = s + localPosX * (t - s);
float b = u + localPosX * (v - u);
return lerp(a, b, localPosY);
}
The function calculateInfluence has the job of generating the random gradient vector and distance vector for one of the corner points of the current grid cell and returning the dot product of these. It is implemented as:
float calculateInfluence(int xGrid, int yGrid, float x, float y)
{
// Calculate gradient vector
float gradientXComponent = dist(rdEngine);
float gradientYComponent = dist(rdEngine);
// Normalize gradient vector
float magnitude = sqrt( pow(gradientXComponent, 2) + pow(gradientYComponent, 2) );
gradientXComponent = gradientXComponent / magnitude;
gradientYComponent = gradientYComponent / magnitude;
magnitude = sqrt(pow(gradientXComponent, 2) + pow(gradientYComponent, 2));
// Calculate distance vectors
float dx = x - (float)xGrid;
float dy = y - (float)yGrid;
// Compute dot product
return (dx * gradientXComponent + dy * gradientYComponent);
}
Here, dist is a random number generator from C++11:
std::mt19937 rdEngine(1);
std::normal_distribution<float> dist(0.0f, 1.0f);
And lerp is simply implemented as:
float lerp(float v0, float v1, float t)
{
return ( 1.0f - t ) * v0 + t * v1;
}
To implement the algorithm, I primarily made use of the following two resources:
Perlin Noise FAQ
Perlin Noise Pseudo Code
It's difficult for me to pinpoint exactly where I seem to be messing up. It could be that I am generating the gradient vectors incorrectly, as I'm not quite sure what type of distribution they should have. I have tried with a uniform distribution, however this seemed to generate repeating patterns in the texture!
Likewise, it could be that I am averaging the influence values incorrectly. It has been a bit difficult to discern exactly how it should be done from from the Perlin Noise FAQ article.
Does anyone have any hints as to what might be wrong with the code? :)
It seems like you are only generating a single octave of Perlin Noise. To get a result like the one shown, you need to generate multiple octaves and add them together. In a series of octaves, each octave should have a grid cell size double that of the last.
To generate multi-octave noise, use something similar to this:
float multiOctavePerlinNoise2D(float x, float y, int octaves)
{
float v = 0.0f;
float scale = 1.0f;
float weight = 1.0f;
float weightTotal = 0.0f;
for(int i = 0; i < octaves; i++)
{
v += perlinNoise2D(x * scale, y * scale) * weight;
weightTotal += weight;
// "ever-increasing frequencies and ever-decreasing amplitudes"
// (or conversely decreasing freqs and increasing amplitudes)
scale *= 0.5f;
weight *= 2.0f;
}
return v / weightTotal;
}
For extra randomness you could use a differently seeded random generator for each octave. Also, the weights given to each octave can be varied to adjust the aesthetic quality of the noise. If the weight variable is not adjusted each iteration, then the example above is "pink noise" (each doubling of frequency carries the same weight).
Also, you need to use a random number generator that returns the same value each time for a given xGrid, yGrid pair.

Missing vertices in sphere model

I'm learning OpenGL and I'm working on creating my own sphere model. I was able to draw a complete sphere, although with some puzzling results. I'm wondering if someone can explain (and possibly correct) my code.
The rationale: build triangles using carthesian coordinates calculated from polar coordinates. The number of subdivisions tells me the steps in phi or theta radians to generate the sphere point. From a particular point P(phi, theta), I build the other edges of the sector for [phi, delta_phi], [theta, delta_tetha], with phi varying from [0, pi] (180 degrees) and tetha from [0, 2*pi] (360 degrees).
This is the code I came up with (I'm using QT objects, but it should be pretty straitghforward):
QVector3D polarToCarthesian(float rho, float phi, float theta)
{
float r = qSin(phi) * rho;
float y = qCos(phi) * rho;
float x = r * qSin(theta);
float z = r * qCos(theta);
return QVector3D{x, y, z};
}
void make_sector(QVector<QVector3D>& mesh, float phi, float theta, float rho, float deltaPhi, float deltaTheta)
{
QVector3D p1 = polarToCarthesian(rho, phi, theta);
QVector3D p2 = polarToCarthesian(rho, phi, theta + deltaTheta);
QVector3D p3 = polarToCarthesian(rho, phi + deltaPhi, theta);
QVector3D p4 = polarToCarthesian(rho, phi + deltaPhi, theta + deltaTheta);
// First Triangle
mesh.push_back(p1);
mesh.push_back(p1); // Normal
mesh.push_back(p3);
mesh.push_back(p3); // Normal
mesh.push_back(p2);
mesh.push_back(p2); // Normal
// Second Triangle
mesh.push_back(p2);
mesh.push_back(p2); // Normal
mesh.push_back(p3);
mesh.push_back(p3); // Normal
mesh.push_back(p4);
mesh.push_back(p4); // Normal
}
void build_sphere(QVector<QVector3D>& mesh, int ndiv)
{
const float PHI_MAX = static_cast<float>(M_PI);
const float THETA_MAX = static_cast<float>(M_PI) * 2;
const float delta_phi = PHI_MAX / ndiv;
const float delta_theta = THETA_MAX / ndiv;
for (int i = 0; i < ndiv; ++i) {
float phi = i * delta_phi;
for (int j = 0; j < ndiv; ++j) {
float theta = j * delta_theta;
make_sector(mesh, phi, theta, 1.0f, delta_phi, delta_theta);
}
}
}
// Then I can generate the sphere with
build_sphere(sphere_mesh, 10);
However, I cannot get a complete sphere unless I change the iteration for phi from ndiv iterations to 3 * ndiv iterations. I don't understand why! Phi should vary from 0 to PI to cover the whole Y axis while Theta from 0 to 2 * pi should cover the XZ plane.
Can somebody explain what's happening and why 3 * ndiv works?
phi should go from -π/2 to +π/2, not from 0 to π:
float phi = i * delta_phi - (M_PI / 2);
Also, you appear to have your r and y calculations the wrong way around. You want r to be maximum at the equator (when phi == 0).
I think your code may have worked (albeit producing twice as many polygons as it should have) if you had stuck at 2 * ndiv. As it is, going from 0 to π only puts polygons in the northern hemisphere, so you have to keep going beyond that to have any polygons in the southern hemisphere.
p.s. there's no 'h' in cartesian ;)

glRotate divide-by-zero

I think I understand why calling glRotate(#, 0, 0, 0) results in a divide-by-zero. The rotation vector, a, is normalized: a' = a/|a| = a/0
Is that the only situation glRotate could result in a divide-by-zero? Yes, I know glRotate is deprecated. Yes, I know the matrix is on the OpenGL manual. No, I don't know linear algebra enough to confidently answer the question from the matrix. Yes, I think it would help. Yes, I asked this already in #opengl (can you tell?). And no, I didn't get an answer.
I would say yes. And I would say that you are right about the normalization step as well. The matrix shown in the OpenGL manual only consists of multiplications. And multiplying a vector would result into the same. Of course, it would do strange things if you result in a vector of (0,0,0). OpenGL states in the same manual that |x,y,z|=1 (or OpenGL will normalize).
So IF it wouldn't normalize, you would end up with a very empty matrix of:
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 1
Which will implode your object in the strangest ways. So DON'T call this function with a zero-vector. If you would like to, tell me why.
And I recommend using a library like GLM to do your matrix calculations if it gets too complicated for some simple glRotates.
Why should it divide by zero when you can check for that?:
/**
* Generate a 4x4 transformation matrix from glRotate parameters, and
* post-multiply the input matrix by it.
*
* \author
* This function was contributed by Erich Boleyn (erich#uruk.org).
* Optimizations contributed by Rudolf Opalla (rudi#khm.de).
*/
void
_math_matrix_rotate( GLmatrix *mat,
GLfloat angle, GLfloat x, GLfloat y, GLfloat z )
{
GLfloat xx, yy, zz, xy, yz, zx, xs, ys, zs, one_c, s, c;
GLfloat m[16];
GLboolean optimized;
s = (GLfloat) sin( angle * DEG2RAD );
c = (GLfloat) cos( angle * DEG2RAD );
memcpy(m, Identity, sizeof(GLfloat)*16);
optimized = GL_FALSE;
#define M(row,col) m[col*4+row]
if (x == 0.0F) {
if (y == 0.0F) {
if (z != 0.0F) {
optimized = GL_TRUE;
/* rotate only around z-axis */
M(0,0) = c;
M(1,1) = c;
if (z < 0.0F) {
M(0,1) = s;
M(1,0) = -s;
}
else {
M(0,1) = -s;
M(1,0) = s;
}
}
}
else if (z == 0.0F) {
optimized = GL_TRUE;
/* rotate only around y-axis */
M(0,0) = c;
M(2,2) = c;
if (y < 0.0F) {
M(0,2) = -s;
M(2,0) = s;
}
else {
M(0,2) = s;
M(2,0) = -s;
}
}
}
else if (y == 0.0F) {
if (z == 0.0F) {
optimized = GL_TRUE;
/* rotate only around x-axis */
M(1,1) = c;
M(2,2) = c;
if (x < 0.0F) {
M(1,2) = s;
M(2,1) = -s;
}
else {
M(1,2) = -s;
M(2,1) = s;
}
}
}
if (!optimized) {
const GLfloat mag = SQRTF(x * x + y * y + z * z);
if (mag <= 1.0e-4) {
/* no rotation, leave mat as-is */
return;
}
x /= mag;
y /= mag;
z /= mag;
/*
* Arbitrary axis rotation matrix.
*
* This is composed of 5 matrices, Rz, Ry, T, Ry', Rz', multiplied
* like so: Rz * Ry * T * Ry' * Rz'. T is the final rotation
* (which is about the X-axis), and the two composite transforms
* Ry' * Rz' and Rz * Ry are (respectively) the rotations necessary
* from the arbitrary axis to the X-axis then back. They are
* all elementary rotations.
*
* Rz' is a rotation about the Z-axis, to bring the axis vector
* into the x-z plane. Then Ry' is applied, rotating about the
* Y-axis to bring the axis vector parallel with the X-axis. The
* rotation about the X-axis is then performed. Ry and Rz are
* simply the respective inverse transforms to bring the arbitrary
* axis back to its original orientation. The first transforms
* Rz' and Ry' are considered inverses, since the data from the
* arbitrary axis gives you info on how to get to it, not how
* to get away from it, and an inverse must be applied.
*
* The basic calculation used is to recognize that the arbitrary
* axis vector (x, y, z), since it is of unit length, actually
* represents the sines and cosines of the angles to rotate the
* X-axis to the same orientation, with theta being the angle about
* Z and phi the angle about Y (in the order described above)
* as follows:
*
* cos ( theta ) = x / sqrt ( 1 - z^2 )
* sin ( theta ) = y / sqrt ( 1 - z^2 )
*
* cos ( phi ) = sqrt ( 1 - z^2 )
* sin ( phi ) = z
*
* Note that cos ( phi ) can further be inserted to the above
* formulas:
*
* cos ( theta ) = x / cos ( phi )
* sin ( theta ) = y / sin ( phi )
*
* ...etc. Because of those relations and the standard trigonometric
* relations, it is pssible to reduce the transforms down to what
* is used below. It may be that any primary axis chosen will give the
* same results (modulo a sign convention) using thie method.
*
* Particularly nice is to notice that all divisions that might
* have caused trouble when parallel to certain planes or
* axis go away with care paid to reducing the expressions.
* After checking, it does perform correctly under all cases, since
* in all the cases of division where the denominator would have
* been zero, the numerator would have been zero as well, giving
* the expected result.
*/
xx = x * x;
yy = y * y;
zz = z * z;
xy = x * y;
yz = y * z;
zx = z * x;
xs = x * s;
ys = y * s;
zs = z * s;
one_c = 1.0F - c;
/* We already hold the identity-matrix so we can skip some statements */
M(0,0) = (one_c * xx) + c;
M(0,1) = (one_c * xy) - zs;
M(0,2) = (one_c * zx) + ys;
/* M(0,3) = 0.0F; */
M(1,0) = (one_c * xy) + zs;
M(1,1) = (one_c * yy) + c;
M(1,2) = (one_c * yz) - xs;
/* M(1,3) = 0.0F; */
M(2,0) = (one_c * zx) - ys;
M(2,1) = (one_c * yz) + xs;
M(2,2) = (one_c * zz) + c;
/* M(2,3) = 0.0F; */
/*
M(3,0) = 0.0F;
M(3,1) = 0.0F;
M(3,2) = 0.0F;
M(3,3) = 1.0F;
*/
}
#undef M
matrix_multf( mat, m, MAT_FLAG_ROTATION );
}

Creating a linear gradient in 2D array

I have a 2D bitmap-like array of let's say 500*500 values. I'm trying to create a linear gradient on the array, so the resulting bitmap would look something like this (in grayscale):
(source: showandtell-graphics.com)
The input would be the array to fill, two points (like the starting and ending point for the Gradient tool in Photoshop/GIMP) and the range of values which would be used.
My current best result is this:
alt text http://img222.imageshack.us/img222/1733/gradientfe3.png
...which is nowhere near what I would like to achieve. It looks more like a radial gradient.
What is the simplest way to create such a gradient? I'm going to implement it in C++, but I would like some general algorithm.
This is really a math question, so it might be debatable whether it really "belongs" on Stack Overflow, but anyway: you need to project the coordinates of each point in the image onto the axis of your gradient and use that coordinate to determine the color.
Mathematically, what I mean is:
Say your starting point is (x1, y1) and your ending point is (x2, y2)
Compute A = (x2 - x1) and B = (y2 - y1)
Calculate C1 = A * x1 + B * y1 for the starting point and C2 = A * x2 + B * y2 for the ending point (C2 should be larger than C1)
For each point in the image, calculate C = A * x + B * y
If C <= C1, use the starting color; if C >= C2, use the ending color; otherwise, use a weighted average:
(start_color * (C2 - C) + end_color * (C - C1))/(C2 - C1)
I did some quick tests to check that this basically worked.
In your example image, it looks like you have a radial gradient. Here's my impromtu math explanation for the steps you'll need. Sorry for the math, the other answers are better in terms of implementation.
Define a linear function (like y = x + 1) with the domain (i.e. x) being from the colour you want to start with to the colour your want to end with. You can think of this in terms of a range the within Ox0 to OxFFFFFF (for 24 bit colour). If you want to handle things like brightness, you'll have to do some tricks with the range (i.e. the y value).
Next you need to map a vector across the matrix you have, as this defines the direction that the colours will change in. Also, the colour values defined by your linear function will be assigned at each point along the vector. The start and end point of the vector also define the min and max of the domain in 1. You can think of the vector as one line of your gradient.
For each cell in the matrix, colours can be assigned a value from the vector where a perpendicular line from the cell intersects the vector. See the diagram below where c is the position of the cell and . is the the point of intersection. If you pretend that the colour at . is Red, then that's what you'll assign to the cell.
|
c
|
|
Vect:____.______________
|
|
I'll just post my solution.
int ColourAt( int x, int y )
{
float imageX = (float)x / (float)BUFFER_WIDTH;
float imageY = (float)y / (float)BUFFER_WIDTH;
float xS = xStart / (float)BUFFER_WIDTH;
float yS = yStart / (float)BUFFER_WIDTH;
float xE = xEnd / (float)BUFFER_WIDTH;
float yE = yEnd / (float)BUFFER_WIDTH;
float xD = xE - xS;
float yD = yE - yS;
float mod = 1.0f / ( xD * xD + yD * yD );
float gradPos = ( ( imageX - xS ) * xD + ( imageY - yS ) * yD ) * mod;
float mag = gradPos > 0 ? gradPos < 1.0f ? gradPos : 1.0f : 0.0f;
int colour = (int)( 255 * mag );
colour |= ( colour << 16 ) + ( colour << 8 );
return colour;
}
For speed ups, cache the derived "direction" values (hint: premultiply by the mag).
There are two parts to this problem.
Given two colors A and B and some percentage p, determine what color lies p 'percent of the way' from A to B.
Given a point on a plane, find the orthogonal projection of that point onto a given line.
The given line in part 2 is your gradient line. Given any point P, project it onto the gradient line. Let's say its projection is R. Then figure out how far R is from the starting point of your gradient segment, as a percentage of the length of the gradient segment. Use this percentage in your function from part 1 above. That's the color P should be.
Note that, contrary to what other people have said, you can't just view your colors as regular numbers in your function from part 1. That will almost certainly not do what you want. What you do depends on the color space you are using. If you want an RGB gradient, then you have to look at the red, green, and blue color components separately.
For example, if you want a color "halfway between" pure red and blue, then in hex notation you are dealing with
ff 00 00
and
00 00 ff
Probably the color you want is something like
80 00 80
which is a nice purple color. You have to average out each color component separately. If you try to just average the hex numbers 0xff0000 and 0x0000ff directly, you get 0x7F807F, which is a medium gray. I'm guessing this explains at least part of the problem with your picture above.
Alternatively if you are in the HSV color space, you may want to adjust the hue component only, and leave the others as they are.
void Image::fillGradient(const SColor& colorA, const SColor& colorB,
const Point2i& from, const Point2i& to)
{
Point2f dir = to - from;
if(to == from)
dir.x = width - 1; // horizontal gradient
dir *= 1.0f / dir.lengthQ2(); // 1.0 / (dir.x * dir.x + dir.y * dir.y)
float default_kx = float(-from.x) * dir.x;
float kx = default_kx;
float ky = float(-from.y) * dir.y;
uint8_t* cur_pixel = base; // array of rgba pixels
for(int32_t h = 0; h < height; h++)
{
for(int32_t w = 0; w < width; w++)
{
float k = std::clamp(kx + ky, 0.0f, 1.0f);
*(cur_pixel++) = colorA.r * (1.0 - k) + colorB.r * k;
*(cur_pixel++) = colorA.g * (1.0 - k) + colorB.g * k;
*(cur_pixel++) = colorA.b * (1.0 - k) + colorB.b * k;
*(cur_pixel++) = colorA.a * (1.0 - k) + colorB.a * k;
kx += dir.x;
}
kx = default_kx;
ky += dir.y;
}
}