I am working on a sketch that is basically a very simple particle system, but the trails are drawn as lines. I wrote one version of the code in processing and one in openFrameworks. The code is nearly identical, but the output isn’t. While I expected some differences, the processing version does not render smooth lines, whenever the particles are moving quickly, there are some clearly visible stripes/artifacts/gaps in the lines. Also, the aliasing is quite noticeable (I enabled smooth(16)).
Is there anything I can do to improve the output of the processing sketch? I’d prefer to work in an environement without xcode and with shorter compile-time, but if the difference in quality is that noticeable, I can’t justify not choosing openFrameworks.
Welcome to the community. Is it possible to see the source code that you are using in Processing? It would also be nice to see screen shots of the two outputs so that we can compare them.
Sorry, I didn’t scroll down. I can see that there is a clear difference and I would be surprised if we can’t match it, but will defer to others with more experience for their comments.
Yes, that does improve the overall look, but sadly the lines still seem to be a series of spaced points rather than continuous lines. Is there anything I can to about that?
Well spotted @glv I have slightly modified your code to avoid creating new objects in the draw method. The copy is now made in setup and the pos[] contents is copied to prevPos[] in draw.
PVector[] pos;
PVector[] vel;
PVector[] prevPos;
float[] mass;
int number;
void setup() {
size(800, 600);
surface.setResizable(true);
background(0);
stroke(255, 255, 255, 5);
strokeWeight(1);
smooth(16);
number = 500;
pos = new PVector[number];
vel = new PVector[number];
prevPos = new PVector[number];
mass = new float[number];
for (int i = 0; i < number; i++) {
pos[i] = PVector.random2D().mult(5.0);
vel[i] = PVector.random2D();
prevPos[i] = pos[i].copy();
mass[i] = random(0.3);
}
}
void draw() {
PVector mouse = new PVector(mouseX, mouseY);
for (int i = 0; i < number; i++) {
PVector acc = PVector.sub(mouse, pos[i]);
acc.mult(0.01);
acc.mult(mass[i]);
vel[i].add(acc);
pos[i].add(vel[i]);
line(prevPos[i].x, prevPos[i].y, pos[i].x, pos[i].y);
vel[i].mult(0.97);
prevPos[i].set(pos[i]);
}
}
Thank you, now it looks almost as good as the OF version. I checked the references you linked, but I still don’t quite unterstand, why the original code is slower.
Does
prevPos[i] = pos[i]
point prevPos to the vector in pos instead of copying it and that makes it slower?
Both prevPos[i] and pos[i] are references to a PVector object. When you execute the code above you copy the reference not the object. It means that both prevPos[i] and pos[i] refer to the same PVector object.
The statement prevPos[i] = pos[i].copy();
creates a new PVector object with its own reference which is stored in prevPos[i]. So now both prevPos[i] and pos[i] refer to different objects.
