cannot render perfectly 45 degree lines

[NemoOrange]

This is in reference to the problem I was having this thread:

http://forums.winamp.com/showthread....hreadid=184860

So I'm working on a scope that basically moves in 45 degrees, point by point. So, its at one spot, then moves a certain distance 45 degrees away from that spot.

But, due to what seems to be a rounding error in AVS, every couple of frames, the point will move off by one pixel, & then back to the correct per-pixel path.

& the real kicker is that when you set the width of the AVS window to 284 or 756 (height does not matter), this doesn't happen at all.

I'm quite perplexed & if anyone could help, it would be of great assistance.

[vanderphunck]

I suppose you do something like this:


[frame]
speed=1/w;


[beat]
rr=rand(8)*$pi/4;
vx=cos(rr);
vy=sin(rr);

x=0;y=0;

[point]
x=x+speed*vx;
y=y+speed*vy;


There's definitely some rounding errors in this... My first suggestion would be to round vx and vy like this:

vx=floor(cos(rr)+.5);
vy=floor(sin(rr)+.5);

That helped for me... But is ofcourse not exactly the same...

[vanderphunck]

Actually the above does not work... but this does:

[init]
n=1000;rr=1;
[frame]
vx=floor(cos(rr*$pi/4)+.5);
vy=floor(sin(rr*$pi/4)+.5);

[beat]
rr=rr+1;
xp=floor(w/2);yp=floor(h/2);

[point]
xp=xp+vx;
yp=yp+vy;

x=(xp+.5)*2/w-1;
y=(yp+.5)*2/h-1;



---------

here i'm calculating the position in pixels. The weird thing is that it does not work unless i add the .5 as in the last 2 lines. I hope this helps...

[Rovastar]

Yeah don't worry to much about the .5 thing I have seen similar things in MilkDrop and R4 visualizations among other and I can imagine AVS being no different.

This coudl well be due to a grid like system and positioning the 'centre' point of teh grid. you will have to shift the 'centre' along.

I explain more here

http://www.rabidhamster.org/phpBB2/v...&highlight=map

but it is in a bit of R4 pseudo-code speak though. and talking about mapping....umm I suppose DM's would be a similar thing. MAybe not the same thing but maybe gives some ideas about the theory behind the 0.5 bit.

[NemoOrange]

Thanks so much Phunck!

The pixel positioning solution is exactly what I was looking for. Works perfectly.