Expensive fix for Cannon aiming?

[ILMTitan]

The aiming direction of the Cannon weapon type is wrong when there is a significant y difference between the weapon and the target. I have a solution, changing this code :

Code: Select all

if(highTrajectory){
			if(predict>maxPredict)
				predict=maxPredict;
			if(predict<minPredict)
				predict=minPredict;

			float3 dif=targetPos-weaponPos;
			float k1=dif.SqLength2D();
			float h=projectileSpeed*predict;
			h=h*h;

			float k2=h-k1;
			if(k2>0){
				k2=sqrt(k2);
				k2+=weaponPos.y-targetPos.y;
				k2/=-gs->gravity*0.5f;
				if(k2>0)							//why is this needed really (i no longer understand my equations :) )
					predict=sqrt(k2);
			}
			wantedDir=targetPos-float3(0,5,0)-weaponPos;
			wantedDir.y-=predict*predict*gs->gravity*0.5f;
			wantedDir.Normalize();
		} else {
			float3 dif=targetPos-weaponPos;
			dif.y-=predict*predict*gs->gravity*0.5f;
			if (projectileSpeed == 0)
				predict = maxPredict;
			else {
				predict=dif.Length()/projectileSpeed;
				if(predict>maxPredict)
					predict=maxPredict;
			}
			wantedDir=dif;
			wantedDir.Normalize();
		}

To this:

Code: Select all

	float Dsq = diff.SqLength();
	float DFsq = diff.SqLength2D();
	float g = gs->gravity;
	float v = projectileSpeed;
	float dy = diff.y;
	float dxz = sqrt(DFsq);
	float Vxz;
	float Vy;
	if(Dsq == 0) {
		Vxz = 0;
		Vy = highTrajectory ? v : -v;
	} else {
		float root1 = v*v*v*v + 2*v*v*g*dy-g*g*DFsq;
		if(root1 >= 0) {
			float root2 = 2*DFsq*Dsq*(v*v + g*dy + (highTrajectory ? -1 : 1)
				* sqrt(root1));
			if(root2 >= 0) {
				Vxz = sqrt(root2)/(2*Dsq);
				Vy = (dxz == 0 || Vxz == 0) ? v : (Vxz*dy/dxz - dxz*g/(2*Vxz));
			} else {
				Vxz = 0;
				Vy = v;
			}
		} else {
			Vxz = 0;
			Vy = v;
		}
	}
	float3 dir(diff.x, 0, diff.z);
	dir.Normalize();
	dir *= Vxz;
	dir.y = Vy;
	dir.Normalize();
        wantedDir = dir;

(there are also a few other non-pertinent changes.)

Because this code will run every emg shot, I'm worried that the two sqrt functions will hurt performance. Maybe I'm just being silly here. Anyone want to alleviate/confirm my fears?

[rattle]

Doesn't the EMG weapon have it's own type?

[ILMTitan]

.... yes. Ok, I'm going to commit this.

[Zpock]

Wow, this fixes aiming for turrets on aircraft?

[KDR_11k]

What's the difference between EMG and Cannon?

[ILMTitan]

EMG projectiles are not affected by gravity, Cannon projectiles are. In a different sense, EMGs are weapons with lineofsight = 1, renderType = 4 and color = 2, while cannons are the default weapon type (any weapon that does not otherwise describe a valid weapon).

This does not fix anything specifically related to aircraft, but if an aircraft had a cannon weapon as a turret, it was probably not aiming very well because the aircraft was above the target, and this will fix that.

[KDR_11k]

I think arced (TrajectoryHeight=1) missiles have a similar problem (they ignore the y distance), do you feel like looking into that?

[Decimator]

Arced missiles definitely have a problem with height differences. I think it may have been fixed in svn, so I would suggest trying a recent beta build.

[KDR_11k]

Nope, it's not fixed in the SVN.

[jcnossen]

Because this code will run every emg shot, I'm worried that the two sqrt functions will hurt performance. Maybe I'm just being silly here. Anyone want to alleviate/confirm my fears?

You could do sqrtf() for lower precision, but other than that i doubt its going to be such a problem. Vectors are being normalized all the time as well, much more often than the cannon aiming probably.

[ILMTitan]

The Gundam Federation Type 61 tank, which I'm pretty sure has an arched missile weapon, has no problem hitting ground far above and far below itself in my version (Latest svn).

[Snipawolf]

testing? With gundam!? Oh how happy Smoth will be...

[ILMTitan]

I did all of my limited arc/unit rotation testing on Gundam. And it had the only arced missile unit I could think of, other than the AA arm rocket artillery bot that does not have enough accuracy to test accuracy on.

[smoth]

ILM, while you are look at that, would you mind looking at the "accuracy" tag in the weapon TDF?

The shots from indirect(trajectory height weapons) do not drift like other weapons can. So they always hit the same spot no matter what accuracy I give them.

[ILMTitan]

Sounds like a new project. Will do. Is that just trajectory height missiles? Cannon weapons do not have a trajectory height, and I know the shellshocker (AA L1 artillery) has some inaccuracy.

[smoth]

Just the trajectory height weapons. To see what I mean.. give the zaku2 and accuracy of 2000 and watch how it's shots go all over the place and the guntank with the same accuracies all land in the same spot.

[KDR_11k]

I've patched that, remember? The guntank's shots go all over the place in the SVN. The missile launcher weapon was just missing the code that adds the inaccuracy to the weapon vector.

The T61 aims properly but the Guntank doesn't.

[smoth]

OH WOW! no, kdr, I had forgotten. Sorry about that.