Various changes made to support Σκοπός over the past few years #1
Conversation
| => (ant.CanTarget && ant.ToTarget != Vector3.zero) ? PointingLoss(Vector3.Angle(origin - ant.Position, ant.ToTarget), ant.Beamwidth) : 0; | ||
| => (ant.CanTarget && !ant.IsTracking) ? PointingLoss(Vector3.Angle(origin - ant.Position, ant.ToTarget), ant.Beamwidth) : 0; |
There was a problem hiding this comment.
Just be careful here, as the pointingloss for a poorly defined angle (ie direction to target is somehow near 0,0,0) will produce bad results. Unsure how much this particular library code is exercised now, especially with the conversion to jobs.
There was a problem hiding this comment.
Well, this won’t affect cases where ToTarget is near zero, only those where it is zero.
It seems unlikely that it would be zero in cases where it should be, since it is computed as difference in double precision where one of the terms is Position, which is PrecisePosition, which is a position in double precision (the other term of the difference, Target.transform.position, is regrettably in single precision, but the result, while perhaps garbage, shouldn’t come out to 0).
(The difference is then converted to single precision but this matters not. All that switching between precisions is neither a recipe for performance nor correctness, but let’s not get distracted.)
There was a problem hiding this comment.
Could be irrelevant if calculating garbage data anyway [what is ToTarget with a somehow broken/invalid target?]. Other source of potentially bad result is maybe no longer possible for reason you described [PrecisePosition is double]. Main indicator might be vessels & targets that are extremely far away from current origin [ie, focus on a vessel past Pluto, what are the calculations for vessels/targets in similar orbit near Earth?]
| @@ -88,7 +90,7 @@ public struct CalcAntennaBodyNoise : IJobParallelForDefer | |||
| [WriteOnly] public NativeArray<float> bodyNoise; | |||
| public void Execute(int index) | |||
| { | |||
| bodyNoise[index] = rxHome[index] ? Precompute.NoiseFromOccluders(rxPos[index], rxGain[index], rxDir[index], rxFreq[index], rxBeamwidth[index], occluders) : rxPrecalcNoise[index]; | |||
| bodyNoise[index] = rxTracking[index] ? Precompute.NoiseFromOccluders(rxPos[index], rxGain[index], rxDir[index], rxFreq[index], rxBeamwidth[index], occluders) : rxPrecalcNoise[index]; | |||
| } | |||
There was a problem hiding this comment.
I wonder about this particular change. The purpose of this job was to defer the noise calculation for antennas whose pointing was not determined earlier, specifically: the ground station antennas that are allowed to "simultaneously" point at any possible peer (as a gameplay simplification).
But isTracking is also changing, from:
public virtual bool CanTarget => Shape != AntennaShape.Omni && (ParentNode == null || !ParentNode.isHome);
to:
public bool IsTracking => Shape != AntennaShape.Omni && Target == null;
I don't know if those are the same, ie do only ground stations have Target == null?
But this is the only purpose of isTracking -- to indicate this non-physical idea of how a given antenna is pointing.
There was a problem hiding this comment.
To your question
I don't know if those are the same, ie do only ground stations have Target == null?
I think the answer is yes, because this is how Target is initialized, where we see some of the logic that was in CanTarget:
if (config.HasNode("TARGET"))
Target = Targeting.AntennaTarget.LoadFromConfig(config.GetNode("TARGET"), this);
else if (Shape != AntennaShape.Omni && (ParentNode == null || !ParentNode.isHome) && !(Target?.Validate() == true) && HighLogic.LoadedSceneHasPlanetarium)
Target = Targeting.AntennaTarget.LoadFromConfig(SetDefaultTarget(), this);The only intended change in behaviour in the changes surrounding the introduction of IsTracking is that you can explicitly configure an antenna to have this kind of track-everything-at-once behaviour using TARGET {}; Σκοπός uses that for its ground stations.
Those ground stations are not home, so that you cannot downlink your science to a random Орбита station (or maybe even someone’s TVRO dish!) using the entire C band.
No description provided.