Scawen,
Have some InSim node issues. All these tests were performed with an MPR replay using single AI driver, on Blackwood, with MCI turned on and running at a 50ms interval.
First issue is that the node indexs for the finish line and splits that are given in the IS_RST packet don't seem to match up with emperical evidence. For Blackwood, the node indexes are as follows:
NodeFinishIndex: 431
NodeSplit1Index: 113
NodeSplit2Index: 288
NodeSplit3Index: 65535 (of course means there isn't a 3rd split)
When running MCI, it produces the following output: http://www.thzero.com/racing/nodesx1.log
The first lap, we can see that the node that was received immediately afterward the Split 1 packet (denoted by Lap 1, Split 1 <<<<<<<<<<<<<<<) is index 110, instead of 113. Second lap it was index 111. For the second split it was indexes 287 and 285 instead of index 288. And for the finish line it was 429 and 428 instead of 431. So we're off a couple of nodes.
So thats a minor issue. Although at the rate the AI driver was running laps (79seconds) and the 50ms interval, I would have expected about 3 MCI packets per node, but I was only getting a smattering of MCI packets that had the same node. I am assuming, as I haven't had chance to examine the path information to determine distances between nodes, that not all nodes are equidistant.
The second issue is what happens when you increase the replay speed. The following log outputs show the nodes at x8 and x16 speeds, respectively.
http://www.thzero.com/racing/nodesx8.log
http://www.thzero.com/racing/nodesx16.log
We see a dramatic drop-off in the number of MCI packets received due to the engine running at x8 and x16 speeds. The AI guy is running at 79 seconds a lap which consists of 476, thats 6.03 nodes per second. So any interval under 166ms, we should get all the nodes via the MCI packets. Once we hit the x4 replay speed, we are would need to below 41ms intervals in order to attempt to capture every node packet.
So in order to capture position and speed statistics, reliably, with replays you either need to run the replay at most x2 speeds unless we are able to get the race position and speed included in the IS_LAP and IS_SPX, as mentioned in the Urgent InSim requests thread.
Why the need for pulling stats off replays as opposed to "live"? Well, one could say its a backup mechanism in case the "live" stat collection fails. It also may be perferable because of the server hosting situation. Or something else. But in a situation where you are collecting stats offline, and you are dealing with needing to wade through several hours of racing, on multiple servers, running the replay at the x8/x16 range is quite beneficial.
Thanks.
Have some InSim node issues. All these tests were performed with an MPR replay using single AI driver, on Blackwood, with MCI turned on and running at a 50ms interval.
First issue is that the node indexs for the finish line and splits that are given in the IS_RST packet don't seem to match up with emperical evidence. For Blackwood, the node indexes are as follows:
NodeFinishIndex: 431
NodeSplit1Index: 113
NodeSplit2Index: 288
NodeSplit3Index: 65535 (of course means there isn't a 3rd split)
When running MCI, it produces the following output: http://www.thzero.com/racing/nodesx1.log
The first lap, we can see that the node that was received immediately afterward the Split 1 packet (denoted by Lap 1, Split 1 <<<<<<<<<<<<<<<) is index 110, instead of 113. Second lap it was index 111. For the second split it was indexes 287 and 285 instead of index 288. And for the finish line it was 429 and 428 instead of 431. So we're off a couple of nodes.
So thats a minor issue. Although at the rate the AI driver was running laps (79seconds) and the 50ms interval, I would have expected about 3 MCI packets per node, but I was only getting a smattering of MCI packets that had the same node. I am assuming, as I haven't had chance to examine the path information to determine distances between nodes, that not all nodes are equidistant.
The second issue is what happens when you increase the replay speed. The following log outputs show the nodes at x8 and x16 speeds, respectively.
http://www.thzero.com/racing/nodesx8.log
http://www.thzero.com/racing/nodesx16.log
We see a dramatic drop-off in the number of MCI packets received due to the engine running at x8 and x16 speeds. The AI guy is running at 79 seconds a lap which consists of 476, thats 6.03 nodes per second. So any interval under 166ms, we should get all the nodes via the MCI packets. Once we hit the x4 replay speed, we are would need to below 41ms intervals in order to attempt to capture every node packet.
So in order to capture position and speed statistics, reliably, with replays you either need to run the replay at most x2 speeds unless we are able to get the race position and speed included in the IS_LAP and IS_SPX, as mentioned in the Urgent InSim requests thread.
Why the need for pulling stats off replays as opposed to "live"? Well, one could say its a backup mechanism in case the "live" stat collection fails. It also may be perferable because of the server hosting situation. Or something else. But in a situation where you are collecting stats offline, and you are dealing with needing to wade through several hours of racing, on multiple servers, running the replay at the x8/x16 range is quite beneficial.
Thanks.