I am over in Ireland at the moment visiting my father but just before I went away I managed to get a couple of things done.
The first thing I attempted was the use of full matrices for the Instancing which has not been going so well. I worked on Implementing this and increased what data needed to be passed across and the size of the input layout and what not but came out with some rather peculiar results. The models where stretched across the screen and some just did not appear at all. At this point I ran the project using PIX to check the data that had been passed across and this is what it came out with:
The first thing I attempted was the use of full matrices for the Instancing which has not been going so well. I worked on Implementing this and increased what data needed to be passed across and the size of the input layout and what not but came out with some rather peculiar results. The models where stretched across the screen and some just did not appear at all. At this point I ran the project using PIX to check the data that had been passed across and this is what it came out with:
As you can see the outcome after the vertex shader is completely wrong. After checking the values in the instance buffer I can see that only the first values are correct, which are the positions. The X,Y and Z axis are completely random values which is the reason that the results are so strange. I will need to have a further in depth look at the reasons behind all this but I'm presuming the byte width of the instance data is being calculated wrong at some point, however this is only a guess.
Terrain Analysis Instancing
I took a break from trying to implement the matrices geometry instancing and decided to add the current position only instancing to the Terrain Analysis Tool of my third year project. This tool only needed the positions as it only had sphere models which do not need rotation and the scale is the same for all. The integration of this took no time at all as I had implemented the instancing in a way that would simply mean adding a "Instancing=true" value to the templates of the XML document and replacing the "RenderAllEntities" function with the "RenderAllEntitiesInstanced" function. Once I had replaced all the files that were needed I created a quick video to show the terrain analysis with the Instancing, which is shown below.
The frame rate on the new project has increased massively. On this map the original frame rate was around 30 FPS when all nodes were in view and with the geometry instancing I am now reaching up to 250 FPS although you cannot really see this as Fraps slows down the program quite considerably.
There are lots of improvements still to be made to the project such as adding more features to the tool allowing for nodes and game zones to be specified as water nodes. More than likely I might add a feature to specify my own types of data that can be assigned on the fly. This would allow so much flexibility to the tool and allow it be used in many more situations and genres. Well that's all for now, I will be continuing work on my projects such as the Road to Eldorado game and implementing the full matrices instancing, although I will also be looking for jobs in the mean time as well.
Peace.
There are lots of improvements still to be made to the project such as adding more features to the tool allowing for nodes and game zones to be specified as water nodes. More than likely I might add a feature to specify my own types of data that can be assigned on the fly. This would allow so much flexibility to the tool and allow it be used in many more situations and genres. Well that's all for now, I will be continuing work on my projects such as the Road to Eldorado game and implementing the full matrices instancing, although I will also be looking for jobs in the mean time as well.
Peace.
RSS Feed
