I was wondering if anyone had looked into the performance of gl arrays with different stride values.
Right now I am doing lerping and the like and generating an array of structs which hold the texture coords, normals and verts and simply using glInterleavedArray(). Pretty simple except that I need to index back into this array. Indexing sucks it cause the cpu to thrash the cache from all the random accessing that takes place.
Was considering storing that lerped data back into the triangle structure (to save the random nature of indexing) which is fine and I just set the stride myself. The problem with this is that i wanted to use VAR for lerped array. I can’t use it for that because I need to index back into it (which is slow) and I can’t put the verts into my triangle structure for almost the same reason. I would need to access other items in that structure which is again reading from VAR. THe only thing I can see to do is to make a copy of the data (in an intelligent manner of course).
Ahhhh. The hell with it I am just gonna do store the vert in 2 places, both in the Triangles and in the Lerped/Packed Array.
This is all because of damn stencil shadow volumes. MultiPass SUCKS !!!