wglAllocateMemoryNV problem

OpenGL OpenGL: Advanced Coding
1

previously:
Vertex= new float [9];

now:
Vertex=(float )wglAllocateMemoryNV(9sizeof(float),.0f,.0f,1.0f);

Everything is OK, it renders but it’s slower!!! Why?

2

Simply getting AGP or video memory, without setting up VAR, will make the CPU read back vertex data from uncached memory, which is slow.

  • Matt
3

But been able to get the proc address of wglAllocateMemoryNV doesn’t mean that you have the VAR extension… or does it? If it does how do I use it? My code is like this:

Vertex1=wglAllocateMemoryNV(…);
Vertex2=wglAllocateMemoryNV(…);
while(1)
{
glVertexPointer(…,Vertex1);
glDrawElements(…);

glVertexPointer(…,Vertex2);
glDrawElements(…);
}

Where do I put to calls to VAR?

4

You should allocate only one block of memory, not two.

Add a call to VertexArrayRangeNV at the start of the program.

The existence of NV_vertex_array_range implies that wglAllocateMemoryNV must be available. wglAllocateMemoryNV by itself is utterly useless.

  • Matt
5

OK thanks!!!

6

You may have already done this, but…

Please check out the VAR/fence whitepaper and demo on the NVIDIA Developer Site (www.nvidia.com/Developer). There are a number of “gotchas” related to data formats and alignment that can seriously impair your VAR performance, and the paper covers all these and more…