In the buisiness forum : An Open Source pro math lib . Please discuss more technical stuff here. Ask for precisions if you want. [edited by - Charles B on March 5, 2004 12:27:19 AM]

No sorry it is not public now because there are still some questions concerning what we will be set as Open Source and what we may keep for us.

For instance we could make the lowest level (the virtual SIMD layer and a bit more) Open Source and we could keep some geometric, col. det. routines for ourself at least for a while.

But we will start a web site very soon. However we need to have a first feedback from the community.

To give you an idea, some code sample that reflects the current state of the technology :

// {x,y,z,w} vectors and AoS (array of structures) style :
// The same code for ANSI, 586, 3DNOW, SSE, Altivec, etc...
// High perfs and less efforts.

vDecl(A,B,C); // Declare these vectors as registers
vLoad(A, *pInp1);
vLoad(B, *pInp2);
vCross(C,A,B); // Higher level than SSE here.
...
vStore(*pOut, C);


// SoA (structures of arrays) and C++ style :
// 1(ANSI), 2(3DNOW), or 4(SSE, Altivec) channels
mfloat Nx,Ny,Nz,Lx,Ly,Lz, Diffuse;
Nx=...; Ny=...; Nz=...;
Diffuse = mfPow(Nx*Lx+Ny*Ly+Nz*Lz, 32); // Lighting eq.

// Note this class is similar to F32Vec
// of the Intel compiler.

With this last example, on SSE or Altivec you'll get an output (asm) code of perfect quality , while the same source code will still work on non SSE capable hardware.


[edited by - Charles B on March 8, 2004 5:12:21 PM]