[Solved] Storing Vectormath SSE Matrix4 in a float[16] ?

[dfalcone]

So since I want to use Bullet Physics for my project, I thought it would be nice to also use the included Vectormath library. The only problem I'm having with the library is that there seems to be a lack of an efficient way to get the float data out of a Matrix4 to pass to OpenGL/DirectX for transformation. This works fine with the Scalar library, but for the SSE library it stoes the values in a _m128, so you can't just pass the memory address of the matrix.

Other libraries have a simple method to get this data, for example M$oft's Math library uses the 'StoreFloat4x4' functions for passing the float values. Based on that implementation I've started doing this:

Code: Select all

static float rotY = 0.0f;
rotY += .0005f;
Vectormath::Aos::Matrix4 modelMatrix = Vectormath::Aos::Matrix4::rotationY(rotY); //Rotating Model Matrix for example

float tmpMtx[16];

_mm_store_ps( &tmpMtx[0] , modelMatrix .getCol(0).get128() );
_mm_store_ps( &tmpMtx[4] , modelMatrix .getCol(1).get128() );
_mm_store_ps( &tmpMtx[8] , modelMatrix .getCol(2).get128() );
_mm_store_ps( &tmpMtx[12] , modelMatrix .getCol(3).get128() );

unsigned int locModelMat = glGetUniformLocation(m_programID, "modelMat"); //Model Matrix Location
glUniformMatrix4fv(locModelMat, 1, GL_FALSE, &tmpMtx[0]);                 //Copy Float[16] model matrix to OpenGL Shader

Is there a better way to do this? If not, would it be possible to get functions that do this integrated into the library?

[dfalcone]

After a lot of reading and re-reading, I finally figured this out.

All SSE data must be 16 byte aligned for the memory address. That means the pointer to the data must have a memory address dividable by 16. Since there is no code I can find on the net anywhere for using the Vectormath library with directX or OpenGL, I'll post both samples. And you can pass the SIMD Matrix4 (which is a struct of 4 _m128 Vectors) directly to the GPU to read in as multiple floats.

OpenGL:

Code: Select all

unsigned int locProjMat = glGetUniformLocation(m_programID, "projMat");
ATTRIBUTE_ALIGNED16(Vectormath::Aos::Matrix4 proj) = Vectormath::Aos::Matrix4::perspective(45.0f, 4.0f/3.0f, 0.1f, 100.0f);
glUniformMatrix4fv(locProjMat, 1, GL_FALSE, (GLfloat*)&proj);

DirectX 11:

Code: Select all

m_d3dDevice->CreateBuffer( &constDesc, 0, &m_cbChangeOnResize );
ATTRIBUTE_ALIGNED16(Vectormath::Aos::Matrix4 proj) = Vectormath::Aos::Matrix4::perspective(45.0f, 4.0f/3.0f, 0.1f, 100.0f);
m_d3dDeviceContext->UpdateSubresource( m_cbChangeOnResize, 0, 0, &proj, 0, 0 );