OpenMP and geometric transformation

[open_devel]

Hi all. I am working on geometric transformations using openmp. I am using GCC 4.4 with OpenMp support and Qt toolkit. My application reads 3ds files and loads object to memory part by part. Then draws them onto the windows using QPainter object. (It seems simplistic for now, but I want to see the fact before moving onto 3D API). The object parts are loaded into global float typed matrices. It does rotating for 3 axis. But my application does not divide task so I cannot accelarate rotating process. Could you recommend me any of openmp code for matrix multiply for 3d transformations? And which 3D API for use with openmp?

Sincerely.

[gaston-hillar]

Quoting - open_devel

Hi all. I am working on geometric transformations using openmp. I am using GCC 4.4 with OpenMp support and Qt toolkit. My application reads 3ds files and loads object to memory part by part. Then draws them onto the windows using QPainter object. (It seems simplistic for now, but I want to see the fact before moving onto 3D API). The object parts are loaded into global float typed matrices. It does rotating for 3 axis. But my application does not divide task so I cannot accelarate rotating process. Could you recommend me any of openmp code for matrix multiply for 3d transformations? And which 3D API for use with openmp?

Sincerely.

Hi open_devel,

Firstly, I'm not an expert on OpenMP. However, I've read your post and I was wondering. What about using Vectorization (SIMD, SSE2, SSE3 and so) to perform a matrix multiply. I do believe that you'll find sample codes about matrix multiplication in this software network. I think that you could also check Intel Math Kernel Library: http://software.intel.com/en-us/intel-mkl/- and its forums: http://software.intel.com/en-us/forums/intel-math-kernel-library/

Just my opinion as I always try to use vectorization when working with matrices and vectors. BTW, I love vectorization.

I'm sure there are experts that are going to be able to provide you more info with your specific question on OpenMP.