boost::multi-array as a tensor library

Basically, I am a scientist looking at boost::multi-array as a tensor container. If it can be used to create code on many architectures (GPU/Intel MIC etc.) than it could be useful to me.

Positive Features:

• It's an N-dimensional container with slicing, necessary.
• I can call other libraries (BLAS, cuBLAS, etc.)
• OpenMP/MPI will be trivial.

Next Feature Tests:

• Slicing on the GPU?
• MPI/OpenMP examples
• Speed vs. standard C arrays

Building it

Need:

• cmake
• boost::multi-array

Optional, but basically the whole point of this repository:

• OpenMP
• CBLAS
• CUDA
• OpenACC: I haven't tested this yet

git clone https://github.com/barrymoo/multi-array-testing.git
cd multi-array-testing
mkdir build
cd build
cmake ..
make