CUDAdillo

A static c++ library to complete basic matrix operations on Armadillo matrices with the intent to speed up runtime through the use of the CUDA and cuBLAS GPU libraries.

#Qt Creator Integration

• Have the cudadillo.h file in the same working directory
• Have the compiled libCUDAdillo.a in the project, preferably a Libraries folder
• Then include the following in the project's .pro file

macx{
    LIBS += -L/Developer/NVIDIA/CUDA-7.5/lib/ -lcudart -lcublas
    LIBS += $$PWD/../../Libraries/libCUDAdillo.a
}

• And in the files using CUDAdillo,

#include "armadillo.h"
#include "cudadillo.h"

#Examples

mat matA = randu<mat>(5,5);
mat matB = randu<mat>(5,5);

CUDAdillo::init();

mat * sum = CUDAdillo::addMat<double>(&matA, &matB); //equivalent to matA + matB
mat * mult = CUDAdillo::multMat<double>(&matA, &matB); //equivalent to matA * matB
mat * trans = CUDAdillo::transposeMat<double>(&matA); // equivalent to matA.t()
mat * cov = CUDAdillo::covMat<double>(&matA,&matB); //equivalent to matA * matB.t()
//the above work for fmats and floats in a similar fashion

CUDAdillo::destroy();
delete sum;
delete mult;
delete trans;
delete cov;

Note that these functions allocate heap memory to create these new matrices, so they must be freed by the user, and the functions assume that the inputs share the same dimensions.

#References cuBLAS Documentation

Specific Functions to note are:

• cublasXtDgemm()
• cublasDgeam()

The Test subdirectory makes use of google-benchmark to benchmark running time of these functions and the default CPU operations.

#Future Work

• Transpose currently fails on non-square inputs, and covMat fails to get any correct output.

• Matrix multiply and addition is still slower than CPU version alone. Is memory bandwidth an issue?

• Some links worth looking at: http://www.cs.cmu.edu/afs/cs/academic/class/15869-f11/www/readings/fatahalian04_gpumatmat.pdf

http://stackoverflow.com/questions/2952277/when-is-a-program-limited-by-the-memory-bandwidth

http://pertsserver.cs.uiuc.edu/~mcaccamo/papers/private/IEEE_TC_journal_submitted_C.pdf

• Implement a function to multiply a matrix by its transpose
• look at other benchmarking tools to gather more information on memory bandwidth

• STREAM
• NVIDIA Visual Profiler

#Copyright Notice CUDAdillo is free software: you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

CUDAdillo is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.

You should have received a copy of the GNU Lesser General Public License along with CUDAdillo. If not, see http://www.gnu.org/licenses/.