An Open Source, GPU Accelerated, Quantum Computer Simulator
- Uses the parallel programing library ArrayFire, which supports CUDA and OpenCL kernels.
- Runs Cross Platform (x86, ARM, CUDA, and OpenCL devices)
This is currently a work in progress, not suitable for use.
- A C++11 compiler, like gcc or clang
- CMake 3.0.0 or newer
- ArrayFire 3.0.1 or newer via. pre-built binaries
To build this project simply:
mkdir build
cd build
cmake ..
make
Download the project and files into your working directory. For simplicity,
lets assume the project is at C:\workspace\qcgpu.
Open the CMake GUI.
In the source directory field, enter the working directory
C:\workspace\qcgpu. In the build directory field, enter the build
directory C:\workspace\qcgpu\build.
Click Configure.
If there are no errors, click Generate in the CMake GUI.
Now, the solution files will be created in the build directory. There will be different projects created for CPU, CUDA and OpenCL. Setting one of these projects as start project and then running it will run the example with the specified backend.
The minimum filling time of a matrix with (n_1 * n_2) X (m_1 * m_2) is O(n_1 * n_2 * m_1 * m_2). This is relevent to the kronecker product. Some other considerations for the future development of this software, and the speedup of this function is:
- Parallelize
- Store on the fly (i.e. store M as kron of A and B, and calculate only when single element is needed)
- Could Use ArrayFire Tile Function
Can also max at 30 qubits, or the number storage will have to be changed
grammar.pegjs needs to be formatted and parts rewritten, conditionals not implemented yet.
This software is licensed under the MIT licence (see LICENSE.md)
Copyright (c) 2018 Adam Kelly
@misc{Yalamanchili2015,
abstract = {ArrayFire is a high performance software library for parallel computing with an easy-to-use API. Its array based function set makes parallel programming simple. ArrayFire's multiple backends (CUDA, OpenCL and native CPU) make it platform independent and highly portable. A few lines of code in ArrayFire can replace dozens of lines of parallel computing code, saving you valuable time and lowering development costs.},
address = {Atlanta},
author = {Yalamanchili, Pavan and Arshad, Umar and Mohammed, Zakiuddin and Garigipati, Pradeep and Entschev, Peter and Kloppenborg, Brian and Malcolm, James and Melonakos, John},
publisher = {AccelerEyes},
title = {{ArrayFire - A high performance software library for parallel computing with an easy-to-use API}},
url = {https://github.com/arrayfire/arrayfire},
year = {2015}
}