Peter Wittek

Conjugation and Hermitian variables

10

I cannot possibly estimate the difficulty of this, but it would be great to define a conjugation or adjoint operation over the variables and monomials. The first interesting case is...

Reimplement GPU kernels in CUDA to replace Thrust

2

The Thrust kernels are somewhat inefficient for the multidimensional data structures. Testing is in the [cuda_opt](https://github.com/peterwittek/somoclu/tree/cuda_opt) branch.

Clang OpenMP and CUDA

2

Support for OpenMP [improved significantly](https://openmp.llvm.org/) in Clang. Furthermore, now it [can compile CUDA](https://llvm.org/docs/CompileCudaWithLLVM.html) code. This needs a revision of the `configure`, `Makefile`, and `setup.py` scripts, and it will be especially...

In principle, it is straightforward to pass Python sparse matrices, for instance, `scipy.sparse.lil_matrix`, to C/C++. The problem is that Somoclu's internal sparse representation does not match any of them. First...

R interface with CUDA support on Windows cannot be properly built. The CUDA code that requires VC to compile, whereas Rcpp requires GCC to compile, and linking cannot be done...

Only 1D works at the moment, and without the Fourier transformation to fully exploit block diagonal structures. See [arXiv:1307.1002](https://arxiv.org/abs/1307.1002).

Finish example on quantum steering

1

The [hierachy for quantum steering](http://dx.doi.org/10.1103/physrevlett.115.210401) is supported, but a working example exampel is missing. The core idea is simple, it follows the NPA hierarchy: ``` python from ncpol2sdpa import SteeringHierarchy,...