Optimizing Wilson-Dirac Operator and Linear Solvers for Intel® KNL

Lattice Quantumchromodynamics (QCD) is a powerful tool to numerically access the low energy regime of QCD in a straightforward way with quantifyable uncertainties. In this approach, QCD is discretized on a four dimensional, Euclidean space-time grid with millions of degrees of freedom. In modern lattice calculations, most of the work is still spent in solving large, sparse linear systems. This part has two challenges, i.e. optimizing the sparse matrix application as well as BLAS-like kernels used in the linear solver. We are going to present performance optimizations of the Dirac operator (dslash) with and without clover term for recent Intel (R) architectures, i.e. Haswell and Knights Landing (KNL). We were able to achieve a good fraction of peak performance for the Wilson-Dslash kernel, and Conjugate Gradients and Stabilized BiConjugate Gradients solvers. We will also present a series of experiments we performed on KNL, i.e. running MCDRAM in different modes, enabling or disabling hardware prefetching as well as using different SoA lengths. Furthermore, we will present a weak scaling study up to 16 KNL nodes.