We propose a new updating scheme to perform constrained hybrid Monte Carlo on lattices with different scales. Starting from a coarse lattice, the scheme preserves the long distance physics and the evolution fills in the short distance physics on the fine lattice. Methods of tuning lattice parameters to follow the renormalization group transformation are explored. With this scheme we expect to redu ... More

I will present a method for calculating eigenvectors of the staggered Dirac operator based on the Golub-Kahan-Lanczos bidiagonalization algorithm. Instead of using orthogonalization during the bidiagonalization procedure to increase stability, we choose to stabilize the method by combining it with an outer iteration that refines the approximate eigenvectors obtained from the inner bidiagona ... More

Owing to its success in removing the critical slowing down of Dirac linear systems, adaptive multigrid is now a standard solver in the arsenal of tools that the lattice field theorist expects. In this work we report on the latest progress in improving the strong scaling of adaptive multigrid algorithms when running on GPU-accelerated architectures using the QUDA library. Techniques include Schwa ... More

Several parallel machines on which Lattice LQCD applications are being run utilize a new fabric, Intel Omni-Path. We present an overview of Omni-Path, comparing it to the well-known competitor InfiniBand. In the process of adding support for Omni-Path to our communication library pMR we discovered several insights which we discuss along some general usage recommendations. We substantiate our findi ... More

We report recent efforts by CLS to generate an ensemble with physical light- and strange-quark masses in a lattice volume of $96^3x192$ at $\beta=3.55$ corresponding to a lattice spacing of 0.064fm. This ensemble is being generated as part of the CLS 2+1 flavor effort with improved Wilson fermions. Our simulations currently cover 5 lattice spacings ranging from 0.039fm to 0.086fm at various pion m ... More

The RBC and UKQCD Collaborations are generating coarse lattices with volumes of (4.8 fm)$^3$, (6.4 fm)$^3$ and (9.6 fm)$^3$. These lattices have physical values for the light and strange quark masses and $1/a = 1.0$ GeV. An important feature of these lattices is that measured observables show very small $O(a^2)$ corrections. Physical properties of these lattices will be presented, along with de ... More

In recent years adaptive smoothed aggregation algebraic multigrid ($\alpha$SA-AMG) methods have been developed and subsequently adapted for use in lattice quantum chromodynamics (QCD). The purpose of these efforts has been to reduce the critical slowdown that occurs in lattice QCD algorithms when working on state-of-the-art problems. Convergence theorems can establish the robustness of such meth ... More

RBC/UKQCD has recently finished implementing the exact one flavor algorithm (EOFA), which allows for HMC simulations of single quark flavors without taking a square root of the fermion determinant. In this talk we elaborate on the details of our implementation, including a novel preconditioning scheme for the exact one flavor Dirac operator that has been shown to significantly accelerate the algor ... More

The LLR method is a novel algorithm that allows the evaluation of the density of states in lattice gauge theory. In this talk I will present our study of the ergodicity properties of the LLR algorithm for the model of Yang Mills SU(3). I will focus on the use of the replica exchange method as tool to alleviate the topological freeze-out of the algorithm.

Using the t-V model, we show how the fermion bag idea can be applied to develop algorithms to Hamiltonian lattice field theories. We argue that fermion world lines suggest an alternative method to the traditional SVD techniques for calculating ratios of determinants in a stable manner. We show the power behind these ideas by extracting the physics of the t-V model on large lattices.

We perform Monte Carlo simulations of the CP$^{N-1}$ model on the square lattice for $N=10$, $21$, and $41$. Our focus is on the severe slowing down related to instantons. To fight this problem we employ open boundary conditions as proposed by Lüscher and Schaefer for lattice QCD. Furthermore we test the efficiency of parallel tempering of a line defect. Our results for open boundary condit ... More

Intel Xeon Phi processors deliver a huge amount of computational power. When used efficiently, they improve application performance and shorten the time-to-solution. Our talk will describe the processor architecture and discuss on how to efficiently use the hardware resources. We will cover different mechanisms to address this efficient usage. Among the mechanisms, we will introduce QPhiX. QPhiX i ... More

While Lattice QCD presents itself as a problem with an obvious parallelization over lattice sites, it still needs to face the trend in HPC to use wider processors. In addition, maximizing locality becomes increasingly important as wider processors have led to a scenario where available FLOPS has grown faster than the available memory bandwidth. Block Krylov space solvers, which combine solves for ... More

I discuss the status and performance of Grid software with emphasis on Fermion solver performance on both single and multiple nodes of common CPU technologies, with Cray Aries, Intel Omnipath and Mellanox Infiniband interconnects. Support includes Domain Wall, Wilson and Staggered Fermions. Prospects for GPU support are discussed.

With recent developments in parallel supercomputing architecture, many core, multi-core, and GPU processors are now commonplace resulting in more levels of parallelism, memory hierarchy, and programming complexity. It has been necessary to adapt the MILC code to these new processors starting with NVIDIA GPUs and more recently the Intel Xeon Phi processors. We report on our efforts to port and opti ... More

Simulations at physical quark masses are affected by the critical slowing down of the solvers.Multigrid preconditioning has proved to effectively deal with this problem. The ETM collaboration is performing multigrid accelerated simulations at the physical point to generate Nf = 2 and Nf = 2 + 1 + 1 gauge ensembles. The adaptive aggregation-based domain decomposition multigrid solver, referred to a ... More

Experiences with optimizing the matrix-times-vector application of the Brillouin operator on the Intel KNL processor are reported. Without any adjustments to the memory layout, performances figures of 300 Gflop/s in sp and 230 Gflop/s in dp are observed. This is with Nc=3 colors, Nv=12 right-hand-sides, Nthr=256 threads, on lattices of size 32^3*64, using exclusively OMP pragmas. Interestingly, th ... More

We present the tuning strategy for the generation of an ensemble of $N_f=2+1+1$ twisted mass fermions with a clover term at maximal twist that ensures automatic ${\cal O}(a)$ -improvement. The target lattice is of size $64^3\times 128$ with a lattice spacing of $a\sim 0.08$~fm. We show preliminary results on the pseudoscalar masses and decay constants.

We present the preliminary tests on two modifications of the Hybrid Monte Carlo (HMC) algorithm. Both algorithms are designed to travel much farther in the Hamiltonian phase space for each trajectory and reduce the autocorrelations among physical observables thus tackling the critical slowing down towards the continuum limit. We present a comparison of costs of the new algorithms with the stand ... More

A significant obstacle facing the approach to the continuum limit in lattice gauge theory calculations is the phenomena of critical slowing down in the fermion sector. The method of adaptive multigrid ($\alpha$-MG) methods offer a permanent solution to the superlinear growth in the cost of iterative Dirac matrix inversions. An exascale-ready implementation of $\alpha$-MG as a preconditioner is sui ... More