No abstract

The introduction of accelerator devices such as graphics processing units (GPUs) has had profound impact on molecular dynamics simulations and has enabled order-of-magnitude performance advances using commodity hardware. To fully reap these benefits, it has been necessary to reformulate some of the most fundamental algorithms, including the Verlet list, pair searching, and cutoffs. Here, we present the heterogeneous parallelization and acceleration design of molecular dynamics implemented in the GROMACS codebase over the last decade. The setup involves a general cluster-based approach to pair lists and non-bonded pair interactions that utilizes both GPU and central processing unit (CPU) single instruction, multiple data acceleration efficiently, including the ability to load-balance tasks between CPUs and GPUs. The algorithm work efficiency is tuned for each type of hardware, and to use accelerators more efficiently, we introduce dual pair lists with rolling pruning updates. Combined with new direct GPU–GPU communication and GPU integration, this enables excellent performance from single GPU simulations through strong scaling across multiple GPUs and efficient multi-node parallelization.

The recently developed Symanzik-improved staggered-quark discretization allows unquenched lattice-QCD simulations with much smaller (and more realistic) quark masses than previously possible. To test this formalism, we compare experiment with a variety of nonperturbative calculations in QCD drawn from a restricted set of "gold-plated" quantities. We find agreement to within statistical and systematic errors of 3% or less. We discuss the implications for phenomenology and, in particular, for heavy-quark physics.

The semileptonic process, B −→ π lν, is studied via full QCD lattice simulations. We use unquenched gauge configurations generated by the MILC collaboration. These include the effect of vacuum polarization from three quark flavors: the s quark and two very light flavors (u/d) of variable mass allowing extrapolations to the physical chiral limit. We employ Nonrelativistic QCD to simulate the b quark and a highly improved staggered quark action for the light sea and valence quarks. We calculate the form factors f+(q 2 ) and f0(q 2 ) in the chiral limit for the range 16 GeV 2 ≤ q 2 < q 2 max and obtain q 2 (2006)). Results for the form factor f+(q 2 ) in the chiral limit have changed significantly. The last two sentences in this abstract should now read; "We calculate the form factors f+(q 2 ) and f0(q 2 ) in the chiral limit for the range 16 GeV 2 ≤ q 2 < q 2 max and obtain q 2 max 16GeV 2 [ dΓ/dq 2 ] dq 2 / |V ub | 2 = 2.07(57) ps −1 . Combining this with a preliminary average by the Heavy Flavor Averaging Group (HFAG'05) of recent branching fraction data for exclusive B semileptonic decays from the BaBar, Belle and CLEO collaborations, leads to |V ub | = 3.55(25)(50) × 10 −3 ."

We obtain a new value for the QCD coupling constant by combining lattice QCD simulations with experimental data for hadron masses. Our lattice analysis is the first to (1) include vacuum polarization effects from all three light-quark flavors (using MILC configurations), (2) include third-order terms in perturbation theory, (3) systematically estimate fourth and higher-order terms, (4) use an unambiguous lattice spacing, and (5) use an [symbol: see text](a2)-accurate QCD action. We use 28 different (but related) short-distance quantities to obtain alpha((5)/(MS))(M(Z)) = 0.1170(12).