ADEPT: a domain independent sequence alignment strategy for gpu architectures.

BACKGROUND: Bioinformatic workflows frequently make use of automated genome assembly and protein clustering tools. At the core of most of these tools, a significant portion of execution time is spent in determining optimal local alignment between two sequences. This task is performed with the Smith-Waterman algorithm, which is a dynamic programming based method. With the advent of modern sequencing technologies and increasing size of both genome and protein databases, a need for faster Smith-Waterman implementations has emerged. Multiple SIMD strategies for the Smith-Waterman algorithm are available for CPUs. However, with the move of HPC facilities towards accelerator based architectures, a need for an efficient GPU accelerated strategy has emerged. Existing GPU based strategies have either been optimized for a specific type of characters (Nucleotides or Amino Acids) or for only a handful of application use-cases. RESULTS: In this paper, we present ADEPT, a new sequence alignment strategy for GPU architectures that is domain independent, supporting alignment of sequences from both genomes and proteins. Our proposed strategy uses GPU specific optimizations that do not rely on the nature of sequence. We demonstrate the feasibility of this strategy by implementing the Smith-Waterman algorithm and comparing it to similar CPU strategies as well as the fastest known GPU methods for each domain. ADEPT's driver enables it to scale across multiple GPUs and allows easy integration into software pipelines which utilize large scale computational systems. We have shown that the ADEPT based Smith-Waterman algorithm demonstrates a peak performance of 360 GCUPS and 497 GCUPs for protein based and DNA based datasets respectively on a single GPU node (8 GPUs) of the Cori Supercomputer. Overall ADEPT shows 10x faster performance in a node-to-node comparison against a corresponding SIMD CPU implementation. CONCLUSIONS: ADEPT demonstrates a performance that is either comparable or better than existing GPU strategies. We demonstrated the efficacy of ADEPT in supporting existing bionformatics software pipelines by integrating ADEPT in MetaHipMer a high-performance denovo metagenome assembler and PASTIS a high-performance protein similarity graph construction pipeline. Our results show 10% and 30% boost of performance in MetaHipMer and PASTIS respectively.

Exascale applications: skin in the game.

As noted in Wikipedia, skin in the game refers to having 'incurred risk by being involved in achieving a goal', where 'skin is a synecdoche for the person involved, and game is the metaphor for actions on the field of play under discussion'. For exascale applications under development in the US Department of Energy Exascale Computing Project, nothing could be more apt, with the skin being exascale applications and the game being delivering comprehensive science-based computational applications that effectively exploit exascale high-performance computing technologies to provide breakthrough modelling and simulation and data science solutions. These solutions will yield high-confidence insights and answers to the most critical problems and challenges for the USA in scientific discovery, national security, energy assurance, economic competitiveness and advanced healthcare. This article is part of a discussion meeting issue 'Numerical algorithms for high-performance computational science'.

GW100: Benchmarking G0W0 for Molecular Systems.

We present the GW100 set. GW100 is a benchmark set of the ionization potentials and electron affinities of 100 molecules computed with the GW method using three independent GW codes and different GW methodologies. The quasi-particle energies of the highest-occupied molecular orbitals (HOMO) and lowest-unoccupied molecular orbitals (LUMO) are calculated for the GW100 set at the G0W0@PBE level using the software packages TURBOMOLE, FHI-aims, and BerkeleyGW. The use of these three codes allows for a quantitative comparison of the type of basis set (plane wave or local orbital) and handling of unoccupied states, the treatment of core and valence electrons (all electron or pseudopotentials), the treatment of the frequency dependence of the self-energy (full frequency or more approximate plasmon-pole models), and the algorithm for solving the quasi-particle equation. Primary results include reference values for future benchmarks, best practices for convergence within a particular approach, and average error bars for the most common approximations.

An explicit formula for optical oscillator strength of excitons in semiconducting single-walled carbon nanotubes: family behavior.

The sensitive structural dependence of the optical properties of single-walled carbon nanotubes, which are dominated by excitons and tunable by changing diameter and chirality, makes them excellent candidates for optical devices. Because of strong many-electron interaction effects, the detailed dependence of the optical oscillator strength f(s) of excitons on nanotube diameter d, chiral angle θ, and electronic subband index P (the so-called family behavior), however, has been unclear. In this study, based on results from an extended Hubbard Hamiltonian with parameters derived from ab initio GW plus Bethe-Salpeter equation (GW-BSE) calculations, we have obtained an explicit formula for the family behavior of the oscillator strengths of excitons in semiconducting single-walled carbon nanotubes (SWCNTs), incorporating environmental screening. The formula explains recent measurements well and is expected to be useful in the understanding and design of possible SWCNT optical and optoelectronic devices.

First-principles calculations of quasiparticle excitations of open-shell condensed matter systems.

We develop a Green's function approach to quasiparticle excitations of open-shell systems within the GW approximation. It is shown that accurate calculations of the characteristic multiplet structure require a precise knowledge of the self-energy and, in particular, its poles. We achieve this by constructing the self-energy from appropriately chosen mean-field theories on a fine frequency grid. We apply our method to a two-site Hubbard model, several molecules, and the negatively charged nitrogen-vacancy defect in diamond and obtain good agreement with experiment and other high-level theories.

An atlas of carbon nanotube optical transitions.

Electron-electron interactions are significantly enhanced in one-dimensional systems, and single-walled carbon nanotubes provide a unique opportunity for studying such interactions and the related many-body effects in one dimension. However, single-walled nanotubes can have a wide range of diameters and hundreds of different structures, each defined by its chiral index (n,m), where n and m are integers that can have values from zero up to 30 or more. Moreover, one-third of these structures are metals and two-thirds are semiconductors, and they display optical resonances at many different frequencies. Systematic studies of many-body effects in nanotubes would therefore benefit from the availability of a technique for identifying the chiral index of a nanotube based on a measurement of its optical resonances, and vice versa. Here, we report the establishment of a structure-property 'atlas' for nanotube optical transitions based on simultaneous electron diffraction measurements of the chiral index and Rayleigh scattering measurements of the optical resonances of 206 different single-walled nanotube structures. The nanotubes, which were suspended across open slit structures on silicon substrates, had diameters in the range 1.3-4.7 nm. We also use this atlas as a starting point for a systematic study of many-body effects in the excited states of single-walled nanotubes. We find that electron-electron interactions shift the optical resonance energies by the same amount for both metallic and semiconducting nanotubes, and that this shift (which corresponds to an effective Fermi velocity renormalization) increases monotonically with nanotube diameter. This behaviour arises from two sources: an intriguing cancellation of long-range electron-electron interaction effects, and the dependence of short-range electron-electron interactions on diameter.

Simple approximate physical orbitals for GW quasiparticle calculations.

Generating unoccupied orbitals within density functional theory (DFT) for use in GW calculations of quasiparticle energies becomes prohibitive for large systems. We show that, without any loss of accuracy, the unoccupied orbitals may be replaced by a set of simple approximate physical orbitals made from appropriately prepared plane waves and localized basis DFT orbitals that represent the continuum and resonant states of the system, respectively. This approach allows for accurate quasiparticle calculations using only a very small number of unoccupied DFT orbitals, resulting in an order of magnitude gain in speed.

Many-body interactions in quasi-freestanding graphene.

The Landau-Fermi liquid picture for quasiparticles assumes that charge carriers are dressed by many-body interactions, forming one of the fundamental theories of solids. Whether this picture still holds for a semimetal such as graphene at the neutrality point, i.e., when the chemical potential coincides with the Dirac point energy, is one of the long-standing puzzles in this field. Here we present such a study in quasi-freestanding graphene by using high-resolution angle-resolved photoemission spectroscopy. We see the electron-electron and electron-phonon interactions go through substantial changes when the semimetallic regime is approached, including renormalizations due to strong electron-electron interactions with similarities to marginal Fermi liquid behavior. These findings set a new benchmark in our understanding of many-body physics in graphene and a variety of novel materials with Dirac fermions.

Excitonic effects on the optical response of graphene and bilayer graphene.

We present first-principles calculations of many-electron effects on the optical response of graphene, bilayer graphene, and graphite employing the GW-Bethe Salpeter equation approach. We find that resonant excitons are formed in these two-dimensional semimetals. The resonant excitons give rise to a prominent peak in the absorption spectrum near 4.5 eV with a different line shape and significantly redshifted peak position from those of an absorption peak arising from interband transitions in an independent quasiparticle picture. In the infrared regime, our calculated optical absorbance per graphene layer is approximately a constant, 2.4%, in agreement with recent experiments; additional low frequency features are found for bilayer graphene because of band structure effects.

Electron-hole interaction in carbon nanotubes: novel screening and exciton excitation spectra.

The optical response of single-walled carbon nanotubes is dominated by exciton states with unusually large binding energies. We show that screening in semiconducting tubes enhances rather than reduces the electron-hole interaction for separations larger than the tube diameter. This "antiscreening" region deepens the relative energy level of the higher exciton states yielding unconventional excitation spectra. The effect explains the discrepancy in the current experimentally extrapolated exciton binding energies (deduced using conventional model spectra) and those obtained from ab initio calculations on isolated tubes.

Bound excitons in metallic single-walled carbon nanotubes.

We extend previous ab initio calculations on excitonic effects in metallic single-walled carbon nanotubes to more experimentally realizable larger diameter tubes. Our calculations predict bound exciton states in both the (10,10) and (12,0) tubes with binding energies of approximately 50 meV providing experimentally verifiable changes to the absorption line shape in each case. The second and third van Hove singularities in the joint density of states also give rise to a single optically active bound or resonant excitonic state.

Observation of excitons in one-dimensional metallic single-walled carbon nanotubes.

Excitons are generally believed not to exist in metals because of strong screening by free carriers. Here we demonstrate that excitonic states can in fact be produced in metallic systems of a one-dimensional character. Using metallic single-walled carbon nanotubes as a model system, we show both experimentally and theoretically that electron-hole pairs form tightly bound excitons. The exciton binding energy of 50 meV, deduced from optical absorption spectra of individual metallic nanotubes, significantly exceeds that of excitons in most bulk semiconductors and agrees well with ab initio theoretical predictions.