Skip to content

BUDE - Bristol University Docking Engine

BUDE is a general purpose molecular docking program written in C++, OpenMP and OpenCL that uses GPU acceleration to perform:

  1. Virtual screening by docking of millions of small-molecule ligands;
  2. Ligand binding site identification on protein surfaces;
  3. Protein-protein docking in real space.

Interaction energies are calculated using an empirical free energy forcefield developed at the University of Bristol. These energies approximate binding free energies in units of kJ/mol and are used both for pose and affinity prediction. The force-field comprises very soft-core potentials to accommodate geometrical approximations inherent in the docking approach and is designed to give a better balance between energetic interactions and shape matching than more traditional methods.

Principal Investigator:

Simon McIntosh-Smith simonm@compsci.bristol.ac.uk

Other application users/developers:

University of Bristol – School of Biochemistry

Scientific area:

Molecular Docking

Scalability:

Good scaling up to thousands of GPU cores on SMP clusters or similar hybrid computing architectures.

Versions:

OpenMP
CUDA
OpenCL

Other

C++ language

Tested on platforms:

SMP cluster based on NVIDIA, AMD GPU, ARM CPUs and Mali GPUs, Intel Xeon Phi and Xeon processors.

URL: BUDE

Authors: Richard Sessions, Amaurys Avila Ibarra, Simon McIntosh-Smith, James Price, Debbie Shoemark and Tony Clarke.

More information: 
http://hpc.sagepub.com/content/early/2014/05/13/1094342014528252

Hadoop/Map-Reduce clustered bioinformatics applications

As stated in the MontBlanc2 DoW for the WP3, wherever applicable, we will “… test the prototype platforms configured as a Map-Reduce/Hadoop distributed engine on top of which data-driven applications like those in the area of genomics may be implemented and bioinformatics tools to query, align and collect DNA sequences properly benchmarked”. CINECA is currently opening a multi-tiered server and storage infrastructure able to provide top-class solutions to high performance data-handling applications. At the centre of the software stack is an OpenStack virtualization engine (NUBES) on top of which several IaaS/PaaS services will be released and among these, Hadoop/MapReduce (H/MR) hypervised clusters will be experimented. With the Hypervised H/MR solution implemented in NUBES and the bioinformatics benchmarks setup, we will have at disposal a finely tuneable virtualized cluster, which could be as well, closely adaptable to the physics infrastructure eventually deployed in MontBlanc2 and capable to host such a kind of HTC applications. Among other, bioinformatics applications and workflow will be implemented in the H/MR cluster over NUBES to model the HTC pipeline and when ready, it will be deployed over the capable Mont-Blanc2 prototypes. Examples of tester applications will cover many bionformatics sectors of interest with focus on those for Next Generation Sequencing and epigenomics like bowtie, pattern search, multiple alignement just to cite a few.

Principal Investigator:

At CINECA:
G. Fiameni, M. Rosati.
g.fiameni,m.rosati@cineca.it

For Mont-Blanc project:
Nico Sanna, CINECA
n.sanna@cineca.it

Other application users/developers:

  • CNR Flagship Project EPIGEN
  • Industrial partners and Health services

Scientific area:

Bioinformatics

Scalability:

Typical H/MR infrastructures have good scaling up to hundreds of node on Infiniband clusters and/or similar computing architectures.

Other:

HTC 10 Gbps and Infiniband environment

Tested on platforms:

Infiniband and 10 Gbps clusters.

LBC – LATTICE BOLTZMANN CODE

LBC (Lattice-Boltzmann Code) is a solver for the Boltzmann Equation using the Lattice-Boltzmann Method (LBM) written in Fortran95. The Lattice-Boltzmann Method is based on a discretization of the Boltzmann Equation, which describes the time-evolution of the phase-space density of a fluid or an ensemble of particles through advection and collisions, respectively. The numerical model is conceptually very simple: a single equation is solved on an orthogonal, (usually) equidistant grid by basic stencil operations without relying on complex matrix operations or iterative processes. The LBM method can be implemented in a highly efficient manner based on vector operations making it very suitable both for vector architectures as the NEC SX series and for modern SSE instruction sets available to all x86- based processors. LBC is a sister code of BEST (Boltzmann Equation Solver Tool), this code is highly optimiced and developed for several different architektures. Basically, due to a complex system of pre-processor macros, porting work on BEST is very error-prone. Both codes are very similar in terms of the basic code structure, i.e. data organization, etc., and mathematical model.

LBC and its sister code BEST are regularly used for performance evaluation during procurements.

Principal Investigator:

At HLRS-USTUTT:
Dr. José Gracia
gracia@hlrs.de

For Mont-Blanc project:
Mathias Nachtmann, HLRS
nachtmann@hlrs.de

Other application users/developers:

  • German Research School for Simulation Sciences (Aachen)
  • High-Performance Computing Center Stuttgart (HLRS)

Scientific area:

Computational Fluid Dynamics

Scalability:

Good scaling up to ten thousands cores on Infiniband clusters, Cray XE6 and/or similar computing architectures.

Versions:

  • MPI
  • OpenMP

Tested on platforms:

Infiniband clusters, Cray XE6, Cray XC30, NEC SX-9

ROTORSIM – A multi-grid, multi-block CFD code

The ROTORSIM CFD code is a multi-grid and multi-block CFD code written in C, MPI and OpenCL, the first GPU port that supports these features that we are aware of. We have measured the performance of ROTORSIM across a diverse range of hardware platforms, and observed a very good level of performance portability. ROTORSIM is a structured grid code, and as with other structured grid codes it is memory bandwidth limited, and so we have measured ROTORSIM’s performance portability by recording the fraction of peak memory bandwidth that the code sustains across the different target hardware platforms. A paper detailing these experimental results for ROTORSIM was published at ISC 2014. The absolute performance of the memory bandwidth-bound ROTORSIM code on a range of many-core devices as well as its performance relative to peak memory bandwidth and peak double precision floating point, has been carefully benchmarked. These results shown that ROTORSIM is also exhibiting a good degree of performance portability, sustaining a similar fraction of peak memory bandwidth across most devices, even when those devices have very low double precision floating point performance thus making it ideal as testbed of the upcoming architectures in Mont-Blanc 2 project.

Principal Investigator:

Christian Allen C.B.Allen@bristol.ac.uk

Simon McIntosh-Smith (for Mont Blanc)
simonm@compsci.bristol.ac.uk

Other application users/developers:

University of Bristol – Faculty of Engineering

Scientific area:

Computational Fluid Dynamics

Scalability:

Good scaling up to thousands of GPU cores on SMP clusters or similar hybrid computing architectures.

Versions:

  • MPI
  • OpenMP
  • CUDA
  • OpenCL

Other:

C language

Tested on platforms:

SMP cluster based on NVIDIA, AMD GPU, Intel Xeon and Xeon Phi and ARM processors

URL: Rotorsim