Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · astro-ph/0606464

pMatlab Parallel Matlab Library

Nadya Bliss·Jeremy Kepner
View original source ↗Read source document ↗Export metadata ↓

Abstract

MATLAB has emerged as one of the languages most commonly used by scientists and engineers for technical computing, with ~1,000,000 users worldwide. The compute intensive nature of technical computing means that many MATLAB users have codes that can significantly benefit from the increased performance offered by parallel computing. pMatlab (www.ll.mit.edu/pMatlab) provides this capability by implementing Parallel Global Array Semantics (PGAS) using standard operator overloading techniques. The core data structure in pMatlab is a distributed numerical array whose distribution onto multiple processors is specified with a map construct. Communication operations between distributed arrays are abstracted away from the user and pMatlab transparently supports redistribution between any block-cyclic-overlapped distributions up to four dimensions. pMatlab is built on top of the MatlabMPI communication library (www.ll.mit.edu/MatlabMPI) and runs on any combination of heterogeneous systems that support MATLAB, which includes Windows, Linux, MacOSX, and SunOS. Performance is validated by implementing the HPC Challenge benchmark suite and comparing pMatlab performance with the equivalent C+MPI codes. These results indicate that pMatlab can often achieve comparable performance to C+MPI at usually 1/10th the code size. Finally, we present implementation data collected from a sample of 10 real pMatlab applications drawn from the ~100 users at MIT Lincoln Laboratory. These data indicate that users are typically able to go from a serial code to a well performing pMatlab code in about 3 hours while changing less than 1% of their code.

Explore related subjects

astro-ph

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Nadya Bliss, Jeremy Kepner. 2006-06-20. pMatlab Parallel Matlab Library. https://doi.org/10.1177/1094342007078446

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2023

Dark Energy is the Cosmological Quantum Vacuum Energy of Light Particles-The Axion and the Lightest Neutrino

We uncover the general mechanism producing the dark energy(DE). This is only based on well known quantum physics and cosmology. We show that the observed DE originates from the cosmological quantum vacuum of light particles which provides a continuous energy distribution able to reproduce the data. Bosons give positive contributions to the DE while fermions yield negative contributions. As usual in field theory, ultraviolet divergences are subtracted from the physical quantities. The subtractions respect the symmetries of the theory and we normalize the physical quantities to be zero for the Minkowski vacuum. The resulting finite contributions to the energy density and the pressure from the quantum vacuum grow as log a(t) where a(t) is the scale factor, while the particle contributions dilute as 1/a^3(t), as it must be for massive particles. The DE equation of state P = w(z)H turns to be w(z)<-1 with w(z) asymptotically reaching the value -1 from below.A scalar particle can produce the observed DE through its quantum cosmological vacuum provided:(i)its mass is of the order of 10^{-3} eV = 1 meV,(ii) it is very weakly coupled and (iii) it is stable on the time scale of the age of the universe. The axion vacuum thus appears as a natural candidate. The neutrino vacuum (especially the lightest mass eigenstate) can give negative contributions to the DE. We find that w(z=0) is slightly below -1 by an amount ranging from [-1.5 10^{-3}] to [-8 10^{-3}] and we predict the axion mass to be in the range between 4 and 5 meV. We find that the universe will expand in the future faster than the de Sitter universe, as an exponential in the square of the cosmic time. DE arises from the quantum vacua of light particles in FRW cosmological space time in an analogous way to the Casimir effect in Minkowski spacetime with non trivial boundaries.

astro-ph↗
arXiv2023

Scaling of Black Hole Accretion Discs from Gamma-Ray Bursts and Black Hole X-Ray Binaries to Active Galactic Nuclei

I consider how physical processes scale over eight orders of magnitude in black hole mass, from stellar masses in gamma-ray bursts (GRB) and black-hole X-ray binaries (BHXRB) to supermassive active galactic nuclei (AGN). Accretion rates onto stellar mass black holes range over more than sixteen orders of magnitude, from the lower luminosity BHXRB to GRB. These enormous parameter ranges correspond to qualitative as well as quantitative differences in behavior. The fundamental questions involve the balance between nonequilibrium and thermalized plasmas. When energy fluxes exceed a critical value $\sim 10^{29}$ erg/cm$^2$s, as in GRB, a black-body equilibrium pair plasma forms. At the lower fluxes found in AGN, BHXRB and microquasars, accretion power electrodynamically accelerates a small number of very energetic particles, explaining their non-thermal spectra and the high energy gamma-ray emission of blazars. Ultra-high energy cosmic rays may be accelerated by massive black holes, otherwise undetectable, with very low thermal luminosities. New-born fast high-field pulsars may be in the black-body equilibrium regime, resembling SGR in permanent outburst. I also consider the question, significant for the acceleration of nonthermal particles in GRB outflows, of whether collisionless plasmas interpenetrate rather than forming hydrodynamic shocks, and propose this as an alternative to internal shock models of GRB. A new appendix attempts to explain why AGN are, proportionally, more efficient accelerators of energetic particles than stellar mass black holes.

astro-ph↗