Published April 16, 2019 | Version v.3.1.5

gprMax/gprMax: v.3.1.5 (Big Smoke)

  • 1. Northumbria University
  • 2. Imaginary Industries
  • 3. School of Engineering, The Univeristy of Edinburgh
  • 4. Columbia University

Description

This is the v.3.1.5, codenamed Big Smoke, release of gprMax.

Version 3.1.5 contains general usability and performance improvements, as well as the following specific enhancements and bug fixes since v.3.1.5:

  • Changed rounding method (to ceil) used to calculate number of iterations. Now guarantees simulation will run beyond specified time window.
  • Open input files with utf-8 encoding to allow comments in Asian languages
  • Added an input file command #output_dir to specify the directory of output files other than the directory of the input files.
  • Improved memory reporting when using GPU-based solver
  • Modified detection and selection of GPUs: -gpu by itself uses a default device ID of 0, otherwise a list of GPU device IDs can be given, e.g. -gpu [0 1 2]
  • Cleaned up MPI task farm handling: default is to use MPI Spawn mechanism, but an alternate implementation can be specified to avoid this using --mpi-no-spawn flag
  • Corrected bug with snapshots on GPU when snapsgpu2cpu flag active.
  • Corrected path to custom pulse for optimised GSSI 1.5GHz antenna model.
  • Corrected handling of custom waveform less than length of time window.
  • Implemented new (optional) MRIPML formulation with both CPU and GPU-based solvers.
  • Corrected bug with PML on GPU-based solver when using different thicknesses of PML in a model.

v.3.1 continues our whisky-based naming, and is also a reference to the cities of Edinburgh (Scotland) and San Francisco (USA). Why? Because the development of v.3.1 was funded, through a research project, by Google.

The most significant feature of this release is the ability for simulations to utilise general-purpose computing using graphics processing units (GPGPU). We have used NVIDIA's Compute-Unified Device Architecture (CUDA). Our testing on both consumer and data centre NVIDIA GPU cards has shown dramatic performance increases over our parallelised CPU (OpenMP) implementation.

You can read about how to use the GPU functionality and find all the features of gprMax described in detail in the User Guide (http://docs.gprmax.com)

Please report any bugs with code via Issues on GitHub.

For general help and questions about using gprMax visit our Google Group forum (http://www.gprmax.com/forum.shtml)

Files

gprMax/gprMax-v.3.1.5.zip

Files (34.8 MB)

Name Size Download all
md5:73ead1aa94a2044a390cc58fd1d27a5a
34.8 MB Preview Download

Additional details

Related works