flamegpu-1.5-readme-v1.md

FLAME GPU Version 1.5

http://www.flamegpu.com/

Copyright 2015 University of Sheffield (see LICENCE)

Authors

• Paul Richmond
  • [email protected]
  • http://paulrichmond.shef.ac.uk/
• Mozhgan K. Chimeh : [email protected]

Usage

# Console Mode
<exe> <path/to/0.xml> <iterations> <device_id>

# Visualisation mode
<exe> <path/to/0.xml> <device_id>

Compilation

Windows (Visual Studio 2015)

Linux (Makefile)

To build executables under linux, the CUDA toolkit, an nvcc compatible host compiler such as g++, xmllint and xsltproc must be installed an accessible to make.

All examples can be built in batch via make all from the examples directory.

Individual examples can be built using make build in each example directory.

If an alternate CUDA version or CUDA intallation location is required, modify the CUDA_PATH environmental variable. I.e.

export CUDA_PATH=/path/to/CUDA

The PedestrianNavigation and PedestrianLOD examples show custom visualisations which is reflected in the relevant makefiles.

The CirclesBruteForce_double, CirclesBruteForce_float, CirclesPartitioning_double, CirclesPartitioning_float & StableMarriage examples are console only examples.

Further information can be found via make help in the relevant directories.

Requirements

Changelog

• 1.5.0
  • CUDA 8.0 and SM60 support
  • Removed SM20 support
  • Visual Studio 2015
  • @todo
• 1.4.2
  • @todo ?
• 1.4.1
  • @todo ?
• 1.4.0
  • Improved sorting using atomics
  • Support for simultaneous execution of function on same layer (using CUDA Streams)
  • Bug fixes and improvements
• 1.3
  • CUDPP replaced with Thrust for scan and sort primitives
  • Minor bug fixes in templates
  • Built for CUDA 6.0
  • Visual Studio Project and FLAMEGPU build rule updated
  • 32 bit configurations removed from visual studio projects
• 1.0.4
  • @todo ?