Problems encountered when using FLAME GPU about flamegpu HOT 6 OPEN

Hi,

I don't think that FLAME GPU 1 is currently compatible with CUDA 11.

If you can, using CUDA 10.2 should work.

CUDA 11 includes an updated version of the Thrust library, which includes a full copy of Thrust, which conflict with the verison we include (supporting older versions of CUDA).
We've not found a simple solution to support CUDA 10 and CUDA 11 without changing changing the build system or duplicating work.

Visual Studio intellisense also generates a lot of incorrect errors/warnigns with CUDA. If you hide intellisense errors from the log it will greatly reduce the numbers of errors reported, though CUDA 11.2 will still not work.

This is related to #284

from flamegpu.

Thanks so much for your help Ptheywood!

I successfully used CUDA10.2 to build the solution 'Boids_BruteForce' in Visual Studio. I want to run the console mode and visualization mode of the project on Windows with my local GPU. What should I do next?

I tried to call Visual Studio's nmake.exe to run the makefile in the directory, but it gave me such an error message. This was confusing.

How can I get the output XML files and generate a visualization window?

Thanks again for your help!

Haoyang

from flamegpu.

The Makefile provided is for building under linux, it's not needed in windows.

In Visual Studio 4 build configurations are provided, Release_Console, Debug_Console, Release_Visualisation and Debug_Visualisation. You can either select configuration from the drop down at the top, or use the batch build process to build multiple configurations.
You'll want to build the Release_Console and Release_Visualisation configurations.

Once the builds are complete, the executables will be placed into FLAMEGPU/bin/x64/Release_Console/ and FLAMEGPU/bin/x64/Release_Visualisation. Batch files are provided in the FLAMEGPU/bin/x64 folder for each example to show how they can be used.

The console version will place output .xml files in the same directory as the input xml file (probably FLAMEGPU/examples/Boids_BruteForce/iterations/, with 0.xml used as the input file, and 1.xml, 2.xml, ... as the output file for each iteration. I don't believe the visualisation executables will generate output files.

from flamegpu.

Hi Ptheywood, thank you so much that with your help, I could now modify the FLAME GPU models to a certain extent according to my wishes.

I am currently interested in an example model PedestrianNavigation. I would like to ask, in the XML document of this model, how are the directions of 256*256 Navmap agents pointing to different exit definitions calculated? Is there an introduction to their calculation methods?

Thanks again for your help!

Haoyang

from flamegpu.

I believe that the following paper describes the pedestrian navigation model, with some discussion of the force vector fields (FVF).
http://paulrichmond.shef.ac.uk/cv/publications/eguk2010_vectorfields.pdf

In addition the following paper describes teh development of a tool to generate force vector fields using an interactive tool. Section 3.5 and Figure 5 describe how the FVF can be calculated using a flood fill algorithm.
I'm not sure if this tool is publicly available anywhere.

http://diglib.eg.org/bitstream/handle/10.2312/cgvc20161304/105-112.pdf

Section 3.1 of Twin Karmakharms' thesis provides a more detailed description of the vector field generation:

http://etheses.whiterose.ac.uk/20938/

from flamegpu.

Thanks a lot, Ptheywood! These articles really helped. I didn't find the source of the tool Pedestrian Environment Designer (PED) on the internet, so I wonder whether it is not available for the public anymore. Could you please help me to check if it is publicly available at present, and if so how can I access it?

Thanks again for your help!

Haoyang

from flamegpu.