govertb/gpugraphlayout

An experimental GPU accelerated implementation of ForceAtlas2

📌 Pinned

• GPUGraphLayout

An experimental GPU accelerated implementation of ForceAtlas2.

gpugraphlayout's Issues

[question] optimal parameters

Just a curiosity... can you share your experience with FA2 parameters in large networks?
When I use Gephi, my best results are with:

Gravity: 0.01
Scale: [2-5]
linlog
BH: aprox
strong gravity

document format for edge list input

please add an example input file and document edges/weights input format.

Error running，ubuntu18.04 and cuda10

when i run this project use ./graph_viewer gpu 10000 100 wg 5 1 approximate ./result.csv ./graph.csv csv, then throw me a Error is error: CUDA was requested, but not compiled for. I don't know how to solve this problem . please help me ~~

Can't export the layout

Hi there,

I'm trying to use the code, but it won't create a layout file after finished running. I can create a file in the same output folder using tools like nano without sudo access.

[maxzhang@cdr352 linux]$ ./graph_viewer gpu 50 0 wg 2 1 approximate edgelist output csv
Loading edgelist at 'edgelist'...done.
    fetched 4 nodes and 4 edges.
Started Layout algorithm...
Starting iteration 3 (6.00%).
Starting iteration 6 (12.00%).
Starting iteration 9 (18.00%).
Starting iteration 12 (24.00%).
Starting iteration 15 (30.00%).
Starting iteration 18 (36.00%).
Starting iteration 21 (42.00%).
Starting iteration 24 (48.00%).
Starting iteration 27 (54.00%).
Starting iteration 30 (60.00%).
Starting iteration 33 (66.00%).
Starting iteration 36 (72.00%).
Starting iteration 39 (78.00%).
Starting iteration 42 (84.00%).
Starting iteration 45 (90.00%).
Starting iteration 48 (96.00%).
[maxzhang@cdr352 linux]$ cd output/
[maxzhang@cdr352 output]$ ls
[maxzhang@cdr352 output]$

saving node positions

Hi @govertb,

Is it possible to output the (x,y) positions of the nodes instead of just a .png file?

Thanks!

Crashes on graphs with disconnected components when using weak gravity

E.g. using ./graph_viewer gpu 500 0 wg 5 1 approximate edgelist . results in

error: (GPUassert) an illegal memory access was encountered (error 77). ../../src/RPGPUForceAtlas2.cu:232

Possibly because gravitational force doesn't overcome repulsive force that displaces small disconnected components away from the largest component, perhaps beyond the bounds of the layout frame.

Allow linlog to be used.

Sorry,I think your code has a bug.It's serious problem.

In RPGraph.cpp file.
adjacency_list[std::min(s_mapped, t_mapped)].push_back(std::max(s_mapped, t_mapped));
This code is not build directed edge correctly.It's should be as adjacency_list[s_mapped].push_back(t_mapped)

Can not produce layout

Hi I have followed every steps. But I am getting stuck at the last part and can not produce the layout. Can you please tell me what I am doing wrong here? Thanks!

Add a GUI

Working on this.

Dev under C# on Windows 10 using RTX using CUDA and needing streaming visualization

The title pretty much says it all. My dev platform is VS2019. Also under W10.

I have a need to visualize streaming data (financial)

Is this feasible? If not, any suggestions?

Thanks,
Merlin

Stalls on Ampere

Running on a GeForce RTX 3090 (Ampere architecture) the code stalls in the SortKernel. Possibly because certain threads no longer co-reside on the same multiprocessor during shared memory reductions in the SummarizationKernel and/or SortKernel.

repulsive force coefficient k_r is not valuation

I have a quesition，your class ForceAltas2 have a protected k_r，but your are not give that value。it's so inconceivability。

Segfaults on edgelist_path in current directory

For example ./graph_viewer cpu 2 1 wg 5 1 approximate "edges.txt" .

Will crash before printing any pictures, (but after running for a while) which a segfault.

This appears to be because the program searches for the character "/" in edgelist_path, but fails if there is no slash.

Output on MacOS

Starting iteration 1 (50.00%).
libc++abi.dylib: terminating with uncaught exception of type std::out_of_range: basic_string
Abort trap: 6

On Ubuntu it simply reports segmentation fault.

circle constrained layout

Hi @govertb,

Great work on this project. Is there any way to alter the calculation such that the layout is not constrained to a circle?

I plan to use this for a research project of mine and would love to get in touch to ask a few more detailed questions. Is there an email address I can reach you at? You can also send me a message to [email protected].

Thanks!

I have a question about your paper and code.

One question, Your coding's performance very good but why your paper don't publish the good academic journal? Another question, I read carefully your this paper but I can't find 'pipeline' in the code which usually presents in your paper context. The 'pipeline' as your important contribution I don't find it.

ubuntu 16.04 error compilation

I have this error while compiling

paco@W35:~/GPUGraphLayout/builds/linux$ make graph_viewer
mkdir -p obj
nvcc -c -D NO_FREETYPE  -O3 -std=c++11 -o obj/graph_viewer.o ../../src/graph_viewer.cpp
mkdir -p obj
nvcc -c -D NO_FREETYPE  -O3 -std=c++11 -o obj/RPGraph.o ../../src/RPGraph.cpp
mkdir -p obj
nvcc -c -D NO_FREETYPE  -O3 -std=c++11 -o obj/RPCPUForceAtlas2.o ../../src/RPCPUForceAtlas2.cpp
mkdir -p obj
nvcc -c -D NO_FREETYPE  -O3 -std=c++11 -o obj/RPForceAtlas2.o ../../src/RPForceAtlas2.cpp
mkdir -p obj
nvcc -c -D NO_FREETYPE  -O3 -std=c++11 -o obj/RPGraphLayout.o ../../src/RPGraphLayout.cpp
mkdir -p obj
nvcc -c -D NO_FREETYPE  -O3 -std=c++11 -o obj/RPCommon.o ../../src/RPCommon.cpp
mkdir -p obj
nvcc -c -D NO_FREETYPE  -O3 -std=c++11 -o obj/RPBarnesHutApproximator.o ../../src/RPBarnesHutApproximator.cpp
mkdir -p obj
nvcc -c -D NO_FREETYPE  -O3 -std=c++11 -o obj/RPLayoutAlgorithm.o ../../src/RPLayoutAlgorithm.cpp
mkdir -p obj/pngwriter/src
nvcc -c -D NO_FREETYPE  -O3 -std=c++11 -o obj/pngwriter/src/pngwriter.o ../../lib/pngwriter/src/pngwriter.cc
mkdir -p obj
nvcc --device-c  -O3 -std=c++11  -D NO_FREETYPE -o obj/RPBHKernels.o ../../src/RPBHKernels.cu
/usr/include/string.h: In function ‘void* __mempcpy_inline(void*, const void*, size_t)’:
/usr/include/string.h:652:42: error: ‘memcpy’ was not declared in this scope
   return (char *) memcpy (__dest, __src, __n) + __n;
                                          ^
Makefile:72: recipe for target 'obj/RPBHKernels.o' failed
make: *** [obj/RPBHKernels.o] Error 1

Ubuntu 16.04
cuda 7.5