10000 GitHub - siegfriedgreg/gunrock: High-Performance Graph Primitives on GPUs
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

siegfriedgreg/gunrock

 
 

Repository files navigation

The versions noted in the script are the only ones that work after "make"!!!!

  1. run "script_mods.sh", or at least note the order. helps when in the interact environment to install modules in the same order as compilation.
  2. clone the github repository in your bridges folder.
  3. from the gunrock folder, "mkdir build && cd build/"
  4. run "cmake .."
  5. run "make" (many warnings present, but still finishes.) running "make help" will also give a list of options.
  6. you should have all of the algorithms in your build/bin/ folder. Algorithms are set to find my folders on bridges, find the appropriate files from the main gunrock to use their directories, or edit for your own.
  7. to run a graph from the build/bin directory, in an "interact --gpu" session, try: "./bfs market ../../../maps/gh0**.mtx" where the ** is 1-16.
  8. dont forget to reload the modules after the "interact --gpu", in the same order as compiled!

####################################################################################################################################################################################################################################################################################################################################################################################################################################################################################################

SAMPLE RUN (gh012.mtx):

Loading Matrix-market coordinate-formatted graph ... Reading from ../../../maps/gh012.mtx: Parsing MARKET COO format (12057441 nodes, 36164358 directed edges)... Done parsing (6s). Converting 12057441 vertices, 36164358 directed edges (unordered tuples) to CSR format... Done converting (1s).

Degree Histogram (12057441 vertices, 36164358 edges): Degree 0: 0 (0.00%) Degree 2^0: 0 (0.00%) Degree 2^1: 12057441 (100.00%)

Converting 12057441 vertices, 36164358 directed edges (unordered tuples) to CSR format... Done converting (2s). Using 1 GPU: [ 0 ]. Using traversal-mode TWC



iteration 0 elapsed: 278.289080 ms, src = 0, #iteration = 4511 Computing reference value ... CPU BFS finished in 1857.490112 msec. cpu_search_depth: 4511

Label Validity: CORRECT

First 40 labels of the GPU result: [0:0 1:3 2:6 3:9 4:12 5:15 6:18 7:21 8:3189 9:137 10:141 11:3117 12:3082 13:123 14:127 15:3064 16:3063 17:113 18:117 19:171 20:172 21:167 22:168 23:349 24:87 25:83 26:60 27:64 28:73 29:77 30:57 31:61 32:43 33:46 34:187 35:188 36:193 37:194 38:3153 39:18 ]

Memory Usage(B)	 #keys0,0	 #keys0,1

GPU_0 1579155456 78373368 78373368 queue_sizing = 6.500000 6.500000

[BFS] finished. avg. elapsed: 278.2891 ms iterations: 4511 min. elapsed: 278.2891 ms max. elapsed: 278.2891 ms rate: 129.9525 MiEdges/s src: 0 nodes_visited: 12057441 edges_visited: 36164358 nodes queued: 12057441 edges queued: 36164358 load time: 28722.4212 ms preprocess time: 130.4500 ms postprocess time: 179.8768 ms total time: 31208.3521 ms

######################################################################################################################### ######################################################################################################################### ########################################################################################################################

Loading Matrix-market coordinate-formatted graph ... Reading from ../../../maps/gh002.mtx: Parsing MARKET COO format (50912018 nodes, 108109320 directed edges)... Done parsing (18s). Converting 50912018 vertices, 108109320 directed edges (unordered tuples) to CSR format... Done converting (4s).

Degree Histogram (50912018 vertices, 108109320 edges): Degree 0: 0 (0.00%) Degree 2^0: 2055578 (4.04%) Degree 2^1: 47863091 (94.01%) Degree 2^2: 993289 (1.95%) Degree 2^3: 60 (0.00%)

Converting 50912018 vertices, 108109320 directed edges (unordered tuples) to CSR format... Done converting (5s). Using 1 GPU: [ 0 ]. Using traversal-mode TWC



iteration 0 elapsed: 1036.210060 ms, src = 0, #iteration = 17346 Computing reference value ... CPU BFS finished in 7562.914062 msec. cpu_search_depth: 17346

Label Validity: CORRECT

First 40 labels of the GPU result: [0:0 1:1 2:8837 3:8836 4:8835 5:8834 6:8833 7:8832 8:8831 9:8830 10:8829 11:8828 12:8827 13:8826 14:8825 15:8824 16:8823 17:8822 18:8821 19:8820 20:8819 21:8818 22:8817 23:8816 24:8815 25:8814 26:8813 27:8812 28:8811 29:8810 30:8809 31:8808 32:8807 33:8806 34:8805 35:8804 36:8803 37:8802 38:8801 39:8800 ]

Memory Usage(B)	 #keys0,0	 #keys0,1

GPU_0 6289358848 330928119 330928119 queue_sizing = 6.500000 6.500000

[BFS] finished. avg. elapsed: 1036.2101 ms iterations: 17346 min. elapsed: 1036.2101 ms max. elapsed: 1036.2101 ms rate: 104.3315 MiEdges/s src: 0 nodes_visited: 50912018 edges_visited: 108109320 nodes queued: 50912018 edges queued: 108109320 load time: 30994.4069 ms preprocess time: 253.9389 ms postprocess time: 679.3540 ms total time: 40767.7522 ms

About

High-Performance Graph Primitives on GPUs

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Cuda 80.5%
  • HTML 10.2%
  • C++ 3.0%
  • Shell 2.4%
  • CMake 1.5%
  • Makefile 0.8%
  • Other 1.6%
0