Strong Scaling Test on Stampede

  • Run Time
    1. With hypre
Num of Cores Run Time (secs)
1024 7963.7
2048 5862.3
4096 4005.8

2.Without hypre

Num of Cores Run Time (secs)
1024 7401.0
2048 5436.6
4096 4126.2/4025.2
  • Scaling Test Result
Runtime http://www.pas.rochester.edu/~bliu/Stampede/strongScalingOnStampedeLog.png
Runtime Considering Efficiency http://www.pas.rochester.edu/~bliu/Stampede/SC_withEff_Stampede.png
Cell Updates Per Second http://www.pas.rochester.edu/~bliu/Stampede/SC_CellUpStampede.png
  • Standard output of last advance
    1. 1024 cores: 
      Info allocations    =    79.8 gb  110.0 mb
 message allocations =   ------     32.0 mb
 sweep allocations   =   ------     29.9 mb
 filling fractions   =   0.017  0.597  0.855  0.000
 Current efficiency  =  66%  31%  97% 
 Cell updates/second =        437      1215  36%
 Wall Time Remaining =   ------   
 AMR Speed-Up Factor =       0.3331E+03
    2. 2048 cores 
      Info allocations    =   106.9 gb   85.5 mb
 message allocations =   ------     64.0 mb
 sweep allocations   =   ------     30.7 mb
 filling fractions   =   0.017  0.591  0.848  0.000
 Current efficiency  =  58%  39%  97% 
 Cell updates/second =        298      1011  29%
 Wall Time Remaining =   ------   
 AMR Speed-Up Factor =       0.2305E+03
    3. 4096 cores 
       Info allocations    =   147.4 gb   61.2 mb
 message allocations =   ------    128.0 mb
 sweep allocations   =   ------     20.1 mb
 filling fractions   =   0.016  0.619  0.846  0.000
 Current efficiency  =  47%  50%  97% 
 Cell updates/second =        187       785  24%
 Wall Time Remaining =   ------   
 AMR Speed-Up Factor =       0.1753E+03
  • Standard output of last advance (No self-gravity)
    1. 1024 
       Info allocations    =    67.0 gb   93.0 mb
 message allocations =   ------     32.0 mb
 sweep allocations   =   ------     25.2 mb
 filling fractions   =   0.017  0.597  0.852  0.000
 Current efficiency  =  69%
 Cell updates/second =        466      1299  36%
 Wall Time Remaining =   ------
 AMR Speed-Up Factor =       0.4099E+03
  1. 2048 
     Info allocations    =    92.0 gb   71.2 mb
 message allocations =   ------     64.0 mb
 sweep allocations   =   ------     22.5 mb
 filling fractions   =   0.016  0.620  0.848  0.000
 Current efficiency  =  61%
 Cell updates/second =        322      1097  29%
 Wall Time Remaining =   ------
 AMR Speed-Up Factor =       0.2873E+03
  1. 4096 
     Info allocations    =   125.3 gb   51.0 mb
 message allocations =   ------    128.0 mb
 sweep allocations   =   ------     19.9 mb
 filling fractions   =   0.016  0.616  0.849  0.000
 Current efficiency  =  50%
 Cell updates/second =        206       865  24%
 Wall Time Remaining =   ------
 AMR Speed-Up Factor =       0.1884E+03

Info allocations    =   125.6 gb   54.7 mb
 message allocations =   ------    128.0 mb
 sweep allocations   =   ------     19.9 mb
 filling fractions   =   0.016  0.620  0.845  0.000
 Current efficiency  =  51% 
 Cell updates/second =        211       882  24%
 Wall Time Remaining =   ------   
 AMR Speed-Up Factor =       0.1919E+03
  • CPU hours
    1. 1 frame: 4600 SUs
    2. 50 frames: 230,000 SUs
    3. 4~5 runs: 1,150,000 SUs (on stampede)
    4. Current Allocation: 416,000 SUs (on stampede), 1,138,234 SUs (on Kraken)

Comments

No comments.