Difference between revisions of "Slicer:benchmarks"

From NAMIC Wiki
Jump to: navigation, search
m (Update from Wiki)
 
Line 5: Line 5:
 
The machines used were:
 
The machines used were:
  
* localhost.localdomain: an 8 processor Opteron system with 128 Gig of RAM made available for testing by [http://www.pcsforeveryone.com PC's for Everyone]
+
* localhost.localdomain: an 8 processor Opteron system with 128 Gig of RAM made available for testing by [http://www.thinkmate.com/article.php?art_id=156 Thinkmate]
 
* mimer.local: a 2 processor G5 PowerMac
 
* mimer.local: a 2 processor G5 PowerMac
 
* forest: a 10 processor SunFire
 
* forest: a 10 processor SunFire

Revision as of 16:24, 2 March 2007

Home < Slicer:benchmarks

Introduction

The timings below were generated on various machines using a simple benchmarking utility in slicer (bench.tcl). The goal was to look at performance scaling for multi-threaded processing.

The machines used were:

  • localhost.localdomain: an 8 processor Opteron system with 128 Gig of RAM made available for testing by Thinkmate
  • mimer.local: a 2 processor G5 PowerMac
  • forest: a 10 processor SunFire
  • vision: a 6 processor Sun
  • rock: a 3.4 gigahertz pentium 4 dell XPS laptop (one processor - two thread timings use 'hyper threading')
  • deuce: a dual processor opteron
  • glider: a dual processor xeon (with hyperthreading to get 4 threads)

Times in the boxes are in seconds.


Results generated with blur.tcl. Gaussian smoothing with local memory access.

blur

blur
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 % 2 Threads % Max Threads
deuce 10.07 5.16 51.27 51.27
forest 73.14 35.51 24.04 18.26 14.83 12.33 10.66 10.02 8.22 8.17 48.55 11.17
glider.bwh.harvard.edu 15.26 9.48 8.42 7.88 62.14 51.67
localhost.localdomain 15.41 8.11 5.76 4.86 4.02 3.61 3.53 3.37 52.67 21.85
mimer.local 24.94 15.45 61.94 61.94
rock 13.61 11.69 85.88 85.88
vision 62.86 30.03 20.00 15.38 12.43 11.20 47.77 17.81
pcs4everyone16.localdomain 9.59 7.45 4.76 3.58 3.11 2.91 2.64 2.48 2.70 2.33 2.53 2.19 2.29 2.34 2.06 2.49 77.70 25.98

axflip

Results generated with axflip.tcl. Image reslicing with cubic interpolation and non-local memory access.

axflip
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 % 2 Threads % Max Threads
deuce 13.64 6.95 50.91 50.91
forest 108.78 52.59 45.67 32.97 28.55 23.23 20.45 17.57 15.79 14.26 48.35 13.10
glider.bwh.harvard.edu 16.72 8.72 9.45 7.72 52.17 46.15
localhost.localdomain 13.98 7.29 6.93 5.11 4.61 4.33 3.68 2.75 52.18 19.69
mimer.local 32.76 17.31 52.82 52.82
rock 20.09 17.36 86.39 86.39
vision 92.14 44.53 38.64 27.92 24.20 20.02 48.33 21.73
pcs4everyone16.localdomain 6.56 3.28 3.03 2.15 1.92 1.57 1.33 1.17 1.09 1.01 0.95 0.87 0.81 0.78 0.76 0.75 49.94 11.50