Marianna: 1 revision

2008-05-15T18:52:26Z

1 revision

Pieper: /* Introduction */

2007-03-02T16:24:11Z

Introduction

New page

= Introduction =

The timings below were generated on various machines using a simple benchmarking utility in slicer ([http://splweb.bwh.harvard.edu:8000/cgi-bin/develop/viewcvs.cgi/slicer2/Testing/Attic/bench.tcl?rev=1.1.2.8&content-type=text/vnd.viewcvs-markup&only_with_tag=release-2-6&cvsroot=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 [http://www.thinkmate.com/article.php?art_id=156 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.

<br /> Results generated with [http://splweb.bwh.harvard.edu:8000/cgi-bin/develop/viewcvs.cgi/~checkout~/slicer2/Testing/Benchmarks/Attic/blur.tcl?rev=1.1.2.2&only_with_tag=release-2-6&cvsroot=slicer blur.tcl]. Gaussian smoothing with local memory access.

== blur ==

{| border="2"
|+ blur
|-
!
! 1
! 2
! 3
! 4
! 5
! 6
! 7
! 8
! 9
! 10
! 11
! 12
! 13
! 14
! 15
! 16
! % 2 Threads
! % Max Threads
|- align="right"
! deuce
| 10.07
| 5.16
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51.27
| 51.27
|- align="right"
! 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
|- align="right"
! glider.bwh.harvard.edu
| 15.26
| 9.48
| 8.42
| 7.88
|
|
|
|
|
|
|
|
|
|
|
|
| 62.14
| 51.67
|- align="right"
! localhost.localdomain
| 15.41
| 8.11
| 5.76
| 4.86
| 4.02
| 3.61
| 3.53
| 3.37
|
|
|
|
|
|
|
|
| 52.67
| 21.85
|- align="right"
! mimer.local
| 24.94
| 15.45
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61.94
| 61.94
|- align="right"
! rock
| 13.61
| 11.69
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85.88
| 85.88
|- align="right"
! vision
| 62.86
| 30.03
| 20.00
| 15.38
| 12.43
| 11.20
|
|
|
|
|
|
|
|
|
|
| 47.77
| 17.81
|- align="right"
! 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 [http://splweb.bwh.harvard.edu:8000/cgi-bin/develop/viewcvs.cgi/~checkout~/slicer2/Testing/Benchmarks/Attic/axflip.tcl?rev=1.1.2.3&only_with_tag=release-2-6&cvsroot=slicer axflip.tcl]. Image reslicing with cubic interpolation and non-local memory access.

{| border="2"
|+ axflip
|-
!
! 1
! 2
! 3
! 4
! 5
! 6
! 7
! 8
! 9
! 10
! 11
! 12
! 13
! 14
! 15
! 16
! % 2 Threads
! % Max Threads
|- align="right"
! deuce
| 13.64
| 6.95
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50.91
| 50.91
|- align="right"
! 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
|- align="right"
! glider.bwh.harvard.edu
| 16.72
| 8.72
| 9.45
| 7.72
|
|
|
|
|
|
|
|
|
|
|
|
| 52.17
| 46.15
|- align="right"
! localhost.localdomain
| 13.98
| 7.29
| 6.93
| 5.11
| 4.61
| 4.33
| 3.68
| 2.75
|
|
|
|
|
|
|
|
| 52.18
| 19.69
|- align="right"
! mimer.local
| 32.76
| 17.31
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52.82
| 52.82
|- align="right"
! rock
| 20.09
| 17.36
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 86.39
| 86.39
|- align="right"
! vision
| 92.14
| 44.53
| 38.64
| 27.92
| 24.20
| 20.02
|
|
|
|
|
|
|
|
|
|
| 48.33
| 21.73
|- align="right"
! 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
|}

← Older revision	Revision as of 18:52, 15 May 2008
(No difference)

Slicer:benchmarks - Revision history

Marianna: 1 revision

Pieper: /* Introduction */