Benchmarking spatial decomposition performance in CHARMM

Benchmarking spatial decomposition performance in CHARMM

Data

First number is timing in seconds (lower number is better)
Second number in () is speedup from previous number of CPUs
number in brackets <> is speedup over single CPU timing
OS is GNU/Linux-2.6.11-gentoo-r6, Gentoo distribution
CHARMM is c32a2 modified for spatially distributed parallel computation
pref.dat was used
AMD Opteron (x86_64): 2 X Opteron 242 (1.6GHz) per box, gigE, inexpensive 24 port switch

CPUs spatial standard NAMD
276483 atoms 276483 atoms 276483 atoms
1 1034.1 1049.2 1189.0
2 547.3(1.89)<1.89> 552.5(1.90)<1.90> 651.37(1.83)<1.83>
4 283.9(1.93)<3.64> 301.9(1.83)<3.48> 338.38(1.92)<3.22>
8 158.6(1.79)<6.52> 176.3(1.71)<5.95> 168.84(2.00)<6.45>
16 90.0(1.76)<11.49> 115.0(1.53)<9.12> 86.2(1.96)<12.63>
32 53.8(1.67)<19.22> 84.1(1.37)<12.48> 45.0(1.92)<24.2>

Notes:


NAMD: version 2.5, binary taken from NAMD_2.5_Linux-i686-TCP.tar.gz

Compile options:
  • gcc-x86_64-3.4.3:g77 -O3 -mcmodel=medium -ffast-math -fno-backslash -fugly-complex -fno-globals -Wno-globals

See also other benchmarks.
In order to repeat the timings in standard column follow the instructions on this page
Milan Hodoscek
Last modified: Fri July 15 09:47:27 CEST 2005