~2007-10-15: Comparison of a matlab benchmark and some 
transcoding, on several computers (mainly for reason of
the new `diagsim' and upgraded `gnu').


Matlab 7.5.0.338 (R2007b) with "nbench" (a modified, 
non-graphical version of an older matlab "bench").

LU    FFT   ODE   Sparse   Total

0.99  0.94  0.66  0.16   2.74,   AthlonXP2000+ (at 1250MHz), 512MB RAM  (box.ntmm.org)
0.26  0.30  0.40  0.07   1.03,   P4-HT,f15,m2,s9  (at 2600MHz), 1024MB RAM (cecill.ets...)
0.19  0.14  0.13  0.04   0.51,   Athlon64-X2,f15,m67,s3 (at 3000MHz), 2048MB RAM (gnu.ets...)
0.20  0.29  0.33  0.05   0.87,   Xeon-HT,f15,m4,s1 (at 3000MHz), 2048MB RAM (magsim.ets...)
0.11  0.22  0.17  0.04   0.53,   Xeon-E5345,f6,m15,s7 (at 2333MHz), 4096MB RAM (diagsim.ets...)

The above were with maxNumCompThreads(1);
Now, on gnu, magsim, diagsim (all multiprocessor), try more:


0.20  0.14  0.13  0.04   0.51,    (gnu: maxthreads:1)
0.12  0.14  0.14  0.05   0.44,    (gnu: maxthreads:2)
0.12  0.14  0.15  0.08   0.48,    (gnu: maxthreads:4 [note: variation])

0.20  0.29  0.33  0.05   0.87,    (magsim: maxthreads:1)
0.14  0.29  0.33  0.07   0.82,    (magsim: maxthreads:2)
0.15  0.30  0.33  0.09   0.87,    (magsim: maxthreads:4)

0.11  0.22  0.17  0.04   0.53,    (diagsim: maxthreads:1)
0.10  0.23  0.17  0.04   0.54,    (diagsim: maxthreads:1)
0.10  0.25  0.17  0.05   0.56,    (diagsim: maxthreads:4)
0.08  0.24  0.17  0.05   0.55,    (diagsim: maxthreads:8)



rip audio, normalise, oggenc

serv
real    4m2.992s
user    2m35.754s
sys     0m10.333s

diagsim
real    2m24.817s
user    2m3.680s
sys     0m6.724s

gnu
real    3m2.395s
user    2m2.162s
sys     0m12.054s

magsim
real    4m32.5s
user    3m25.0s
sys     0m17s


2-pass video x264 encode, mkv merge

serv
real    109m31.182s
user    106m39.328s
sys     0m51.671s

diagsim
real    68m47.273s
user    90m26.651s
sys     0m24.874s

gnu
real    60m12.388s
user    80m9.524s
sys     0m33.723s

magsim
real    119m59.574s
user    177m34.442s
sys     1m36.218s