# 2011-08-11, 01:31. ---------------------------------------------------------------- # On 'gnu.ets.kth.se'. 8GB RAM (DDRIII, 1.6GHz), 3.2GHz Phenom X6, Linux-2.6.37, Gentoo sys. >> for n=1:8; maxNumCompThreads(n); nbench(4,1.2); end Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code. > In maxNumCompThreads at 24 Matlab 7.10.0.499 (R2010a) Benchmark, running 4 runs with parameter 1.20 LU FFT ODE Sparse TOTAL Machine 0.15 4.69 0.13 0.04 5.01 (this computer, run 1) 0.14 4.64 0.17 0.04 4.98 (this computer, run 2) 0.15 4.66 0.13 0.04 4.98 (this computer, run 3) 0.15 4.57 0.19 0.04 4.94 (this computer, run 4) 0.15 4.64 0.16 0.04 4.98 (this computer, mean) Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code. > In maxNumCompThreads at 24 Matlab 7.10.0.499 (R2010a) Benchmark, running 4 runs with parameter 1.20 LU FFT ODE Sparse TOTAL Machine 0.12 3.13 0.17 0.05 3.47 (this computer, run 1) 0.09 3.08 0.13 0.05 3.35 (this computer, run 2) 0.10 3.01 0.13 0.05 3.29 (this computer, run 3) 0.11 3.01 0.15 0.05 3.33 (this computer, run 4) 0.11 3.06 0.15 0.05 3.36 (this computer, mean) Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code. > In maxNumCompThreads at 24 Matlab 7.10.0.499 (R2010a) Benchmark, running 4 runs with parameter 1.20 LU FFT ODE Sparse TOTAL Machine 0.09 2.36 0.17 0.06 2.68 (this computer, run 1) 0.08 2.41 0.17 0.06 2.71 (this computer, run 2) 0.08 2.35 0.15 0.06 2.64 (this computer, run 3) 0.08 2.40 0.17 0.06 2.71 (this computer, run 4) 0.08 2.38 0.16 0.06 2.68 (this computer, mean) Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code. > In maxNumCompThreads at 24 Matlab 7.10.0.499 (R2010a) Benchmark, running 4 runs with parameter 1.20 LU FFT ODE Sparse TOTAL Machine 0.08 2.50 0.17 0.06 2.81 (this computer, run 1) 0.08 2.36 0.17 0.06 2.68 (this computer, run 2) 0.08 2.38 0.17 0.07 2.68 (this computer, run 3) 0.09 2.47 0.13 0.07 2.76 (this computer, run 4) 0.08 2.43 0.16 0.06 2.73 (this computer, mean) Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code. > In maxNumCompThreads at 24 Matlab 7.10.0.499 (R2010a) Benchmark, running 4 runs with parameter 1.20 LU FFT ODE Sparse TOTAL Machine 0.08 2.00 0.12 0.06 2.26 (this computer, run 1) 0.06 1.90 0.12 0.07 2.15 (this computer, run 2) 0.06 1.89 0.15 0.06 2.16 (this computer, run 3) 0.06 1.91 0.13 0.06 2.17 (this computer, run 4) 0.06 1.93 0.13 0.06 2.18 (this computer, mean) Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code. > In maxNumCompThreads at 24 Matlab 7.10.0.499 (R2010a) Benchmark, running 4 runs with parameter 1.20 LU FFT ODE Sparse TOTAL Machine 0.08 2.54 0.12 0.07 2.81 (this computer, run 1) 0.06 2.51 0.12 0.08 2.76 (this computer, run 2) 0.06 2.51 0.12 0.07 2.76 (this computer, run 3) 0.06 2.51 0.18 0.07 2.83 (this computer, run 4) 0.06 2.52 0.13 0.07 2.79 (this computer, mean) Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code. > In maxNumCompThreads at 24 Matlab 7.10.0.499 (R2010a) Benchmark, running 4 runs with parameter 1.20 LU FFT ODE Sparse TOTAL Machine 0.06 2.29 0.12 0.07 2.54 (this computer, run 1) 0.06 2.18 0.12 0.07 2.43 (this computer, run 2) 0.06 2.19 0.15 0.07 2.47 (this computer, run 3) 0.06 2.30 0.17 0.07 2.59 (this computer, run 4) 0.06 2.24 0.14 0.07 2.51 (this computer, mean) Warning: maxNumCompThreads will be removed in a future release. Please remove any instances of this function from your code. > In maxNumCompThreads at 24 Matlab 7.10.0.499 (R2010a) Benchmark, running 4 runs with parameter 1.20 LU FFT ODE Sparse TOTAL Machine 0.06 2.28 0.12 0.07 2.52 (this computer, run 1) 0.06 2.20 0.12 0.07 2.44 (this computer, run 2) 0.06 2.15 0.12 0.07 2.40 (this computer, run 3) 0.06 2.19 0.12 0.07 2.43 (this computer, run 4) 0.06 2.20 0.12 0.07 2.45 (this computer, mean) ---------------------------------------------------------------- r = [... % numthread, LU, FFT, ODE, Sparse, Total 1 0.15 4.64 0.16 0.04 4.98 2 0.11 3.06 0.15 0.05 3.36 3 0.08 2.38 0.16 0.06 2.68 4 0.08 2.43 0.16 0.06 2.73 5 0.06 1.93 0.13 0.06 2.18 6 0.06 2.52 0.13 0.07 2.79 7 0.06 2.24 0.14 0.07 2.51 8 0.06 2.20 0.12 0.07 2.45 ]; The 'sparse' actually rises in time with number of threads, but it is a very short calculation compared to the others, so the behaviour with, say, a 1s problem, may be very different. All the others get little or no benefit after 4 or 5 threads, and strongly sublinear speed-up in any case. As usual, if one has many parameters and enough memory, it would be better use of a multiprocessor system to run separate single-threaded matlab sessions.