[Gretl-devel] matrix_perf

Am 16.06.2014 21:10, schrieb Allin Cottrell: Nice to hear that new features are implemented. I ran it on a 5 year old AMD machine with 2 cores on ubuntu 13.10. See the output below. Please note that there seems to be an error in the "printvec" function. <OUTPUT> ? matrix_perf(1234) dgemm experiment 1, variant 1, speed in Gflops m n k vanilla openmp netlib 128 128 128 1.0131 1.6128 1.2421 128 128 256 1.0918 1.6047 2.8438 128 128 512 1.1172 1.5613 2.9728 128 128 1024 0.56444 0.90985 2.9641 128 128 2048 0.55022 0.81304 3.0153 result: netlib dominates for mnk >= 4194304 openmp dominates for mnk < 4194304 dgemm experiment 1, variant 2, speed in Gflops m n k vanilla openmp netlib 128 128 128 1.1262 1.7132 2.7780 256 256 128 0.87032 1.4749 2.8563 512 512 128 0.65816 1.2828 3.0365 1024 1024 128 0.57000 0.88802 3.1381 2048 2048 128 0.55844 0.85323 2.7625 result: netlib dominates dgemm experiment 1, variant 3, speed in Gflops m n k vanilla openmp netlib 128 128 128 1.0705 1.6740 2.7019 256 256 256 0.79600 1.3527 3.0087 512 512 512 0.53325 0.94332 3.1471 1024 1024 1024 0.46140 0.86064 3.3736 2048 2048 2048 0.48132 0.87184 3.5175 result: netlib dominates dgemm experiment 2, variant 1, speed in Gflops m n k vanilla openmp netlib 8 8 8 0.27082 0.16645 0.27986 16 8 8 0.36053 0.28264 0.39668 32 8 8 0.50332 0.45099 0.58258 64 8 8 0.65486 0.57687 0.74851 128 8 8 0.74979 0.75263 0.88244 256 8 8 0.77793 0.82707 0.93674 512 8 8 0.79804 0.80055 0.89166 1024 8 8 0.72645 0.84095 0.84162 2048 8 8 0.66676 0.77530 0.83087 4096 8 8 0.38886 0.53804 0.63493 result: netlib dominates dgemm experiment 2, variant 2, speed in Gflops m n k vanilla openmp netlib 10 2 1000 0.68929 0.43634 1.1583 20 2 1000 0.80831 0.44405 1.4244 40 2 1000 0.97165 0.70141 1.4610 80 2 1000 0.58981 0.68580 0.39415 160 2 1000 0.54345 0.70875 0.36567 320 2 1000 0.50763 0.71322 0.52442 640 2 1000 0.48572 0.74920 0.64036 1280 2 1000 0.47973 0.81962 0.33120 2560 2 1000 0.47045 0.80442 0.27437 5120 2 1000 0.47159 0.77478 0.26306 result: openmp dominates for mnk >= 160000 netlib dominates for mnk < 160000 dgemm experiment 2, variant 3, speed in Gflops m n k vanilla openmp netlib 10 10 1000 0.73219 1.0472 1.2821 20 10 1000 0.83210 1.3098 1.3693 40 10 1000 1.0183 1.5722 1.1212 80 10 1000 0.62957 0.95673 1.2608 160 10 1000 0.56578 0.72985 1.2352 320 10 1000 0.54269 0.84510 1.1248 result: netlib dominates for mnk >= 800000 Operating system: Linux (64-bit) BLAS library: Netlib Number of processors: 2 OpenMP enabled: yes Performance summary: vanilla - dominates outright in 0 out of 6 tests openmp - dominates outright in 0 out of 6 tests dominates in 1 test(s) for mnk >= > n = veclen( The symbol 'veclen' is undefined *** error in function printvec, line 1 called by function analyse_score called by function matrix_perf </OUTPUT> Artur