Hi, > This is strange, we are hoping to improve R performance through high clock > speed Power CPU(4.0 GHz above), > Now I think we should take more considerations, RAM is also cheaper for X86 > than Power :)
power is no good CPU. $ lsdev -C | grep proc proc0 Available 00-00 Processor proc2 Available 00-02 Processor $ lsattr -El proc0 frequency 4204000000 Processor Speed False smt_enabled true Processor SMT enabled False smt_threads 2 Processor SMT threads False state enable Processor state False type PowerPC_POWER6 Processor type False $ lsattr -El proc2 frequency 4204000000 Processor Speed False smt_enabled true Processor SMT enabled False smt_threads 2 Processor SMT threads False state enable Processor state False type PowerPC_POWER6 Processor type False As a result of DGEMM by GotoBLAS(http://prs.ism.ac.jp/~nakama/SurviveGotoBLAS2). $ GOTO_NUM_THREADS=1 ./bm 2000 12.954 GFLOPS (N x N : N=2000 1.23517sec) 12.719 GFLOPS (N x T : N=2000 1.25796sec) 13.118 GFLOPS (T x N : N=2000 1.21965sec) 12.726 GFLOPS (T x T : N=2000 1.25732sec) $ GOTO_NUM_THREADS=2 ./bm 2000 25.259 GFLOPS (N x N : N=2000 0.633444sec) 24.050 GFLOPS (N x T : N=2000 0.665272sec) 25.710 GFLOPS (T x N : N=2000 0.622316sec) 24.075 GFLOPS (T x T : N=2000 0.664595sec) $ GOTO_NUM_THREADS=4 ./bm 2000 21.311 GFLOPS (N x N : N=2000 0.750802sec) 25.778 GFLOPS (N x T : N=2000 0.620694sec) 26.398 GFLOPS (T x N : N=2000 0.60611sec) 25.826 GFLOPS (T x T : N=2000 0.619536sec) It's fast with the structure (L2 shared cache) of the CPU to 2CPU. Best Regards, -- EI-JI Nakama <nakama (a) ki.rim.or.jp> "\u4e2d\u9593\u6804\u6cbb" <nakama (a) ki.rim.or.jp> ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel