b_eff = 527.805 MB/s = 32.988 * 16 PEs with 256 MB/PE on HI-UX/MPP hitachi 02-03 0 SR2201 SECTION-HEAD-BEGIN b_eff.c, Revision 3.2 from Nov. 07, 1999 MEMORY_PER_PROCESSOR = 256 MBytes [1M = 1024*1024] 1-dim-paterns: size = 16 1-dim-paterns: size = 16 2-dim-paterns: size = 4 * 4 3-dim-paterns: size = 4 * 2 * 2 Used message lengths (and used loop length for each message lengths): msglng= 1 (300), 2 (300), 4 (300), 8 (300), 16 (300), 32 (300), 64 (300), 128 (300), 256 (300), 512 (300), 1024 (300), 2048 (300), 4096 (300), 8933 (234), 19484 (107), 42495 (49), 92682 (22), 202141 (10), 440872 (4), 961548 (2), 2097152 (1), SECTION-HEAD-END SECTION-LOOP-BEGIN measurment loop: elapsed time = 656.901 sec sum of max elapsed time per entries above = 657.279 sec difference = -0.378 sec = 0.1% The difference is less than 5 % SECTION-LOOP-END SECTION-PATTERN-BEGIN Pattern parameters: p00 ring-8*2fix => 1 sendrecv_calls with 16 messages, i.e. msgs/used node, all nodes are used p01 ring-4*4fix => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p02 ring-2*8fix => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p03 ring-1*16fix => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p04 ring-1*16fix => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p05 ring-1*16fix => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p06 random-cyc-1dim => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p07 random-cyc-1dim => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p08 random-cyc-1dim => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p09 random-cyc-1dim => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p10 random-cyc-1dim => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p11 random-cyc-1dim => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p12 random-cyc-1dim => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p13 random-cyc-1dim => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p14 random-cyc-1dim => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p15 random-cyc-1dim => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p16 worst-cyc-1dim => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p17 best bi-section => 2 sendrecv_calls with 16 messages, i.e. msgs/used node, all nodes are used p18 worst bi-section => 2 sendrecv_calls with 16 messages, i.e. msgs/used node, all nodes are used p19 acyclic-1dim-all => 2 sendrecv_calls with 30 messages, i.e. msgs/used node, all nodes are used p20 acyclic-2dim-all => 4 sendrecv_calls with 48 messages, i.e. msgs/used node, all nodes are used p21 acyclic-3dim-all => 6 sendrecv_calls with 56 messages, i.e. msgs/used node, all nodes are used p22 cyclic-1dim-all => 2 sendrecv_calls with 32 messages, i.e. msgs/used node, all nodes are used p23 cyclic-2dim-all => 4 sendrecv_calls with 64 messages, i.e. msgs/used node, all nodes are used p24 cyclic-3dim-all => 4 sendrecv_calls with 64 messages, i.e. msgs/used node, all nodes are used SECTION-PATTERN-END Printing accumulated bandwidth / number of processes = bandwidth per process (in MBytes/sec = 1e6 Bytes/sec) Additionally the accumulated value processes is printed in the last column Only the cyclic and random patterns are used to compute b_eff (except p00), the other patterns are only as information. SECTION-BY-METHODS-BEGIN 2nd analysis path, only as information, last row: for all mthd: logavg_pat ( avg_L ( max_rep ( b(L) ) ) ) || logavg_pat ( max_mthd ( avg_L ( max_rep ( b(L) ) ) ) ) pattern / methods: Sndrcv Alltoal non-blk -> best method -> accumulated p00 ring-8*2fix : 38.631 8.007 34.717 -> 38.631 -> 618.097 MByte/s p01 ring-4*4fix : 32.766 11.365 32.506 -> 32.766 -> 524.248 MByte/s p02 ring-2*8fix : 32.720 10.018 32.646 -> 32.720 -> 523.528 MByte/s p03 ring-1*16fix : 31.912 8.922 32.676 -> 32.676 -> 522.810 MByte/s p04 ring-1*16fix : 32.142 8.931 32.237 -> 32.237 -> 515.791 MByte/s p05 ring-1*16fix : 32.050 8.930 32.076 -> 32.076 -> 513.215 MByte/s p06 random-cyc-1dim : 30.640 8.949 30.932 -> 30.932 -> 494.909 MByte/s p07 random-cyc-1dim : 30.743 8.775 30.737 -> 30.743 -> 491.890 MByte/s p08 random-cyc-1dim : 30.730 8.767 31.293 -> 31.293 -> 500.687 MByte/s p09 random-cyc-1dim : 30.295 8.592 30.467 -> 30.467 -> 487.473 MByte/s p10 random-cyc-1dim : 30.148 8.779 30.996 -> 30.996 -> 495.939 MByte/s p11 random-cyc-1dim : 30.971 8.748 31.452 -> 31.452 -> 503.227 MByte/s p12 random-cyc-1dim : 31.387 8.742 31.037 -> 31.387 -> 502.185 MByte/s p13 random-cyc-1dim : 30.313 8.895 30.303 -> 30.313 -> 485.003 MByte/s p14 random-cyc-1dim : 30.677 8.647 30.776 -> 30.776 -> 492.413 MByte/s p15 random-cyc-1dim : 30.404 8.929 30.701 -> 30.701 -> 491.217 MByte/s -- additional patterns that are not used to compute b_eff: p16 worst-cyc-1dim : 32.181 8.860 32.440 -> 32.440 -> 519.046 MByte/s p17 best bi-section : 31.697 8.044 36.808 -> 36.808 -> 588.924 MByte/s p18 worst bi-section : 31.562 22.389 36.417 -> 36.417 -> 582.672 MByte/s p19 acyclic-1dim-all : 30.674 8.618 30.668 -> 30.674 -> 490.784 MByte/s p20 acyclic-2dim-all : 24.608 15.056 26.084 -> 26.084 -> 417.352 MByte/s p21 acyclic-3dim-all : 27.645 15.634 29.337 -> 29.337 -> 469.390 MByte/s p22 cyclic-1dim-all : 32.397 8.950 32.407 -> 32.407 -> 518.506 MByte/s p23 cyclic-2dim-all : 33.162 14.931 31.896 -> 33.162 -> 530.595 MByte/s p24 cyclic-3dim-all : 35.308 17.751 31.598 -> 35.308 -> 564.923 MByte/s log_avg of all rings : 33.292 9.304 32.798 || 33.444 -> 535.108 MByte/s log_avg of all random : 30.629 8.782 30.868 || 30.904 -> 494.460 MByte/s log_avg(ring,random) : 31.933 9.039 31.818 ||( 32.149 -> 514.383)MByte/s SECTION-BY-METHODS-END SECTION-BY-REPETITIONS-BEGIN 3rd analysis path, only as information, last row: for all rep.:logavg_pat ( avg_L ( max_mthd ( b(L) ) ) ) || logavg_pat ( max_rep ( avg_L ( max_mthd ( b(L) ) ) ) ) pattern / repetition: rep.0 rep.1 rep.2 -> best repetition -> accumulated p00 ring-8*2fix : 37.537 38.665 38.738 -> 38.738 -> 619.813 MByte/s p01 ring-4*4fix : 33.504 34.401 34.250 -> 34.401 -> 550.412 MByte/s p02 ring-2*8fix : 32.822 33.532 33.599 -> 33.599 -> 537.578 MByte/s p03 ring-1*16fix : 32.202 32.534 33.039 -> 33.039 -> 528.631 MByte/s p04 ring-1*16fix : 32.434 32.902 33.016 -> 33.016 -> 528.261 MByte/s p05 ring-1*16fix : 32.898 32.523 32.892 -> 32.898 -> 526.372 MByte/s p06 random-cyc-1dim : 31.125 31.191 31.418 -> 31.418 -> 502.688 MByte/s p07 random-cyc-1dim : 30.906 31.608 31.167 -> 31.608 -> 505.723 MByte/s p08 random-cyc-1dim : 31.143 31.501 31.300 -> 31.501 -> 504.020 MByte/s p09 random-cyc-1dim : 30.618 31.193 30.783 -> 31.193 -> 499.088 MByte/s p10 random-cyc-1dim : 30.500 31.080 30.786 -> 31.080 -> 497.276 MByte/s p11 random-cyc-1dim : 31.283 32.054 31.854 -> 32.054 -> 512.864 MByte/s p12 random-cyc-1dim : 31.348 31.732 31.996 -> 31.996 -> 511.936 MByte/s p13 random-cyc-1dim : 30.697 30.912 30.763 -> 30.912 -> 494.598 MByte/s p14 random-cyc-1dim : 30.976 31.239 31.156 -> 31.239 -> 499.820 MByte/s p15 random-cyc-1dim : 30.494 31.392 31.095 -> 31.392 -> 502.278 MByte/s -- additional patterns that are not used to compute b_eff: p16 worst-cyc-1dim : 32.284 33.270 32.640 -> 33.270 -> 532.317 MByte/s p17 best bi-section : 36.460 36.266 36.346 -> 36.460 -> 583.366 MByte/s p18 worst bi-section : 36.217 36.132 36.184 -> 36.217 -> 579.469 MByte/s p19 acyclic-1dim-all : 31.279 31.752 30.759 -> 31.752 -> 508.036 MByte/s p20 acyclic-2dim-all : 26.394 26.715 26.579 -> 26.715 -> 427.438 MByte/s p21 acyclic-3dim-all : 29.631 29.914 29.952 -> 29.952 -> 479.234 MByte/s p22 cyclic-1dim-all : 32.924 33.036 32.953 -> 33.036 -> 528.582 MByte/s p23 cyclic-2dim-all : 34.239 34.157 34.177 -> 34.239 -> 547.832 MByte/s p24 cyclic-3dim-all : 35.586 35.751 35.654 -> 35.751 -> 572.024 MByte/s log_avg of all rings : 33.519 34.029 34.198 || 34.224 -> 547.580 MByte/s log_avg of all random : 30.908 31.388 31.229 || 31.437 -> 502.998 MByte/s log_avg(ring,random) : 32.187 32.682 32.680 ||( 32.801 -> 524.815)MByte/s SECTION-BY-REPETITIONS-END SECTION-BY-MTHD-MSGLNG-BEGIN 4th analysis path, only as information, last 3 rows: for all methods and some msglng L: logavg_pat ( max_rep ( b(L) ) ) for all methods: logavg_pat ( avg_L ( max_rep ( b(L) ) ) ) pattern & method / msg: 1 16 256 4096 92682 2097152 -> average -> accumulated p00 ring-8*2fix p00 method 0 : 0.015 0.234 2.803 32.204 102.881 107.779 -> 38.631 -> 618.097 MByte/s p00 method 1 : 0.001 0.016 0.248 3.546 21.704 27.176 -> 8.007 -> 128.114 MByte/s p00 method 2 : 0.017 0.262 2.978 33.583 83.727 99.208 -> 34.717 -> 555.465 MByte/s p01 ring-4*4fix p01 method 0 : 0.015 0.233 2.826 32.438 78.494 100.929 -> 32.766 -> 524.248 MByte/s p01 method 1 : 0.002 0.032 0.477 6.364 30.970 35.587 -> 11.365 -> 181.844 MByte/s p01 method 2 : 0.018 0.285 3.166 34.896 83.722 86.826 -> 32.506 -> 520.097 MByte/s p02 ring-2*8fix p02 method 0 : 0.015 0.234 2.819 32.298 81.812 95.715 -> 32.720 -> 523.528 MByte/s p02 method 1 : 0.002 0.032 0.467 6.133 27.267 30.502 -> 10.018 -> 160.289 MByte/s p02 method 2 : 0.018 0.284 3.174 35.047 85.525 86.752 -> 32.646 -> 522.328 MByte/s p03 ring-1*16fix p03 method 0 : 0.015 0.234 2.822 32.313 81.289 88.160 -> 31.912 -> 510.596 MByte/s p03 method 1 : 0.002 0.032 0.456 5.889 24.172 26.693 -> 8.922 -> 142.756 MByte/s p03 method 2 : 0.018 0.283 3.184 34.998 83.277 89.066 -> 32.676 -> 522.810 MByte/s p04 ring-1*16fix p04 method 0 : 0.015 0.234 2.812 32.468 79.994 93.207 -> 32.142 -> 514.279 MByte/s p04 method 1 : 0.002 0.032 0.456 5.887 24.243 26.730 -> 8.931 -> 142.904 MByte/s p04 method 2 : 0.018 0.284 3.183 35.006 82.760 85.403 -> 32.237 -> 515.791 MByte/s p05 ring-1*16fix p05 method 0 : 0.015 0.232 2.810 32.593 80.757 90.109 -> 32.050 -> 512.802 MByte/s p05 method 1 : 0.002 0.032 0.455 5.881 24.235 26.671 -> 8.930 -> 142.873 MByte/s p05 method 2 : 0.018 0.283 3.180 34.939 83.266 84.843 -> 32.076 -> 513.215 MByte/s p06 random-cyc-1dim p06 method 0 : 0.015 0.238 2.873 32.844 77.070 82.393 -> 30.640 -> 490.235 MByte/s p06 method 1 : 0.002 0.032 0.458 5.908 24.352 26.714 -> 8.949 -> 143.191 MByte/s p06 method 2 : 0.018 0.288 3.231 35.180 78.677 79.667 -> 30.932 -> 494.909 MByte/s p07 random-cyc-1dim p07 method 0 : 0.015 0.238 2.894 33.008 78.156 84.481 -> 30.743 -> 491.890 MByte/s p07 method 1 : 0.002 0.032 0.455 5.886 23.939 26.062 -> 8.775 -> 140.398 MByte/s p07 method 2 : 0.018 0.288 3.217 35.145 79.242 80.857 -> 30.737 -> 491.791 MByte/s p08 random-cyc-1dim p08 method 0 : 0.015 0.238 2.881 32.915 77.352 87.918 -> 30.730 -> 491.681 MByte/s p08 method 1 : 0.002 0.032 0.457 5.901 24.011 25.857 -> 8.767 -> 140.278 MByte/s p08 method 2 : 0.018 0.286 3.237 35.201 79.644 81.255 -> 31.293 -> 500.687 MByte/s p09 random-cyc-1dim p09 method 0 : 0.015 0.237 2.878 32.889 75.780 85.708 -> 30.295 -> 484.720 MByte/s p09 method 1 : 0.002 0.032 0.456 5.890 23.986 24.838 -> 8.592 -> 137.471 MByte/s p09 method 2 : 0.018 0.291 3.214 35.033 77.912 80.907 -> 30.467 -> 487.473 MByte/s p10 random-cyc-1dim p10 method 0 : 0.015 0.238 2.873 33.003 75.655 83.881 -> 30.148 -> 482.361 MByte/s p10 method 1 : 0.002 0.032 0.458 5.906 24.135 25.868 -> 8.779 -> 140.469 MByte/s p10 method 2 : 0.018 0.289 3.215 35.128 77.551 80.697 -> 30.996 -> 495.939 MByte/s p11 random-cyc-1dim p11 method 0 : 0.015 0.239 2.890 32.925 76.555 85.845 -> 30.971 -> 495.537 MByte/s p11 method 1 : 0.002 0.032 0.457 5.873 23.972 25.750 -> 8.748 -> 139.972 MByte/s p11 method 2 : 0.018 0.290 3.216 35.160 80.961 81.711 -> 31.452 -> 503.227 MByte/s p12 random-cyc-1dim p12 method 0 : 0.015 0.238 2.865 32.968 78.123 90.338 -> 31.387 -> 502.185 MByte/s p12 method 1 : 0.002 0.031 0.455 5.894 23.990 25.787 -> 8.742 -> 139.869 MByte/s p12 method 2 : 0.018 0.287 3.211 35.084 79.498 80.999 -> 31.037 -> 496.586 MByte/s p13 random-cyc-1dim p13 method 0 : 0.015 0.236 2.813 32.361 76.634 87.576 -> 30.313 -> 485.003 MByte/s p13 method 1 : 0.002 0.032 0.457 5.903 24.243 26.485 -> 8.895 -> 142.316 MByte/s p13 method 2 : 0.018 0.284 3.122 34.590 77.880 79.486 -> 30.303 -> 484.853 MByte/s p14 random-cyc-1dim p14 method 0 : 0.015 0.238 2.883 32.748 77.383 85.504 -> 30.677 -> 490.835 MByte/s p14 method 1 : 0.002 0.032 0.456 5.884 23.953 25.308 -> 8.647 -> 138.355 MByte/s p14 method 2 : 0.018 0.288 3.227 35.204 78.874 83.567 -> 30.776 -> 492.413 MByte/s p15 random-cyc-1dim p15 method 0 : 0.015 0.237 2.829 32.513 76.466 84.189 -> 30.404 -> 486.459 MByte/s p15 method 1 : 0.002 0.032 0.455 5.872 24.339 26.688 -> 8.929 -> 142.868 MByte/s p15 method 2 : 0.018 0.285 3.173 34.784 77.725 78.161 -> 30.701 -> 491.217 MByte/s -- additional patterns that are not used to compute b_eff: p16 worst-cyc-1dim p16 method 0 : 0.015 0.232 2.839 32.561 81.337 91.835 -> 32.181 -> 514.894 MByte/s p16 method 1 : 0.002 0.031 0.448 5.810 24.080 26.607 -> 8.860 -> 141.768 MByte/s p16 method 2 : 0.018 0.283 3.195 34.922 84.342 82.700 -> 32.440 -> 519.046 MByte/s p17 best bi-section p17 method 0 : 0.015 0.235 2.273 26.431 78.327 103.819 -> 31.697 -> 507.154 MByte/s p17 method 1 : 0.001 0.016 0.244 3.482 21.660 27.542 -> 8.044 -> 128.698 MByte/s p17 method 2 : 0.013 0.209 2.477 29.288 99.357 108.966 -> 36.808 -> 588.924 MByte/s p18 worst bi-section p18 method 0 : 0.015 0.233 2.379 26.626 77.623 100.829 -> 31.562 -> 504.988 MByte/s p18 method 1 : 0.001 0.016 0.245 3.785 49.885 100.965 -> 22.389 -> 358.219 MByte/s p18 method 2 : 0.013 0.209 2.578 29.838 96.402 105.682 -> 36.417 -> 582.672 MByte/s p19 acyclic-1dim-all p19 method 0 : 0.014 0.220 2.652 30.565 76.458 88.949 -> 30.674 -> 490.784 MByte/s p19 method 1 : 0.002 0.029 0.429 5.592 23.958 24.114 -> 8.618 -> 137.884 MByte/s p19 method 2 : 0.017 0.270 3.005 32.933 78.987 76.650 -> 30.668 -> 490.694 MByte/s p20 acyclic-2dim-all p20 method 0 : 0.012 0.182 2.136 24.362 59.542 75.392 -> 24.608 -> 393.734 MByte/s p20 method 1 : 0.003 0.047 0.696 9.234 41.317 45.648 -> 15.056 -> 240.903 MByte/s p20 method 2 : 0.014 0.229 2.476 26.834 68.194 63.992 -> 26.084 -> 417.352 MByte/s p21 acyclic-3dim-all p21 method 0 : 0.014 0.214 2.335 26.224 67.892 85.603 -> 27.645 -> 442.328 MByte/s p21 method 1 : 0.003 0.055 0.791 10.173 42.507 47.216 -> 15.634 -> 250.146 MByte/s p21 method 2 : 0.014 0.230 2.605 29.015 76.293 74.658 -> 29.337 -> 469.390 MByte/s p22 cyclic-1dim-all p22 method 0 : 0.015 0.237 2.845 32.632 80.894 90.061 -> 32.397 -> 518.354 MByte/s p22 method 1 : 0.002 0.032 0.456 5.900 24.276 26.660 -> 8.950 -> 143.193 MByte/s p22 method 2 : 0.018 0.289 3.210 35.035 82.848 87.015 -> 32.407 -> 518.506 MByte/s p23 cyclic-2dim-all p23 method 0 : 0.015 0.236 2.839 32.417 80.471 100.367 -> 33.162 -> 530.595 MByte/s p23 method 1 : 0.004 0.063 0.900 11.544 40.264 42.684 -> 14.931 -> 238.893 MByte/s p23 method 2 : 0.019 0.302 3.284 35.722 81.090 81.858 -> 31.896 -> 510.343 MByte/s p24 cyclic-3dim-all p24 method 0 : 0.015 0.237 2.851 32.638 89.029 102.672 -> 35.308 -> 564.923 MByte/s p24 method 1 : 0.004 0.063 0.901 11.722 48.616 52.437 -> 17.751 -> 284.022 MByte/s p24 method 2 : 0.019 0.301 3.292 35.700 81.225 79.766 -> 31.598 -> 505.564 MByte/s log_avg of all rings - ring, method 0 : 0.015 0.233 2.815 32.385 83.826 95.757 || 33.292 -> 532.672 MByte/s - ring, method 1 : 0.002 0.028 0.416 5.518 25.268 28.721 || 9.304 -> 148.867 MByte/s - ring, method 2 : 0.018 0.280 3.143 34.741 83.708 88.555 || 32.798 -> 524.768 MByte/s log_avg of all random - random, method 0 : 0.015 0.238 2.868 32.817 76.913 85.756 || 30.629 -> 490.059 MByte/s - random, method 1 : 0.002 0.032 0.456 5.892 24.091 25.930 || 8.782 -> 140.508 MByte/s - random, method 2 : 0.018 0.288 3.206 35.050 78.790 80.719 || 30.868 -> 493.881 MByte/s log_avg(ring,random) - average, method 0 : 0.015 0.236 2.841 32.600 80.295 90.618 || 31.933 -> 510.921 MByte/s - average, method 1 : 0.002 0.030 0.436 5.702 24.673 27.290 || 9.039 -> 144.627 MByte/s - average, method 2 : 0.018 0.284 3.174 34.895 81.212 84.546 || 31.818 -> 509.090 MByte/s SECTION-BY-MTHD-MSGLNG-END SECTION-BY-MSGLNG-BEGIN 5th analysis path only as information: logavg_pat ( max_mthd ( max_rep ( b(L) ) ) ) and for all methods: logavg_pat ( max_rep ( b(L) ) ) msg length|accumulated| effective bandwidth per process: msg length| effective| average rings random methd_1 methd_2 methd_3 | bandwidth| crt,rnd only only Sendrcv Alltoall non-blk 1 0.287 0.018 0.018 0.018 0.015 0.002 0.018 2 0.571 0.036 0.035 0.036 0.030 0.004 0.036 4 1.138 0.071 0.070 0.072 0.059 0.007 0.071 8 2.280 0.142 0.140 0.144 0.118 0.015 0.142 16 4.542 0.284 0.280 0.288 0.236 0.030 0.284 32 9.033 0.565 0.558 0.571 0.468 0.060 0.565 64 17.827 1.114 1.101 1.128 0.923 0.119 1.114 128 34.766 2.173 2.142 2.204 1.787 0.238 2.173 256 50.791 3.174 3.143 3.206 2.841 0.436 3.174 512 95.503 5.969 5.915 6.023 5.371 0.840 5.969 1024 169.128 10.571 10.478 10.664 9.651 1.560 10.571 2048 351.724 21.983 21.790 22.177 20.108 3.193 21.983 4096 558.325 34.895 34.741 35.050 32.600 5.702 34.895 8933 726.234 45.390 45.589 45.191 44.023 8.799 45.390 19484 799.453 49.966 52.588 47.474 44.119 13.447 49.769 42495 1137.074 71.067 74.251 68.019 65.321 20.125 69.632 92682 1321.889 82.618 86.632 78.790 80.295 24.673 81.212 202141 1401.063 87.566 91.970 83.374 87.519 26.760 84.535 440872 1473.462 92.091 97.160 87.287 92.004 28.209 87.202 961548 1471.002 91.938 97.452 86.735 91.938 28.122 84.569 2097152 1451.129 90.696 95.920 85.756 90.618 27.290 84.546 SECTION-BY-MSGLNG-END SECTION-BY-PATTERN-MSGLNG-BEGIN official (1st) analysis path -- see last column: (only columns of some message lengths are printed) logavg_pat ( avg_L ( max_mthd ( max_rep ( b(L) ) ) ) ) pattern / msg-length: 1 16 256 4096 92682 2097152 -> average -> accumulated p00 ring-8*2fix : 0.017 0.262 2.978 33.583 102.881 107.779 -> 38.864 -> 621.826 MByte/s p01 ring-4*4fix : 0.018 0.285 3.166 34.896 83.722 100.929 -> 34.437 -> 550.990 MByte/s p02 ring-2*8fix : 0.018 0.284 3.174 35.047 85.525 95.715 -> 33.900 -> 542.403 MByte/s p03 ring-1*16fix : 0.018 0.283 3.184 34.998 83.277 89.066 -> 33.066 -> 529.049 MByte/s p04 ring-1*16fix : 0.018 0.284 3.183 35.006 82.760 93.207 -> 33.313 -> 533.006 MByte/s p05 ring-1*16fix : 0.018 0.283 3.180 34.939 83.266 90.109 -> 33.170 -> 530.713 MByte/s p06 random-cyc-1dim : 0.018 0.288 3.231 35.180 78.677 82.393 -> 31.660 -> 506.564 MByte/s p07 random-cyc-1dim : 0.018 0.288 3.217 35.145 79.242 84.481 -> 31.692 -> 507.077 MByte/s p08 random-cyc-1dim : 0.018 0.286 3.237 35.201 79.644 87.918 -> 31.796 -> 508.733 MByte/s p09 random-cyc-1dim : 0.018 0.291 3.214 35.033 77.912 85.708 -> 31.292 -> 500.672 MByte/s p10 random-cyc-1dim : 0.018 0.289 3.215 35.128 77.551 83.881 -> 31.168 -> 498.693 MByte/s p11 random-cyc-1dim : 0.018 0.290 3.216 35.160 80.961 85.845 -> 32.137 -> 514.186 MByte/s p12 random-cyc-1dim : 0.018 0.287 3.211 35.084 79.498 90.338 -> 32.245 -> 515.912 MByte/s p13 random-cyc-1dim : 0.018 0.284 3.122 34.590 77.880 87.576 -> 31.235 -> 499.767 MByte/s p14 random-cyc-1dim : 0.018 0.288 3.227 35.204 78.874 85.504 -> 31.622 -> 505.956 MByte/s p15 random-cyc-1dim : 0.018 0.285 3.173 34.784 77.725 84.189 -> 31.489 -> 503.817 MByte/s -- additional patterns that are not used to compute b_eff: p16 worst-cyc-1dim : 0.018 0.283 3.195 34.922 84.342 91.835 -> 33.352 -> 533.635 MByte/s p17 best bi-section : 0.015 0.235 2.477 29.288 99.357 108.966 -> 36.823 -> 589.172 MByte/s p18 worst bi-section : 0.015 0.233 2.578 29.838 96.402 105.682 -> 36.427 -> 582.834 MByte/s p19 acyclic-1dim-all : 0.017 0.270 3.005 32.933 78.987 88.949 -> 31.798 -> 508.767 MByte/s p20 acyclic-2dim-all : 0.014 0.229 2.476 26.834 68.194 75.392 -> 26.834 -> 429.336 MByte/s p21 acyclic-3dim-all : 0.014 0.230 2.605 29.015 76.293 85.603 -> 30.014 -> 480.217 MByte/s p22 cyclic-1dim-all : 0.018 0.289 3.210 35.035 82.848 90.061 -> 33.422 -> 534.748 MByte/s p23 cyclic-2dim-all : 0.019 0.302 3.284 35.722 81.090 100.367 -> 34.316 -> 549.057 MByte/s p24 cyclic-3dim-all : 0.019 0.301 3.292 35.700 89.029 102.672 -> 35.825 -> 573.207 MByte/s log_avg of all rings : 0.018 0.280 3.143 34.741 86.632 95.920 || 34.402 -> 550.432 MByte/s log_avg of all random : 0.018 0.288 3.206 35.050 78.790 85.756 || 31.632 -> 506.108 MByte/s log_avg(ring,random) : 0.018 0.284 3.174 34.895 82.618 90.696 || 32.988 -> 527.805 MByte/s SECTION-BY-PATTERN-MSGLNG-END SECTION-BEFF-BEGIN The effective bandwidth is b_eff = 527.805 MByte/s on 16 processes ( = 32.988 MByte/s * 16 processes) system parameters : 16 nodes, 256 MB/node system name: HI-UX/MPP hostname : hitachi OS release : 02-03 OS version : 0 machine : SR2201 SECTION-BEFF-END b_eff = 527.805 MB/s = 32.988 * 16 PEs with 256 MB/PE on HI-UX/MPP hitachi 02-03 0 SR2201