---------- X-Sun-Data-Type: text X-Sun-Data-Description: text X-Sun-Data-Name: text X-Sun-Content-Lines: 38 Dear John: The Mathematics Department at Purdue is trying to decide how to improve its computational resources, and I thought that we might try to use your streams benchmark to help us make this decision. We have a pair of Sun 670 MPs that we got several years ago very cheaply, and which we upgraded to four 55 Mhz Ross Hypersparc processors about 18 months ago. One option we have is to upgrade again to 125 Mhz processors, which offer reasonably higher SPEC '92 performance, but we were concerned about the memory subsytem and so I ran your stream benchmark, which I'm having a little trouble interpreting. Perhaps you can help us, and perhaps you'd be interested in the new "data", although these machines are pretty old now. I've included the results from the parallel run below. Little else was running on the machine when I ran these tests yesterday (the load was 0.0 before I started). We're running Sun OS 4.1.3_U1, I compiled stream_d.c with 61 gcc -O2 -Wall -mv8 -o stream_d_c stream_d.c -lm The results I've included seem to indicate that running one job on the hypersparcs, one achieves 33% higher streams performance than with the previous Sparc II class processors. Since other CPU performance measures indicate that the underlying processor is at least twice as fast as the previous ones, this would seem to indicate that the bus is saturated. However, I don't know how to interpret the parallel results. Do I just add together the rates for each subproblem to get numbers to compare to the numbers in your paper? One further question: To get accurate results, is it necessary to bring the machine to single user mode, disable all NFS daemons, etc? (For example, my tests on my Sparc 10/41 (in newton.out) yield < 40 Mbytes/second sustained transfer rate, while the results in your table indicate transfer rates of 48 Mbytes/second.) Or are these type of quiescent machine tests useful? Brad Lucier ---------- X-Sun-Data-Type: default X-Sun-Data-Description: default X-Sun-Data-Name: P1.out X-Sun-Content-Lines: 9 Timing calibration ; time = 3489201.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 43.673 367834.031 366362.000 373511.000 Scaling : 40.256 401034.031 397458.000 422575.000 Summing : 43.841 557433.500 547438.000 626977.000 SAXPYing : 42.356 574233.312 566623.000 623785.000 ---------- X-Sun-Data-Type: default X-Sun-Data-Description: default X-Sun-Data-Name: P2.out X-Sun-Content-Lines: 18 Timing calibration ; time = 6309942.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 32.307 502636.344 495246.000 527032.000 Scaling : 29.087 561208.938 550078.000 592672.000 Summing : 29.545 932572.375 812313.000 1508775.000 SAXPYing : 41.805 839788.062 574095.000 923675.000 Timing calibration ; time = 6178729.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 32.744 521861.500 488643.000 618685.000 Scaling : 28.914 574465.312 553369.000 666630.000 Summing : 28.538 850739.625 840982.000 896688.000 SAXPYing : 30.079 837648.500 797889.000 978143.000 ---------- X-Sun-Data-Type: default X-Sun-Data-Description: default X-Sun-Data-Name: P4.out X-Sun-Content-Lines: 36 Timing calibration ; time = 14932151.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 16.073 1817596.000 995483.000 4652936.000 Scaling : 29.007 1116938.750 551596.000 1695155.000 Summing : 21.954 1949647.750 1093219.000 4286635.000 SAXPYing : 22.267 1410062.625 1077806.000 1529206.000 Timing calibration ; time = 15863464.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 15.570 2104903.500 1027642.000 5792541.000 Scaling : 21.665 1083741.750 738524.000 1584219.000 Summing : 22.234 1940417.375 1079446.000 4378564.000 SAXPYing : 16.596 1491695.250 1446119.000 1737388.000 Timing calibration ; time = 16079398.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 16.719 2167839.750 957016.000 6120144.000 Scaling : 24.418 1097300.375 655244.000 1661332.000 Summing : 22.122 1878149.000 1084876.000 4061789.000 SAXPYing : 16.629 1521825.750 1443237.000 1941887.000 Timing calibration ; time = 13071292.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 27.544 1051737.125 580896.000 1382926.000 Scaling : 24.618 992224.125 649941.000 1184233.000 Summing : 16.677 2843083.250 1439134.000 7796645.000 SAXPYing : 15.621 1674183.500 1536430.000 2316219.000 ---------- X-Sun-Data-Type: default X-Sun-Data-Description: default X-Sun-Data-Name: P6.out X-Sun-Content-Lines: 54 Timing calibration ; time = 14597399.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 15.959 1670010.625 1002547.000 2331279.000 Scaling : 15.726 1600371.250 1017451.000 2505315.000 Summing : 15.403 2086875.250 1558106.000 2976800.000 SAXPYing : 16.429 2071930.250 1460791.000 3114843.000 Timing calibration ; time = 23633580.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 15.604 1571975.250 1025405.000 2286747.000 Scaling : 15.503 1441393.125 1032028.000 1754699.000 Summing : 43.019 2534891.250 557889.000 3421722.000 SAXPYing : 37.930 2612828.000 632741.000 4014828.000 Timing calibration ; time = 24784980.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 15.738 1970514.375 1016629.000 3077572.000 Scaling : 15.780 1288116.750 1013933.000 2057700.000 Summing : 15.844 2227711.500 1514801.000 3001760.000 SAXPYing : 16.105 2157945.500 1490266.000 3114408.000 Timing calibration ; time = 18295492.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 14.701 2128765.250 1088351.000 3201056.000 Scaling : 16.161 1711751.750 990059.000 3046855.000 Summing : 18.724 1965844.625 1281758.000 2647557.000 SAXPYing : 12.990 2378481.500 1847630.000 3375782.000 Timing calibration ; time = 24569478.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 16.145 1703813.500 991044.000 3439936.000 Scaling : 15.696 1822379.750 1019371.000 3295343.000 Summing : 16.328 2230204.000 1469897.000 3828358.000 SAXPYing : 14.040 2528162.250 1709350.000 3544786.000 Timing calibration ; time = 11009414.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 13.856 1784978.750 1154768.000 2879661.000 Scaling : 15.170 1768089.375 1054701.000 2743248.000 Summing : 16.073 2735669.250 1493164.000 3785164.000 SAXPYing : 16.388 3045451.500 1464498.000 5546422.000 ---------- X-Sun-Data-Type: default X-Sun-Data-Description: default X-Sun-Data-Name: P8.out X-Sun-Content-Lines: 72 Timing calibration ; time = 12786794.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 10.987 2407757.500 1456255.000 4625817.000 Scaling : 15.301 2429649.250 1045654.000 3833217.000 Summing : 11.349 2842166.750 2114664.000 3483937.000 SAXPYing : 13.564 2912651.000 1769415.000 4030713.000 Timing calibration ; time = 32922990.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 15.040 2085762.750 1063839.000 2991677.000 Scaling : 15.019 2006983.250 1065351.000 3012632.000 Summing : 14.858 2893017.250 1615312.000 3814367.000 SAXPYing : 15.438 2943197.500 1554617.000 4218842.000 Timing calibration ; time = 24466192.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 10.477 2720866.750 1527185.000 3949640.000 Scaling : 15.975 1903383.625 1001585.000 3180432.000 Summing : 11.436 3139212.500 2098669.000 5217259.000 SAXPYing : 12.482 2969305.250 1922772.000 4219873.000 Timing calibration ; time = 26835332.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 15.054 1987658.000 1062816.000 2792654.000 Scaling : 15.581 1957739.500 1026870.000 2895903.000 Summing : 15.749 3826697.000 1523911.000 5932614.000 SAXPYing : 12.759 3335477.750 1880979.000 4560868.000 Timing calibration ; time = 35010288.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 24.439 2321359.250 654689.000 3750231.000 Scaling : 15.611 2153849.000 1024902.000 3638259.000 Summing : 13.263 2979385.750 1809596.000 3827073.000 SAXPYing : 42.200 3307918.250 568722.000 4820795.000 Timing calibration ; time = 22933168.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 11.295 2265903.250 1416536.000 4570460.000 Scaling : 12.916 2813092.500 1238743.000 4143343.000 Summing : 16.331 2769477.750 1469629.000 4226658.000 SAXPYing : 12.343 3181613.750 1944393.000 4512895.000 Timing calibration ; time = 9775972.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 10.718 2548397.000 1492813.000 4402665.000 Scaling : 14.395 2605616.750 1111522.000 4513584.000 Summing : 10.879 3713404.250 2206114.000 6056553.000 SAXPYing : 14.088 3936789.250 1703564.000 6531615.000 Timing calibration ; time = 16601088.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 15.434 2331702.250 1036671.000 3966096.000 Scaling : 13.490 2535274.250 1186066.000 3207983.000 Summing : 10.046 3696889.000 2388982.000 6810296.000 SAXPYing : 16.072 3070401.000 1493278.000 5877023.000 ---------- X-Sun-Data-Type: default X-Sun-Data-Description: default X-Sun-Data-Name: newton.out X-Sun-Content-Lines: 9 Timing calibration ; time = 2565495.000000 usec. Increase the size of the arrays if this is < 300000 and your clock precision is =< 1/100 second. --------------------------------------------------- Function Rate (MB/s) RMS time Min time Max time Assignment: 39.054 417377.938 409690.000 473710.000 Scaling : 37.428 433634.312 427489.000 469830.000 Summing : 39.339 613015.188 610081.000 627881.000 SAXPYing : 38.605 622815.438 621688.000 625769.000