CreateTable NOTHEADS Threads RealTime CPUTIME RealTime CPUTIME Realtime CPUTIME 100 * 100 * 100 Base Engine 0: 02.54
0: 01.48 0: 02.06
0: 02.06 0: 01.09
0: 03.39 SPD ENGINE 0: 05.29
0: 00.87 0: 01.98
0: 1.98 0: 01.81
0: 03.43 100 * 100 * 500 Base Engine 0: 16.75
0: 08.36 0: 09.56
0: 09.56 0: 04.65
0: 15.98 SPD ENGINE 0: 23.81
0: 04.35 0: 09.10
0: 09.07 0: 04.85
0: 16.11 100 * 100 * 2000 Base Engine 0: 54.93
0: 31.73 0: 53.90
0: 37.65 0: 37.53
1: 09.43 SPD ENGINE 1: 37.51
0: 20.03 1: 09.51
0: 35.60 0: 33.23
1: 09.17
in conclusion:
1. Threads effect is obvious;
2. The SPD Engine target is not as good as BaseEngine;
3. Under normal SCSI / IDE storage, SPDENGINE is no effect or effect is not obvious;
Note: The following Summary is equivalent to calculating 21 statistics in SQL:
1 (count) 4 (SUM) 4 (MIN) 4 4 (Mean) 4 (STD)
LibName SPD SPDE 'E: /' DataPath = ('E: /' 'f: /' 'g: /') Partsize = 32;
/ ************* SPD Engine *************** /
Options cpuCount = 4;
Data SPD.SPDTEST;
DO i = 1 to 100; do j = 1 to 100; do k = 1 to 2000;
x = sin (i); y = x; z = x;
Output; end; end;
Options threads;
Proc Summary Data = SPD.SPDTEST NWAY MISSING;
Class I j; var x y z k; output out = o; RUN;
Options Nothreads;
Proc Summary Data = SPD.SPDTEST NWAY MISSING;
Class I j; var x y z k; output out = o; RUN;
/ ************* BASE engine *************** /
Data SPDTEST;
DO i = 1 to 100; do j = 1 to 100; do k = 1 to 2000;
x = sin (i); y = x; z = x;
Output; end; end;
Options threads;
Proc Summary Data = SPDTest Nway Missing;
Class I j; var x y z k; output out = o; RUN;
Options Nothreads;
Proc Summary Data = SPDTest Nway Missing;
Class I j; var x y z k; output out = o; RUN;
