4 The AROS Development Team
\r
6 This is an implementation of Michael Sinz's DiskSpped and
\r
7 ScsiSpeed packages. The code has been re written to work on the
\r
8 AROS Amiga-Like environment, with a few enhancements on the original
\r
9 version. It was designed to give the most accurate results of the
\r
10 true disk performance in the system. For this reason many of
\r
11 DiskSpeed's results may look either lower or higher than current disk
\r
14 These key words are also available from the TOOLTYPES of the icon.
\r
16 * DRIVE/K - select drive (Default is current directory)
\r
17 You can use either 'DRIVE <path>' or 'DRIVE=<path>'
\r
19 * COMMENT/K - set comment string
\r
20 You can use either 'COMMENT <comment>' or 'COMMENT=<comment>'
\r
22 * ALL/S - select all tests
\r
23 This turns on all of the tests below
\r
25 * DIR/S - setect DIR tests
\r
26 * SEEK/S - select SEEK tests
\r
28 * CHIP/S - select CHIP memory buffer tests
\r
29 * FAST/S - select FAST memory buffer tests
\r
30 You can select both and DiskSpeed will then do a test pass
\r
31 with each type of memory.
\r
33 * LONG/S - select LONG aligned tests
\r
34 * WORD/S - select WORD aligned tests
\r
35 * BYTE/S - select BYTE aligned tests
\r
36 You can select any combinations of the above. DiskSpeed
\r
37 will do a test pass with each one. These combine with
\r
38 the memory type above to create up to 6 test passes.
\r
40 * BUF1/K/N - select buffer size 1 (Default = 512)
\r
41 * BUF2/K/N - select buffer size 2 (Default = 4096)
\r
42 * BUF3/K/N - select buffer size 3 (Default = 32768)
\r
43 * BUF4/K/N - select buffer size 4 (Default = 262144)
\r
44 Will let you select the buffer sizes for the tests.
\r
45 To eliminate a buffer test, set the buffer to 0.
\r
46 You can use either 'BUF1 <num>' or 'BUF1=<num>'
\r
48 * WINDOW/S - use the GUI even though started from the CLI
\r
50 * MINTIME/K/N - Selects the number of seconds to run each test. (1 to 500)
\r
51 New keyword that lets you select the minimum amount of time any test takes.
\r
52 This applies to all tests except for the Directory entry create and delete
\r
53 tests. Also, note that after the file create speed test, a full 256K file
\r
54 is created and this can, on slow systems take some time.
\r
56 * NOCPU/S - Turns off the CPU available task.
\r
57 New keyword that lets you turn off CPU percentage collection. This is so
\r
58 that the secondary task is not created. Seems that just having this task
\r
59 around can be enough to throw the performance of the system way off. The
\r
60 difference in time it takes to task-switch from STOP to the harddisk driver
\r
61 and from a background task and the harddisk driver is sometimes just enough
\r
62 to cause a rotation on the disk to be missed. This feature may well be
\r
63 removed, but the difference in the numbers is rather interesting. (The
\r
64 background task is at -127 priority...)
\r
66 ******************************************************************************
\r
70 ScsiSpeed was written due to the demand for more details on the raw
\r
71 performance of the drives connected to the system. ScsiSpeed uses
\r
72 low-level device I/O to read the disk starting at block 0 and working up.
\r
73 ScsiSpeed, with a reasonable test time such as 20 seconds, will show the
\r
74 true sustained performance of the drive/interface combination without the
\r
75 overhead of the filesystem and AROS.
\r
77 Basically, the usage of ScsiSpeed is the same as DiskSpeed except for options
\r
78 which do not apply. (Such as DIR and SEEK tests, etc)
\r
80 Also, the device/drive specification is different. You must give the device
\r
81 name (such as scsi.device or trackdisk.device) and the unit number as follows:
\r
83 scsi.device:6 <- This would be unit 6 of scsi.device (default)
\r
84 trackdisk.device:0 <- This would be DF0:
\r
86 Note that due to some controller limitations, only long-aligned read tests
\r
87 are done. Also, ScsiSpeed does not write to the disk and thus will not
\r
88 destroy any data on the disk. This also means that it can test devices
\r
89 such as CD-ROM and WORM disks.
\r