home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Collection of Tools & Utilities
/
Collection of Tools and Utilities.iso
/
comm
/
para13.zip
/
PARASPD.DOC
< prev
next >
Wrap
Text File
|
1993-12-25
|
33KB
|
862 lines
Parallel Port Performance Tester ═══════════════════════════════ User's Manual
╔═════════════════════════════════════════════════════════════════════════╗
║ ║
║ ParaSpd.Exe Parallel Technologies, Inc. ║
║ Parallel Port Information Utility PO Box 7 ║
║ (C) Copyright 1993 Jay Lowe Redmond, WA 98073-0007 ║
║ ║
║ This program may be used Lab 206-869-1136 ║
║ free of charge subject to the Fax 206-869-9767 ║
║ Restricted License Agreement Compuserve 76640,203 ║
║ as stated below Internet 76640.203@compuserve.com ║
║ - or - jaylowe@halcyon.com ║
║ ║
╚═════════════════════════════════════════════════════════════════════════╝
Version 1.00
25 December 1993
Parallel Port Performance Tester ═══════════════════════════════ User's Manual
---------------------------------------------------
Welcome to the Parallel Port Performance Tester
---------------------------------------------------
ParaSpd is a small utility which may be used to test the relative
performance of different types of parallel ports on a given system.
How does it do this?
Paraspd contains typical examples of IO code as commonly used in
parallel port peripherals such as network adapters, disks, tape
drives, and so forth. Examples of 4BIT, 8BIT, and EPP/ECP IO code are
included, which allows you to compare the performance of different
types of parallel ports.
ParaSpd times the execution of the benchmark code samples as
displays the results on one of three forms: Raw Data, Chart, Table.
ParaSpd includes a database of it's results when run on various other
computers, which again allows you to compare your system and ports to
results attainable in other configurations.
ParaSpd is a companion to Parallel, the Parallel Port Information System
and many of the terms used in this document are described more completely
in Parallel.Doc.
Note that ParaSpd does not identify your parallel ports... it simply
reports the relative speeds of different types of ports that might be
used on your PC. For example, ParaSpd produces EPP/ECP results even
if you don't have an enhanced port.
So, use Parallel to determine your PC's parallel port capabilities and
configuration. This will help you understand and apply ParaSpd's results.
If you like, you can send us your system's ParaSpd benchmark results for
inclusion in the ParaSpd database. Just capture ParaSpd's output to a
file and Email it to us via Compuserve or Internet at the above addresses.
PARASPD.DOC ─────────────────────────────────────────────────────────── Page 2
Parallel Port Performance Tester ═══════════════════════════════ User's Manual
---------------------------------------------------
Program Operation
---------------------------------------------------
To run ParaSpd, first disconnect any parallel port peripherals.
ParaSpd only does input operations and only to the specified port, but
it is nevertheless best to disconnect any devices you may have
attached to the parallel port being tested. ParaSpd should cause no
system side effects when used with it's standard parameters, but it
can hang your system if you use it to test ports other than LPT ports.
There is a warning dialog which tells you when this might happen.
Then type ParaSpd at the command line, followed by two parameters:
C:\> ParaSpd 378 Chart
Format for Results
│ │ Raw ... The raw measurements
│ └───────── Chart ... Vertical Bar Chart
│ Table ... Tabular Comparison
│
│ Parallel Port to be Tested
│ 3BC
└──────────── 378
278
other addresses may be tested after a
warning message is confirmed. For example
address 37C is the EPP data port.
C:\> ParaSpd By itself produces a help display
C:\> ParaSpd ? Produces more help
Just as you'd expect, the ParaSpd results are extremely sensitive
to variations in bus timings, since the IO instruction times are the
major factor in transfer rate on most systems. Some PC buses have user
selectable timings ... these yield some interesting results with ParaSpd.
In general, ISA bus PC's will have a maximum of 1000 Kbytes/sec I/O
speed for programmed I/O. EISA and MC buses have considerably better
performance, but that may not translate into ParaSpd results in all
cases. We have not tested any PCI or VLB parallel ports...
PARASPD.DOC ─────────────────────────────────────────────────────────── Page 3
Parallel Port Performance Tester ═══════════════════════════════ User's Manual
---------------------------------------------------
Interpreting ParaSpd's Results
---------------------------------------------------
ParaSpd displays four benchmark results: 4BIT, 8BIT, EPP/ECP, and Total.
The values are in Kilobytes per Second (KBytes/Sec or KBps).
4BIT -- IO speed when performing 4BIT input operations
8BIT -- IO speed when performing 8BIT input or output
EPP/ECP -- IO speed when performing EPP input or output
also applies to programmed I/O using ECP ports
Total -- simply, the sum of 4BIT, 8BIT, and EPP/ECP.
The Table display is sorted in order of descending Total transfer rate,
an entirely arbitrary approach.
ParaSpd's benchmarks reflect the IO programming practices commonly used
in existing commerical parallel port peripherals. Paraspd does not always
test the MAXIMUM transfer rate of the parallel port - rather it tests the
MAXIMUM EFFECTIVE transfer rate that data can travel from a program in
the PC to an external peripheral. For technical junkies, this means that
the test is not just an IO bandwidth test, but a mix of input accesses
and memory references.
So, because ParaSpd is reporting effective transfer rates, it's results
are more useful for predicting the potential performance of peripherals.
Most actual parallel peripherals will have lower transfer rates than the
ParaSpd benchmark, because:
(1) not all parallel peripherals use optimal IO methods
(2) the performance of these devices is often affected by other
system factors (network performance, disk latency, etc).
However, since the operation of most parallel peripherals is "IO bound",
it is often possible to predict the performance of device X on System A
by measuring it's performance on System B and applying the ratio of the
ParaSpd benchmarks on the two systems:
ParaSpd Result System A
Device X on System A = ----------------------- * Device X on System B
ParaSpd Result System B
Remember to compare Apples to Apples (or rather PCs to PCs in this case).
Don't use the fairly arbitrary "Total" performance number, but rather use
the result for the appropriate port type on each system. Also remember
that as the IO performance improves - on faster port types or faster
systems - the device becomes less IO bound and you may begin to see
diminishing returns as other factors begin to