TESTCOM Performance Tester Version 1.2 (c) 1995 Telesys GmbH, CH-8306 Brttisellen This is a short manual. Please read the file README.TXT, to see, what TESTCOM does in general. FEATURES: - tests serial link throughput at up to 30'000 cps or more in full duplex mode on a 486DX/50 - up to 115'200 Baud (460'400 Baud with emulation card) - COM1 to COM4 - IRQ2 to IRQ7 can be specified individually - Fully bufferd and interrupt driven receive and transmit - Full 16550 support from assembler level - Maximize mode to test maximum speed - Minimize mode to test optimum performance/speed INSTALLATION: If you have received TESTCOM in an archive, such as a ZIP file, place the contents of the archive in a new directory. If you have received TESTCOM on a floppy directly from Telesys GmbH, simply copy the files into a new directory. COMMAND LINE REFERENCE: Format: TESTCOM [COMx [IRQx []]] You can not use the max.cps parameter, without using the COMx and IRQx parameter too. : 300 - 115200 (effective upto 460kBaud) [COMx]: COM1 - COM4, Default: COM3 [IRQx]: IRQ2 - IRQ7 (not PCI-bus), Default depends on COMx []: 0 - 50000 cps, Default 24000cps Examples: TESTCOM (shows help) TESTCOM 19200 (tests at COM3 IRQ4) TESTCOM 115200 COM1 IRQ4 (maximize up to 24'000cps) TESTCOM 9600 COM4 IRQ7 500 (maximize up to 500cps) TESTCOM 115200 COM2 IRQ : You must make sure yourself, that the Baudrate is not too fast for your system, such as when running from Windows or OS/2 as a DOS task with limited CPU capacity. WARNING! THROUGHPUT TESTS AT PERFORMANCE LIMITS CAN CAUSE OPEN FILES TO GO LOST, WHEN RUN FROM BEHIND NON TRUE MULTITASKING ENVIRONEMENTS. IF YOU DON'T KNOW WHICH FILE IS LOST, THE WHOLE APPLICATION RUNNING IN PARALLEL OR EVEN THE OPERATING SYSTEM MAY CRASH. THEREFORE YOU WILL HAVE TO USE THE TOOL CAREFULLY. IN GENERAL YOU ARE SAVE, IF YOU NEVER DO BACKGROUND TASK TESTING ON SYSTEMS, WHICH CAN NOT BE FULLY RESTORED FROM BACKUP TAPES IN AN ACCEPTABLE WAY. COM port emulation cards, such as the Telelink TUSSI, can multiply and fully ignore the Baud rate indicated in this parameter. COMx: You have to know, on which port you are testing. If you have an old serial port with an UART 8250, you can not use baud rates above 9600 baud. IRQx: You will also need to know, on which hardware interrupt your com port is operating, if it is non standard. Standard IRQs for COM1 to COM4 are: COM1 : IRQ4 COM2 : IRQ3 COM3 : IRQ4 COM4 : IRQ3 : This parameter is only needed, if you plan to test throughputs above 24'000 cps, or if you need to optimize speeds in minimize mode. See the paragraph on TEST MODES later in this chapter. Another good reason is to do decent testing, by starting at low upper boudery limits. This however will not prevent you from crashing overloaded systems, because the parameter is adaptive and does not show his effect immediately. Use the Baud parameter to limit speeds savely. INTERNAL OPERATION: TESTCOM sends one byte after the other by simply counting from 0 to 255. The reception of continuous counting bytes is used to evaluate the characters per second (cps) received and the errors. TESTCOM is a fully ISA-interrupt driven DOS task. It can measure the top level performance. The reason for the actual bottleneck however can have many possible sources and will need many tests with different environements. TEST MODES: TESTCOM can run in two modes. In maximize mode to test for maximum performance. And in minimize mode to test for optimum error free performance. Maximize mode is the normal mode and needs no special command line parameter. In maximize mode (max.cps > 0) the CPU delay is decreased as long as max.cps is not reached. This measures the performance of serial channels in optimal system environements. In minimize mode (max.cps = 0) a smooth CPU delay is encreased as long as there are errors. This measures the possible performance in a practically given system environement. You must specify all parameters to use minimize mode (see command line syntax). TEST RESULTS: TESTCOM reports the type of serial UART chip found. It will distinguish between bufferd UART (16550) and non buffered UART (8250 or 16450). The cps values in this test is "full duplex". You might find, your serial port or modem to support two times as much, but this is only simplex, such as in a ZMODEM file transfer. You will also see any errors, that occur. The throughput and the errors are reported for each second separately. Therefore you might find your cps rate to toggle, which is a buffer effect of regular reporting and not affecting measurement by heavy calculating and screen printing. The final report reflects the whole measurement. SERIAL PORTS: TESTCOM needs to directly operate on a true hardware compatible COM port. This can be any UART 16552, 16550, 16450, 8250 on a serial ISA or PCMCIA bus card, modem card or on serial ISDN adapter cards or on any equipement emulating such serial ports, such as the Telelink TUSSI serial card for speeds up to 460'400 Baud. Software interface com port emulation such as fossil, EBIOS or ISDN-cFOS drivers are currently not supported. Much of the power of this tool is in operating correctly together with the hardware at high speeds, and using interrupts correctly, therefore it would loose a lot of its capabilities, when running on software interfaces. CABLING AND FLOW CONTROL: The standard method of RTS/CTS flow control is used. TESTCOM only needs the RS232 signal lines TxD, RxD, RTS, CTS and Ground and is therefore compatible with the situation given in normal highspeed communications. You can't use a normal link cable to test two PC's directly, because Link Software doesn't normally use RTS/CTS flow control. Therefore a Null Modem Cable will most likely not be good. You can order TESTCOM PACK with a link cable and a full set of adapters, if you want to make sure, to test properly. TEST ENVIRONEMENTS: 1) PC --- RS232-Loop-Connector (Rx-Tx und RTS-CTS) 2) PC --- Link-Cabel --- PC 3) PC --- Modem- or RS232-ISDN-link --- PC 4) PC --- Modem- or RS232-ISDN-link --- Loop-Connector 1) Test the serial card. If the CPU is slower, the CPU is tested. 2) Test a given system environement, before a communications link is added. Can also test half duplex by clamping RTS on one side only. 3) Test a communications link in duplex or half duplex mode. 4) Test any remote communications link in duplex mode.