Digital Equipment DECmpp 12000/Sx 100, Руководство по обслуживанию оборудования

Страница: 99 / 156

pe_macro(1)
5. Router instruction tests
•
rt0a
— 16-bit router send test
•
rt0b
— 16-bit router fetch test
• rt0c — Router send test (1, 8, 16, 32, 64 bits
• rt0d — Router fetch test (1, 8, 16, 32, 64 bits)
• rt0e — Router send and fetch test
•
rt0f
— Router send and fetch (while storing) test
•
rt0g
— Router open/send and fetch/close test
•
rt1a
— Single PE per cluster router send & fetch test
•
rt2a
— All PEs send/fetch to self
•
rt2b
— All PEs r0send/rfetchc to self
6. Full backplane router instruction tests
•
rt2c
— Forward router send test
• rt2d — Forward router fetch test
• rt2e — Reverse router send test
• rt2f — Reverse router fetch test
•
rt2g
— All PEs send/fetch to self (M-machine busy)
Basic Test Strategy
Each of the following tests employ the same structure and control method.
Since it is nearly impossible for a short error message to convey the full context
including the events preceding the error, it is important that you be able to trace
this back using the program listings.
The diagnostic consists of a common front-end program (HDB script) which
downloads and starts the assembly language program (macro program), which
runs on the array control unit (ACU). The macro program performs various
operations and returns results to the front end program using the FRBEQ. The
front-end program has access to a golden file which contains all the correct values
to be returned using the FRBEQ.
The golden file also contains an error message with each value and when the
actual value does not compare, the front-end program prints the error message as
well as the error message number.
If you want to delve more deeply into the matter, refer to the macro program
listing. This listing contains a comment at each statement sending data down the
FRBEQ. These comments are numbered, and the number corresponds to the error
message number. Knowing this, you can quickly locate the place in the macro
program where the error occurred.
Sometimes when testing a faulty system, a FRBEQ timeout is a common problem.
The front-end program loads and starts the macro program, then waits for data
to appear on the FRBEQ. If none appears after a reasonable length of time, the
front-end program aborts the test. However, it also reports the contents of the
Halt Code Status Register, which tells why the macro program halted. This can
give a clue to the difficulty.
Data Parallel Unit Reference Pages B–15