The CPU Expert Tool provides sophisticated troubleshooting for
problems associated with a specific
Processor in a Multiprocessor environment. The ability to
deactivate, reactivate, and assign a process to a processor allows
on-line control of processor functionality.
The CPU Expert Tool is fully interactive, with its own set of
menu commands.
Caution
The CPU Expert Tool contains commands which will cause severe performance degradation as a result of processor deallocation,it is assumed that users are aware of the dangers of its utilization.
To use the CPU Expert Tool safely, you should be trained and knowledgeable about CPU modules.
The CPU Expert Tool allows on-line control of processor functionality in a multi-processor environment by issuing processor specific commands.
This product provides individual operations that enable the user to identify a malfunctioning processor. The goal of the Expert Tool is to provide the user with an interactive tool that allows processor deactivating or reactivating of individual processor modules without requiring reset of the SPU or restart of the operating system. The ability to assign a specific process to an individual processor is an another capability of this product.
This tool should allow the multiprocessor system under test to be reduced in scope via Programmatic elimination of one or more modules. Thus, problems caused by a CPU module or specific process can be isolated without off-line testing.
The CPU Expert Tool requires that you install a HP-Only license at the main STM window. Consult a Mesa Adminstrator for the password.
If you experience problems, consult the Troubleshooting Tool-Related Problems help topic.
To run the CPU Expert Tool on a device:
- At the main STM window, select a CPU device.
The CPU Expert Tool will access all CPU devices on the Unit Under Test.
- If desired, change the options for the CPU Expert Tool at the main STM window.
Options include the level of logging for the tool.
For more information, see online help for STM as a whole
(use the Help menu in the STM main window).
- Run the CPU Expert Tool.
- To select the desired commands, use the pull-down menus in the main window of the tool.
For more information, see:
- To run any option commands, pull down Options Menu and make a selection.
- To run exercise commands, pull down CPU_exercise Menu and make a selection.
- If you run into problems, see the Troubleshooting Tool-Related Problems help topic.
- Exit the tool by selecting the Exit command from the File function keys.
To troubleshoot a suspect processor:
- Try running a specific process (identified by a PID Number) on a Processor to identify Process specific Problems.
- Deactivate all processors that may not have any problems, and assign a PID to activated processor that is suspected to be malfunctioning.
- Continue Deactivating and Reactivating Processors untill a malfunctioning processor is identified.
- If you get confused on which processors are diabled/enabled run the Display command, which will indicate the status of all Processors.
If you run into problems, see the Troubleshooting Tool-Related Problems help topic.
A failure or other problem with the CPU Expert Tool:
Consult the Failure Log (for failures) and the Test Activity Log, accessible from the STM main window.
You start a CPU Expert Tool command but nothing seems to happen :
The CPU Expert Expert Tool may be waiting for you to respond to a query message it is displaying.
Look underneath the windows on your workstation for a query message.
Also, de-iconize any windows related to stm or the expert tool.
When you try to run the CPU Expert Tool from the main STM interface, you do not see it listed:
The CPU Expert Tool requires a license to operate. Install the license at the main STM screen and try to run the tool again.
The function Exercise applied on a specified cpu to run stress tests on data cache, data tlb and float point unit in turn.
Data cache exerciser
exercises the data cache by forcing misses, replacements and flushes.
Data TLB exerciser
exercises the data TLB by forcing page faults which forces TLB replacements.
Floating point exerciser
exercises the floating point unit by providing a series of known inputs and
checks for corresponding outputs. The benchmark algorithm used here is "tomcatv".
The parameter Execution Time Limit can be specified by user as option for limit the cpu exercise execution time in minutes.
The default value is 10 minutes.
The parameter Error Allowed Number can be specified by user as option for limit the allowed number of errors before the cpu exercise test termination.
The default value is 10.
The CPU Expert Tool offers the following commands and Menu:
- File Menu
Save As
Exit
- Option Menu
Active
Assign
Deactive
Display
- CPU-Exercise
Run
Termination
- Help Menu
General Help
Version
- Buttons
Abort Button
Interrupt Button
Save As
The Save As command in the File pulldown menu saves the contents of tool's main window. ( Only the most recent 200 lines can be saved.)
The Print command in the the File pulldown menu prints the contents of tool's main window. ( Only the most recent 200 lines can be printed. )
Exit
Use the Exit command to terminate the execution of the CPU Expert Tool and return to the main STM window. If the user tries to exit without reactivating deactivated processor(s), a warning message will appear. The message will indicate that all processors have not been reactivated, and reactivation of all processors will be done upon exiting.
Activate
The Activate command will only return a processor that has been previously Deactivated to normal system use. A deactivated processor that is also marked for deconfiguration may not be reactivated until it is marked for reconfiguration. Will return error message if activation routine does not return Successful or wrong pdev # was selected. The Activate command will display in the main window the SPU and Physical Device number along with the current number of activated processors after execution.
Assign
Forces(and locks) a process to a specific processor module identified by the physical device number. PID and pdev number arguments are required. If a wrong physical device number in entered, an error message will display. The user is required to find an active PID numbers with a different application(top, ps ef).
Deactivate
Deactivates a processor from system use. A pdev number argument is required. The Expert Tool will display a warning message if there is only one processor currently activated or on any attempt to deaotivate the Monarch. The Deactivate command will display in the main window the SPU and Physical Device number along with the current number of activated processors after execution.
DisplayThe Display command will display a listing of the current processor status and identify information for all system processors:
SPU Number:
Physical Device Number:
Status: (Enabled/Disabled)
Monarch: (Yes/No)
Run
The Run command will launch Cpu Exercise for a selected cpu. On the popup dialog box, the user can modify parameters of the exerciser such as Execution Time Limit and Error Allowed Number. By design, only one cpu exerciser can run at any time.
To monitor the running processes, the user may use the Top utility to obtain execution information on the running process.
The running record can be found on cpu expert activity log.
Termination
The Termination command terminates running the exerciser as user desires to stop the exercise earlier than the Cpu Exercise. The exercise termination message will be indicated on main window of this expert tool.
General Help
The "Generalhelp" command displays online help for the CPU Expert Tool (the help system you are currently viewing). For online help on STM as a whole, use the "Help" command at the STM main window.
Version
The "Version" command displays version and copyright information for the CPU Expert Tool.
Abort Button
The Abort button allows you to abort the CPU Expert Tool, halting its execution and returning to the STM main window.
The tool will be requested to abort and the state of the tool execution in the system map will be set to ABORT_PENDING. When the tool actually exits, its state will be set to ABORTED.
Interrupt Button
The Interrupt button allows you to interrupt the tool. When the tool receives the interrupt, it will stop whatever it is doing and prompt the user whether they want to continue with the current peration. If the choice is "Continue", the tool will continue with the current operation. If the choice is "Abort", the tool will abort the current operation and return to the window for the CPU Expert Tool.
Control-C (Abort or Interrupt)
Pressing Control-C (CNTL-C) allows you to interrupt or abort execution of tool, or to escape to the user interface. After you press Control-C, the following function keys are displayed:
ESCAPE TO UI
The ESCAPE TO UI function key allows you to disconnect from the currently executing tool and return to the main User Interface (UI) command level. To reconnect to the tool, select the Attach to Tool menu item in the Tools->Tool Management pulldown menu.INTERRUPT
The INTERRUPT function key allows you to interrupt the tool.
When the tool receives the interrupt, it stops whatever it is doing and asks you if you want to continue with the operation. If the reply is "Continue", the tool resumes what it was doing at the time of the interrupt. If the reply is "Abort", it terminates the operation and returns to the CPU Expert Tool main menu.
The tool may not immediately receive the interrupt as it could be in the middle of an operation that should not or could not be interrupted. The interrupt is received as soon as the tool completes the operation.
The tool will receive and ignore user interrupts when it is already in the process of returning to the main function keys of the tool.ABORT
The ABORT function key allows you to abort the tool.
The tool will be requested to abort and the state of the tool execution in the system map will be set to ABORT_PENDING. When the tool actually exits, its state will be set to ABORTED.CONTINUE
The CONTINUE function key allows you to continue with tool execution. The tool will continue execution just as if the Control-C keys had not been pressed.