Overview
This is the central location for all things PowerTOP testing-related.
Status
- 06-11-08 - Bugs/issues with PowerTOP are still being discovered.
- 06-11-08 - A preliminary test plan for PowerTOP has been developed but is in its infancy and will be changed in the weeks to come.
- 06-11-08 - An initial attempt at assertions is underway and a non-exhaustive list should be presented soon.
Test Plan
Section I - Introduction
- ===== Author(s):
- ===== Revision History|=Date (yyyy-mm-dd)|=Revision|=Comments|=Status
| 2008-03-06 | 0.1 | Initial draft | Complete |
| 2008-06-11 | 0.2 | Revised in light of newer features to PowerTOP and add references to manual testing. | Complete |
| 2008-06-23 | 0.3 | Fixed typos, add dependency information, clarify term definitions. | Complete |
- ===== References
- ===== Glossary
P-state - A level of frequency in which the processor operates, which increments and decrements depending upon demand. Specific to x86.
residency - A percentage of time how long a processor stays in a given C-state.
wakeup - The action of a CPU being brought out of a C-state
Section 2 - Operational Factors
- ===== Project Dependencies
cyclic-{start,end} and cpu-change-speed probes
DTrace(7D) user privileges
cpu_info kstat(3KSTAT) module - ===== Risks
- ===== Assumptions
Section 3 - Test Program Operational Plan
- ===== Functional Testing
- ===== Stress Testing
- ===== Performance Regression Testing
- ===== Performance Benefit Resting
- ===== Functional Regression Testing
- ===== Test Development|=Test Name|=Automated/Manual|=Test Type|=Project Component|=Description
| ptop_test | Automated | Functional | Command Line Interface | This test suite shall verify the functionality of the various (combined with -d) options PowerTOP utilizes and that the output is correct. |
| ptop_manual | Manual | Functional/Stress | Command Line Interface | Without the -d option, PowerTOP will display information that updates every seconds, so this will need to be monitored manually. Also, some behavior (i.e. multiple instances of PowerTOP) is better tested manually to be efficient. |
Section 5 - Areas of Testing
- ===== Functional Testing
- ===== Regression Testing
STC2 cpupm - ===== Conformance Testing
- ===== Stress/Robustness Testing
psrset stress
togproc (online/offline) stress - ===== Performance Testing
- ===== Zones Testing
- ===== Interoperability Testing
- ===== Documentation Testing
Section 6 - Test Execution
- ===== Hardware Test Configurations
- ===== Software Test Configurations
- ===== Pre-Integration (PIT) Run
- ===== Test Execution Matrix
Observations
- 01-29-08 - There is no documentation for PowerTOP save a sparse man page available off of lesswatts.org. Resolved
- 01-29-08 - Executing PowerTOP with -d and -time results in no output under "Top causes for wakeups" and skewed statistics. Resolved 03-05-08, usage error
- 01-29-08 - PowerTOP "works" with inappropriate options such as -1, -1d, -1l, *, ]. Resolved
# powertop -1 -d
powertop: illegal option ~-- 1
Solaris PowerTOP 1.0 (C) 2007 Intel Corporation
Collecting data for 5 seconds
Cn Avg residency
C0 (cpu running) (48.4%)
C1 2.4ms (51.6%)
P-states (frequencies)
2191 Mhz 100.0%
Wakeups-from-idle per second: 440.4 interval: 5.0s
Top causes for wakeups:
22.8% (100.2) :unix`cbe_hres_tick
.
.
- Updated 03-05-08. Resolved
- 01-29-08 - PowerTOP accepts large values but doesn't really do much and wraps. Resolved
# ./powertop -t 100000000000000
Solaris PowerTOP 1.0 (C) 2007 Intel Corporation
Collecting data for -2147483648 seconds
- 01-29-08 - zero or characters used as sub-options to -t or --time causes PowerTOP to rapidly spew illegible data. Resolved
- 03-05-08 - pressing the "P" key to enable power management seems to do nothing (may be expected at this time). Resolved
- 03-05-08 - pressing any key during powertop activity causes interval time to shrink and doesn't return to time specified by user or the default time interval. Resolved
- 05-22-08 - PowerTop works with options such as "2", "patb", "l", "-", and so on. It doesn't appear to do anything with them.
- 05-22-08 - any keystroke, with the exception of 'q', causes PowerTOP to refresh.
- 05-22-08 - "Successfull" should be "Successful" in the man page under "EXIT STATUS."
- 05-22-08 - The man page indicates that not all wakeups are listed by PowerTOP but doesn't specify which wakeups are being displayed. Is there a maximum number?
- 05-23-08 - PowerTOP executed with no options should use 5 seconds for a time interval. It does for the first iteration, but then uses 8 seconds for each iteration thereafter.
- 05-23-08 - Multiple PowerTOP instances experience time intervals fluctuating between < 1 sec. and 11-16 seconds.
