blob: 35a9f28cb9e75a9941252b45fbe8e54acb2ef437 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
|
* there is a filedescriptor leak!
* move the 2 data files somewhere sane
* check various tunables (see powertop 1.x) into a tunable checklist
* pretty print estimate to only do 2 decimal places instead of 3
* inode dirty trace point
* have a "show me my process" page
* fix memory leaks
* find out why CPU %age and C state C0 %age don't mesh sometimes
* device events/sec into the process view
* USB urbs as events/sec counter
* investigate tick_sched_timer and why it wakes up
* find if we can do more detailed reporting about tasklet(softirq)
* deal with tasklets -- needs kernel patch
* audio calibration? Need appropriate sample
* "report" mode
* CSV-output mode
* delta mode (Tony)
done
----
* on NHM, cpu P state stats accumulate and don't reset at measurment time
* rescan for USB devices each measurement
* brightness calibration
* rfkill calibration
* wakeup calibration
* disk calibration
* in learning, find the biggest delta and only adjust that parameter, not
all of them as they go
* in learning, try slightly lower base power
* display device power in power top 20
* find out why CPU power and sum of components doesn't add up
* exclude xorg/etc from wakeup blame backcounting
* GPU ops in interrupt context -> do not account special ?
* try to see if we can account graphics ops done by X to the guy who woke up
X
* don't save measurements < 5 seconds
* auto save every 10 or so measurements
* between 500 - 1000 measurements, start replacing (at random) existing ones
with a hard cap at 1000
* network interfaces as device
* network interfaces -- packet counters as events
from /proc/net/dev
* identify kernel threads in the process list for visual differentiation
* deal with the "no measurements" case better with default values
* add device runtime PM stats for pci, spi and i2c
* looks like events/sec keep adding up for processes
* Readable PCI names for runtime PM pci devices
* debug why wakeups sometimes go to all 0's
-looks like I fixed it, keep an eye on it
* pretty print estimates to only show 2 significant places cleanly
* pretty print various kernel pieces
* deal with debugfs not being in /sys/kernel/debug
* if you plug in the AC during a run, the battery code is not detecting that
* status bar at the bottom
* health report (reporting mode, not interactive) needs to include devices that don't support runtime PM
* for unknown USB devices, list VID/PID
|