Hi Aubrey,
On Dec 15, 2007, at 12:01 AM, Li, Aubrey wrote:
> Hi Pat,
>
> Please post the plain text format thread on the mailing list.
Apologies as I was using my MacBook and didn't realize the default
isn't plain text. Hopefully, this is better.
> Pat Bredenberg wrote:
>> Hi folks,
>
>> I took a few minutes to just play with powertop and had a couple
> of comments/suggestions for general consumption.
>> I conferred briefly with Eric ahead of this email but we figured this
> would be good to pick up some other ideas if they exist.
>
>> First, I noticed that the "(frequencies)" string in the header
> bleeds over from the column with which it is associated
>> to the interval column. This is confusing as it looks like the
> interval column's header is actually "frequencies." I suggest
>> either omit "frequencies" altogether or significantly abbreviate
>> it to
> avoid this confusion.
>
> Thanks to point this style out.
> How about change
>
> P-states (frequencies)
> 2000MHz 100%
>
> To
>
> P-states (frequencies)
> 2000MHz 100%
>
> ?
I'm not sure I see the difference between those two, but if the 100%
correlates to CPU utilization, I think that'd be great. Is
"frequencies" necessary? Kindly forgive my ignorance if it's running
rampant. From the new-user standpoint, if it extends beyond the
column with which it's associated, deciphering the output can be
arduous. I'm thinking of vmstat(1) and mpstat(1) when the elements
in their sundry columns begin to reach into neighboring columns,
reading the output is nearly impossible.
>
>
>> Further, it might not be a bad idea
>> to put "interval" or something more appropriate over the interval
> column.
>
> Good suggestion, let's try to find if something useful can be added,
> like cpu utilization...
>
>> Second, would it be worthwhile to put some kind of vertical
> divider between the c-state and p-state information at the
>> top of the powertop output? I believe it would be because the two
> concepts are distinct and the lack of a divider in the
>> output can be misleading and cause the user to incorrectly assume
>> they
> are related. I put myself in the that category
>> because I'm still in the early stages of learning this stuff. Of
> course, like I told Eric, we don't let people do brain surgery
>> without going to medical school, so maybe the second point is
>> overkill.
>
> Actually C-state and P-state information are related.
> If the processor keeps in C0, the frequency should be at the highest
> speed.
> So here, I don't think some kind of vertical divider is necessary.
OK, I think that just means I need to read up more and become more
familiar with how c-states and p-states coexist. If a separation of
the two isn't appropriate, we'll just nix that idea altogether.
Thanks much,
Pat
>
> Thanks,
> -Aubrey
