Li, Aubrey wrote:
> Darrin.Johnson wrote:
>
>   
>> Quick question...since it looks like additional dtrace
>> probe(s) will be
>> added...will it be possible to do "versioning" of powertop
>> such that it
>> does the write things if the new probes are or are not available? This
>> is especially important since powertop is unbundled.
>>
>> Darrin
>>
>>     
>
> That's a good suggestion. we can add it in our TODO list.
>   
I thought powertop was versioned already and that the current version 
was Solaris PowerTop v 1.0. Or is the suggestion to version the kernel's 
support of powertop? It seems to me that this would probably be 
necessary to address Darrin's concern, but it also seems to me that the 
presence/absence of particular dtrace probes would be informative in itself?

As for the support for "Turbo Mode" average frequency - if I understand 
the request correctly, we're considering whether to report an average C0 
frequency for P0 when in "Turbo Mode", correct? That might be doable. I 
went back and read the Intel System Programmer's Guide description of 
APERF/MPERF. Apparently, those counters only increment when in the C0 
state (I was under the misconception that they incremented even when not 
in C0). So, if both APERF and MPERF are only incrementing while the 
processors are in the C0 state, then we don't need to take C1E into 
account. However, there is a bullet in the Programmer's Guide that says 
"Only the IA32_APERF/IA32_MPERF ratio is architecturally defined; 
software should not attach meaning to the content of the individual of 
IA32_APERF or IA32_MPERF MSRs." So, can we reliably determine the 
operating frequency using the two registers? My guess is yes, that the 
ratio is all that we care about.

Mark

Reply via email to