New OCC sensors - performance mode and memory power down

[geissonator]

New OCC sensors being defined.

• Maximum Performance Mode vs Nominal Performance Mode
  -- Just a 0/1 type value, default is 0 (nominal), Maximum is 1
• Memory Power Down settings
  -- Just a 0x00 type value, TBD on what the values mean

[vishwabmc]

TASK LIST

• Write yaml file for Performance Mode

[ ] Write yaml for Power down setting --> Not needed per discussion with Todd

• Implement the above 2

or probably have just one org/open_power/OCC/Setting.inerface.yaml that defines these 2 ( need to talk )

• Make changes in IPMI yaml file to target new D-Bus object on sensor calls

[vishwabmc]

STATUS

• 9/8:
  • Decided to write 2 YAML files and 2 implementations based on the discussion. Have not started writing yaml
• 9/11:
  • Read the Thermal document and saw the below mentioned mode. Will talk to OCC folks on this
• 9/12:
  • Wrote the Performance mode yaml and posted for review
  • Wrote the implementation for perf mode and posted for review
  • Need to talk to Todd to see if the mode applies to all or just the master

[vishwabmc]

I looked at the Power Thermal document and it defines following Modes.

• Nominal
• Nominal dynamic performance
• Maximum dynamic performance
• Dynamic Power - Favor Performance ( not sure what this is )
• Fixed max frequency

Will talk to OCC folks and see what is really needed from this doc.

[vishwabmc]

Discussed with Todd on this and here is what it is :

• Only Nominal and Maximum are needed since the rest are only for PowerVM
• Boolean 0/1 for the Performance Mode
• No need of the Memory Power Down sensor

Todd Rosedahl [3:06 AM] 
Hi.  So I checked around and since Witherspoon will default this to 0x00, there is no need to implement anything here actually.  So that saves you some work.

[3:07] 
We do not need to override it to other values.  And no need for an OpenBMC sensor for this at all```

[mdmillerii]

Using a single boolean to control power / performance trade-offs seems like a tatical decision for witherspoon versus a design for many open power (or even other power9) systems.

@bradbishop @geissonator Agree?

[vishwabmc]

@geissonator please cancel this issue per the email I forwarded last week.

Todd indicated that we do not do any work on this for now.

[geissonator]

Closing out per discussion with Todd