PROCESSOR-SDK-J721S2: SCIClient Clock Config

Hi,

Please refer this document https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-j721s2/09_02_00_05/exports/docs/pdk_j721s2_09_02_00_30/docs/userguide/j721s2/family_cfg/j721s2/index_modules_sciclient_j721s2.html# let us know if this helps.

Regards,

Karthik 

Thank you for the resource! I read through it and while it provided useful information, it poses more questions for me.

So there is always a sciserver running on one of the R5Fs in the MCU domain. If I'm trying to update clock settings on the C7 dsp and I use sci client, then the request is forwarded to the mcu where the server is running? But how does the mcu have access to the MMRs in the MAIN domain? Wouldn't it be easier/simpler to just update the MMR responsible for the clock setting on the DSP that has access to the MAIN domain?

Is sciserver a requirement? Must it always be run on the R5F or does it need to run on any core that will update device config? Would there be any issue in not using SCIClient for device configuration, i.e., just accessing MMRs directly?

I understand its usefulness in acting as a proxy for security related functions, but I'm still confused as to its role for device configuration.

Thanks!

Hi,

Amandio Neves said:

I understand its usefulness in acting as a proxy for security related functions, but I'm still confused as to its role for device configuration.

I'll look into it and get back to you.

Regards,

Karthik

Hi Karthikeyan,

Any updates?

Thanks!

Will you be able to provide an update today?

Thank you!

Can you please let me know what the status of this is?

Thank you!

Hi,

Amandio Neves said:

Can you please let me know what the status of this is?

Sorry for the delayed response,

Amandio Neves said:

Is sciserver a requirement? Must it always be run on the R5F or does it need to run on any core that will update device config? Would there be any issue in not using SCIClient for device configuration, i.e., just accessing MMRs directly?

I understand its usefulness in acting as a proxy for security related functions, but I'm still confused as to its role for device configuration.

Yes ,you are right without sciclient is unsafe and not recommended by TI.

Sciserver which runs on mcu1_0 acts as Device Manager which controls all PM and RM operations similar to how TIFS runs on m4 controls all security operations. This is the software architecture we are using.

Regards,

Karthik

How is mcu1_0 able to access MMRs in the MAIN domain to setup clocks/plls? Why is it necessary to route a message across cores instead of just locally modifying the necessary MMRs?

Hi,

Amandio Neves said:

How is mcu1_0 able to access MMRs in the MAIN domain to setup clocks/plls? Why is it necessary to route a message across cores instead of just locally modifying the necessary MMRs

Can you please refer this documents let us know if this helps.

1.TISCI User Guide - TISCI User Guide

2.https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/08_06_01_03/exports/docs/pdk_jacinto_08_06_01_03/docs/userguide/jacinto/family_cfg/jacinto/index_modules_sciclient_jacinto.html

Regards,

Karthik

Hi,

Thanks for the documentation. I've reviewed them but they all seem to repeat a lot of bullet points. Can you point me to something that answers my previous question:

Amandio Neves said:

How is mcu1_0 able to access MMRs in the MAIN domain to setup clocks/plls? Why is it necessary to route a message across cores instead of just locally modifying the necessary MMRs?

Ultimately I am trying to understand if there will be an issue NOT using SCIClient to control clock settings.

Thanks!

Amandio Neves said:

How is mcu1_0 able to access MMRs in the MAIN domain to setup clocks/plls? Why is it necessary to route a message across cores instead of just locally modifying the necessary MMRs?

The mcu1_0 core can access MAIN domain MMRs and set up clocks and PLLs as necessary. This is the fundamental software architecture of the J7 family of devices.  We do not recommend deviating from this architecture by locally modifying these MMRs.

Thanks.

Understood, thank you. So then why expose the MMRs in the first place?

Understood, thank you for your time!