Smcsrind/Sscsrind dependency?

[mark-honman]

AIA and the CSRind extensions were developed at about the same time.

Thus AIA provides a CSrind mechanism that is almost identical to Smcsrind/Sscsrind, and the AIA spec has warnings along the lines of "If Smcsrind is also implemented..." usually to do with errror handling.

So... should ACLIC require Smcsrind/Sscsrind even though AIA does not?

I don't have an opinion on which is the better way of addressing this....

[christian-herber-nxp]

In my mental model, AIA depends on Smcsrind. Technically, that is not the case. So you are right to state that this should be an explicit requirement.

[christian-herber-nxp]

Reopening as this requires an adjustment in the spec.

[mark-honman]

On further thought - SXaclic does require SXcsrind, because the source configuration uses mireg2/mireg3 which are part of Sxcsrind but not AIA. Most AIA references to SXcsrind say something like "if SXcsrind is also implemented then when Xiselect has a value in the range 0x30-0x3F or 0x70-0xFF, attempts to access Xireg2 through Xireg6 raise an illegal instruction exception".

So... proposed spec change would be to do the same for mireg4-mireg6 for the new address range introduced by ACLIC.