jakob-engblom-i
commented
3 years ago To clarify the point about CCI parameters and the kernel. This is an actual question raised during our implementation of CCI for SystemC inside of Simics.
Should it be allowed for a callback to a CCI parameter change to affect the kernel state - for example by posting events? I.e., what is the limit for side-effects from CCI parameters? Can they be used as a way to make "anything" happen in the simulation, or is there a stricter limitation? Can you use CCI parameters to do things like start a test operation in the simulated system, instead of using an explicit transaction interface for that?
I am fine with either, but we need to explicitly document it so that implementers know what to expect.
Review current version of LRM
Review code comments and code documentation
In addition, review the pdf quality - we have seen problems when searching in the pdf version of the LRM
Consider also updates based on comments from WG meetings e.g. CCI parameters should not be allowed to interfere with the SystemC kernel (as pointed out by Jakob Engblom)
DoD: a change proposal for the LRM, to be implemented by the CCI WG