Open tjaychen opened 2 years ago
@sriyerg @jdonjdon
let me know if there are others you noticed. I admittedly didn't really look at the scoreboard and focused mostly on the driver / monitor / base_seq
@jdonjdon i think we can postpone this to M3. Because this issue is really about all the things in the i2c agent that need to be cleaned up, but perhaps it cannot be done in time for M2.
Let me know what you think.
@jdonjdon i think we can postpone this to M3. Because this issue is really about all the things in the i2c agent that need to be cleaned up, but perhaps it cannot be done in time for M2.
Let me know what you think.
Agree
actually @jdonjdon were some of these items already addressed? do you want to update the description to be more accurate?
actually @jdonjdon were some of these items already addressed? do you want to update the description to be more accurate?
This issues are only applied to "host mode' rtl test bench. Unfortunately none of these are addressed during recent "target mode" rtl verification.
do you think this is even worth fixing in V3?
I am neutral at this point. Can we revisit after v2?
This is a V3/icebox item really, hence moved to P4 and removed the PROD Candidate label.
This issue attempts to describe all the known issues with the existing
i2c_agent
and how they should be fixed.The
i2c_monitor
directly interprets the line content and creates response items (acknowledgement, read data) instead of just passively observing the line. Thei2c_driver
, instead of just receiving data and driving it onto the pins, has a method of directly creating randomized dataTo address these issues, the following should be done:
base_seq
if possible. The base sequence can then be extended to emulate a memory model for more advanced kinds of testing, in addition to the "random" response it does now. This also means the base sequence should become responsible for creating acknowledgements.