[i2c/dv] issues with i2c_agent

tjaychen commented 2 years ago

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. The i2c_driver, instead of just receiving data and driving it onto the pins, has a method of directly creating randomized data

To address these issues, the following should be done:

[ ] the "interpretation" done by the monitor should be passed onto the 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.
[ ] the logic inside the i2c driver that creates return data should be removed
[ ] There are many req_analysis_fifo's inside i2c agent, and those probably need to be consolidated (need to look at the interaction between monitor and scoreboard).

tjaychen commented 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

tjaychen commented 1 year ago

@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 commented 1 year ago

@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

tjaychen commented 1 year ago

actually @jdonjdon were some of these items already addressed? do you want to update the description to be more accurate?

jdonjdon commented 1 year ago

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.

tjaychen commented 1 year ago

do you think this is even worth fixing in V3?

jdonjdon commented 1 year ago

I am neutral at this point. Can we revisit after v2?

hcallahan-lowrisc commented 3 months ago

This is a V3/icebox item really, hence moved to P4 and removed the PROD Candidate label.

lowRISC / opentitan

[i2c/dv] issues with i2c_agent #14825