Expose force/release to cocotb

[thasti]

I finally got around to looking at #657. I have mostly re-applied the patch prepared by @lukedarnell to current master, fixed small issues with it and gave it a test.

As it stands, a testbench running on a VPI simulator can do:

from cocotb.handle import Force, Release, Deposit
...
dut.q <= 0 # default action: deposit
dut.q <= Deposit(0) # does the same as above
dut.q <= Force(1) # forces the net to 1
dut.q <= Release() # releases the net

What's still to be done:

• add documentation
• implement the mechanisms also for VHPI/FLI
• add a proper unit test/example

I'd already like to hear whether there are some thoughts on how it was implemented to figure out if any significant changes would be required in your opinion.

[imphil]

This PR has a large number of conflicts. Can you try to rebase on top of the current master?

[thasti]

Indeed, I was a bit behind current master. Should be better now :)

[imphil]

Thanks for the rebase, much easier to read now. I'm fine with the general approach and the implementation. Things do to before this can be merged:

• Please add documentation: this needs doc comments in the SetAction classes, and a couple paragraphs of text in the general user guide.
• Please also add a newsfragment, please see https://github.com/cocotb/cocotb/blob/master/documentation/source/newsfragments/README.rst for details.
• Please combine the commits into one and add a descriptive commit message. You can reset the author to be you, but please mention Julius Baxter and Luke Darnell as co-authors.
• Ensure that you have a Fixes #657 in the commit message as well.

[thasti]

Any chance someone can test the VHPI implementation? I do not have access to a simulator implementing VHPI at the moment.

[eric-wieser]

@themperek can probably test that

[thasti]

Many thanks everyone for the constructive review yesterday. Today I was able to check the patch also compiles in a Python 3.6.8 environment and the testcase correctly works for Cadence Xcelium and Incisive. On my other machine (Python 3.8) I can test against iverilog (works fine) and ghdl (runs, but force/release is not implemented).

Let me know in case there's something else you'd like to see happening for this PR.

Edit: I realized that I can test VHPI by using TOPLEVEL_LANG=vhdl in conjunction with incisive and xcelium. Both worked fine.

[ktbarrett]

"Force or release action not supported for FLI."

This isn't true, mti_ForceSignal has the ability to force, deposit, freeze, (and drive!) a signal, and mti_ReleaseSignal can release a signal. They are just not what is currently used to write to signal handles. The drive method is particularly interesting and is probably worth another SetValueAction, even if only Questa will be able to do it right now.

[thasti]

I'm aware of that. Perhaps 'not implemented' is the wording I should have used instead. In case I can establish access to Questa before this is merged I may try also adding an FLI implementation.

[ktbarrett]

Ah, if you don't have access to Questa, and have no way to test it, it's probably best to leave it out of this PR. I might take a look at adding a more thorough FLI implementation after this PR is in, it's definitely something we want. Thank you for the work.

[imphil]

Great work! Thanks for all reviews everyone, and thanks @thasti for following up on those.