Adding waveform generation

[BjornOlsson]

It would be nice if it was possible to start a simulation and pass along a (pre defined) waveform set () flaga and attritbue to modelsim, like:

python .py --wave-gen

This would make modelsim generate a .wlf file för the simulation and the signals defined in wave.do, and put this file under /msim in the test case lib. If needed, these wave forms could then be analyzed in the GUI after the simulation has ended. (without actually tying up a ModelSim license).

[kraigher]

What we already have

A .wlf file is already saved into vunit_out/tests/lib.entity.test_name/msim/vsim.wlf There is already a --gui=load flag for launching a GUI with the design loaded.

There are two ways to achieve what you want with the existing functionality:

Method 1

Start with the --gui=load flag and do:

log -r /*
vunit_run
quit -f

After this manually launch ModelSim GUI to view vunit_out/tests/lib.entity.test_name/msim/vsim.wlf

Method 2

Launch with --gui=load and do:

log -r /*
vunit_run
dataset save sim sim.wlf
quit -sim
dataset open sim sim.wlf

How can we make it easier to use?

Method 1

We could automate this by adding a --gui=view flag to run a test case and open the .wlf with a viewer license afterwards (Maybe even support GTKWave). The problem is how to specify what to log, we could add a flag for that or a flag to point out a .do file.

Method 2

We could wrap the three last TCL commands of this method to save a dataset and re-open it without a simulation license as a pre-loaded TCL procedure just like vunit_load and vunit_run. It could be named vunit_view

@BjornOlsson What do you think about the above? Any method you prefer?

[sthenc]

Hi,

I think I have a similar issue as Bjorn had.

I am working on a project where I have a bunch of Modelsim projects for simulation of our new IP (written in VHDL), and I have a window of opportunity to refactor it to use something more automatic for build & test.

Our current workflow is:

1. compile files with modelsim / TCL script
2. run TCLscript to show the saved waveform
3. visually inspect waveform & output files (mostly images)
4. add more signals, save wave.do, rerun simulation, stare at the waveform again

I used OSVVM before, and wrote a couple of automatic tests, and it worked ok with scripting everything with TCL & occasionally visually inspecting the waveform.

That being said, I hate TCL and would very much like to replace it with Python, especially if there is a prospect of integrating some unit tests for our code.

I tried Vunit before, but my issue is that I didn't figure out a smart way to save the waveform so I can rerun testcases while looking at the same signals of interest, instead of having to add them again manually.

I would like to use Vunit to add some automatic tests, but I need to support the old way of doing things in Modelsim (10.5b) because full test automation is just not feasible at this point. But I would like to set up a full regression test suite as soon as possible.

I know python & git so I might be able to help if you point me in the right direction.

P.S. Should I open a new issue, or is it OK to continue this discussion here?

[sthenc]

I reread your suggested methods 1 & 2, and I think I understood them this time. I have two problems with them:

1. they don't filter the signals that are logged into the .wlf
2. I would have to enter some additional commands manually into the modelsim command line interface, instead of just using "up-arrow + enter" to recompile and show the waveform as the simulation is running (we have some very long simulations)

This makes the approaches inferior to just having a modelsim TCL file that automates everything.

[kraigher]

@sthenc The contents of this old issue is outdated by now. A more recent discussion is found in https://github.com/VUnit/vunit/issues/223 Your use case can most likely be solved by one of the existing modelsim TCL-hooks we offer such as modelsim.init_files.after_load that you can read about here Let us know if you need more hooks and we will consider it, preferably in a new issue.

[kraigher]

@sthenc You can also join the Gitter Chat if you just want to discuss.

[sthenc]

@kraigher Thanks, that works. I was already lurking in the gitter chat, I will join.

For the benefit of others who might want to do the same thing, example of what to put into run.py:

vu = VUnit.from_argv()
lib = vu.add_library("lib")
lib.add_source_files(... ... )
...
vu.set_sim_option("modelsim.init_file.gui", "wave.do")
vu.main()

Then run the run script like this: python run.py --gui