[ci] Drop non-HyperDebug CW310 bitstream for testing

[jwnrt]

This PR moves CW310 execution environments to the HyperDebug bitstream and disables the non-HyperDebug bitstream build in CI.

Only PMOD tests need the non-HyperDebug bitstream. They run in nightlies and build the bitstream there.

[jwnrt]

I ran a nightly for this branch, it looks okay

https://github.com/lowRISC/opentitan/actions/runs/16322328725

[pamaury]

This looks good to me, it will save a lot of time in CI :)

[a-will]

Were you thinking that the non-hyperdebug variant would essentially not be in the bitstream cache anymore?

(Keep in mind that bitstream cache management grabs whatever is in a given commit's entry, so this bitstream will not generally be present.)

[jwnrt]

Were you thinking that the non-hyperdebug variant would essentially not be in the bitstream cache anymore?

Yes, more or less. The nightly CI will still upload the bitstream every 24 hours, but the only execution environment using it is the custom one in sw/device/tests/pmod.

I think it’s only lowRISC running PMOD tests so we should be able to deal with the potentially outdated bitstream.

Longer term we’re thinking of making a board that sits on top of the HyperDebug/Nucleo board with PMOD ports. The idea would be to continue routing MIOs from the Xilinx chip to HyperDebug and only deasserting reset for the PMOD connections for PMOD tests.

[pamaury]

I think @a-will has a point though that because the bitstream cache always assume that each commit has either all bitstreams or none, it could gte confused, not just for the PMOD but for the other bitstreams as well. It's not hard to fix the cache to use different commits per design though. I can look into it.

[jwnrt]

Good point, let's block merging this until the Bazel rules can deal with a partial cache

[a-will]

Good point, let's block merging this until the Bazel rules can deal with a partial cache

This should already be possible. I had set the rules up to point to the vivado targets when the cache is missing one of the known builds.

opentitan/rules/scripts/bitstreams_workspace.py

Line 474 in 1b1945f

if design_name not in designs:

The repo label might be problematic when OT is not the main module, though. 🤔

In any case, I didn't mean to say there was a problem. I only wanted to make sure dropping this board target for general use was a known effect. It's our least expensive option, and it doesn't rely on special (and unreliable, due to the connector mating...) boards that aren't sold anywhere.

...but I am okay with getting rid of it, personally.

[jwnrt]

Thanks @a-will, me and Amaury discussed and think it's okay. The nightly does not push bitstreams so we shouldn't get into a situation where there's a cache entry with only the non-HyperDebug bitstream and the others need building.

[johngt]

@moidx - could you please take a look at this on behalf of @timothytrippel