wasmtime / PR #11711 Test out s390x native CI · git-wasmtime

The UI isn't great, but timings are:

job [qemu] [native]

misc bucket 1 14m23s 9m40s

misc bucket 2 8m38s 7m3s

misc bucket 3 15m51s 13m18s

wasmtime 7m17s 4m16s

wasmtime-cli 21m51s 13m26s

wasmtime-wasi 9m54s 8m58s

[qemu]: https://github.com/bytecodealliance/wasmtime/actions/runs/17840492728/job/50728482380
[native]: https://github.com/bytecodealliance/wasmtime/actions/runs/17842722261/job/50735874566

Consequences:

Need to install rustup on the s390x runner

Need to duplicate the ci/run-tests.py step for s390x since it doesn't have pwsh installed (kind of the fault of Windows, not necessarily s390x...)

Looks like IPv6 is disabled so trying to get tests to pass by requesting they don't test ipv6

The s390x wasmtime-cli job is the second slowest on average for our CI with the slowest being the Windows C API build which shouldn't be too hard to speed up with sharding further or something like that. This unfortunately won't really bring a ton of benefit though because our next two slowest jobs are riscv64 and armv7 emulation of the wasmtime-cli job as well, and they're neck-and-neck with s390x at this time.

So while this isn't really a slam dunk for performance it does help get QEMU out of the picture and is definitely faster for s390x. Given that I'd say let's give it a whirl and see how reliable the hardware ends up being. If we have issues it's pretty easy to swap back to QEMU-based emulation.

job	[qemu]	[native]
misc bucket 1	14m23s	9m40s
misc bucket 2	8m38s	7m3s
misc bucket 3	15m51s	13m18s
wasmtime	7m17s	4m16s
wasmtime-cli	21m51s	13m26s
wasmtime-wasi	9m54s	8m58s

Wasmtime GitHub notifications bot (Sep 18 2025 at 22:36):

alexcrichton updated PR #11711.

Wasmtime GitHub notifications bot (Sep 18 2025 at 22:36):

alexcrichton has marked PR #11711 as ready for review.

Wasmtime GitHub notifications bot (Sep 18 2025 at 22:36):

alexcrichton requested wasmtime-wasi-reviewers for a review on PR #11711.

Wasmtime GitHub notifications bot (Sep 18 2025 at 22:36):

alexcrichton requested wasmtime-core-reviewers for a review on PR #11711.

Wasmtime GitHub notifications bot (Sep 23 2025 at 16:43):

fitzgen submitted PR review.

Wasmtime GitHub notifications bot (Sep 23 2025 at 20:40):

alexcrichton merged PR #11711.

Wasmtime GitHub notifications bot (Sep 24 2025 at 16:17):

uweigand commented on PR #11711:

Hi @alexcrichton , thanks for trying out our runners!

As to these issues:

Need to install rustup on the s390x runner

Looks like IPv6 is disabled so trying to get tests to pass by requesting they don't test ipv6

I think we might want to fix those on our side to make the environment more similar to the default one. Would you mind opening issues for these in the https://github.com/IBM/actionspz/issues repo?

Not sure if there's anything we can do about the pwsh issue - how does this even work on the other Linux runners?

Wasmtime GitHub notifications bot (Sep 24 2025 at 16:25):

alexcrichton commented on PR #11711:

Sure! I filed https://github.com/IBM/actionspz/issues/45 for rustup and https://github.com/IBM/actionspz/issues/46 for ipv6. For pwsh I think that's fine to ignore. I think it works on Linux since it's distinct from powershell which, as far as I know, is Windows-only. The reason we're using pwsh though has everything to do with mingw though and nothing to do with other platforms. It's just some really minor duplication in the CI config. (I don't actually know what pwsh vs powershell is myself, it just happened to work when I was fixing mingw awhile back)

Last updated: Dec 06 2025 at 07:03 UTC