Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

@conda-forge-admin, please rerender

Hi! This is the friendly automated conda-forge-linting service.

I was trying to look for recipes to lint for you, but it appears we have a merge conflict.
Please try to merge or rebase with the base branch to resolve this conflict.

Please ping the 'conda-forge/core' team (using the @ notation in a comment) if you believe this is a bug.

Wow, does this work already @jaimergp ?

No clue, I recall there were some precision issues in last attempt. Let's see if the new OpenMM has fixes for this.

So both ARM and PPC compile, but there are some runtime linking problems with OpenCL. This might be due to a missing dependency and I will debug that locally.

I see that the OpenMM CI does not have that problem (nor the "large difference in forces" PPC has here), so I guess there's a compiler/library situation we need to resolve anyway, OpenCL involved or not.

but there are some runtime linking problems with OpenCL.

This is a bug with the POCL conda package. I'll fix it this weekend.

Fixed in conda-forge/pocl-feedstock@cb201c2

Confirmed, GCC 8.5 as provided by apt reproduces the large differences in PowerPC too. Check logs for attempt here.

@peastman Do you think this difference in behavior qualifies as a bug that must be addressed at openmm/openmm?

We definitely need to track down what's causing it.

Can openmm use clang? We have clang compilers for linux too.

Can openmm use clang? We have clang compilers for linux too.

Yes! I think @peastman even prefers clang.

We use clang in the omnia recipe, for example.

I restarted the builds. Looks like CPU works fine on aarch64 and POCL works fine on ppc64le and aarch64.

Currently blocking issue is discussed here: openmm/openmm#2722

Switching to clang fails with,

In file included from /home/conda/feedstock_root/build_artifacts/openmm_1594059750456/work/openmmapi/include/openmm/internal/vectorize.h:38:

/home/conda/feedstock_root/build_artifacts/openmm_1594059750456/work/openmmapi/include/openmm/internal/vectorize_ppc.h:286:20: error: use of overloaded operator '+' is ambiguous (with operand types 'fvec4' and '__vector float' (vector of 4 'float' values))

    fvec4 temp = r + vec_sld(r.val, r.val, 8);

                 ~ ^ ~~~~~~~~~~~~~~~~~~~~~~~~

Yeah, we had tried that in #26 and arrived to the same conclusion :/

Thanks to latest changes in master, there's no precision issues with CPU anymore! Thank you so much @peastman and @isuruf!

Huzzah! Thanks so much, everybody!

Will be superseded by #36