Handle custom_ops like any other ET lib in Android and examples/models/lla{m,v}a/CMakeLists.txt

[swolchok]

The old way was a holdover from when custom_ops lived in examples/models/llama (and was copy/pasted to llava and Android).

Needed for #8947.

[swolchok]

Stack from ghstack (oldest at bottom):

• Make serial parallel_for "polyfill" iterate backwards in debug builds #9044
• Split & remove extension_parallel #8983
• Migrate thread_parallel.cpp to ET_CHECK_OR_RETURN_FALSE #8977
• Properly export ET_USE_THREADPOOL from the threadpool extension #8947
• cmake-format executorch-config.cmake #9015
• -> Handle custom_ops like any other ET lib in Android and examples/models/lla{m,v}a/CMakeLists.txt #8946

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/8946

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit b478275 with merge base 484231c ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[swolchok]

linker errors say reduce_util.cpp is in both portable_lib and custom_ops. indeed, executorch_srcs.cmake says it's there too! can't get rid of that file (executorch_srcs.cmake) soon enough, but until then I'll have to fix it.

[swolchok]

looks like we have to handle the Android demo as well

[swolchok]

wondering how high we can bump this and if we can raise it globally

[swolchok]

as I said on Discord:

the concern would be if there are projects that would want to take an ExecuTorch dep, but actually need an older CMake version. seems like we can raise it, but should avoid doing so unnecessarily.

so, i'm going to make this change in this PR but not roll it out to the top-level ExecuTorch CMakeLists.txt.

[kimishpatel]

is that the only way to achieve whole-archive? I dont know much on the implication of requiring this but I suppose non-trivial number of folks using et may come to use for llama

[swolchok]

we pip install latest cmake anyway. people who follow our directions will get 3.31.

[larryliu0820]

It would be good if we can do a codemod to change all CMakeLists.txt to have the same requirement.

[swolchok]

change all CMakeLists.txt to have the same requirement.

@larryliu0820 I wouldn't mind doing this, I just thought it might be beneficial to only increase the requirement for the llama example use cases that need this stuff and leave the core/embedded cases lower. What do you think?

[swolchok]

noting that CI is 100% green in case I rebase

[swolchok]

ping reviewers @lucylq @jackzhxng @GregoryComer @kirklandsign @manuelcandales @kimishpatel

[kimishpatel]

I think @larryliu0820 should review this but left some comments