Compiler FE: List up unused projects

[hseok-oh]

Let's list unused compiler frontend projects to remove or disable build.

[seanshpark]

remove or disable build.

+1 to disable build

@lemmaa , your opinion ?

[seanshpark]

I'll work on this task :) With excluding unused projects while normal build, CI build.

[seanshpark]

11848 Draft implementation:

• exclude module (compiler/something) when there exist exclude.me file in that folder
• if -DBUILD_WHITELIST=... is given, ignore this feature and build all modules in the list
• if -DENABLE_EXCLUDE_ME=OFF is given, ignore this feature and build all modules
• excluded module may fail in build in the future but we won't know that
• we can re-enable excluded module when ever needed again
  • any build failure should be fixed at that time

@lemmaa , any comments?

[seanshpark]

@lemmaa PTAL

[lemmaa]

Let's list unused compiler frontend projects to remove or disable build.

@hseok-oh , what's the purpose of this work?

My guess is that there are many possible reasons for this, such as

• Optimizing CI build resources,
• Reducing the difficulty of source code maintenance,
• Etc.

In any case, the reason, purpose, etc. must be clear so that there is no confusion in the discussion and conclusion.

And, what do you think are excluded? :)

[lemmaa]

11848 Draft implementation:

• exclude module (compiler/something) when there exist exclude.me file in that folder
• if -DBUILD_WHITELIST=... is given, ignore this feature and build all modules in the list
• if -DENABLE_EXCLUDE_ME=OFF is given, ignore this feature and build all modules
• excluded module may fail in build in the future but we won't know that
• we can re-enable excluded module when ever needed again
  • any build failure should be fixed at that time

@lemmaa , any comments?

@seanshpark , I agree with your draft design. However, it would be nice to check what @hseok-oh's original intention was and organize the details.

[hseok-oh]

Let's list unused compiler frontend projects to remove or disable build.

@hseok-oh , what's the purpose of this work?

My guess is that there are many possible reasons for this, such as

* Optimizing CI build resources,

* Reducing the difficulty of source code maintenance,

* Etc.

In any case, the reason, purpose, etc. must be clear so that there is no confusion in the discussion and conclusion.

And, what do you think are excluded? :)

Reason to remove: maybe not used any more

• ex) projects for caffe tools

Reason to disable build: reduce CI build resource

• ex) coco and related, ann-api and related, etc...

[lemmaa]

Reason to disable build: reduce CI build resource

• ex) coco and related, ann-api and related, etc...

In this case, isn't it necessary to automatically and regularly check the following cases among the target modules?

• excluded module may fail in build in the future but we won't know that

• we can re-enable excluded module when ever needed again

  • any build failure should be fixed at that time

For example, even if in frequent PR check build or commit check build are excluded, it is checked regularly on a daily or weekly basis and an issue is registered when a build break or test failure occurs.

• if -DBUILD_WHITELIST=... is given, ignore this feature and build all modules in the list

I think this option would be appropriate for such support.

Taking this into consideration, it would be a good idea to categorize and organize the target list according to the original purpose of the issue.

• Required build target
• Subject to regular inspection
• Fade out target out of interest

[seanshpark]

I think this option would be appropriate for such support.

Yup, -DENABLE_EXCLUDE_ME=OFF is for regular checkup to build all modules :)

[seanshpark]

I'll begin posting related changes.

Modules that have `add_executable()`

``` ./caffe2circle/CMakeLists.txt ./caffegen/CMakeLists.txt ./circle2circle/CMakeLists.txt ./circlechef/tools/console/CMakeLists.txt ./circlechef/tools/file/CMakeLists.txt ./circlechef/tools/reverse/CMakeLists.txt ./circledump/CMakeLists.txt ./circle-eval-diff/CMakeLists.txt ./circle-execution-plan/CMakeLists.txt ./circle-inspect/CMakeLists.txt ./circle-interpreter/CMakeLists.txt ./circle-mpqsolver/CMakeLists.txt ./circle-operator/CMakeLists.txt ./circle-opselector/CMakeLists.txt ./circle-part-driver/CMakeLists.txt ./circle-partitioner/CMakeLists.txt ./circle-quantizer/CMakeLists.txt ./circle-tensordump/CMakeLists.txt ./circle-verify/CMakeLists.txt ./common-artifacts/CMakeLists.txt ./dalgona/CMakeLists.txt ./embedded-import-value-test/CMakeLists.txt ./enco/cli/CMakeLists.txt ./enco/core/CMakeLists.txt ./encodump/CMakeLists.txt ./enco/frontend/caffe/CMakeLists.txt ./enco/frontend/tflite/CMakeLists.txt ./hermes/CMakeLists.txt ./i5diff/CMakeLists.txt ./luci-eval-driver/CMakeLists.txt ./luci/tester/CMakeLists.txt ./minmax-embedder/CMakeLists.txt ./minmax-embedder-value-test/gen/CMakeLists.txt ./mio-circle04/CMakeLists.txt ./mio-circle05/CMakeLists.txt ./mio-circle06/CMakeLists.txt ./mio-circle/CMakeLists.txt ./mio-tflite2121/CMakeLists.txt ./mio-tflite260/CMakeLists.txt ./mio-tflite280/CMakeLists.txt ./mio-tflite/CMakeLists.txt ./moco-tf/CMakeLists.txt ./morph/CMakeLists.txt ./nest/core/CMakeLists.txt ./nnc/CMakeLists.txt ./nnc/cmake/utils.cmake ./nnc/passes/CMakeLists.txt ./nnc/tests/acl_soft_backend/artifact_cmake/CMakeLists.txt ./nnc/tests/import/CMakeLists.txt ./nnc/tests/soft_backend/CMakeLists.txt ./nnc/utils/caffe2_dot_dumper/CMakeLists.txt ./nnc/utils/caffe_dot_dumper/CMakeLists.txt ./nnc/utils/input_gen/CMakeLists.txt ./nnc/utils/tflite_dot_dumper/CMakeLists.txt ./nnkit/tools/benchmark/CMakeLists.txt ./nnkit/tools/run/CMakeLists.txt ./nnop/CMakeLists.txt ./nnsuite/conv/nnkit-caffe/CMakeLists.txt ./nnsuite/conv/nnkit-tflite/CMakeLists.txt ./one-cmds/dummy-driver/CMakeLists.txt ./oneco/CMakeLists.txt ./onnx2circle/CMakeLists.txt ./onnx2tflite/CMakeLists.txt ./onnxkit/CMakeLists.txt ./pp/CMakeLists.txt ./q-implant/CMakeLists.txt ./record-minmax/CMakeLists.txt ./tf2circle/CMakeLists.txt ./tf2nnpkg/CMakeLists.txt ./tf2tflite/CMakeLists.txt ./tfkit/CMakeLists.txt ./tflchef/tools/console/CMakeLists.txt ./tflchef/tools/file/CMakeLists.txt ./tflchef/tools/reverse/CMakeLists.txt ./tfldump/CMakeLists.txt ./tfl-inspect/CMakeLists.txt ./tflite2circle/CMakeLists.txt ./tfl-verify/CMakeLists.txt ./vconone/CMakeLists.txt ```

[seanshpark]

These are the candidate modules that "are"/"will be" excluded. Tool level modules are treated first and then related libraries that follows. Agreement or objection can be discussed in each PR for that module.

Tools

• [x] caffe2circle
  • mir_caffe_importer (mir)
  • mir2loco
  • exo
• [x] caffegen
  • cli
  • caffe
• [x] nnc
• [x] tf2circle-conversion-test
• [x] tf2circle-dredd-pb-test
• [x] tf2circle-dredd-pbtxt-test
• [x] tf2circle-model-test
• [x] tf2circle-ui-check
• [x] tf2circle-value-pbtxt-remote-test
  • nnkit
  • tfkit
• [x] tf2circle
  • moco-tf
  • exo
• [x] tf2tflite-dredd-pb-test
• [x] tf2tflite-dredd-pbtxt-test
• [x] tf2tflite-value-pb-test
• [x] tf2tflite-value-pbtxt-test
• [x] tf2tflite
  • moco-log
  • moco-tf
  • exo
  • tfinfo ?
• [x] tfts
• [x] moco-value-pbtxt-test
• [x] tfkit
  • cli
  • mio_tf
• [x] onnx2circle
• [x] onnx2tflite-integration-test
• [x] onnx2tflite
• [ ] ...

Libraries

• [x] nnop
• [x] nnsuite
• [x] enco-intf
• [x] encodump
• [x] enco
• [x] mir2loco
• [x] mir-interpreter
• [x] mir
• [x] mio-circle04
• [x] mio-circle05
• [x] mio-circle
• [x] mio-tf
• [x] mio-tflite260
• [x] mio-tflite280
• [x] mio-tflite
• [x] moco-tf
• [x] moco-log
• [x] moco
• [x] plier-tf
• [x] exo
• [x] cli
• [x] nnkit-intf
• [x] nnkit-caffe
• [x] nnkit-misc
• [x] nnkit-mocotf
• [x] nnkit-onnxrt
• [x] nnkit-tf
• [x] nnkit-tflite
• [x] nnkit
• [x] aditas
• [x] ann-ref
• [x] ann-api
• [x] bino
• [x] fipe
• [x] kuma
• [x] nest
• [x] nike
• [x] oneco-value-pbtxt-test
• [x] oneco
• [x] onnxkit
• [x] coco
• [ ] ...

[seanshpark]

ping @jinevening , @mhs4670go

[seanshpark]

• [x] pepper-assert
• [x] pepper-env
• [x] tfinfo-v2
• [ ] ...

---

tfinfo is used in tf2nnpkg v4tf has only README.md which can be ignored

[seanshpark]

done for now. @hseok-oh , ok to close?