[mlgo] Don't provide default model URLs

Pointed out in Issue #56432: the current reference models may not be
quite friendly to open source projects. Their purpose is only
illustrative - the expectat

[mlgo] Don't provide default model URLs

Pointed out in Issue #56432: the current reference models may not be
quite friendly to open source projects. Their purpose is only
illustrative - the expectation is that projects would train their own.
To avoid unintentionally pulling such a model, made the URL cmake
setting require explicit user setting.

Differential Revision: https://reviews.llvm.org/D129342

show more ...

TensorFlowCompile: Add object file to list of sources rather than LINK_LIBS

Differential Revision: https://reviews.llvm.org/D126736

Use cmake Python3_EXECUTABLE variable instead of hardcoding

[MLGO] Improved support for AOT cross-targeting scenarios

The tensorflow AOT compiler can cross-target, but it can't run on (for
example) arm64. We added earlier support where the AOT-ed header and

[MLGO] Improved support for AOT cross-targeting scenarios

The tensorflow AOT compiler can cross-target, but it can't run on (for
example) arm64. We added earlier support where the AOT-ed header and object
would be built on a separate builder and then passed at build time to
a build host where the AOT compiler can't run, but clang can be otherwise
built.

To simplify such scenarios given we now support more than one AOT-able
case (regalloc and inliner), we make the AOT scenario centered on whether
files are generated, case by case (this includes the "passed from a
different builder" scenario).
This means we shouldn't need an 'umbrella' LLVM_HAVE_TF_AOT, in favor of
case by case control. A builder can opt out of an AOT case by passing that case's
model path as `none`. Note that the overrides still take precedence.

This patch controls conditional compilation with case-specific flags,
which can be enabled locally, for the component where those are
available. We still keep an overall flag for some tests.

The 'development/training' mode is unchanged, because there the model is
passed from the command line and interpreted.

Differential Revision: https://reviews.llvm.org/D117752

show more ...

[NFC][mlgo]Make the test model generator inlining-specific

When looking at building the generator for regalloc, we realized we'd
need quite a bit of custom logic, and that perhaps it'd be easier to

[NFC][mlgo]Make the test model generator inlining-specific

When looking at building the generator for regalloc, we realized we'd
need quite a bit of custom logic, and that perhaps it'd be easier to
just have each usecase (each kind of mlgo policy) have it's own
stand-alone test generator.

This patch just consolidates the old `config.py` and
`generate_mock_model.py` into one file, and does away with
subdirectories under Analysis/models.

show more ...

[NFC][mlgo]Rename a variable in TensorFlowCompile.cmake

Remaining var that had 'inlining' in name, despite being general-purpose

[NFC][mlgo] Rename some TensorFlowCompile internal vars

They were referring to 'inlining' albeit being generic

Fail gracefully if no inlining model is available to download.

Differential Revision: https://reviews.llvm.org/D104829

Remove ML inlining model artifacts.

They are not conducive to being stored in git. Instead, we autogenerate
mock model artifacts for use in tests. Production models can be
specified with the cmake f

Remove ML inlining model artifacts.

They are not conducive to being stored in git. Instead, we autogenerate
mock model artifacts for use in tests. Production models can be
specified with the cmake flag LLVM_INLINER_MODEL_PATH.

LLVM_INLINER_MODEL_PATH has two sentinel values:
- download, which will download the most recent compatible model.
- autogenerate, which will autogenerate a "fake" model for testing the
model uptake infrastructure.

Differential Revision: https://reviews.llvm.org/D104251

show more ...

Remove redundant environment variable XLA_FLAGS.

If the flag is not set, the script saved_model_aot_compile.py in tensorflow will
default it to the correct value. However, in TF 2.5, the way the val

Remove redundant environment variable XLA_FLAGS.

If the flag is not set, the script saved_model_aot_compile.py in tensorflow will
default it to the correct value. However, in TF 2.5, the way the value is set in
TensorFlowCompile.cmake file triggers a build error.

Reviewed By: mtrofin

Differential Revision: https://reviews.llvm.org/D103972

show more ...

[mlgo] Skip AOT-compiling a model if a header/object pair is provided

This allows one to cross-compile the header/object for a model in a
setup where the compiler is built on a system that cannot ho

[mlgo] Skip AOT-compiling a model if a header/object pair is provided

This allows one to cross-compile the header/object for a model in a
setup where the compiler is built on a system that cannot host the AOT
compiler. For example, if arm-hostable clang is desired, while the AOT
Tensorflow compiler can cross-compile to arm, it can't currently run on
arm.

The only alternative in that scenario would be to cross-compile clang
itself, but that gets complicated when trying to run tests after that.

Differential Revision: https://reviews.llvm.org/D99992

show more ...

[mlgo] Fetch models from path / URL

Allow custom location for pre-trained models used when AOT-compiling
policies.

Differential Revision: https://reviews.llvm.org/D96796

[llvm] Release-mode ML InlineAdvisor

Summary:
This implementation uses a pre-trained model which is statically
compiled into a native function.

RFC: http://lists.llvm.org/pipermail/llvm-dev/2020-Ap

[llvm] Release-mode ML InlineAdvisor

Summary:
This implementation uses a pre-trained model which is statically
compiled into a native function.

RFC: http://lists.llvm.org/pipermail/llvm-dev/2020-April/140763.html

Reviewers: davidxl, jdoerfert, dblaikie

Subscribers: mgorny, eraman, hiraditya, arphaman, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D81515

show more ...