commit | 17212af0a32c365ae66359e59662e57a21563a24 | [log] [tgz] |
---|---|---|
author | Ingo Müller <github.com@ingomueller.net> | Thu Feb 02 11:50:34 2023 +0100 |
committer | GitHub <noreply@github.com> | Thu Feb 02 11:50:34 2023 +0100 |
tree | 264140fe9edaf1d6fde7e466325944ce587ab134 | |
parent | 6a59ff602da58739f3fbfcc2cfa1b5d10c1b6295 [diff] | |
parent | 77faee25efc45aec31d72775dca2bd9d2abbfa22 [diff] |
Add a few missing CMake dependencies required for building shared libs. (#12022) This PR adds a few missing CMake dependencies. IREE is built by default with [`BUILD_SHARED_LIBS=OFF`](https://cmake.org/cmake/help/latest/variable/BUILD_SHARED_LIBS.html), so most libraries are built as static libraries. This means that not all symbols referenced by a particular library need to be present -- it is sufficient if some other library that is used in the same final executable also imports them. Some missing dependencies can therefore go unnoticed for some time. Building IREE with `BUILD_SHARED_LIBS=ON` reveals (some of) them. I have recently fixed these dependencies in `mlir-hlo` (see [PR](https://github.com/tensorflow/tensorflow/pull/59421) and [integration commit](https://github.com/tensorflow/mlir-hlo/commit/ed069b6e)) and the dependency has now been updated, such that most of IREE now builds with `BUILD_SHARED_LIBS=ON`. However, the build is still broken: `IREELinalgExtTransforms` and `IREELinalgExtPasses` respectively require symbols from the other one, and cyclic dependencies are not possible in CMake, so there is no way to build (with shared libs) currently. I guess the only solution is to move code around such that one of the two does not use any symbols from the other one. I'll leave this for someone else to fix if desired.
IREE (Intermediate Representation Execution Environment, pronounced as “eerie”) is an MLIR-based end-to-end compiler and runtime that lowers Machine Learning (ML) models to a unified IR that scales up to meet the needs of the datacenter and down to satisfy the constraints and special considerations of mobile and edge deployments.
See our website for project details, user guides, and instructions on building from source.
IREE is still in its early phase. We have settled down on the overarching infrastructure and are actively improving various software components as well as project logistics. It is still quite far from ready for everyday use and is made available without any support at the moment. With that said, we welcome any kind of feedback on any communication channels!
See our website for more information.
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.