5e83a5b475
Since the very first commits, the Python and C MLIR APIs have had mis-placed registration/load functionality for dialects, extensions, etc. This was done pragmatically in order to get bootstrapped and then just grew in. Downstreams largely bypass and do their own thing by providing various APIs to register things they need. Meanwhile, the C++ APIs have stabilized around this and it would make sense to follow suit.
The thing we have observed in canonical usage by downstreams is that each downstream tends to have native entry points that configure its installation to its preferences with one-stop APIs. This patch leans in to this approach with `RegisterEverything.h` and `mlir._mlir_libs._mlirRegisterEverything` being the one-stop entry points for the "upstream packages". The `_mlir_libs.__init__.py` now allows customization of the environment and Context by adding "initialization modules" to the `_mlir_libs` package. If present, `_mlirRegisterEverything` is treated as such a module. Others can be added by downstreams by adding a `_site_initialize_{i}.py` module, where '{i}' is a number starting with zero. The number will be incremented and corresponding module loaded until one is not found. Initialization modules can:
* Perform load time customization to the global environment (i.e. registering passes, hooks, etc).
* Define a `register_dialects(registry: DialectRegistry)` function that can extend the `DialectRegistry` that will be used to bootstrap the `Context`.
* Define a `context_init_hook(context: Context)` function that will be added to a list of callbacks which will be invoked after dialect registration during `Context` initialization.
Note that the `MLIRPythonExtension.RegisterEverything` is not included by default when building a downstream (its corresponding behavior was prior). For downstreams which need the default MLIR initialization to take place, they must add this back in to their Python CMake build just like they add their own components (i.e. to `add_mlir_python_common_capi_library` and `add_mlir_python_modules`). It is perfectly valid to not do this, in which case, only the things explicitly depended on and initialized by downstreams will be built/packaged. If the downstream has not been set up for this, it is recommended to simply add this back for the time being and pay the build time/package size cost.
CMake changes:
* `MLIRCAPIRegistration` -> `MLIRCAPIRegisterEverything` (renamed to signify what it does and force an evaluation: a number of places were incidentally linking this very expensive target)
* `MLIRPythonSoure.Passes` removed (without replacement: just drop)
* `MLIRPythonExtension.AllPassesRegistration` removed (without replacement: just drop)
* `MLIRPythonExtension.Conversions` removed (without replacement: just drop)
* `MLIRPythonExtension.Transforms` removed (without replacement: just drop)
Header changes:
* `mlir-c/Registration.h` is deleted. Dialect registration functionality is now in `IR.h`. Registration of upstream features are in `mlir-c/RegisterEverything.h`. When updating MLIR and a couple of downstreams, I found that proper usage was commingled so required making a choice vs just blind S&R.
Python APIs removed:
* mlir.transforms and mlir.conversions (previously only had an __init__.py which indirectly triggered `mlirRegisterTransformsPasses()` and `mlirRegisterConversionPasses()` respectively). Downstream impact: Remove these imports if present (they now happen as part of default initialization).
* mlir._mlir_libs._all_passes_registration, mlir._mlir_libs._mlirTransforms, mlir._mlir_libs._mlirConversions. Downstream impact: None expected (these were internally used).
C-APIs changed:
* mlirRegisterAllDialects(MlirContext) now takes an MlirDialectRegistry instead. It also used to trigger loading of all dialects, which was already marked with a TODO to remove -- it no longer does, and for direct use, dialects must be explicitly loaded. Downstream impact: Direct C-API users must ensure that needed dialects are loaded or call `mlirContextLoadAllAvailableDialects(MlirContext)` to emulate the prior behavior. Also see the `ir.c` test case (e.g. ` mlirContextGetOrLoadDialect(ctx, mlirStringRefCreateFromCString("func"));`).
* mlirDialectHandle* APIs were moved from Registration.h (which now is restricted to just global/upstream registration) to IR.h, arguably where it should have been. Downstream impact: include correct header (likely already doing so).
C-APIs added:
* mlirContextLoadAllAvailableDialects(MlirContext): Corresponds to C++ API with the same purpose.
Python APIs added:
* mlir.ir.DialectRegistry: Mapping for an MlirDialectRegistry.
* mlir.ir.Context.append_dialect_registry(MlirDialectRegistry)
* mlir.ir.Context.load_all_available_dialects()
* mlir._mlir_libs._mlirAllRegistration: New native extension that exposes a `register_dialects(MlirDialectRegistry)` entry point and performs all upstream pass/conversion/transforms registration on init. In this first step, we eagerly load this as part of the __init__.py and use it to monkey patch the Context to emulate prior behavior.
* Type caster and capsule support for MlirDialectRegistry
This should make it possible to build downstream Python dialects that only depend on a subset of MLIR. See: https://github.com/llvm/llvm-project/issues/56037
Here is an example PR, minimally adapting IREE to these changes: https://github.com/iree-org/iree/pull/9638/files In this situation, IREE is opting to not link everything, since it is already configuring the Context to its liking. For projects that would just like to not think about it and pull in everything, add `MLIRPythonExtension.RegisterEverything` to the list of Python sources getting built, and the old behavior will continue.
Reviewed By: mehdi_amini, ftynse
Differential Revision: https://reviews.llvm.org/D128593
Introduction
Warning The Bazel build is experimental and best-effort, supported in line with the policy for LLVM's peripheral support tier. LLVM's official build system is CMake. If in doubt use that. If you make changes to LLVM, you're expected to update the CMake build but you don't need to update Bazel build files. Reviewers should not ask authors to update Bazel build files unless the author has opted in to support Bazel. Keeping the Bazel build files up-to-date is on the people who use the Bazel build.
Bazel is a multi-language build system focused on reproducible builds to enable dependency analysis and caching for fast incremental builds.
The main motivation behind the existence of an LLVM Bazel build is that a number of projects that depend on LLVM use Bazel, and Bazel works best when it knows about the whole source tree (as opposed to installing artifacts coming from another build system). Community members are also welcome to use Bazel for their own development as long as they continue to maintain the official CMake build system. See also, the proposal for adding this configuration.
Quick Start
git clone https://github.com/llvm/llvm-project.git; cd llvm-projectif you don't have a checkout yet.- Install Bazel at the version indicated by .bazelversion, following the official instructions, if you don't have it installed yet: https://docs.bazel.build/versions/main/install.html.
cd utils/bazelbazel build --config=generic_clang @llvm-project//...(if building on Unix with Clang).--config=generic_gccand--config=msvcare also available.
Configuration
The repository .bazelrc will import user-specific settings from a
user.bazelrc file (in addition to the standard locations). Adding your typical
config setting is recommended.
build --config=generic_clang
You can enable disk caching, which will cache build results
build --disk_cache=~/.cache/bazel-disk-cache
You can instruct Bazel to use a ramdisk for its sandboxing operations via --sandbox_base, which can help avoid IO bottlenecks for the symlink stragegy used for sandboxing. This is especially important with many inputs and many cores (see https://github.com/bazelbuild/bazel/issues/11868):
build --sandbox_base=/dev/shm
Bear in mind that this requires that your ramdisk is of sufficient size to hold any temporary files. Anecdotally, 1GB should be sufficient.
Coverage
The LLVM, MLIR, and Clang subprojects have configurations for Linux (Clang and GCC), Mac (Clang and GCC), and Windows (MSVC). Configuration options that are platform-specific are selected for in defines. Many are also hardcoded to the values currently used by all supported configurations. If there is a configuration you'd like to use that isn't supported, please send a patch.
Continuous Testing
A Buildkite pipeline runs the full Bazel build on every commit to the main branch. Notifications of failures are sent to the llvm-bazel-alerts google group, which anyone is free to join. Currently, the behavior is just to send an email on each failure using Buildkite's built-in notification system, so if you subscribe, it is highly recommended that you set up email filters or some other mechanism to not flood your inbox. More sophisticated notifications, e.g. only on status change or routed based on blamelist are TODO (contributions welcome).
Pre-merge Testing
A Buildkite pipeline runs the full Bazel build as a pre-merge test using the LLVM pre-merge testing. It is triggered on all changes to the utils/bazel directory and when the patch author is a member of the Bazel Phabricator project. If you use or benefit from the Bazel build, please join the project so that you can help keep it green. As a bonus, it runs in under 5 minutes, much faster than any of the other pre-merge builds.
Usage in Downstream Projects
To use in dependent projects using Bazel, you can import LLVM and then use the
provided configuration rule. See example usage in the examples/ directory.