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
199 lines
8.6 KiB
C
199 lines
8.6 KiB
C
//===-- mlir-c/Dialect/Quant.h - C API for LLVM -------------------*- C -*-===//
|
|
//
|
|
// Part of the LLVM Project, under the Apache License v2.0 with LLVM
|
|
// Exceptions.
|
|
// See https://llvm.org/LICENSE.txt for license information.
|
|
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#ifndef MLIR_C_DIALECT_QUANT_H
|
|
#define MLIR_C_DIALECT_QUANT_H
|
|
|
|
#include "mlir-c/IR.h"
|
|
|
|
#ifdef __cplusplus
|
|
extern "C" {
|
|
#endif
|
|
|
|
MLIR_DECLARE_CAPI_DIALECT_REGISTRATION(quant, quant);
|
|
|
|
//===---------------------------------------------------------------------===//
|
|
// QuantizedType
|
|
//===---------------------------------------------------------------------===//
|
|
|
|
/// Returns `true` if the given type is a quantization dialect type.
|
|
MLIR_CAPI_EXPORTED bool mlirTypeIsAQuantizedType(MlirType type);
|
|
|
|
/// Returns the bit flag used to indicate signedness of a quantized type.
|
|
MLIR_CAPI_EXPORTED unsigned mlirQuantizedTypeGetSignedFlag();
|
|
|
|
/// Returns the minimum possible value stored by a quantized type.
|
|
MLIR_CAPI_EXPORTED int64_t mlirQuantizedTypeGetDefaultMinimumForInteger(
|
|
bool isSigned, unsigned integralWidth);
|
|
|
|
/// Returns the maximum possible value stored by a quantized type.
|
|
MLIR_CAPI_EXPORTED int64_t mlirQuantizedTypeGetDefaultMaximumForInteger(
|
|
bool isSigned, unsigned integralWidth);
|
|
|
|
/// Gets the original type approximated by the given quantized type.
|
|
MLIR_CAPI_EXPORTED MlirType mlirQuantizedTypeGetExpressedType(MlirType type);
|
|
|
|
/// Gets the flags associated with the given quantized type.
|
|
MLIR_CAPI_EXPORTED unsigned mlirQuantizedTypeGetFlags(MlirType type);
|
|
|
|
/// Returns `true` if the given type is signed, `false` otherwise.
|
|
MLIR_CAPI_EXPORTED bool mlirQuantizedTypeIsSigned(MlirType type);
|
|
|
|
/// Returns the underlying type used to store the values.
|
|
MLIR_CAPI_EXPORTED MlirType mlirQuantizedTypeGetStorageType(MlirType type);
|
|
|
|
/// Returns the minimum value that the storage type of the given quantized type
|
|
/// can take.
|
|
MLIR_CAPI_EXPORTED int64_t mlirQuantizedTypeGetStorageTypeMin(MlirType type);
|
|
|
|
/// Returns the maximum value that the storage type of the given quantized type
|
|
/// can take.
|
|
MLIR_CAPI_EXPORTED int64_t mlirQuantizedTypeGetStorageTypeMax(MlirType type);
|
|
|
|
/// Returns the integral bitwidth that the storage type of the given quantized
|
|
/// type can represent exactly.
|
|
MLIR_CAPI_EXPORTED unsigned
|
|
mlirQuantizedTypeGetStorageTypeIntegralWidth(MlirType type);
|
|
|
|
/// Returns `true` if the `candidate` type is compatible with the given
|
|
/// quantized `type`.
|
|
MLIR_CAPI_EXPORTED bool
|
|
mlirQuantizedTypeIsCompatibleExpressedType(MlirType type, MlirType candidate);
|
|
|
|
/// Returns the element type of the given quantized type as another quantized
|
|
/// type.
|
|
MLIR_CAPI_EXPORTED MlirType
|
|
mlirQuantizedTypeGetQuantizedElementType(MlirType type);
|
|
|
|
/// Casts from a type based on the storage type of the given type to a
|
|
/// corresponding type based on the given type. Returns a null type if the cast
|
|
/// is not valid.
|
|
MLIR_CAPI_EXPORTED MlirType
|
|
mlirQuantizedTypeCastFromStorageType(MlirType type, MlirType candidate);
|
|
|
|
/// Casts from a type based on a quantized type to a corresponding typed based
|
|
/// on the storage type. Returns a null type if the cast is not valid.
|
|
MLIR_CAPI_EXPORTED MlirType mlirQuantizedTypeCastToStorageType(MlirType type);
|
|
|
|
/// Casts from a type based on the expressed type of the given type to a
|
|
/// corresponding type based on the given type. Returns a null type if the cast
|
|
/// is not valid.
|
|
MLIR_CAPI_EXPORTED MlirType
|
|
mlirQuantizedTypeCastFromExpressedType(MlirType type, MlirType candidate);
|
|
|
|
/// Casts from a type based on a quantized type to a corresponding typed based
|
|
/// on the expressed type. Returns a null type if the cast is not valid.
|
|
MLIR_CAPI_EXPORTED MlirType mlirQuantizedTypeCastToExpressedType(MlirType type);
|
|
|
|
/// Casts from a type based on the expressed type of the given quantized type to
|
|
/// equivalent type based on storage type of the same quantized type.
|
|
MLIR_CAPI_EXPORTED MlirType
|
|
mlirQuantizedTypeCastExpressedToStorageType(MlirType type, MlirType candidate);
|
|
|
|
//===---------------------------------------------------------------------===//
|
|
// AnyQuantizedType
|
|
//===---------------------------------------------------------------------===//
|
|
|
|
/// Returns `true` if the given type is an AnyQuantizedType.
|
|
MLIR_CAPI_EXPORTED bool mlirTypeIsAAnyQuantizedType(MlirType type);
|
|
|
|
/// Creates an instance of AnyQuantizedType with the given parameters in the
|
|
/// same context as `storageType` and returns it. The instance is owned by the
|
|
/// context.
|
|
MLIR_CAPI_EXPORTED MlirType mlirAnyQuantizedTypeGet(unsigned flags,
|
|
MlirType storageType,
|
|
MlirType expressedType,
|
|
int64_t storageTypeMin,
|
|
int64_t storageTypeMax);
|
|
|
|
//===---------------------------------------------------------------------===//
|
|
// UniformQuantizedType
|
|
//===---------------------------------------------------------------------===//
|
|
|
|
/// Returns `true` if the given type is a UniformQuantizedType.
|
|
MLIR_CAPI_EXPORTED bool mlirTypeIsAUniformQuantizedType(MlirType type);
|
|
|
|
/// Creates an instance of UniformQuantizedType with the given parameters in the
|
|
/// same context as `storageType` and returns it. The instance is owned by the
|
|
/// context.
|
|
MLIR_CAPI_EXPORTED MlirType mlirUniformQuantizedTypeGet(
|
|
unsigned flags, MlirType storageType, MlirType expressedType, double scale,
|
|
int64_t zeroPoint, int64_t storageTypeMin, int64_t storageTypeMax);
|
|
|
|
/// Returns the scale of the given uniform quantized type.
|
|
MLIR_CAPI_EXPORTED double mlirUniformQuantizedTypeGetScale(MlirType type);
|
|
|
|
/// Returns the zero point of the given uniform quantized type.
|
|
MLIR_CAPI_EXPORTED int64_t mlirUniformQuantizedTypeGetZeroPoint(MlirType type);
|
|
|
|
/// Returns `true` if the given uniform quantized type is fixed-point.
|
|
MLIR_CAPI_EXPORTED bool mlirUniformQuantizedTypeIsFixedPoint(MlirType type);
|
|
|
|
//===---------------------------------------------------------------------===//
|
|
// UniformQuantizedPerAxisType
|
|
//===---------------------------------------------------------------------===//
|
|
|
|
/// Returns `true` if the given type is a UniformQuantizedPerAxisType.
|
|
MLIR_CAPI_EXPORTED bool mlirTypeIsAUniformQuantizedPerAxisType(MlirType type);
|
|
|
|
/// Creates an instance of UniformQuantizedPerAxisType with the given parameters
|
|
/// in the same context as `storageType` and returns it. `scales` and
|
|
/// `zeroPoints` point to `nDims` number of elements. The instance is owned
|
|
/// by the context.
|
|
MLIR_CAPI_EXPORTED MlirType mlirUniformQuantizedPerAxisTypeGet(
|
|
unsigned flags, MlirType storageType, MlirType expressedType,
|
|
intptr_t nDims, double *scales, int64_t *zeroPoints,
|
|
int32_t quantizedDimension, int64_t storageTypeMin, int64_t storageTypeMax);
|
|
|
|
/// Returns the number of axes in the given quantized per-axis type.
|
|
MLIR_CAPI_EXPORTED intptr_t
|
|
mlirUniformQuantizedPerAxisTypeGetNumDims(MlirType type);
|
|
|
|
/// Returns `pos`-th scale of the given quantized per-axis type.
|
|
MLIR_CAPI_EXPORTED double mlirUniformQuantizedPerAxisTypeGetScale(MlirType type,
|
|
intptr_t pos);
|
|
|
|
/// Returns `pos`-th zero point of the given quantized per-axis type.
|
|
MLIR_CAPI_EXPORTED int64_t
|
|
mlirUniformQuantizedPerAxisTypeGetZeroPoint(MlirType type, intptr_t pos);
|
|
|
|
/// Returns the index of the quantized dimension in the given quantized per-axis
|
|
/// type.
|
|
MLIR_CAPI_EXPORTED int32_t
|
|
mlirUniformQuantizedPerAxisTypeGetQuantizedDimension(MlirType type);
|
|
|
|
/// Returns `true` if the given uniform quantized per-axis type is fixed-point.
|
|
MLIR_CAPI_EXPORTED bool
|
|
mlirUniformQuantizedPerAxisTypeIsFixedPoint(MlirType type);
|
|
|
|
//===---------------------------------------------------------------------===//
|
|
// CalibratedQuantizedType
|
|
//===---------------------------------------------------------------------===//
|
|
|
|
/// Returns `true` if the given type is a CalibratedQuantizedType.
|
|
MLIR_CAPI_EXPORTED bool mlirTypeIsACalibratedQuantizedType(MlirType type);
|
|
|
|
/// Creates an instance of CalibratedQuantizedType with the given parameters
|
|
/// in the same context as `expressedType` and returns it. The instance is owned
|
|
/// by the context.
|
|
MLIR_CAPI_EXPORTED MlirType
|
|
mlirCalibratedQuantizedTypeGet(MlirType expressedType, double min, double max);
|
|
|
|
/// Returns the min value of the given calibrated quantized type.
|
|
MLIR_CAPI_EXPORTED double mlirCalibratedQuantizedTypeGetMin(MlirType type);
|
|
|
|
/// Returns the max value of the given calibrated quantized type.
|
|
MLIR_CAPI_EXPORTED double mlirCalibratedQuantizedTypeGetMax(MlirType type);
|
|
|
|
#ifdef __cplusplus
|
|
}
|
|
#endif
|
|
|
|
#endif // MLIR_C_DIALECT_QUANT_H
|