ace42cf063
`InitAll***` functions are used by `opt`-style tools to init all MLIR dialects/passes/extensions. Currently they are implemeted as inline functions and include essentially the entire MLIR header tree. Each file which includes this header (~10 currently) takes 10+ sec and multiple GB of ram to compile (tested with clang-19), which limits amount of parallel compiler jobs which can be run. Also, flang just includes this file into one of its headers. Move the actual registration code to the static library, so it's compiled only once. Discourse thread https://discourse.llvm.org/t/rfc-moving-initall-implementation-into-static-library/87559
Standalone Transform Dialect Interpreter
This is an example of using the Transform dialect interpreter functionality standalone, that is, outside of the regular pass pipeline. The example is a
binary capable of processing MLIR source files similar to mlir-opt and other
optimizer drivers, with the entire transformation process driven by a Transform
dialect script. This script can be embedded into the source file or provided in
a separate MLIR source file.
Either the input module or the transform module must contain a top-level symbol
named __transform_main, which is used as the entry point to the transformation
script.
mlir-transform-opt payload_with_embedded_transform.mlir
mlir-transform-opt payload.mlir -transform=transform.mlir
The name of the entry point can be overridden using command-line options.
mlir-transform-opt payload-mlir -transform-entry-point=another_entry_point
Transform scripts can reference symbols defined in other source files, called libraries, which can be supplied to the binary through command-line options. Libraries will be embedded into the main transformation module by the tool and the interpreter will process everything as a single module. A debug option is available to see the contents of the transform module before it goes into the interpreter.
mlir-transform-opt payload.mlir -transform=transform.mlir \
-transform-library=external_definitions_1.mlir \
-transform-library=external_definitions_2.mlir \
-dump-library-module
Check out the Transform dialect tutorial as well as documentation to learn more about the dialect.