interactive: mlir-gui version 1

pyproject.toml

@@ -52,6 +52,8 @@ xdsl-gui = "xdsl.interactive.app:main"
 xdsl-stubgen = "xdsl.utils.dialect_stub:make_all_stubs"
 xdsl-tblgen = "xdsl.tools.tblgen_to_py:main"
 
+mlir-gui = "xdsl.interactive.mlirapp:main"
+
 [tool.setuptools]
 platforms = ["Linux", "Mac OS-X", "Unix"]
 zip-safe = false

xdsl/interactive/mlir_helper.py

@@ -0,0 +1,81 @@
+import subprocess
+from collections.abc import Callable
+
+from xdsl.context import MLContext
+from xdsl.dialects import builtin
+from xdsl.ir import Dialect
+from xdsl.passes import ModulePass
+from xdsl.transforms.mlir_opt import MLIROptPass
+
+
+def iterate_and_extract_options() -> list[str]:
+    """
+    Fetch MLIR passes by calling `mlir-opt --help` and extract options starting with `--`.
+
+    Returns:
+        list: A list of strings containing the extracted options.
+    """
+    options: list[str]
+    options = []
+
+    # Step 1: Call `mlir-opt --help` and capture the output
+    try:
+        result = subprocess.run(
+            ["mlir-opt", "--help"], capture_output=True, text=True, check=True
+        )
+        help_output = result.stdout
+    except FileNotFoundError:
+        print(
+            "Error: 'mlir-opt' not found. Make sure it is installed and added to your PATH."
+        )
+        return []
+    except subprocess.CalledProcessError as e:
+        print(f"Error while running 'mlir-opt --help': {e}")
+        return []
+
+    # Step 2: Parse the options from the output
+    for line in help_output.splitlines():
+        line = line.strip()
+        if "--" in line:
+            start_index = line.find("--")
+            option = line[start_index:].split()[
+                0
+            ]  # Extract the word directly after `--`
+            options.append(option)
+
+    return sorted(options)
+
+
+def get_mlir_pass_list() -> tuple[str, ...]:
+    """
+    Function that returns the condensed pass list for a given ModuleOp, i.e. the passes that
+    change the ModuleOp.
+    """
+
+    return tuple(pass_name for pass_name in iterate_and_extract_options())
+
+
+def get_new_registered_context(
+    all_dialects: tuple[tuple[str, Callable[[], Dialect]], ...],
+) -> MLContext:
+    """
+    Generates a new MLContext, registers it and returns it.
+    """
+    ctx = MLContext(True)
+    for dialect_name, dialect_factory in all_dialects:
+        ctx.register_dialect(dialect_name, dialect_factory)
+    return ctx
+
+
+def generate_mlir_pass(args: tuple[str, ...]) -> MLIROptPass:
+    return MLIROptPass("mlir-opt", arguments=args)
+
+
+def apply_mlir_pass_with_args_to_module(
+    module: builtin.ModuleOp, ctx: MLContext, module_pass: ModulePass
+) -> builtin.ModuleOp:
+    """
+    Function that takes a ModuleOp, an MLContext and a pass_pipeline (consisting of a type[ModulePass] and PipelinePassSpec), applies the pass(es) to the ModuleOp and returns the new ModuleOp.
+    """
+    module_pass.apply(ctx, module)
+    return module