ml_switcheroo.compiler.backends.extras¶

Extra Backends (Standardized implementations for Visual/DSL Targets).

Provides CompilerBackend implementations for: - TikZ (Visualization) - HTML (Visualization) - LaTeX DSL (MIDL) - MLIR (Graph to MLIR Text) - StableHLO (Graph to StableHLO-flavored MLIR Text)

These backends consume the LogicalGraph IR and emit target-specific textual representations.

Classes¶

`HtmlBackend`	Orchestrates the conversion of Logical Graphs to the HTML visual DSL.
`TikzBackend`	Orchestrates the conversion of a LogicalGraph to TikZ source code.
`LatexBackend`	Transforms Logical Graphs into MIDL LaTeX format.
`MlirBackend`	Back-end for generating MLIR text from a LogicalGraph.
`StableHloBackend`	Back-end for generating StableHLO text from a LogicalGraph.

Module Contents¶

class ml_switcheroo.compiler.backends.extras.HtmlBackend(semantics: Any | None = None)[source]¶

Bases: ml_switcheroo.compiler.backend.CompilerBackend

Orchestrates the conversion of Logical Graphs to the HTML visual DSL.

ROW_HEIGHT = 80¶

GAP_HEIGHT = 0¶

compile(graph: ml_switcheroo.compiler.ir.LogicalGraph) → str[source]¶

Compiles the graph into an HTML string document.

Parameters:: graph – The logical graph.
Returns:: HTML string.

class ml_switcheroo.compiler.backends.extras.TikzBackend(semantics: Any | None = None, y_spacing: float = 2.5, x_spacing: float = 3.0)[source]¶

Bases: ml_switcheroo.compiler.backend.CompilerBackend

Orchestrates the conversion of a LogicalGraph to TikZ source code. Verified to use Rank-Based Layout.

y_spacing = 2.5¶

x_spacing = 3.0¶

compile(graph: ml_switcheroo.compiler.ir.LogicalGraph) → str[source]¶: Compile graph to TikZ LaTeX code.

class ml_switcheroo.compiler.backends.extras.LatexBackend(semantics: Any | None = None)[source]¶

Bases: ml_switcheroo.compiler.backend.CompilerBackend

Transforms Logical Graphs into MIDL LaTeX format.

compile(graph: ml_switcheroo.compiler.ir.LogicalGraph) → str[source]¶: Compiles graph to LaTeX.

class ml_switcheroo.compiler.backends.extras.MlirBackend(semantics: Any | None = None)[source]¶

Bases: ml_switcheroo.compiler.backend.CompilerBackend

Back-end for generating MLIR text from a LogicalGraph.

This implementation provides a direct Graph-to-MLIR conversion path for scenarios bypassing the high-level Python CST Rewriter (e.g. source is RDNA).

semantics = None¶

compile(graph: ml_switcheroo.compiler.ir.LogicalGraph) → str[source]¶

Compiles the LogicalGraph into an MLIR string representation.

It emits a simple module structure with a single block, mapping each logical node to an sw.op operation. Inputs are generated as sw.constant if metadata values exist, or sw.op {type=”Input”}.

Parameters:: graph – The logical graph IR.
Returns:: The generated MLIR code.
Return type:: str

class ml_switcheroo.compiler.backends.extras.StableHloBackend(semantics: Any | None = None)[source]¶

Bases: ml_switcheroo.compiler.backend.CompilerBackend

Back-end for generating StableHLO text from a LogicalGraph.

This implementation provides a direct Graph-to-StableHLO conversion path used when an ISA is the source format (e.g. SASS -> StableHLO).

semantics = None¶

compile(graph: ml_switcheroo.compiler.ir.LogicalGraph) → str[source]¶

Compiles the graph to StableHLO-flavored MLIR.

Parameters:: graph – The logical graph.
Returns:: MLIR code string using stablehlo dialect.
Return type:: str