The constant propagation framework, which is a modified version of sparse conditional constant propagation of Wegman et al.

Constant propagation analysis state.

Data-flow analysis framework. 'L is a lattice, 'V is a vertex data type of a graph.

Either forward or backward analysis.

The core interface of a Constant Propagation (CP) algorithm.

Reaching definition analysis with a LowUIR-based CFG.

Thunk and Pointer are the only special kind of SCPValue, which should not be invalidated across function calls. This is to correctly track GOT pointers.

The most basic constant propagation algorithm, which can track stack-based memory objects and GOT pointers. The reader is to enable reading data from external sections, e.g., rodata. If the reader is not given, we simply ignore such global data.

StackPointerPropagation values.

An ID of an SSA memory instance.

This is a variant of the SparseConstantPropagation, which only tracks the stack pointer used in a function. We initiate the stack pointer with a constant first, and check how it propagates within the function. StackPointerPropagation is generally much faster than SparseConstantPropagation due to its simplicity.

Classic data-flow analysis with topological worklist algorithm.

This is a variant of the SparseConstantPropagation, which computes which registers or memory cells are not re-defined (i.e., are untouched) within a function. This algorithm assumes that the SSA has been promoted.

Untouched value propagation value.

Defined variable.

Program point of a defined variable.