Semantic Optimization Pipeline
10 sections
0%
Section 1 Hygiene & Targeted Fixes Section 2 LLVM Metadata Infrastructure Section 3 AIMS State Export to Codegen Section 4 TBAA, Range, and Invariant Metadata Section 5 Structural Derive Normalization Section 6 Runtime Identity Fixes Section 7 Algebra Law Schema Section 8 Algebra Law Adoption Section 9 Algebraic Normalization Pass Section 10 Advanced Rewrites & Verification
Overview
Make proven facts survive into optimization — including algebraic structure. Ori's compiler already proves rich semantic facts about types, ownership, purity, and structure through the trait system, AIMS lattice, ARC pipeline, and type checker. But too many of these facts are consumed internally and discarded before they can benefit LLVM's optimizer. This plan extends existing infrastructure so that facts the compiler already proves travel one phase farther in the pipeline, and adds algebraic law metadata so that user-defined types (matrix, vector, bignum) receive the same algebraic optimizations that built-in types get today.
Planned
10 sections
Section 1 Not Started
Hygiene & Targeted Fixes
0/56 tasks
0/56 tasks
Section 2 Not Started
LLVM Metadata Infrastructure
0/45 tasks
0/45 tasks
Section 3 Not Started
AIMS State Export to Codegen
0/48 tasks
0/48 tasks
Section 4 Not Started
TBAA, Range, and Invariant Metadata
0/46 tasks
0/46 tasks
Section 5 Not Started
Structural Derive Normalization
0/47 tasks
0/47 tasks
Section 6 Not Started
Runtime Identity Fixes
0/40 tasks
0/40 tasks
Section 7 Not Started
Algebra Law Schema
0/86 tasks
0/86 tasks
Section 8 Not Started
Algebra Law Adoption
0/56 tasks
0/56 tasks
Section 9 Not Started
Algebraic Normalization Pass
0/90 tasks
0/90 tasks
Section 10 Not Started
Advanced Rewrites & Verification
0/86 tasks
0/86 tasks