Compiler optimizations: hoisting, inlining, unrolling and coordinate localization across every emitting tier - #872

Open

fuzzie360 wants to merge 9 commits into

feature/compiler-optimizations

Commits on Aug 4, 2026

docs: compiler-optimizations design contract

fuzzie360
and
claude
committed
9f9c6b2
feat(optimizer): the compiler-optimization pass, plus transform H

fuzzie360
and
claude
committed
e9d39e7
feat(optimizer): transform T3 (literal-loop unrolling) and T1 (cpu coordinates)

fuzzie360
and
claude
committed
3811c11
feat(optimizer): transform T2 (helper inlining) across every emitting tier

fuzzie360
and
claude
committed
f927e33
docs(optimizer): README section and the final measured numbers

fuzzie360
and
claude
committed
8939498
fix(optimizer): the ten findings of the build review

fuzzie360
and
claude
committed
84717c4
docs(optimizer): refresh the measured tables after the review fixes

fuzzie360
and
claude
committed
081694f
perf(optimizer): helper inlining is off on cpu, on everywhere else

fuzzie360
and
claude
committed
12d83cd

Commits on Aug 5, 2026

perf(optimizer): T1 coordinate localization ships off -- measured a loss

fuzzie360
and
claude
committed
154cbb0