artdag: Phase 5 optimization — DCE + CSE + adjacent-op fusion + 18 tests
Some checks failed
Test, Build, and Deploy / test-build-deploy (push) Failing after 57s

optimize.sx adds three result-preserving passes: dce (keep outputs + ancestors,
preserve ids), cse (==build; structural sharing is free from content addressing),
and fuse (collapse 1-to-1 fusible unary chains into an artdag/pipeline node fed by
the chain head's input; leaves/fan-out/non-fusible ops never fuse). fusing-runner
replays pipeline stages, output-equivalent to the unfused dag. optimize 18/18,
total 87/87.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-06-07 12:08:12 +00:00
parent a2f4fb5e89
commit 228861215d
6 changed files with 391 additions and 9 deletions

View File

@@ -30,7 +30,7 @@ edges.
## Status (rolling)
`bash lib/artdag/conformance.sh`**69/69** (4 suites: dag, analyze, plan, execute)
`bash lib/artdag/conformance.sh`**87/87** (5 suites: dag, analyze, plan, execute, optimize)
## Ground rules
@@ -116,12 +116,12 @@ lib/artdag/optimize.sx lib/artdag/federation.sx
## Phase 5 — Effect-pipeline optimization
- [ ] `lib/artdag/optimize.sx` — rewrite the DAG before execution: dead-node
- [x] `lib/artdag/optimize.sx` — rewrite the DAG before execution: dead-node
elimination (unreachable from outputs), common-subexpression sharing (free from
content ids), adjacent-op fusion
- [ ] optimizations are content-id-preserving where semantically identical; assert
- [x] optimizations are content-id-preserving where semantically identical; assert
the optimized DAG produces identical results
- [ ] `lib/artdag/tests/optimize.sx` — DCE, CSE dedup, fusion equivalence
- [x] `lib/artdag/tests/optimize.sx` — DCE, CSE dedup, fusion equivalence
- [ ] (optional/later) rule-based optimization via `maude-on-sx`'s rewriting engine —
flag the integration point, don't block on it
@@ -136,6 +136,19 @@ lib/artdag/optimize.sx lib/artdag/federation.sx
## Progress log
- **Phase 5 — Effect-pipeline optimization** (optimize suite 18/18, total 87/87).
`lib/artdag/optimize.sx`: `artdag/dce dag outputs` keeps only the outputs plus
their transitive ancestors (via analyze), preserving surviving content-ids.
`artdag/cse` == build — structural sharing is inherent to content addressing, so
identical subexpressions collapse to one node/id and execute once (verified).
`artdag/fuse entries fusible?` rewrites entries: a maximal 1-to-1 chain of fusible
unary ops (predecessor used only by its single consumer, both fusible) collapses
into one `artdag/pipeline` node carrying ordered `{:op :params}` stages, fed by the
chain head's external input; leaves, fan-out nodes, and non-fusible ops never fuse.
`artdag/fusing-runner` wraps a base runner to replay pipeline stages — output
equivalent to the unfused DAG (asserted). Note: CSE auto-dedup means test fixtures
intended as distinct nodes must use distinct op/params.
- **Phase 4 — Execute (incremental + memoized)** (execute suite 15/15, total 69/69).
`lib/artdag/execute.sx`: `artdag/execute` folds a plan, computing each node via an
injected `runner (op params input-results)` (production = `perform` to JAX/IPFS