john/cell
1
0
Fork 0
Code Issues 31 Pull Requests Actions Packages Projects Releases 2 Wiki Activity

Merge branch 'fix_heap_closure'

Browse Source
This commit is contained in:
John Alanbrook
2026-02-18 12:46:18 -06:00
parent 6f8cad9bb2 4868a50085
commit 91b73f923a
13 changed files with 124048 additions and 121950 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
docs/spec/pipeline.md
View File

@@ -62,12 +62,13 @@ See [Mcode IR](mcode.md) for the instruction format and complete instruction ref
Optimizes the Mcode IR through a series of independent passes. Operates per-function:
1. **Backward type inference**: Infers parameter types from how they are used in typed operators (`add_int`, `store_index`, `load_field`, `push`, `pop`, etc.). Immutable `def` parameters keep their inferred type across label join points.
2. **Type-check elimination**: When a slot's type is known, eliminates `is_<type>` + conditional jump pairs. Narrows `load_dynamic`/`store_dynamic` to typed variants.
3. **Algebraic simplification**: Rewrites identity operations (add 0, multiply 1, divide 1) and folds same-slot comparisons.
4. **Boolean simplification**: Fuses `not` + conditional jump into a single jump with inverted condition.
5. **Move elimination**: Removes self-moves (`move a, a`).
6. **Unreachable elimination**: Nops dead code after `return` until the next label.
7. **Dead jump elimination**: Removes jumps to the immediately following label.
2. **Write-type invariance**: Determines which local slots have a consistent write type across all instructions. Slots written by child closures (via `put`) are excluded (forced to unknown).
3. **Type-check elimination**: When a slot's type is known, eliminates `is_<type>` + conditional jump pairs. Narrows `load_dynamic`/`store_dynamic` to typed variants.
4. **Algebraic simplification**: Rewrites identity operations (add 0, multiply 1, divide 1) and folds same-slot comparisons.
5. **Boolean simplification**: Fuses `not` + conditional jump into a single jump with inverted condition.
6. **Move elimination**: Removes self-moves (`move a, a`).
7. **Unreachable elimination**: Nops dead code after `return` until the next label.
8. **Dead jump elimination**: Removes jumps to the immediately following label.
See [Streamline Optimizer](streamline.md) for detailed pass descriptions.

4
docs/spec/streamline.md
View File

@@ -95,7 +95,9 @@ Write type mapping:
| `move`, `load_field`, `load_index`, `load_dynamic`, `pop`, `get` | T_UNKNOWN |
| `invoke`, `tail_invoke` | T_UNKNOWN |
The result is a map of slot→type for slots where all writes agree on a single known type. Parameter slots (1..nr_args) and slot 0 are excluded.
Before filtering, a pre-pass (`mark_closure_writes`) scans all inner functions for `put` instructions (closure variable writes). For each `put`, the pass traverses the parent chain to find the target ancestor function and marks the written slot as `closure_written`. Slots marked as closure-written are forced to T_UNKNOWN regardless of what the local write analysis infers, because the actual runtime write happens in a child function and can produce any type.
The result is a map of slot→type for slots where all writes agree on a single known type. Parameter slots (1..nr_args) and slot 0 are excluded. Closure-written slots are excluded (forced to unknown).
Common patterns this enables: