Fix to resolve Windows 32-bit Out Of Memory benchmark tests by frostney · Pull Request #57 · frostney/GocciaScript

frostney · 2026-03-10T08:18:05Z

Initialize Souffle GC and fix related memory management issues to resolve Windows OOM in benchmarks.

The Souffle garbage collector was never initialized, causing all Souffle heap objects to accumulate indefinitely and leading to out-of-memory errors. Enabling the GC exposed several bugs in its mark phase, including missing roots and infinite recursion for cyclic references, which were addressed by fixing the marking logic and disabling automatic collection during VM execution, opting for explicit collection at controlled safe points.

Summary by CodeRabbit

Release Notes

Documentation
- Updated garbage collection lifecycle and initialization documentation
Bug Fixes
- Enhanced garbage collection system with additional safeguards to prevent redundant marking operations
- Improved external root coverage in garbage collection traversal

Root cause: TSouffleGarbageCollector.Initialize was never called anywhere in the codebase, so the Souffle GC singleton was nil. All AllocateObject calls were no-ops (guarded by 'if Assigned(FGC)'), and CollectIfNeeded never triggered. Souffle heap objects (closures, arrays, records, blueprints, strings, upvalues) accumulated without limit, causing EOutOfMemory on Windows after ~29s of benchmark execution. Changes: - Goccia.Engine.Backend.pas: Initialize Souffle GC in TGocciaSouffleBackend.Create (with Enabled := False to prevent automatic collection during execution), run full Collect in destructor, add finalization section for Shutdown - Souffle.Compound.pas: Add GCMarked cycle-breaking checks in TSouffleArray.MarkReferences and TSouffleRecord.MarkReferences (prevents infinite recursion on self-referential objects like globalThis.globalThis) - Souffle.VM.Closure.pas: Add GCMarked check in TSouffleClosure.MarkReferences for upvalue marking - Goccia.Runtime.Operations.pas: Add FMapDelegate, FSetDelegate, and FBlueprintSuperValues to MarkExternalRoots (previously unmarked Souffle heap objects) - Update AGENTS.md and docs/souffle-vm.md with GC lifecycle docs Automatic collection (CollectIfNeeded) is disabled because the mark phase does not yet cover all cross-GC roots during bridge calls. The per-backend-destruction Collect approach is sufficient to prevent OOM in multi-file scenarios (benchmarks, test suites). Co-authored-by: Johannes Stein <frostney@users.noreply.github.com>

cursor · 2026-03-10T08:18:06Z

Cursor Agent can help with this pull request. Just @cursor in comments and I'll start working on changes in this branch.
_{Learn more about Cursor Agents}

coderabbitai · 2026-03-10T08:18:25Z

📝 Walkthrough

Walkthrough

This PR implements garbage collection lifecycle management for the Souffle backend system. Changes include: initializing the GC singleton with automatic collection disabled, adding GCMarked guards to prevent redundant marking in Souffle compound objects and closures, integrating collection during backend destruction, expanding external-root marking coverage for super-values and delegates, and updating documentation to reflect the new GC workflow.

Changes

Cohort / File(s)	Summary
Documentation `AGENTS.md`, `docs/souffle-vm.md`	Updated documentation to describe GC lifecycle: singleton initialization with disabled automatic collection, full collection in backend destructor, and finalization shutdown. Also noted test suite expansion (3,433 tests) and current disablement of threshold-based collection.
GC Marking Guards `souffle/Souffle.Compound.pas`, `souffle/Souffle.VM.Closure.pas`	Added GCMarked condition checks before MarkReferences calls to prevent redundant marking of already-marked references in array/record entries and closure upvalues.
Backend Integration `units/Goccia.Engine.Backend.pas`	Integrated SouffleGarbageCollector lifecycle: initialization in constructor with automatic collection disabled, full collection triggered on destruction, and finalization block for shutdown.
Runtime Operations `units/Goccia.Runtime.Operations.pas`	Extended external-root marking in MarkExternalRoots to cover FBlueprintSuperValues and FMapDelegate/FSetDelegate, improving GC root coverage for bridge call paths.

Sequence Diagram

sequenceDiagram
    participant Backend as TGocciaSouffleBackend
    participant GC as TSouffleGarbageCollector
    participant Runtime as MarkExternalRoots
    participant SouffleObj as Souffle Objects<br/>(Compound, Closure)
    
    Backend->>GC: Create() - Initialize singleton
    Backend->>GC: Disable automatic collection
    note over GC: Enabled := False
    
    Runtime->>Runtime: Mark external roots
    Runtime->>SouffleObj: MarkReferences(obj)
    activate SouffleObj
    alt GCMarked = False
        SouffleObj->>SouffleObj: Recursively mark references
    else GCMarked = True
        note over SouffleObj: Skip (already marked)
    end
    deactivate SouffleObj
    
    Backend->>GC: Destroy() - Explicit collection
    activate GC
    GC->>GC: Collect() - Free Souffle heap
    deactivate GC
    
    note over Backend: Module Finalization
    Backend->>GC: Shutdown()

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Poem

🐰 The GC hops with purpose clear,
Marks are guarded, doubles disappear,
From backend birth to final rest,
Memory managed at its best! 🌱✨

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the primary focus of the pull request: resolving Windows 32-bit Out Of Memory issues in benchmark tests through Souffle GC initialization and memory management fixes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests
Post copyable unit tests in a comment
Commit unit tests in branch cursor/windows-benchmark-memory-4b72

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

This reverts commit 898ed1e.

github-actions · 2026-03-10T08:22:16Z

Benchmark Results

254 benchmarks (no baseline)

arraybuffer.js — 14 benchmarks

Benchmark	Interpreted	Bytecode
create ArrayBuffer(0)	422,249	145,291
create ArrayBuffer(64)	412,397	139,482
create ArrayBuffer(1024)	316,893	125,314
create ArrayBuffer(8192)	140,451	84,139
slice full buffer (64 bytes)	508,960	376,971
slice half buffer (512 of 1024 bytes)	428,321	328,982
slice with negative indices	442,373	363,622
slice empty range	494,561	371,496
byteLength access	1,476,389	1,062,730
Symbol.toStringTag access	1,149,794	538,217
ArrayBuffer.isView	694,775	457,297
clone ArrayBuffer(64)	371,582	320,622
clone ArrayBuffer(1024)	281,036	252,027
clone ArrayBuffer inside object	253,407	152,563

arrays.js — 19 benchmarks

Benchmark	Interpreted	Bytecode
Array.from length 100	13,489	12,703
Array.from 10 elements	222,793	150,637
Array.of 10 elements	298,912	214,100
spread into new array	331,314	564,308
map over 50 elements	27,000	22,714
filter over 50 elements	23,251	21,685
reduce sum 50 elements	26,075	19,516
forEach over 50 elements	22,635	24,008
find in 50 elements	33,795	26,512
sort 20 elements	11,714	3,359
flat nested array	116,931	286,571
flatMap	72,078	187,637
map inside map (5x5)	20,630	67,853
filter inside map (5x10)	15,996	13,546
reduce inside map (5x10)	18,823	13,743
forEach inside forEach (5x10)	16,127	15,277
find inside some (10x10)	13,171	10,621
map+filter chain nested (5x20)	5,100	4,494
reduce flatten (10x5)	37,282	4,894

async-await.js — 6 benchmarks

Benchmark	Interpreted	Bytecode
single await	335,971	243,655
multiple awaits	151,449	97,731
await non-Promise value	723,607	695,368
await with try/catch	328,074	235,162
await Promise.all	47,589	41,051
nested async function call	172,273	177,893

classes.js — 31 benchmarks

Benchmark	Interpreted	Bytecode
simple class new	112,080	338,019
class with defaults	91,419	235,280
50 instances via Array.from	5,540	6,231
instance method call	56,943	154,714
static method call	88,902	366,582
single-level inheritance	46,377	150,460
two-level inheritance	37,836	120,179
private field access	53,996	164,787
private methods	59,968	201,952
getter/setter access	62,998	171,058
class decorator (identity)	79,959	52,762
class decorator (wrapping)	48,190	36,543
identity method decorator	57,680	42,963
wrapping method decorator	47,651	39,437
stacked method decorators (x3)	33,058	28,014
identity field decorator	65,469	44,218
field initializer decorator	56,179	40,583
getter decorator (identity)	55,597	41,238
setter decorator (identity)	48,198	35,633
static method decorator	58,511	61,812
static field decorator	68,967	66,389
private method decorator	43,265	37,284
private field decorator	47,115	38,780
plain auto-accessor (no decorator)	86,015	53,731
auto-accessor with decorator	51,890	37,218
decorator writing metadata	40,308	41,574
static getter read	91,955	385,457
static getter/setter pair	70,820	207,175
inherited static getter	52,087	261,898
inherited static setter	60,370	208,366
inherited static getter with this binding	50,450	154,506

closures.js — 11 benchmarks

Benchmark	Interpreted	Bytecode
closure over single variable	121,837	599,398
closure over multiple variables	113,528	390,307
nested closures	111,960	564,883
function as argument	85,645	490,186
function returning function	102,832	592,481
compose two functions	60,536	347,462
fn.call	130,676	131,243
fn.apply	100,280	92,273
fn.bind	123,830	142,153
recursive sum to 50	10,680	38,656
recursive tree traversal	17,580	57,727

collections.js — 12 benchmarks

Benchmark	Interpreted	Bytecode
add 50 elements	6,851	5,716
has lookup (50 elements)	92,742	88,817
delete elements	48,128	36,968
forEach iteration	14,633	15,726
spread to array	27,716	23,856
deduplicate array	37,737	30,822
set 50 entries	5,555	5,777
get lookup (50 entries)	93,575	100,499
has check	137,886	159,128
delete entries	49,240	36,391
forEach iteration	15,053	16,008
keys/values/entries	7,732	102,415

destructuring.js — 22 benchmarks

Benchmark	Interpreted	Bytecode
simple array destructuring	367,698	606,066
with rest element	245,953	454,958
with defaults	356,282	646,820
skip elements	384,771	717,182
nested array destructuring	155,396	349,738
swap variables	454,697	976,512
simple object destructuring	260,559	521,333
with defaults	322,419	326,811
with renaming	275,234	572,795
nested object destructuring	124,896	251,266
rest properties	171,597	231,942
object parameter	81,563	187,369
array parameter	109,030	319,979
mixed destructuring in map	33,674	35,424
forEach with array destructuring	59,446	136,690
map with array destructuring	54,813	165,829
filter with array destructuring	57,837	174,596
reduce with array destructuring	65,453	169,881
map with object destructuring	76,072	65,988
map with nested destructuring	59,655	55,916
map with rest in destructuring	38,796	46,374
map with defaults in destructuring	53,937	36,415

fibonacci.js — 8 benchmarks

Benchmark	Interpreted	Bytecode
recursive fib(15)	301	1,130
recursive fib(20)	27	100
recursive fib(15) typed	332	1,435
recursive fib(20) typed	28	125
iterative fib(20) via reduce	10,615	16,807
iterator fib(20)	8,458	10,480
iterator fib(20) via Iterator.from + take	8,207	8,644
iterator fib(20) last value via reduce	6,823	7,837

for-of.js — 7 benchmarks

Benchmark	Interpreted	Bytecode
for...of with 10-element array	48,420	—
for...of with 100-element array	5,438	—
for...of with string (10 chars)	34,233	—
for...of with Set (10 elements)	42,783	—
for...of with Map entries (10 entries)	25,237	11,459
for...of with destructuring	35,968	—
for-await-of with sync array	38,628	44,916

iterators.js — 20 benchmarks

Benchmark	Interpreted	Bytecode
Iterator.from({next}).toArray() — 20 elements	10,197	—
Iterator.from({next}).toArray() — 50 elements	4,312	—
spread pre-wrapped iterator — 20 elements	9,671	—
Iterator.from({next}).forEach — 50 elements	3,268	—
Iterator.from({next}).reduce — 50 elements	3,299	3,524
wrap array iterator	55,339	—
wrap plain {next()} object	6,863	—
map + toArray (50 elements)	3,098	3,124
filter + toArray (50 elements)	3,224	—
take(10) + toArray (50 element source)	16,332	—
drop(40) + toArray (50 element source)	4,170	—
chained map + filter + take (100 element source)	4,859	4,861
some + every (50 elements)	1,925	—
find (50 elements)	4,161	—
array.values().map().filter().toArray()	3,542	4,318
array.values().take(5).toArray()	54,658	54,506
array.values().drop(45).toArray()	15,540	—
map.entries() chained helpers	5,257	2,156
set.values() chained helpers	8,286	—
string iterator map + toArray	7,016	—

json.js — 20 benchmarks

Benchmark	Interpreted	Bytecode
parse simple object	157,042	—
parse nested object	103,222	—
parse array of objects	50,338	—
parse large flat object	47,884	—
parse mixed types	68,244	—
stringify simple object	140,722	—
stringify nested object	73,131	—
stringify array of objects	36,233	—
stringify mixed types	69,972	—
reviver doubles numbers	43,090	—
reviver filters properties	37,565	—
reviver on nested object	51,218	—
reviver on array	26,933	—
replacer function doubles numbers	44,240	—
replacer function excludes properties	56,984	—
array replacer (allowlist)	106,125	—
stringify with 2-space indent	81,519	—
stringify with tab indent	80,115	—
parse then stringify	46,498	—
stringify then parse	27,571	—

jsx.jsx — 21 benchmarks

Benchmark	Interpreted	Bytecode
simple element	194,182	—
self-closing element	207,590	—
element with string attribute	165,602	—
element with multiple attributes	139,748	—
element with expression attribute	151,066	—
text child	183,182	—
expression child	183,823	—
mixed text and expression	173,959	—
nested elements (3 levels)	68,167	—
sibling children	52,129	—
component element	131,000	—
component with children	83,208	—
dotted component	115,915	—
empty fragment	196,380	—
fragment with children	51,458	—
spread attributes	104,497	—
spread with overrides	85,381	—
shorthand props	138,931	—
nav bar structure	24,778	—
card component tree	30,072	—
10 list items via Array.from	13,146	—

numbers.js — 11 benchmarks

Benchmark	Interpreted	Bytecode
integer arithmetic	504,828	—
floating point arithmetic	553,478	—
number coercion	182,351	—
toFixed	99,129	—
toString	149,344	—
valueOf	201,321	—
toPrecision	130,521	—
Number.isNaN	273,986	—
Number.isFinite	274,944	—
Number.isInteger	274,878	—
Number.parseInt and parseFloat	231,348	—

objects.js — 7 benchmarks

Benchmark	Interpreted	Bytecode
create simple object	448,111	—
create nested object	190,468	—
create 50 objects via Array.from	8,637	—
property read	582,384	—
Object.keys	283,693	—
Object.entries	96,056	—
spread operator	169,435	—

promises.js — 12 benchmarks

Benchmark	Interpreted	Bytecode
Promise.resolve(value)	457,940	—
new Promise(resolve => resolve(value))	159,379	—
Promise.reject(reason)	481,221	—
resolve + then (1 handler)	144,108	—
resolve + then chain (3 deep)	57,243	—
resolve + then chain (10 deep)	18,900	—
reject + catch + then	81,216	—
resolve + finally + then	68,526	—
Promise.all (5 resolved)	27,219	—
Promise.race (5 resolved)	27,322	—
Promise.allSettled (5 mixed)	21,920	—
Promise.any (5 mixed)	25,878	—

strings.js — 11 benchmarks

Benchmark	Interpreted	Bytecode
string concatenation	376,324	—
template literal	407,572	—
string repeat	366,939	—
split and join	132,275	—
indexOf and includes	165,063	—
toUpperCase and toLowerCase	229,067	—
slice and substring	149,675	—
trim operations	169,728	—
replace and replaceAll	227,497	—
startsWith and endsWith	123,754	—
padStart and padEnd	185,157	—

typed-arrays.js — 22 benchmarks

Benchmark	Interpreted	Bytecode
new Int32Array(0)	285,132	—
new Int32Array(100)	266,398	—
new Int32Array(1000)	161,899	—
new Float64Array(100)	250,099	—
Int32Array.from([...])	171,636	—
Int32Array.of(1, 2, 3, 4, 5)	275,273	—
sequential write 100 elements	3,450	—
sequential read 100 elements	3,454	—
Float64Array write 100 elements	3,225	—
fill(42)	45,539	—
slice()	189,281	—
map(x => x * 2)	6,500	—
filter(x => x > 50)	7,432	—
reduce (sum)	7,247	—
sort()	174,546	—
indexOf()	424,188	—
reverse()	321,873	—
create view over existing buffer	360,178	—
subarray()	438,530	—
set() from array	565,965	—
for-of loop	4,760	—
spread into array	15,472	—

_{Measured on ubuntu-latest x64. Changes within ±7% are considered insignificant.}

The Windows CI uses i386-win32 FPC (32-bit), limiting the process to ~2 GB of address space. In bytecode mode, the Souffle VM creates heap objects (closures, arrays, records, blueprints) that were never tracked or collected — TSouffleGarbageCollector.Initialize was never called anywhere in the codebase. On 64-bit platforms this leak is invisible (sufficient address space), but on 32-bit Windows the benchmark runner exhausts memory after ~29 seconds across 254 benchmarks. Initialize the Souffle GC in TGocciaSouffleBackend.Create with automatic collection disabled (Enabled := False). A full Collect runs in the destructor to free all Souffle heap objects between backend instances. Also fix latent bugs in the Souffle GC mark phase (never triggered before because the GC was never initialized): - Add GCMarked cycle-breaking checks in TSouffleArray.MarkReferences, TSouffleRecord.MarkReferences, and TSouffleClosure.MarkReferences - Add FMapDelegate, FSetDelegate, and FBlueprintSuperValues to MarkExternalRoots Mid-execution Collect (like the GocciaScript GC does between benchmark measurement rounds) is not yet possible because the Souffle mark phase does not cover all cross-GC roots in bridge call paths. Per-backend destruction collection is sufficient to bound memory on 32-bit. Co-authored-by: Johannes Stein <frostney@users.noreply.github.com>

coderabbitai

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@AGENTS.md`:
- Line 173: The documented test count is inconsistent: AGENTS.md currently
states "3,433" while docs/souffle-vm.md's "Known Limitations" section still
shows "3,406"; update the numeric string in both documents so they match
(replace "3,406" with "3,433" or vice versa to the authoritative current value)
and ensure the "Known Limitations" heading text remains intact so search/links
still work.

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: c58fa0f1-3c2e-4802-a386-79a6855472e7

📥 Commits

Reviewing files that changed from the base of the PR and between d2aea53 and 795ad62.

📒 Files selected for processing (6)

AGENTS.md
docs/souffle-vm.md
souffle/Souffle.Compound.pas
souffle/Souffle.VM.Closure.pas
units/Goccia.Engine.Backend.pas
units/Goccia.Runtime.Operations.pas

coderabbitai · 2026-03-10T09:08:26Z

 | GocciaScript Runtime | `Goccia.Runtime.Operations.pas` | `TGocciaRuntimeOperations` — GocciaScript semantics for Souffle VM, bridge caches (`FClosureBridgeCache`, `FArrayBridgeCache`, `FArrayBridgeReverse`, `FRecordBridgeCache`, `FBlueprintBridgeCache`), cross-GC rooting, array bridge sync (`SyncCachedGocciaToSouffle` at bridge entry, `SyncSouffleArrayToCache` after native mutations), native delegate methods (array, string, number, Map, Set), `FBlueprintSuperValues` for built-in subclass static property inheritance |

-**Souffle VM known limitations:** The bytecode backend passes 100% of the test suite (3,406 tests). Most language features execute natively in the VM: user-defined classes with constructors, named methods, getters, setters, static members, private fields/methods, computed property names, and instance fields compile to `TSouffleBlueprint` with `__fields__` initializer closures; classes extending built-in constructors (`Array`, `Map`, `Set`, `Promise`, `Object`) compile natively via `FBlueprintSuperValues` and `TGocciaSuperCallHelper`; array and string iteration uses `TGocciaSouffleArrayIterator`/`TGocciaSouffleStringIterator`; Map and Set iteration uses direct `TGocciaMapIteratorValue`/`TGocciaSetIteratorValue` creation; built-in constructors use a `TGocciaBridgedFunction` fast path; Map/Set property access and method calls use native delegates (`FMapDelegate`/`FSetDelegate`). The bridge still delegates to the interpreter for: module imports, async/await (microtask queue), and decorator evaluation. Bridged arrays use a dual-representation model with `SyncCachedGocciaToSouffle` at bridge entry and `SyncSouffleArrayToCache` after native mutations; `FArrayBridgeReverse` preserves reference identity across round-trips. `FBlueprintBridgeCache` persists across bridge calls to maintain `instanceof` identity for blueprint-backed classes. Only classes with decorators are deferred to the interpreter via `FPendingClasses` / `GOCCIA_EXT_EVAL_CLASS`. `.sbc` binary format uses native endianness (not yet portable). ABC-encoded instructions limit constant pool references to 255 per prototype. See [docs/souffle-vm.md § Known Limitations](docs/souffle-vm.md#known-limitations) for the full list.
+**Souffle VM known limitations:** The bytecode backend passes 100% of the test suite (3,433 tests). Most language features execute natively in the VM: user-defined classes with constructors, named methods, getters, setters, static members, private fields/methods, computed property names, and instance fields compile to `TSouffleBlueprint` with `__fields__` initializer closures; classes extending built-in constructors (`Array`, `Map`, `Set`, `Promise`, `Object`) compile natively via `FBlueprintSuperValues` and `TGocciaSuperCallHelper`; array and string iteration uses `TGocciaSouffleArrayIterator`/`TGocciaSouffleStringIterator`; Map and Set iteration uses direct `TGocciaMapIteratorValue`/`TGocciaSetIteratorValue` creation; built-in constructors use a `TGocciaBridgedFunction` fast path; Map/Set property access and method calls use native delegates (`FMapDelegate`/`FSetDelegate`). The bridge still delegates to the interpreter for: module imports, async/await (microtask queue), and decorator evaluation. Bridged arrays use a dual-representation model with `SyncCachedGocciaToSouffle` at bridge entry and `SyncSouffleArrayToCache` after native mutations; `FArrayBridgeReverse` preserves reference identity across round-trips. `FBlueprintBridgeCache` persists across bridge calls to maintain `instanceof` identity for blueprint-backed classes. Only classes with decorators are deferred to the interpreter via `FPendingClasses` / `GOCCIA_EXT_EVAL_CLASS`. `.sbc` binary format uses native endianness (not yet portable). ABC-encoded instructions limit constant pool references to 255 per prototype. The Souffle GC is initialized by `TGocciaSouffleBackend.Create` with automatic collection disabled; a full `Collect` runs in the backend destructor to free Souffle heap objects between files (critical on 32-bit Windows where address space is limited to ~2 GB). See [docs/souffle-vm.md § Known Limitations](docs/souffle-vm.md#known-limitations) for the full list.


⚠️ Potential issue | 🟡 Minor

Keep the test-count docs in sync.

This now says 3,433 tests, but docs/souffle-vm.md still says 3,406 in the Known Limitations section. Please update both sources together so contributors do not end up with conflicting baseline numbers.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed. In `@AGENTS.md` at line 173, The documented test count is inconsistent: AGENTS.md currently states "3,433" while docs/souffle-vm.md's "Known Limitations" section still shows "3,406"; update the numeric string in both documents so they match (replace "3,406" with "3,433" or vice versa to the authoritative current value) and ensure the "Known Limitations" heading text remains intact so search/links still work.

Revert "Fix Windows OOM in bytecode benchmarks: initialize Souffle GC"

4e183e1

This reverts commit 898ed1e.

cursor Bot force-pushed the cursor/windows-benchmark-memory-4b72 branch from a11a150 to 795ad62 Compare March 10, 2026 08:58

frostney marked this pull request as ready for review March 10, 2026 09:02

frostney changed the title ~~Windows benchmark memory~~ Fix to resolve Windows 32-bit Out Of Memory benchmark tests Mar 10, 2026

coderabbitai Bot reviewed Mar 10, 2026

View reviewed changes

frostney merged commit b78b6f9 into main Mar 10, 2026
7 checks passed

frostney deleted the cursor/windows-benchmark-memory-4b72 branch March 10, 2026 09:13

This was referenced Mar 10, 2026

Unify garbage collector #60

Merged

Garbage collector memory optimization #67

Merged

Experiment: 8-byte TSouffleValue with reference pools #97

Closed

coderabbitai Bot mentioned this pull request Mar 26, 2026

Refactor environment bootstrap and simplify bridge setup #135

Closed

frostney added the bug Something isn't working label Apr 9, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix to resolve Windows 32-bit Out Of Memory benchmark tests#57

Fix to resolve Windows 32-bit Out Of Memory benchmark tests#57
frostney merged 3 commits into
mainfrom
cursor/windows-benchmark-memory-4b72

frostney commented Mar 10, 2026 •

edited by coderabbitai Bot

Loading

Uh oh!

cursor Bot commented Mar 10, 2026

Uh oh!

coderabbitai Bot commented Mar 10, 2026 •

edited

Loading

Walkthrough

Changes

Sequence Diagram

Estimated code review effort

Poem

Uh oh!

github-actions Bot commented Mar 10, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot left a comment

Uh oh!

coderabbitai Bot Mar 10, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

frostney commented Mar 10, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary by CodeRabbit

Release Notes

Uh oh!

cursor Bot commented Mar 10, 2026

Uh oh!

coderabbitai Bot commented Mar 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Sequence Diagram

Estimated code review effort

Poem

Uh oh!

github-actions Bot commented Mar 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmark Results

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

coderabbitai Bot Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

frostney commented Mar 10, 2026 •

edited by coderabbitai Bot

Loading

coderabbitai Bot commented Mar 10, 2026 •

edited

Loading

github-actions Bot commented Mar 10, 2026 •

edited

Loading