Bug 186740 - CodeBlockSet wastes 190KB of HashTable capacity on nytimes.com
Summary: CodeBlockSet wastes 190KB of HashTable capacity on nytimes.com
Status: NEW
Alias: None
Product: WebKit
Classification: Unclassified
Component: JavaScriptCore (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Yusuke Suzuki
URL:
Keywords: InRadar
Depends on:
Blocks:
 
Reported: 2018-06-16 12:41 PDT by Simon Fraser (smfr)
Modified: 2019-05-07 02:25 PDT (History)
10 users (show)

See Also:

Attachments
Patch (11.31 KB, patch)
2019-04-28 02:19 PDT, Yusuke Suzuki 		no flags Details | Formatted Diff | Diff
Patch (20.81 KB, patch)
2019-04-28 04:29 PDT, Yusuke Suzuki 		no flags Details | Formatted Diff | Diff
Patch (62.33 KB, patch)
2019-05-07 02:25 PDT, Yusuke Suzuki 		no flags Details | Formatted Diff | Diff
Show Obsolete (2) View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Simon Fraser (smfr) 2018-06-16 12:41:34 PDT 
Using tooling from bug 186698, loading can and running "notifyutil -p com.apple.WebKit.dumpHashTableCapacity" shows:

Wasted capacity: 189656 bytes (used 72488 of 262144 bytes, utilization: 27.65%) - 1 allocations
1   0x116005465 WTF::HashTable<JSC::CodeBlock*, JSC::CodeBlock*, WTF::IdentityExtractor, WTF::PtrHash<JSC::CodeBlock*>, WTF::HashTraits<JSC::CodeBlock*>, WTF::HashTraits<JSC::CodeBlock*> >::HashTable()
2   0x116005445 WTF::HashSet<JSC::CodeBlock*, WTF::PtrHash<JSC::CodeBlock*>, WTF::HashTraits<JSC::CodeBlock*> >::HashSet()
3   0x116001de5 WTF::HashSet<JSC::CodeBlock*, WTF::PtrHash<JSC::CodeBlock*>, WTF::HashTraits<JSC::CodeBlock*> >::HashSet()
4   0x116001da9 JSC::CodeBlockSet::CodeBlockSet()
5   0x116001e05 JSC::CodeBlockSet::CodeBlockSet()
6   0x11600ff64 JSC::Heap::Heap(JSC::VM*, JSC::HeapType)
7   0x1160126b3 JSC::Heap::Heap(JSC::VM*, JSC::HeapType)
8   0x1166b8e72 JSC::VM::VM(JSC::VM::VMType, JSC::HeapType)
Comment 1 Radar WebKit Bug Importer 2018-06-16 12:42:55 PDT 
<rdar://problem/41189289>
Comment 2 Yusuke Suzuki 2019-04-27 23:46:30 PDT 
CodeBlockSet is basically duplicate to IsoSubspace's set feature. But it exists for Debuggers and Sampling Profilers. If they are not enabled, we should not use this.
Comment 3 Yusuke Suzuki 2019-04-28 02:19:41 PDT 
Created attachment 368426 [details]
Patch

WIP
Comment 4 Yusuke Suzuki 2019-04-28 04:25:44 PDT 
I think we can implement CodeBlockSet with IsoCellSet with a bit interesting lock to support SamplingProfiler etc.
Comment 5 Yusuke Suzuki 2019-04-28 04:29:12 PDT 
Created attachment 368428 [details]
Patch

WIP
Comment 6 jsc-armv7 EWS 2019-04-28 06:50:42 PDT 
Comment on attachment 368428 [details]
Patch

Attachment 368428 [details] did not pass jsc-armv7-ews (jsc-only):
Output: https://webkit-queues.webkit.org/results/12022531

New failing tests:
stress/scope-operation-cache-global-property-bump-counter.js.default
apiTests
Comment 7 EWS Watchlist 2019-04-28 06:57:15 PDT 
Comment on attachment 368428 [details]
Patch

Attachment 368428 [details] did not pass jsc-ews (mac):
Output: https://webkit-queues.webkit.org/results/12022539

New failing tests:
stress/scope-operation-cache-global-property-bump-counter.js.default
Comment 8 Yusuke Suzuki 2019-05-07 02:25:27 PDT 
Created attachment 369261 [details]
Patch