Bug 216056 - [bmalloc] Remove code duplication from small bump ranges.
Summary: [bmalloc] Remove code duplication from small bump ranges.
Status: ASSIGNED
Alias: None
Product: WebKit
Classification: Unclassified
Component: bmalloc (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Basuke Suzuki
URL:
Keywords: InRadar
Depends on:
Blocks:
 
Reported: 2020-09-01 15:37 PDT by Basuke Suzuki
Modified: 2020-09-08 15:38 PDT (History)
4 users (show)

See Also:

Attachments
PATCH (827 bytes, patch)
2020-09-02 20:54 PDT, Basuke Suzuki 		no flags Details | Formatted Diff | Diff
PATCH (8.95 KB, patch)
2020-09-02 20:57 PDT, Basuke Suzuki 		no flags Details | Formatted Diff | Diff
Fix debug build (8.26 KB, patch)
2020-09-03 10:35 PDT, Basuke Suzuki 		no flags Details | Formatted Diff | Diff
Refactoring (9.26 KB, patch)
2020-09-03 14:48 PDT, Basuke Suzuki 		ysuzuki: review+
Details | Formatted Diff | Diff
Show Obsolete (3) 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 Basuke Suzuki 2020-09-01 15:37:14 PDT 
Both byMetadata and byObject share same logic and can be refactored.
Comment 1 Basuke Suzuki 2020-09-02 20:54:29 PDT 
Created attachment 407860 [details]
PATCH
Comment 2 Basuke Suzuki 2020-09-02 20:57:17 PDT 
Created attachment 407861 [details]
PATCH
Comment 3 Basuke Suzuki 2020-09-03 10:35:03 PDT 
Created attachment 407890 [details]
Fix debug build
Comment 4 Basuke Suzuki 2020-09-03 14:48:05 PDT 
Created attachment 407916 [details]
Refactoring
Comment 5 Yusuke Suzuki 2020-09-04 12:51:40 PDT 
Comment on attachment 407916 [details]
Refactoring

View in context: https://bugs.webkit.org/attachment.cgi?id=407916&action=review

Can you run MallocBench to ensure that this does not affect on the performance? Because this code is super critical path in bmalloc.

> Source/bmalloc/bmalloc/Heap.cpp:372
> +inline void allocateSmallBumpRangesInline(

Let's put BINLINE since this is super critical path which we should ensure that this is always inlined regardless of compiler's decision.

> Source/bmalloc/bmalloc/Heap.cpp:385
> +            lineCache[page->sizeClass()].push(page);

Let's pass sizeClass and use it.

> Source/bmalloc/bmalloc/Heap.cpp:403
> +    auto sizeClass = page->sizeClass();

Let's use passed sizeClass.

> Source/bmalloc/bmalloc/Heap.h:172
> +        allocateSmallBumpRangesByMetadata(lock, page, allocator, rangeCache, lineCache);
> +    else
> +        allocateSmallBumpRangesByObject(lock, page, allocator, rangeCache, lineCache);

Let's just pass `sizeClass` too.
Comment 6 Radar WebKit Bug Importer 2020-09-08 15:38:16 PDT 
<rdar://problem/68534246>