Bug 154436

Summary:

bmalloc: Remove the concept of medium objects

Product:

WebKit

Reporter:

Geoffrey Garen <ggaren>

Component:

New Bugs

Assignee:

Geoffrey Garen <ggaren>

Status:

RESOLVED FIXED

Severity:

Normal

CC:

barraclough, commit-queue, kling, msaboff, ossy

Priority:

Version:

WebKit Nightly Build

Hardware:

Unspecified

OS:

Unspecified

Attachments:

Description	Flags
Patch	sam: review+

Geoffrey Garen

Reported 2016-02-18 22:24:07 PST

bmalloc: Remove the concept of medium objects

Attachments
Patch (55.07 KB, patch) 2016-02-18 22:47 PST, Geoffrey Garen	sam: review+	Details Formatted Diff Diff
View All Add attachment proposed patch, testcase, etc.

Geoffrey Garen

Comment 1 2016-02-18 22:47:07 PST

Created attachment 271733 [details] Patch

WebKit Commit Bot

Comment 2 2016-02-18 22:47:57 PST

Attachment 271733 [details] did not pass style-queue: ERROR: Source/bmalloc/bmalloc/Heap.cpp:169: Place brace on its own line for function definitions. [whitespace/braces] [4] Total errors found: 1 in 16 files If any of these errors are false positives, please file a bug against check-webkit-style.

Geoffrey Garen

Comment 3 2016-02-18 22:48:24 PST

~/OpenSource/PerformanceTests/MallocBench> ./run-malloc-benchmarks Baseline:~/OpenSource/WebKitBuildBaseline/Release/ Patch:~/OpenSource/WebKitBuild/Release/ Baseline Patch Δ Execution Time: churn 88ms 80ms ^ 1.1x faster list_allocate 88ms 83ms ^ 1.06x faster tree_allocate 88ms 84ms ^ 1.05x faster tree_churn 92ms 90ms ^ 1.02x faster fragment 104ms 99ms ^ 1.05x faster fragment_iterate 68ms 66ms ^ 1.03x faster medium 202ms 203ms ! 1.0x slower big 152ms 152ms facebook 431ms 435ms ! 1.01x slower reddit 89ms 91ms ! 1.02x slower flickr 94ms 99ms ! 1.05x slower theverge 114ms 119ms ! 1.04x slower message_one 225ms 221ms ^ 1.02x faster message_many 121ms 119ms ^ 1.02x faster churn --parallel 40ms 39ms ^ 1.03x faster list_allocate --parallel 73ms 70ms ^ 1.04x faster tree_allocate --parallel 88ms 85ms ^ 1.04x faster tree_churn --parallel 86ms 85ms ^ 1.01x faster fragment --parallel 55ms 52ms ^ 1.06x faster fragment_iterate --parallel 33ms 32ms ^ 1.03x faster medium --parallel 189ms 188ms ^ 1.01x faster big --parallel 135ms 136ms ! 1.01x slower <geometric mean> 102ms 100ms ^ 1.02x faster <arithmetic mean> 121ms 119ms ^ 1.01x faster <harmonic mean> 88ms 86ms ^ 1.03x faster Peak Memory: churn 780kB 784kB ! 1.01x bigger list_allocate 2,100kB 2,100kB tree_allocate 5,536kB 5,512kB ^ 1.0x smaller tree_churn 4,828kB 4,808kB ^ 1.0x smaller fragment 7,048kB 7,024kB ^ 1.0x smaller fragment_iterate 25,824kB 25,712kB ^ 1.0x smaller medium 1,071,080kB 1,071,084kB ! 1.0x bigger big 1,064,808kB 1,065,832kB ! 1.0x bigger facebook 77,932kB 77,704kB ^ 1.0x smaller reddit 15,060kB 14,100kB ^ 1.07x smaller flickr 27,476kB 25,356kB ^ 1.08x smaller theverge 28,840kB 26,764kB ^ 1.08x smaller message_one 4,632kB 4,516kB ^ 1.03x smaller message_many 5,036kB 4,928kB ^ 1.02x smaller churn --parallel 968kB 948kB ^ 1.02x smaller list_allocate --parallel 2,212kB 2,248kB ! 1.02x bigger tree_allocate --parallel 3,364kB 3,308kB ^ 1.02x smaller tree_churn --parallel 3,040kB 3,040kB fragment --parallel 7,196kB 7,176kB ^ 1.0x smaller fragment_iterate --parallel 25,944kB 25,848kB ^ 1.0x smaller medium --parallel 1,069,564kB 1,070,048kB ! 1.0x bigger big --parallel 1,060,860kB 1,057,056kB ^ 1.0x smaller <geometric mean> 16,808kB 16,571kB ^ 1.01x smaller <arithmetic mean> 205,188kB 204,813kB ^ 1.0x smaller <harmonic mean> 4,244kB 4,215kB ^ 1.01x smaller Memory at End: churn 376kB 380kB ! 1.01x bigger list_allocate 412kB 412kB tree_allocate 512kB 488kB ^ 1.05x smaller tree_churn 484kB 464kB ^ 1.04x smaller fragment 500kB 476kB ^ 1.05x smaller fragment_iterate 832kB 720kB ^ 1.16x smaller medium 6,704kB 6,708kB ! 1.0x bigger big 6,692kB 6,692kB facebook 3,484kB 3,672kB ! 1.05x bigger reddit 2,132kB 2,268kB ! 1.06x bigger flickr 3,196kB 3,268kB ! 1.02x bigger theverge 3,280kB 3,340kB ! 1.02x bigger message_one 696kB 680kB ^ 1.02x smaller message_many 1,120kB 1,176kB ! 1.05x bigger churn --parallel 564kB 540kB ^ 1.04x smaller list_allocate --parallel 588kB 588kB tree_allocate --parallel 720kB 688kB ^ 1.05x smaller tree_churn --parallel 1,140kB 1,180kB ! 1.04x bigger fragment --parallel 1,352kB 1,356kB ! 1.0x bigger fragment_iterate --parallel 1,044kB 908kB ^ 1.15x smaller medium --parallel 6,836kB 6,824kB ^ 1.0x smaller big --parallel 6,800kB 6,792kB ^ 1.0x smaller <geometric mean> 1,353kB 1,336kB ^ 1.01x smaller <arithmetic mean> 2,248kB 2,255kB ! 1.0x bigger <harmonic mean> 915kB 892kB ^ 1.03x smaller

Geoffrey Garen

Comment 4 2016-02-18 22:48:42 PST

~/OpenSource/PerformanceTests/MallocBench> ./run-malloc-benchmarks Baseline:~/OpenSource/WebKitBuildBaseline/Release/ Patch:~/OpenSource/WebKitBuild/Release/ Baseline Patch Δ Peak Memory: reddit_memory_warning --runs 0 14,940kB 14,000kB ^ 1.07x smaller flickr_memory_warning --runs 0 27,360kB 25,356kB ^ 1.08x smaller theverge_memory_warning --runs 0 28,504kB 26,500kB ^ 1.08x smaller <geometric mean> 22,670kB 21,110kB ^ 1.07x smaller <arithmetic mean> 23,601kB 21,952kB ^ 1.08x smaller <harmonic mean> 21,650kB 20,188kB ^ 1.07x smaller Memory at End: reddit_memory_warning --runs 0 9,604kB 9,264kB ^ 1.04x smaller flickr_memory_warning --runs 0 12,576kB 12,312kB ^ 1.02x smaller theverge_memory_warning --runs 0 22,344kB 21,252kB ^ 1.05x smaller <geometric mean> 13,923kB 13,433kB ^ 1.04x smaller <arithmetic mean> 14,841kB 14,276kB ^ 1.04x smaller <harmonic mean> 13,135kB 12,700kB ^ 1.03x smaller

Geoffrey Garen

Comment 5 2016-02-19 16:03:52 PST

Committed r196845: <http://trac.webkit.org/changeset/196845>

Note You need to log in before you can comment on or make changes to this bug.