Bug 116066

Summary:

Do not bloat HTMLTokenizer with a giant inline InputStreamPreprocessor::peek

Product:

WebKit

Reporter:

Ryosuke Niwa <rniwa>

Component:

Page Loading

Assignee:

Benjamin Poulain <benjamin>

Status:

RESOLVED FIXED

Severity:

Normal

CC:

benjamin, commit-queue, esprehn+autocc, kling, koivisto

Priority:

Version:

528+ (Nightly build)

Hardware:

Unspecified

OS:

Unspecified

Attachments:

Description	Flags
Patch	rniwa: review+

Description Ryosuke Niwa 2013-05-13 15:59:10 PDT

We should probably merge
https://chromium.googlesource.com/chromium/blink/+/45e0337b2f2db535ab08365f6e763a5015e4d990

Shrunk HTMLTokenizer.o by 28.2%

This CL moves the cold code in InputStreamPreprocessor::peek out-of-line, which
shrunks HTMLTokenizer.o on Linux by 28.2%. The CL also makes the
html-parser.html performance test faster by 1.3% (average of 15 runs).

Comment 1 Alexey Proskuryakov 2013-05-14 09:55:15 PDT

Size of a raw object file is a funny metric - perhaps useful as a tool to discover optimization opportunities, but certainly not a goal in itself :)

> html-parser.html performance test faster by 1.3% (average of 15 runs).

This is of course more interesting.

Comment 2 Benjamin Poulain 2013-05-14 13:41:57 PDT

(In reply to comment #1)
> Size of a raw object file is a funny metric - perhaps useful as a tool to discover optimization opportunities, but certainly not a goal in itself :)

We were as confused as you by that claim.

Comment 3 Benjamin Poulain 2013-05-14 13:42:14 PDT

I'll take this one.

Comment 4 Benjamin Poulain 2013-05-14 19:17:39 PDT

Created attachment 201779 [details]
Patch

Comment 5 Benjamin Poulain 2013-05-15 01:28:15 PDT

Committed r150105: <http://trac.webkit.org/changeset/150105>