screencast on Surface

This is a screencast.

Mozilla community meeting in Tokyo

15th, January, we have community meetup at new Tokyo office.

Also, New Tokyo office for Mozilla Japan will open at Spring. This place will include MozSpaces.

SpiderMonkey on WOA

HTC touch input w/ Fennec

In this week, I bought HTC Desire Z to debug several input issue on Firefox Mobile. HTC touch input is very buggy implementation as IME (Input Method Editor). Although I turn on IME support on hardware keyboard for Firefox Mobile from version 9 (released at Dec 2011), Touch Input with hardware keyboard doesn't work well.

When I investigate this, a articles of stackoverflow.com is found. But this wasn't good article for workaround and details.

When IME support on android application, onKeyPreIme has to return false. Then, InputConnection.commitText() is called per input. But 1st parameter may be incorrect because it has no meta state (alt/fn or shift). In other word, When you input Shift+[A], we expect "A". But HTC touch input calls commitText("a")....

If you doesn't implement full IME support using BaseInputConnection. this issue doesn't occurs. BaseInputConnection.commitText() has a workaround code for this issue. It generates KeyEvent for input, then it uses InputConnection.sendKeyEvent. Since this event has correct state, onKeyDown can handle correct key event.

I will create new fix for this problem then, I will land this to Firefox 9 and 10.

Starting new platform porting

You know, you can build Firefox, Thunderbird, Seamonkey and Xulrunner as Windows x64 binary by my work. Now, Firefox x64 is tier 2. (automation builder builds binary, runs automation test and nightly is available.). I am interested in porting new platforms

Android x86

It already works on gingerbread of Android x86 and will work on Honeybomb. Bug number is 681202. Also, Opera mobile already runs on Android x86

Windows/arm

Required Visual Studio 11. There is a lack of important libraries such as wsock32 to build binaries. I believe that it will be fixed by Visual Studio 11 beta (actually CTP, not beta!). Also, bug number for my status is 686843.

Dino Art by Kojimachi Cafe

From Kojimachi Cafe at Tokyo. It is surprised!

Fennec for Android on VirtualBox

These are screenshots using Fennec for Android on VirtualBox

APK package is here (this is debug build)

Windows x64 status updated

Today, I have landed bug 663748 to mozilla-inbound. (Also this will be integrated to mozilla-central today or tomorrow). It means that Win64 build will pass all automation tests. But, since this is for test tool, no effect on Firefox package.

Next Step

• To analyze crash data easily, we have to improve google breakpad for Win64. dumpsym.exe has to add unwind information to symbol file.
• PGO using VS2010. We consider to use VS2010 as official build. Now although we uses VS2008SP1 with PGO options, switching to VS2010SP1 means that we can get more performance. Here is a data how improve to switch to VS2010. But, unfortunately, due to a bug of VS2010, we cannot use 64-bit toolchain. Of course, there is a workaround.

Status Win64 build

Mochitest-1 failure

This is WebGL issue. Actually, Win64 build has no ANGLE that is D3D -> OpenGL ES wrapper. ANGLE now supports Win64 building by r697, and I and ransGaming fix a Win64 bug by r707. After merging these into m-c, it will pass mochitest-1.

test failures on jstest

I cannot reproduce this on my development machine. So if you know reproduce step, let me know. I think that this is GC issue, but I am still investigating...

But if this is compiler issue, this will be fixed using VS2010SP1. We have a plan to upgrade toolchains to VS2010SP1 as default.

Some browser migrations are gone

Since I landed bug 656205, some old migration (MacIE, Netscape 4 and Firebird / Phoenix) functions is gone. This code was too old and it didn't use async Places APIs.

Also, I am working Chrome migration code. At least, it will be landed on Firefox 7.

Firefox 4 on pandaboard (Ubuntu 10.10)

I got pandaboard to improve ARM code on Firefox. Pandaboard uses TI OMAP4 (Dual-core Cortex-A9 and PowerVR SGX 540), so it is the most powerful ARM board.

This data is sunspider benchmark on Firefox 4 on pandaboard. BTW, my Android phone (HTC Desire HD) is 2800ms.

============================================
RESULTS (means and 95% confidence intervals)
--------------------------------------------
Total:                 1984.1ms +/- 0.7%
--------------------------------------------

  3d:                   354.2ms +/- 1.3%
    cube:               139.8ms +/- 2.0%
    morph:               65.2ms +/- 5.1%
    raytrace:           149.2ms +/- 3.5%

  access:               290.7ms +/- 2.5%
    binary-trees:        47.1ms +/- 3.5%
    fannkuch:           116.6ms +/- 6.9%
    nbody:               47.1ms +/- 6.9%
    nsieve:              79.9ms +/- 2.0%

  bitops:               105.1ms +/- 1.8%
    3bit-bits-in-byte:    5.1ms +/- 8.0%
    bits-in-byte:        50.4ms +/- 1.2%
    bitwise-and:         12.3ms +/- 5.5%
    nsieve-bits:         37.3ms +/- 2.9%

  controlflow:           21.5ms +/- 7.2%
    recursive:           21.5ms +/- 7.2%

  crypto:               181.5ms +/- 1.9%
    aes:                 94.5ms +/- 0.8%
    md5:                 51.8ms +/- 4.7%
    sha1:                35.2ms +/- 5.2%

  date:                 279.9ms +/- 1.9%
    format-tofte:       173.9ms +/- 2.6%
    format-xparb:       106.0ms +/- 2.9%

  math:                 144.1ms +/- 1.2%
    cordic:              39.7ms +/- 1.5%
    partial-sums:        47.5ms +/- 2.3%
    spectral-norm:       56.9ms +/- 1.7%

  regexp:               123.1ms +/- 3.9%
    dna:                123.1ms +/- 3.9%

  string:               484.0ms +/- 1.0%
    base64:              44.3ms +/- 3.3%
    fasta:               96.4ms +/- 3.1%
    tagcloud:           146.4ms +/- 1.5%
    unpack-code:        135.5ms +/- 1.7%
    validate-input:      61.4ms +/- 3.2%

Although I have already landed ARM optimized NSS code (for crypto) into cvs.mozilla.org, it won't used by Firefox 4 or 5. It is for NSS trunk (NSS version 4). But I am investigate ARM specific optimization for Firefox 5.

Running 32-bit plugin on 64-bit Firefox

I am working Bug 595053 and Bug 620888 for 32-bit plugin support on 64-bit Firefox. Today, it is successful that 32-bit plugin (Flash and Sliverlight) runs on 64-bit Firefox.

This screenshot is sliverlight page on Firefox 64-bit (with 32-bit plugin support). Although Firefox process is 64-bit, sliverlight works.

SunSpider on Intel Atom processor

This result of sunspider data is Firefox 4.0b9pre (2010-12-18) vs Google Chrome 10.0.612.1 dev on Ubuntu 10.10 + Intel Atom Z560. This result is interesting. (FROM is Chrome and TO is Firefox)

TEST                   COMPARISON            FROM                 TO             DETAILS

=============================================================================

** TOTAL **:           1.38x as fast     1361.8ms +/- 5.4%   987.8ms +/- 1.0%     significant

=============================================================================

  3d:                  1.41x as fast      224.0ms +/- 11.6%   158.5ms +/- 0.9%     significant
    cube:              ??                  61.7ms +/- 23.3%    64.0ms +/- 1.8%     not conclusive: might be *1.037x as slow*
    morph:             1.48x as fast       47.2ms +/- 8.7%    31.9ms +/- 3.1%     significant
    raytrace:          1.84x as fast      115.1ms +/- 9.3%    62.6ms +/- 1.1%     significant

  access:              -                  130.6ms +/- 12.0%   128.7ms +/- 2.5% 
    binary-trees:      *2.02x as slow*      9.7ms +/- 42.8%    19.6ms +/- 12.8%     significant
    fannkuch:          1.42x as fast       53.3ms +/- 18.4%    37.6ms +/- 2.7%     significant
    nbody:             1.73x as fast       48.4ms +/- 11.9%    27.9ms +/- 2.2%     significant
    nsieve:            *2.27x as slow*     19.2ms +/- 7.6%    43.6ms +/- 2.5%     significant

  bitops:              1.77x as fast       87.6ms +/- 19.7%    49.5ms +/- 4.9%     significant
    3bit-bits-in-byte: 2.79x as fast        7.8ms +/- 5.8%     2.8ms +/- 10.8%     significant
    bits-in-byte:      ??                  17.8ms +/- 37.5%    19.6ms +/- 3.1%     not conclusive: might be *1.101x as slow*
    bitwise-and:       4.19x as fast       30.6ms +/- 22.5%     7.3ms +/- 6.6%     significant
    nsieve-bits:       1.59x as fast       31.4ms +/- 14.6%    19.8ms +/- 8.3%     significant

  controlflow:         1.49x as fast       12.7ms +/- 14.0%     8.5ms +/- 5.9%     significant
    recursive:         1.49x as fast       12.7ms +/- 14.0%     8.5ms +/- 5.9%     significant

  crypto:              1.65x as fast      111.6ms +/- 6.8%    67.6ms +/- 1.1%     significant
    aes:               1.61x as fast       59.7ms +/- 14.7%    37.1ms +/- 2.8%     significant
    md5:               1.64x as fast       26.0ms +/- 7.3%    15.9ms +/- 1.4%     significant
    sha1:              1.77x as fast       25.9ms +/- 7.1%    14.6ms +/- 3.4%     significant

  date:                1.50x as fast      241.3ms +/- 3.7%   161.3ms +/- 1.4%     significant
    format-tofte:      *1.116x as slow*    84.7ms +/- 8.2%    94.5ms +/- 1.2%     significant
    format-xparb:      2.34x as fast      156.6ms +/- 4.6%    66.8ms +/- 4.2%     significant

  math:                ??                  88.1ms +/- 4.4%    91.6ms +/- 1.3%     not conclusive: might be *1.040x as slow*
    cordic:            *1.47x as slow*     19.4ms +/- 10.7%    28.5ms +/- 2.7%     significant
    partial-sums:      1.27x as fast       51.0ms +/- 5.3%    40.1ms +/- 1.0%     significant
    spectral-norm:     *1.30x as slow*     17.7ms +/- 6.0%    23.0ms +/- 2.1%     significant

  regexp:              1.34x as fast       50.2ms +/- 8.4%    37.5ms +/- 7.6%     significant
    dna:               1.34x as fast       50.2ms +/- 8.4%    37.5ms +/- 7.6%     significant

  string:              1.46x as fast      415.7ms +/- 12.6%   284.6ms +/- 1.9%     significant
    base64:            1.44x as fast       30.2ms +/- 6.4%    21.0ms +/- 6.2%     significant
    fasta:             1.25x as fast       60.7ms +/- 5.1%    48.5ms +/- 2.6%     significant
    tagcloud:          1.31x as fast      114.1ms +/- 9.5%    87.2ms +/- 1.4%     significant
    unpack-code:       1.60x as fast      137.3ms +/- 22.0%    85.6ms +/- 2.5%     significant
    validate-input:    1.74x as fast       73.4ms +/- 20.7%    42.3ms +/- 6.5%     significant

If you use Intel Core i7, this score is even. Why is V8 engine slow on Intel Atom?

64bit Browser speed by SunSpider

This data is Firefox 4 x64 vs Internet Explorer 9 x64. Firefox 4 is three times faster than IE9。

TEST                   COMPARISON            FROM                 TO             DETAILS

=============================================================================

** TOTAL **:           *2.90x as slow*   389.8ms +/- 3.0%   1132.2ms +/- 0.8%     significant

=============================================================================

  3d:                  *2.57x as slow*    71.7ms +/- 0.7%    184.0ms +/- 3.9%     significant
    cube:              *1.94x as slow*    27.5ms +/- 1.4%     53.4ms +/- 1.8%     significant
    morph:             *6.11x as slow*    11.0ms +/- 0.0%     67.2ms +/- 2.5%     significant
    raytrace:          *1.91x as slow*    33.2ms +/- 0.9%     63.4ms +/- 9.6%     significant

  access:              *4.97x as slow*    40.5ms +/- 18.4%    201.4ms +/- 1.0%     significant
    binary-trees:      *2.81x as slow*    10.1ms +/- 71.9%     28.4ms +/- 3.4%     significant
    fannkuch:          *4.72x as slow*    17.8ms +/- 1.7%     84.1ms +/- 0.3%     significant
    nbody:             *11.0x as slow*     6.2ms +/- 4.9%     67.9ms +/- 1.3%     significant
    nsieve:            *3.28x as slow*     6.4ms +/- 5.8%     21.0ms +/- 0.0%     significant

  bitops:              *8.29x as slow*    18.9ms +/- 3.3%    156.7ms +/- 0.3%     significant
    3bit-bits-in-byte: *45.1x as slow*     0.8ms +/- 37.7%     36.1ms +/- 0.6%     significant
    bits-in-byte:      *3.98x as slow*    10.3ms +/- 6.6%     41.0ms +/- 0.0%     significant
    bitwise-and:       *23.3x as slow*     1.7ms +/- 20.3%     39.6ms +/- 0.9%     significant
    nsieve-bits:       *6.56x as slow*     6.1ms +/- 3.7%     40.0ms +/- 0.0%     significant

  controlflow:         *8.30x as slow*     4.7ms +/- 7.3%     39.0ms +/- 0.0%     significant
    recursive:         *8.30x as slow*     4.7ms +/- 7.3%     39.0ms +/- 0.0%     significant

  crypto:              *3.04x as slow*    35.4ms +/- 21.1%    107.5ms +/- 0.7%     significant
    aes:               *1.62x as slow*    21.9ms +/- 33.5%     35.4ms +/- 2.0%     significant
    md5:               *3.65x as slow*     9.6ms +/- 3.8%     35.0ms +/- 0.0%     significant
    sha1:              *9.51x as slow*     3.9ms +/- 5.8%     37.1ms +/- 0.6%     significant

  date:                *1.37x as slow*    58.0ms +/- 19.3%     79.7ms +/- 0.6%     significant
    format-tofte:      *1.62x as slow*    27.4ms +/- 1.3%     44.5ms +/- 0.8%     significant
    format-xparb:      ??                 30.6ms +/- 36.1%     35.2ms +/- 0.9%     not conclusive: might be *1.150x as slow*

  math:                *3.46x as slow*    31.8ms +/- 1.4%    109.9ms +/- 0.9%     significant
    cordic:            1.28x as fast      13.4ms +/- 3.7%     10.5ms +/- 3.6%     significant
    partial-sums:      *6.56x as slow*     9.0ms +/- 0.0%     59.0ms +/- 1.0%     significant
    spectral-norm:     *4.30x as slow*     9.4ms +/- 3.9%     40.4ms +/- 1.2%     significant

  regexp:              *2.24x as slow*    13.8ms +/- 4.1%     30.9ms +/- 2.3%     significant
    dna:               *2.24x as slow*    13.8ms +/- 4.1%     30.9ms +/- 2.3%     significant

  string:              *1.94x as slow*   115.0ms +/- 11.6%    223.1ms +/- 0.4%     significant
    base64:            *6.18x as slow*     4.4ms +/- 8.4%     27.2ms +/- 1.1%     significant
    fasta:             *2.06x as slow*    23.5ms +/- 31.5%     48.5ms +/- 1.4%     significant
    tagcloud:          *1.99x as slow*    29.1ms +/- 25.9%     57.9ms +/- 0.4%     significant
    unpack-code:       ??                 46.6ms +/- 21.0%     49.8ms +/- 1.1%     not conclusive: might be *1.069x as slow*
    validate-input:    *3.48x as slow*    11.4ms +/- 76.3%     39.7ms +/- 0.9%     significant

Windows x64 version status

• Since there is no build machine, Mozilla doesn't build nightly version for Win64. More detail is Bug 597919. So if this bug is fixed, nightly for Win64 will be built.
• I have fixed JIT engine bugs for Win64. Although current nightly (2010-09-16?) will be crash by V8 benchmark (from browser only), please wait next nightly. it will be fixed.
• Flash Plugin for Win64 is released as preview!. It works fine on Minefield. (64bit browser that uses NPAPI is Firefox/Minefield only!:-))
• Although Internet Explorer 9 for x64 has released as beta, do you test JavaScript benchmark such as SunSpider or Kraken?. Our JIT engine (TraceMonkey + JaegerMonkey) is too faster than IE9 x64 JIT!

Windows x64 build is available!

Armen announced Windows x64 build available on nightly build.

Also...

• This built on VC++ 2010. (x86 build still uses VC++ 2005SP1)
• installer is 32-bit. So installer path becomes same x86 by default, and registry is 32-bit. I am investigating about workaround for this.
• jemalloc is disable now. Ted is reviewing my patch for VC++ 2005 and VC++ 2008. About for VC++2010, see bug 563316
• comm-central tree is available to build x64 building. I am not owner for Thunderbird and SeaMonkey, so I don't know when they start building x64.

If you found new problem, please file a bug to bugzilla!

structure support on js-ctypes

I posted an article to Japanese Mozilla community site about js-ctypes on Firefox.next (3.7). Firefox.next improves many features such as struct support and callback support for js-ctypes.

This is a sample to use color picker on Windows using js-ctypes. ComDlg32 has color picker API "ChooseColor()". This sample is to access ChooseColor().

Components.utils.import("resource://gre/modules/ctypes.jsm");
 
var custColors_type = ctypes.ArrayType(ctypes.int32_t, 16);
const CHOOSECOLOR = new ctypes.StructType(
 'CHOOSECOLOR',
 [
  {'lStructSize': ctypes.uint32_t},
  {'hwndOwner' : ctypes.uint32_t},
  {'hInstance' : ctypes.uint32_t},
  {'rgbResult' : ctypes.uint32_t},
  {'lpCustColors' : custColors_type.ptr},
  {'Flags' : ctypes.uint32_t},
  {'lCustData' : ctypes.uint32_t},
  {'lpfnHook' : ctypes.uint32_t},
  {'lpTemplateName' : ctypes.uint32_t}
 ]);
 
var comdlg32 = ctypes.open("comdlg32");
var ChooseColor = comdlg32.declare("ChooseColorW",
                                   ctypes.stdcall_abi,
                                   ctypes.int32_t,
                                   CHOOSECOLOR.ptr);
 
var custColors = new custColors_type();
var col = new CHOOSECOLOR (CHOOSECOLOR.size, 0, 0, 0,
                           custColors.address(), 0, 0, 0, 0);
 
ChooseColor(col.address());
alert(col.rgbResult);

JavaScript Performance on Cortex-A8

You know, V8 by Google has the fastest JIT compiler. Many benchmarks get the top score on x86 environment. But how about on other CPU such as ARM? Any web sites don't research the benchmark on ARM.

I have a netbook using Freescale i.MX515. Although this net book "netwalker" is sold ($400?) at Japan market, I have tested Chromium 5.0 and Firefox 3.6 using Ubuntu 10.04 (lucid) beta. This data is on this environment.

This is interesting... V8 on ARM isn't optimized!. They should improve it for mobile market.

Symbian update for xulrunner

Choose Qt as widget

Nokia has a plan to discontinue Avkon (current S60 UI) as default widget. They will move to Qt that Nokia is their owner. For feature, I choose Qt as widget

I consider OpenGL ES for gfx/thebes

As long as I look current port for mobile device such as Windows Mobile and Maemo, graphic performance is poor. Although currently device has Graphics acceleration, xulrunner uses software rendering. I think that this design is bad for mobile. Since mobile device is poor CPU performance, we should use GPU

Compiler

Although RVCT 4.0 is free for Symbian developer, we cannot use for S60 now. S60 SDK only support RVCT 2.2 or GCCE (gcc 3.4). I want to use gcc 4.3 for Symbian build, but I have to modify SDK headers...

64-bit version Firefox 3.6 RC1

I have uploaded 64-bit version Firefox 3.6 RC1 to http://wiki.mozilla-x86-64.com/Firefox:Download.

Patch (Mercurial Queue format) will be uploaded to http://hg.mozilla-x86-64.com/hgwebdir.cgi/mozilla1.9.2-win64-mq/ at this week.