Markus Hübner | 2 Dec 09:08 2002

Re: nglayout.initialpaint.delay

Yves Glodt wrote:
> Hi,
> 
> 
> in user.js
> user_pref("nglayout.initialpaint.delay", 0);
> 
> 
> I found this in the Phoenix forum at Mozillazine.org, and it makes feel 
> Mozilla somewhat faster.
> Maybe it should be set to 0 by default. (apparently it's 1200 by default)
> 
> 
> 
> 

there is a bug about that ... see
http://bugzilla.mozilla.org/show_bug.cgi?id=180241

:: markus

Trevin Beattie | 3 Dec 06:48 2002

Re: why is mozilla so slow?

I have some profiles run from a '486/100MHz using NetBSD.  They have a
lot more detail than my Linux profiles because I was able to use the
profiled libc.  It took an awfully long time however.  Each mozilla
run took 3 minutes just for starting and exiting.  Summing up multiple
profile runs took 8 DAYS on my 380MHz computer, so I gave up doing
that after getting the base profiles.

So, according to my profiles, it took 15.35 seconds (after startup) to
load my test form.  The process was clearly visible as I watched
mozilla slowly lay out the table of text boxes and scrolled lists.
Here are the top 35 functions by time (representing 3 seconds):

  %   cumulative   self              self
 time   seconds   seconds    calls  ms/call  name
  0.00      0.21     0.21   103715     1.99  nsStyleContext::GetStyleData(nsStyleStructID)
  0.00      0.39     0.19     5137    36.21  nsHTMLReflowState::InitConstraints(nsIPresContext *, int, int, nsMargin *,
nsMargin *)
  0.00      0.55     0.16     1183   131.87  nsLineLayout::ReflowFrame(nsIFrame *, unsigned int &, nsHTMLReflowMetrics
*, int &)
  0.00      0.70     0.15    54550     2.79  nsRuleNode::GetStyleData(nsStyleStructID, nsIStyleContext *, int)
  0.00      0.82     0.12     1071     0.11  nsTextFrame::Reflow(nsIPresContext *, nsHTMLReflowMetrics &,
nsHTMLReflowState const &, unsigned int &)
  0.00      0.94     0.12     2592    45.52  RuleProcessorData::RuleProcessorData(nsIPresContext *, nsIContent *,
nsRuleWalker *, nsCompatibility *)
  0.00      1.05     0.11     1763    62.39  nsBlockFrame::ReflowBlockFrame(nsBlockReflowState &,
nsLineList_iterator, int *)
  0.00      1.15     0.10      109   950.64  do_lzw(gif_struct *, unsigned char const *)
  0.00      1.25     0.10     5312    19.20  nsCaseConversionImp2::CaseInsensitiveCompare(unsigned short const *,
unsigned short const *, unsigned int, int *)
  0.00      1.34     0.09     2101    43.78  memset
(Continue reading)

user | 4 Dec 03:55 2002
Picon

Array Performance Question

Why using the forms of the Array constructor that specify a size takes 
more time than the forms of the Array constructor that don't specify a 
size?
I wrote two JavaScript functions, unspecify() and specify(), as 
displayed below and ran them several times in the Mozilla browser.
The results shows that unspecify() has a better runtime than specify().

function unspecify( )
{
	var n = 10000;
	var m = 10;
	var a = new Array();
	for (var i=0; i< n; i++)
	{
		a[i] = new Array();
		for(var j=0; j<m; j++)
			a[i][j] = j;
	}
}

function specify( )
{
	var n = 10000;
	var m = 10;
	var a = new Array(n);
	for (var i=0; i< n; i++)
	{
		a[i] = new Array(m);
		for(var j=0; j<m; j++)
			a[i][j] = j;
(Continue reading)

Olaf Dietsche | 4 Dec 05:29 2002
Picon

Re: why is mozilla so slow?

Trevin Beattie <http://www.xmission.com/~trevin/mail.html? <at> x.y> writes:

> I have some profiles run from a '486/100MHz using NetBSD.  They have a
> lot more detail than my Linux profiles because I was able to use the
> profiled libc.  It took an awfully long time however.  Each mozilla
> run took 3 minutes just for starting and exiting.  Summing up multiple
> profile runs took 8 DAYS on my 380MHz computer, so I gave up doing
> that after getting the base profiles.
>
> So, according to my profiles, it took 15.35 seconds (after startup) to
> load my test form.  The process was clearly visible as I watched
> mozilla slowly lay out the table of text boxes and scrolled lists.
> Here are the top 35 functions by time (representing 3 seconds):
>
>   %   cumulative   self              self
>  time   seconds   seconds    calls  ms/call  name
>   0.00      0.21     0.21   103715     1.99  nsStyleContext::GetStyleData(nsStyleStructID)
>   0.00      0.39     0.19     5137    36.21  nsHTMLReflowState::InitConstraints(nsIPresContext *, int, int, nsMargin
*, nsMargin *)
>   0.00      0.55     0.16     1183   131.87  nsLineLayout::ReflowFrame(nsIFrame *, unsigned int &,
nsHTMLReflowMetrics *, int &)
>   0.00      0.70     0.15    54550     2.79  nsRuleNode::GetStyleData(nsStyleStructID, nsIStyleContext *, int)
[...]

How did you get these numbers? I tried this some time ago on Linux,
but didn't succeed, because Linux doesn't support profiling shared
libraries. The numbers I got, where only for the mozilla main binary,
which is not very useful.

Regards, Olaf.
(Continue reading)

Trevin Beattie | 4 Dec 06:13 2002

Re: why is mozilla so slow?

Olaf Dietsche wrote:

> Trevin Beattie <http://www.xmission.com/~trevin/mail.html? <at> x.y> writes:
>
> > I have some profiles run from a '486/100MHz using NetBSD.  They have a
> > lot more detail than my Linux profiles because I was able to use the
> > profiled libc.  It took an awfully long time however.  Each mozilla
> > run took 3 minutes just for starting and exiting.  Summing up multiple
> > profile runs took 8 DAYS on my 380MHz computer, so I gave up doing
> > that after getting the base profiles.
> >
> > So, according to my profiles, it took 15.35 seconds (after startup) to
> > load my test form.  The process was clearly visible as I watched
> > mozilla slowly lay out the table of text boxes and scrolled lists.
> > Here are the top 35 functions by time (representing 3 seconds):
> >
> >   %   cumulative   self              self
> >  time   seconds   seconds    calls  ms/call  name
> >   0.00      0.21     0.21   103715     1.99  nsStyleContext::GetStyleData(nsStyleStructID)
> >   0.00      0.39     0.19     5137    36.21  nsHTMLReflowState::InitConstraints(nsIPresContext *, int, int, nsMargin
*, nsMargin *)
> >   0.00      0.55     0.16     1183   131.87  nsLineLayout::ReflowFrame(nsIFrame *, unsigned int &,
nsHTMLReflowMetrics *, int &)
> >   0.00      0.70     0.15    54550     2.79  nsRuleNode::GetStyleData(nsStyleStructID, nsIStyleContext *, int)
> [...]
>
> How did you get these numbers? I tried this some time ago on Linux,
> but didn't succeed, because Linux doesn't support profiling shared
> libraries. The numbers I got, where only for the mozilla main binary,
> which is not very useful.
(Continue reading)

Jim Nance | 4 Dec 15:55 2002
Picon

Re: why is mozilla so slow?

In article <askuad$s9goe$1 <at> ID-19472.news.dfncis.de>, Olaf Dietsche wrote:

>>> How did you get these numbers? I tried this some time ago on Linux,
>>> but didn't succeed, because Linux doesn't support profiling shared
>>> libraries.

> Thanks for this hint! Seems, I should take a closer look at *BSD. I
> always thought, profiling is a gcc/glibc thing and is independent of
> the used OS.

It basically is, but I dont think *BSD uses glibc.

You can try profiling mozilla under Linux with jprof
(mozilla/tools/jprof) if you like.  I believe there are some other
profilers available as well, but I dont have any experience with
them.

Thanks,

Jim

Bostjan Vilfan | 5 Dec 10:19 2002
Picon

Screen corruption with LCD monitor

I am not sure about the status of this phenomenon (is it a bug or 
something else) so I am firt inquiring about it in a newsgroup.

I have a Philips BRILLIANCE 180P LCD monitor, and I use a Matrox Millenium 
G550 adapter (driver version 5.86.032). With all versions of the Mozilla 
browser (currently, 1.2.1) I am experiencing frequent screen corruption, 
which is visible as smudged patches of the screen (screen shot on 
request). It must be said that the same phenomenon is found in other 
applications (e.g. Lotus Organizer). Due to this I was inclined to 
attribute it to the monitor; however, recently I tried out the Windows 
Internet Explorer, AND THERE WAS NO SCREEN CORRUPTION. Can anybody explain 
what is the cause of this?

Regards, Bostjan Vilfan

--

-- 
**************************************************************************
Bostjan Vilfan                       Voice  : +386-1-476-8391
University of Ljubljana, FRI         Fax    : +386-1-426-4647
Trzaska 25                           E-mail : bostjan.vilfan <at> fri.uni-lj.si
1000  Ljubljana, Slovenia
**************************************************************************


Jeffrey Siegal | 5 Dec 21:23 2002

Re: Mozilla 1.2b rules !

Vincent Nicolas wrote:
> But what a disapointment for the MacOS X version... it is one hundred times 
> slower than windows version. And my dual-G4 is far more powerful than my 
> PIII-550... :-(((
> 
> I guess people will propose me to try Chimera ?

You might try the mach-o build.

pbergsagel | 6 Dec 03:49 2002
Picon

Re: what's mozilla's major performance problems now?


Vincent Nicolas wrote:
> Wenqing Jiang <jiang_wq <at> hotmail.com> wrote in 

> Very slow on MacOs X (compared to Windows version, not compared to IE)
> Example : http://www.dhtmlcentral.com (was very slow too on windows until 
> version 1.2b I guess)
> 
> Displaying pages with lots of texts and images are slow too (see 
> http://www.macgeneration.com/actu/)
> 

Have you tried out the Mach-O build of Mozilla for OS X. It is about 
20-25% faster than the current CFM build for OS X.  Shortly the Mac OS X 
default download of Mozilla will switch to the Mach-O build. There are 
still one or two issues to solve before the switch.

One result will be that the Mac Classic Mozilla (for OS 8/9) will no 
longer be an offically supported build :(, although I heard a volunteer 
will maintain it. Time to switch to OS X (if your hardware can support it.

Vincent Nicolas | 6 Dec 11:46 2002
X-Face
Picon

Re: Mozilla 1.2b rules !

Jeffrey Siegal <jbs <at> quiotix.com> wrote in
news:3DEFB5B6.8070301 <at> quiotix.com: 

> Vincent Nicolas wrote:
>> But what a disapointment for the MacOS X version... it is one hundred
>> times slower than windows version. And my dual-G4 is far more
>> powerful than my PIII-550... :-(((
>> 
>> I guess people will propose me to try Chimera ?
> 
> You might try the mach-o build.
> 

But where do you find it ?

--

-- 
Le Vintz.
http://membres.tripod.fr/vnicolas/


Gmane