Angus Leeming wrote:
> Jean-Marc Lasgouttes wrote:
>> Jean-Marc> Lars, is it OK nevertheless? It looks like a natural thing
>> Jean-Marc> to do, unless there is something I do not understand about
>> Jean-Marc> how C++ works.
>>
>> OK, since nobody complained, I committed it.
>
> Looking at src/
Jean-Marc Lasgouttes wrote:
> Jean-Marc> Lars, is it OK nevertheless? It looks like a natural thing
> Jean-Marc> to do, unless there is something I do not understand about
> Jean-Marc> how C++ works.
>
> OK, since nobody complained, I committed it.
Looking at src/FontIterator.C, it's an entirely
> "Jean-Marc" == Jean-Marc Lasgouttes <[EMAIL PROTECTED]> writes:
> "Alfredo" == Alfredo Braunstein <[EMAIL PROTECTED]> writes:
>>> I am not sure I understand FontIterator::operator*(), since
>>> inlining it did not improve the situation.
Alfredo> Could the problem be LyXFont's copy const
On Wed, Apr 06, 2005 at 11:21:46AM +0200, Jean-Marc Lasgouttes wrote:
> getFont and friends are a real problem (looking for current font at
> each character of a paragraph)
We would get automatically rid of that if we had font changes as insets.
> and I think that the use of a FontIterator (or di
On Wed, Apr 06, 2005 at 12:15:27PM +0200, Lars Gullik Bj?nnes wrote:
> So, why not just remove "our" caching completely, and just let those
> still using old QTs suffer?
Because there's practically no advantage in terms of maintenance in
removing the cache?
john
On Wed, Apr 06, 2005 at 11:42:54AM +0200, Lars Gullik Bj?nnes wrote:
> | Significantly.
>
> no numbers? :-)
I did numbers way back. Perhaps you can be bothered to search the
archives? :)
john
On Wed, Apr 06, 2005 at 11:21:46AM +0200, Jean-Marc Lasgouttes wrote:
> example RowPainter::paintChars would help. However, the RtL code in
> there scares me :)
I doubt that still works, there's nobody left who understands it
john
> "Alfredo" == Alfredo Braunstein <[EMAIL PROTECTED]> writes:
>> I am not sure I understand FontIterator::operator*(), since
>> inlining it did not improve the situation.
Alfredo> Could the problem be LyXFont's copy constructor?
Sure, we are not supposed to make a copy here. The attached pat
> "Angus" == Angus Leeming <[EMAIL PROTECTED]> writes:
>> Good question. It seems that it makes Qt/Mac suffer, though.
Angus> Bennett's results were so much worse than your's that I think
Angus> that its hard to draw any firm conclusions about performance on
Angus> the Mac.
Indeed. It would
Jean-Marc Lasgouttes wrote:
> | The caching exists since version 3.1.0, which is probably
> | widespread enough. Qt/Win free and Qt/Mac also have this code.
>
> Lars> So, why not just remove "our" caching completely, and just let
> Lars> those still using old QTs suffer?
>
> Good question. It see
> "Lars" == Lars Gullik Bjønnes <[EMAIL PROTECTED]> writes:
| The caching exists since version 3.1.0, which is probably
| widespread enough. Qt/Win free and Qt/Mac also have this code.
Lars> So, why not just remove "our" caching completely, and just let
Lars> those still using old QTs suffer?
Jean-Marc Lasgouttes wrote:
> Each sample counts as 0.01 seconds.
> % cumulative self self total
> time seconds secondscalls ms/call ms/call name
> 7.37 0.16 0.16 1463980 0.00 0.00
> LyXText::getFont(Paragraph const&, int) const
> 6.45
Jean-Marc Lasgouttes <[EMAIL PROTECTED]> writes:
>> "Lars" == Lars Gullik Bjønnes <[EMAIL PROTECTED]> writes:
>
| Lars> John Levon <[EMAIL PROTECTED]> writes:
| Lars> | On Tue, Apr 05, 2005 at 04:49:53PM +0200, Lars Gullik Bj?nnes wrote:
>>>
Also I'd like to know how much the old code is
> "Lars" == Lars Gullik Bjønnes <[EMAIL PROTECTED]> writes:
Lars> John Levon <[EMAIL PROTECTED]> writes:
Lars> | On Tue, Apr 05, 2005 at 04:49:53PM +0200, Lars Gullik Bj?nnes wrote:
>>
>>> Also I'd like to know how much the old code is penalized if we do
>>> it the no-cache way all the time.
>
John Levon <[EMAIL PROTECTED]> writes:
| On Tue, Apr 05, 2005 at 04:49:53PM +0200, Lars Gullik Bj?nnes wrote:
>
>> Also I'd like to know how much the old code is penalized if we do it
>> the no-cache way all the time.
>
| Significantly.
no numbers? :-)
What version of QT got the caching? How wid
> "John" == John Levon <[EMAIL PROTECTED]> writes:
John> On Tue, Apr 05, 2005 at 02:18:58PM +0200, Jean-Marc Lasgouttes
John> wrote:
>> The font caching in qt2/qfont_loader.C uses a std::map which shows
>> on gprof profiles (around 16% with all optimizations on and no
>> assertions). It appea
> "Lars" == Lars Gullik Bjønnes <[EMAIL PROTECTED]> writes:
Lars> Bennett Helm <[EMAIL PROTECTED]> writes: | I might add: in
Lars> general LyX-140 is unacceptably slow for me on a | relatively
Lars> fast Mac (1.25GHz G4 iMac). In editing academic papers of | mine
Lars> (no math, but a fair num
On Tue, Apr 05, 2005 at 04:49:53PM +0200, Lars Gullik Bj?nnes wrote:
> Also I'd like to know how much the old code is penalized if we do it
> the no-cache way all the time.
Significantly.
john
Jean-Marc Lasgouttes <[EMAIL PROTECTED]> writes:
| The font caching in qt2/qfont_loader.C uses a std::map which shows on
| gprof profiles (around 16% with all optimizations on and no
| assertions). It appears that Qt has its own font caching mechanism
| since version 3.1.0. With the attached patch
Bennett Helm <[EMAIL PROTECTED]> writes:
| I might add: in general LyX-140 is unacceptably slow for me on a
| relatively fast Mac (1.25GHz G4 iMac). In editing academic papers of
| mine (no math, but a fair number of bibliographical citations) I can
| easily type faster than characters display on
On Apr 5, 2005, at 8:18 AM, Jean-Marc Lasgouttes wrote:
I would be interested to know what happens with LyX/Mac and LyX/Win,
though.
On MacOS X, the patch makes things worse:
Before:
real0m38.977s
user0m28.770s
sys 0m2.380s
After:
real0m42.792s
user0m33.420s
sys 0m2.100s
I m
On Tue, Apr 05, 2005 at 02:18:58PM +0200, Jean-Marc Lasgouttes wrote:
>
> The font caching in qt2/qfont_loader.C uses a std::map which shows on
> gprof profiles (around 16% with all optimizations on and no
> assertions). It appears that Qt has its own font caching mechanism
Finally! Good news. W
22 matches
Mail list logo
