trinity-devel@lists.pearsoncomputing.net

Message: previous - next
Month: February 2012

Re: [trinity-devel] Trinity qt4 port?

From: Pawel Soltys <sh4dou@...>
Date: Mon, 13 Feb 2012 22:32:40 +0100
On Mon, Feb 13, 2012 at 10:10 PM, /dev/ammo42
<mickeytintincolle@...> wrote:
> On Mon, 13 Feb 2012 14:50:42 -0600
> "Timothy Pearson" <kb9vqf@...> wrote:
>
>> > On Sunday 12 February 2012 02:20:38 Timothy Pearson wrote:
>> >> These are the exact reasons I stopped to the move to Qt4 and
>> >> picked up maintinance of Qt3 (now becoming TQt3).  I gave the
>> >> conversion effort my best shot (which took many, many months of
>> >> effort) and stopped once it became apparent how extensive Qt4's
>> >> limitations were (especially in drawing operations and native X11
>> >> window handling) and also how buggy Qt4
>> >> really is.
>> > May I ask for the benchmarking results? I am seriously interested
>> > here as I
>> > think that I can provide you some valuable help here. My fear is
>> > that you had
>> > an incorrect benchmark (the experts call that phoronixing) and that
>> > your decision is because of that on false grounds.
>> >
>> > Especially important is to consider that graphics cards, drivers,
>> > what you render (widgets vs scenes), which Qt graphicssystem you
>> > use seriously influences the result. E.g. rendering widgets with
>> > raster might be a bad idea.
>> >
>> > Cheers
>> > Martin
>>
>> I do not have any benchmarks handy at the moment, although this is
>> simply due to the length of time from those tests to the writing of
>> this message. The upshot was that Qt4 is significantly slower when it
>> comes to rendering raster graphics (as you mentioned), and also when
>> large numbers of widgets are displayed on-screen.  If I understand
>> some of the Qt changes, the raster performance was improved somewhat
>> in the latest versions of Qt4, but I don't know if the other problems
>> were addressed.
>>
>> As an aside, 3D graphics hardware is not only expensive, it is also
>> one of the most proprietary and least-understood components in a
>> typical computer (it also tends to burn out a lot, at least that is
>> my experience with anything other than an enterprise-grade nVidia
>> Quadro card).  If nVidia and ATI experienced supply shortages (don't
>> laugh, remember the recent hard drive scarcity due to flooding in
>> Thailand) I would still need to be able to use my computers with
>> not-so-great backup graphics hardware, and possibly without good
>> OpenGL support.
> Almost all less than 5 years old PC hardware has either a nVidia, ATI or
> Intel GPU, that is not necessarily powerful but definitely has enough
> power to render standard controls with OpenGL.
>>
Please take a note that there are places on earth, where people use
computers older that 5 years old, simply because they cannot (don't
have means, like money) to upgrade them.
Once linux was considered as a system that could be installed and used
on nearly antic computers. I remember my friend installing it on a PC
that come form 1996, and turning it to fully functional computer that
needed only allow internet access and document handling. It was in
2004 or 2004. I don't really remember what he had installed there, but
probably it was some old (2.x) version of KDE. He ended up with fully
functional PC that a non-expert user (actually his mother, which
couldn't distinguish between PC and monitor) could use it. Right now,
I don't find it possible without need to use a software that only
experienced linux user could use/configure.
If we are to introduce software that demands bu default more and more
powerful hardware, we do nothing other that taking away the
possibility of comfortable usage of computer for those that cannot
obtain modern(ish) computers, and we do what closed-source companies
(I don't want to throw names, but all of you can imagine what I'm
taking about) were doing for years: forcing updates of hardware with
every new version of the software.

>> More practically, even slightly sluggish performance is quite
>> noticeable to power users.  Many applications, upon converting from
>> Qt3 to Qt4, appeared to slow down noticeably.
> Even with a common Qt-internal style such as CDE or Win9x ?
>>
>> These are just my $0.02 and experiences in working with Qt 4.7.  I am
>> open to looking at Qt4 again once Trolltech fixes the raster graphics
>> problem for good.
> It will be "fixed" with Qt5, which will have OpenGL ES 2.0 as the only
> graphics system. It had better work well with llvmpipe...
>>
>> Tim