I'm sorry, I didn't know there was a development list. That's why I am 
off-topic. :)

On May 25, 2012, at 9:29 AM, sam.spilsb...@canonical.com wrote:

> On Fri, 25 May 2012, Ryan Gauger wrote:
> 
>> It doesn't hurt to try. Transparency in Windows 7 usually makes the computer 
>> run slower, but makes too little of a difference to be noticeable. My guess 
>> would be that this same thing happens in Unity. Even though it may be 
>> unnoticeable if this is the case, it still matters, because there are people 
>> who run tests of speed and performance and put those online and make blog 
>> posts about them. I am thinking if anybody or blog would do that, it would 
>> probably be OMG! Ubuntu! Thanks!!!
>> 
> 
> I feel like we're straying off topic (this is the design list, not the 
> development list). However, I want to clarify this before people become
> misinformed.
> 
> Transparency does have a cost, however disabling it wouldn't be of much 
> effect. Disabling it does allow you to do things like out of order rendering, 
> however the kind of rendering work that unity does when it isn't doing things 
> like lighting or gaussian blurs is so cheap that you wouldn't see a 
> difference.
> 
> Transparency in windows comes at a cost because of the fact that you're doing 
> surface blurs (eg, synchronous operation using the DX equavilent for 
> glCopyTexSubImage2D or binding of framebuffer objects mid-render-pipe as well 
> as fragment shaders doing about 9 texture samples per pixel). Especially on 
> low end hardware where you don't have a lot of shader cores and can't 
> parallelize this stuff very easily (eg, intel) this comes at a cost.
> 
> If you want real performance gains, the best way to do it is to profile using 
> tools like callgrind or gprof or the like. For example, a couple of months 
> ago in compiz we found that we were creating about 200,000 regions for paint 
> calculation - first we optimzied by reducing the expensive part of this 
> (memory allocation) and secondly we optimized by reducing the number of 
> regions that needed to be created in order to do paint area calculation.
> 
> If you wish to contribute in this area, there are plenty of guides as to how 
> to use callgrind to get useful analytics - contributions would be much 
> appreciated because the developers don't have access to all the hardware and 
> every possible configuration that there might be. In addition, running under 
> callgrind is time-consuming because it slows down operation about 20x while 
> it looks for hot-spots in the code.
> 
> Cheers,
> 
> Sam
> 
>> On May 25, 2012, at 3:43 AM, Mikkel Kamstrup Erlandsen 
>> <mikkel.kamst...@canonical.com> wrote:
>> 
>>> On 05/24/2012 03:28 PM, Ryan Gauger wrote:
>>>> Hi Team,
>>>> 
>>>> I just had an idea that may speed up Unity even further. We could
>>>> perhaps make transparency behind the launcher an option, so that if
>>>> users choose to disable it, their computer's performance and speed
>>>> should run a bit smoother. Just an idea :) Thanks!
>>> 
>>> Sorry, do you have any data to back up the argument that anything would be 
>>> measurably faster with launcher transparency disabled? Without any hint of 
>>> evidence I don't think this is worth spending time on.
>>> 
>>> Don't get me wrong, I'd love to make everything perform better, but we need 
>>> a way to gauge this rigorously, otherwise we're shooting in the dark.
>>> 
>>> Cheers,
>>> Mikkel
>>> 
>>> --
>>> Mailing list: https://launchpad.net/~unity-design
>>> Post to     : unity-design@lists.launchpad.net
>>> Unsubscribe : https://launchpad.net/~unity-design
>>> More help   : https://help.launchpad.net/ListHelp
>> 
>> -- 
>> Mailing list: https://launchpad.net/~unity-design
>> Post to     : unity-design@lists.launchpad.net
>> Unsubscribe : https://launchpad.net/~unity-design
>> More help   : https://help.launchpad.net/ListHelp
>> 

-- 
Mailing list: https://launchpad.net/~unity-design
Post to     : unity-design@lists.launchpad.net
Unsubscribe : https://launchpad.net/~unity-design
More help   : https://help.launchpad.net/ListHelp

Reply via email to