Hi, today I've begun experimenting a little with the ScummVM theming. I've added a new shading effect that blurs the background when a new dialog appears (instead of the default dimming).
Here are some screenshots:
It's just a proof of concept. It has a configurable radius and it could be mixed with dimming to get a better global effect.
So, now to the problem: the current implementation is very very slow (around 7sec unresponsive waiting to show the new dialog). Is it worth trying to optimize it, or it's very improbable that it would get accepted?
Any feedback will be welcome
Blur shading test
Moderator: ScummVM Team
Seems like a nice idea. But you really have to do it a factor 1000 faster or so . The current dimming code already is far too slow, IMO (due to the fact that it uses colorToRGB and RGBToColor on each pixel -- that's very bad for performance, as it involves calling virtual methods in a tight loop, which the compiler has no chance of optimizing.
maybe you are doing something like that, too? In that case, first grabbing the whole screen content, converting it to a suitable format (16 or 32 bit data) then running your filter and then converting back might already give a big speed boost (at the expense of memory).
Furthermore, consider using an alternate algorithm which maybe is not "mathematically correct", but looks good enough. After all, nobody will care how exactly the "blurry" effected was created
We use that for the aspect ratio correction, for example: The code is not doing 100% precise stretching, which most people never notice, yet is an order of magnitude faster.
maybe you are doing something like that, too? In that case, first grabbing the whole screen content, converting it to a suitable format (16 or 32 bit data) then running your filter and then converting back might already give a big speed boost (at the expense of memory).
Furthermore, consider using an alternate algorithm which maybe is not "mathematically correct", but looks good enough. After all, nobody will care how exactly the "blurry" effected was created
We use that for the aspect ratio correction, for example: The code is not doing 100% precise stretching, which most people never notice, yet is an order of magnitude faster.
Re: Blur shading test
It is a terrible idea, the current dimming effect is bad enough.
The current screen shouldn't be made more difficult to see, just because the GUI is currently been shown.
The current screen shouldn't be made more difficult to see, just because the GUI is currently been shown.
Re: Blur shading test
dimming is useful in the launcher though so you can easily see what the current active dialog is - http://sourceforge.net/tracker/index.ph ... tid=418823Kirben wrote:It is a terrible idea, the current dimming effect is bad enough.
The current screen shouldn't be made more difficult to see, just because the GUI is currently been shown.
Re: Blur shading test
Yes, it can be useful inside the GUI interface. But not when used over the complete screen of a game in progress.clem wrote: dimming is useful in the launcher though so you can easily see what the current active dialog is - http://sourceforge.net/tracker/index.ph ... tid=418823