About mipmapping and further planning

[Tom M]

On Fri, Jul 9, 2010 at 6:43 AM, LukasT.dev at gmail.com
<lukast.dev at gmail.com> wrote:
> Zdrastvuj tavarish,
>
> On Friday, July 09, 2010 15:56:24 Dmitry Kazakov wrote:
>> Hi!
>>
>> I was considering implementation of mipmapping and i've finally come to a
>> conclusion. We really needn't mipmapping! :P

You  are wrong :)

As an extreme example, say you have a 30 GB image vastly zoomed out,
and do a paint stroke with a large brush from one corner to the other,
then rotate the image 45 degrees.

With a properly designed mipmapping

a) you don't need to load the 30GB image into memory at once, only the
tiles and mipmaps needed at any one time and zoom level.
b) during the stroke you update only the top layer of the mipmap, and
store the stroke path/history, from the artists perspective they just
did a paint stroke across a massive image and rotated a huge image,
from the processors perspective it just did a trivial amount of
painting.  You can propogate to lower levels either a) at your leisure
b) when zoomed c) some combination of the former.

Note that the example isn't contrived - Mari is a new 3D painting app
coming out for the animation industry that works on huge amounts of
image data at one time.

http://www.youtube.com/thefoundrychannel#p/p

LetterRip