i see, that's very unfortunate :/
in my case, i will do what i always do: whoever has implementations that generate mipmaps on the cpu, will simply have to get a computer that can play games...
i know it's a hard stance, and probably very unpopular, but i make games for myself =)
i have recently gone the fbo route with one of the generatemipmap instances, which was for blurring the depth end of my scene
mostly because i got to write my own blurring, which is miles better, and yes using hardware bilinear interpolation always helps =)
it's noteworthy that on performance, using fbo to blur vertically and horizontally actually lowered my fps slightly compared to the glGenerateMipmap solution (even with the hardware bilinear tricks, and only 1 step in each direction!)
and with mipmaps i generated 2 levels, which i could interpolate between using texture2DLod(..)
i have to say, i like the mipmap version better, but unfortunately, it doesn't look quite as good!
Show differencesHistory of post edits
#1Kaptein
Posted 16 November 2012 - 06:41 AM
i see, that's very unfortunate :/
in my case, i will do what i always do: whoever has implementations that generate mipmaps on the cpu, will simply have to get a computer that can play games...
i know it's a hard stance, and probably very unpopular, but i make games for myself =)
i have recently gone the fbo route with one of the generatemipmap instances, which was for blurring the depth end of my scene
mostly because i got to write my own blurring, which is miles better, and yes using hardware bilinear interpolation always helps =)
in my case, i will do what i always do: whoever has implementations that generate mipmaps on the cpu, will simply have to get a computer that can play games...
i know it's a hard stance, and probably very unpopular, but i make games for myself =)
i have recently gone the fbo route with one of the generatemipmap instances, which was for blurring the depth end of my scene
mostly because i got to write my own blurring, which is miles better, and yes using hardware bilinear interpolation always helps =)