lwf artifacts?

[jstrob]

1 point about the 32 bit EXR seq mentioned. You really don't always need 32 bit. Most of th etime 16 bit (half float) is enough. 16 float is containing way higher information than 8 bit non float. And try to render a frame in 16 bit and another one in 32 bit and use heavy compositing as you whish on it and try to find a difference, your eyes will never find one.

Some cass you may need 32 bits (full float): motion vector pass and zdepth pass.

[pechart]

Yepp, we use 16bit float by now most of the time. And for the zDepth there are different opinions around. In the docs for frischluft they mention that they don't use higher bit values internal. we just use 16bit for nearly everything. But i'll test 32bit for motion vectors.

[seq]

Totally right. 16bit is enough for nearly most cases.