Im not directly responsible, but im using NormalMappr and manually editing the. We are exporting OBJ files with material files, and creating textures, etc.

I am working on a project and have access to modelers running Soft Image XSI. NSShadow is now only used when image dimensions change. Here's the (UB) build, which is faster, and has some bug fixes:ĮDIT: Implemented caching of shadow in display view. I may have to render the image's shadow into an intermediate NSImage and then just composite that. Now, the real bottleneck to fast updates is no longer my normal map code, but rather the use of NSShadow when displaying the normalmap! NSShadow is slow as molassas. Long mag = lsqrt( v.x*v.x + v.y*v.y + v.z*v.z ) Vec3i_scale_and_normalize_integer( vec3i v, int xscale, int yscale, int zscale ) It uses an integer sqrt I got from codecodex Third I went to a mostly integer math implementation with only 3 int->float->int conversions per pixel. I can still speed that up further by taking the clamping/wrapping edge-cases out of the loop and have the loop only process inner pixels. Second, I cleaned up my loops, caching scanline offsets and elminating a lot of multiplies. Now it caches the intermediate buffers, and only re-creates them when needed. This is a considerable improvement for situations where you're dragging the flatness slider or whatnot. I eliminated the main bottleneck, which proved to the the intermediate rendering of the input image to a greyscale context with up or down sampling according to desired output dimensions. I've sharked and I've done a fair amount of optimization. Now, if gcc or chirpflare are still reading, I've got news on the optimization front. That's a considerable improvement, nice work xype.

Soft shadows in cheetah3d