TUTORIALS.md

Parts of the tutorial code are being ported over. As of now, it is not 1-to-1 in the Tutorial Application architecture, but the section is a work-in-progress.

There are 4 rendering methods, flagged in the HTML file headers.

1. Standard - This ray-traces with 1-ray at a time in a single threaded nature
2. Standard - Multi-Threaded - This raytraces 1-ray at a time, but split into tiles in a multiple worker threads scatter/gather tile code
3. Stream - This uses the Embree3 technique of blocking a whole tile of rays into a single call at once in a WASM call, to avoid context-switching costs per ray
4. Stream - Multi-Threaded - This will combine 2 & 3, launching multiple threads and requesting tiles from each, which independently running a intersect1M stream API call.

Parts of the AffineSpace3f have been ported to the tutorial code, as have parts of the ISPCCamera

Triangle Geometry

An unused vertex color array was in the tutorial, so a flag was added to the file to turn this off or on and switch from face to vertex coloring

Curve Geometry

These were referred to as hairs in the code, but this is just curve rendering in the demo

Forest

Forest requires a heavy amount of memory. 32MB should work for Complexity 0 (250 tree squares) Forest Complexity 1 requires about 48MB heap. Complexity 2 requires about 96MB heap.

Grid Geometry

Requires a mulit-threading enabled build along with a tesselation_cache of at least 1 and ~62MB of memory. Multi-threading is not used for the BVH builder, but is needed for tesselation, since it will wait for a secondary thread to tesselate regardless of device configuration. This results in a lockkup on single-thread compiled libraries.

The Alternate render is a bug where I had the wrong displacement sampling UV values used.