SIMD Initial Implementation

In the previous meeting my tutor pointed out that SIMD is in fact just a form of parallel processing and is perfectly relevant to the project. My main aim for this week is to look at SIMD, and hopefully get some of it running to achieve a speed increase. Then next week I will make what feels like a long overdue start on the second volume rendering technique.

SIMD comes in different forms on various architectures but the most common one seems to be SSE (Streaming SIMD Extensions), originally introduced by Intel and now supported on (most/all?) AMD chips.

The first tutorial I found was this one:
http://www.neilkemp.us/v3/tutorials/SSE_Tutorial_1.html

After spending a while trying to get the assembly to not crash I continued searching for other tutorials, and found this one:
http://www.codeproject.com/KB/recipes/sseintro.aspx

This one resulted in success. I added two vectors together in a small testbed program. Not having to use assembly (which I am in no way familiar with) is a nice bonus and helps me feel more positive about it. Tomorrow I will begin integrating some SIMD in the actual volume renderer.

To finish this post, a picture of the recently improved lighting calculations. It now uses two light sources from opposite directions (based on the same dot product), rather than one, and has higher contrast. The lighting is all precalculated before rendering so there is no extra cost to the rendering time.