This is all keyword mumbo jumbo. What exact code/algorithm are you trying to optimize?

I am looking into parsing a syntax tree in OpenSL on my GPU using compute shaders, so GPGPU abuse, and have excessive pipes lines, or special conditions and only having one global one and a giant parse tree create…