Making a Shading Language for My Offline Renderer(agraphicsguynotes.com)
agraphicsguynotes.com
Making a Shading Language for My Offline Renderer
https://agraphicsguynotes.com/posts/making_a_shading_langauge_for_my_offline_renderer/
4 comments
Technically, we did RTSL before Ingo and then Matt did their stuff at Intel. LLVM wasn't particularly usable in 2007, but Austin and I did try it (this isn't in the paper!). A few years later, I did an actual LLVM JIT for OSL, but we didn't have enough coherence for vectorized shading at the time at Sony.
But none of it was surprising. The original RenderMan shading language was "vectorized" and it even used SIMD instructions on modern processors to run the "interpreter loops". That is, a single "color add" in RSL might have looked like:
The point of ISPC et al. was to give people CUDA-like easy mode for "trust me, just vectorize the whole thing and deal with the masking for me". It goes beyond a hardcoded shading language (easier target!) though didn't reach as far as CUDA with complex structs / C++ capabilities.
But none of it was surprising. The original RenderMan shading language was "vectorized" and it even used SIMD instructions on modern processors to run the "interpreter loops". That is, a single "color add" in RSL might have looked like:
for (int i=0; i < grid_points; i++) {
out_color[i] += foo;
}
The point of ISPC et al. was to give people CUDA-like easy mode for "trust me, just vectorize the whole thing and deal with the masking for me". It goes beyond a hardcoded shading language (easier target!) though didn't reach as far as CUDA with complex structs / C++ capabilities.
Sometimes when I see people taking on the challenge of creating a language, I feel a desire to try it myself. I hope that someday I too can create the language I have in mind.
A valid choice; one might also find re2c[1] or Ragel [2] + Lemon[3] a friendlier if less well-established implementation of essentially the same theory. Moving further away, there’s also ANTLR[4] or spending a week on a hand-coded lexer and parser.
> CPU’s parallelism is not utilized in the same way as GPU since different threads are not synchronized at all, there is simply nothing like Warp or Wavefront. [...] SIMD optimization would be a fairly bad fit for shaders since there are no four instances of shader executions at the same time.
See, however, ISPC[5,6] for how attempting to use SIMD (also known as “90% of your CPU’s compute”) does yield wavefronts, etc. (If I’m reading between the lines of the second reference correctly, the work on it paved the way for LLVM-based shader-compiler backends in AMD and Intel GPU drivers.)
[1] https://re2c.org/
[2] https://www.colm.net/open-source/ragel/
[3] https://sqlite.org/src/doc/trunk/doc/lemon.html
[4] https://www.antlr.org/
[5] https://ispc.github.io/
[6] https://pharr.org/matt/blog/2018/04/18/ispc-origins