First of all, thanks Avinash for sharing this project: the purpose is clear, the code is readable, plus it defers to JS (which is objectively the best programming language) - an all around win for everyone. However, I couldn't help but notice it might be lacking in the matter of optimization.
If you allow the metaphor, as much as V8 makes our regular, run-of-the-mill grazing-horse-of-a-code run as if it were on a whisky-and-crack-infused-wheat diet, I feel like we might be missing out on exploiting the next step in advanced sign-flipping tech.
I believe the generated code could be further optimized by mapping the sign-flipping operation into a texture transformation in GPU space. Have you toyed with this idea, and if so, what did you find out?
Once again, thank you for this awesome piece of tech.
First of all, thanks Avinash for sharing this project: the purpose is clear, the code is readable, plus it defers to JS (which is objectively the best programming language) - an all around win for everyone. However, I couldn't help but notice it might be lacking in the matter of optimization.
If you allow the metaphor, as much as V8 makes our regular, run-of-the-mill grazing-horse-of-a-code run as if it were on a whisky-and-crack-infused-wheat diet, I feel like we might be missing out on exploiting the next step in advanced sign-flipping tech.
I believe the generated code could be further optimized by mapping the sign-flipping operation into a texture transformation in GPU space. Have you toyed with this idea, and if so, what did you find out?
Once again, thank you for this awesome piece of tech.