“Currently the focus is on language design, later it will shift to compiler optimizations with performance benchmarks. Q. Is it OK to deal with performance later? A. … We keep performance in mind during making design decisions. Q. Is it possible to get good performance for a high-level language like LambdaCube 3D? A. … You can see LambdaCube 3D as an experiment achieving this goal.”
Overall, I am skeptical, considering design decisions made by Futhark for high performance. Futhark made innovations in monomorphisation and defunctionalisation and still needed somewhat big restrictions.
Futhark is probably better for high-performance, but writing shaders is still very error prone, so that LambdaCube would definitely solve a problem there. And LambdaCube works on WebGL while Futhark doesn’t. It’s good to see some innovative on shader programming, it’s a hot topic.
From their Q&A:
“Currently the focus is on language design, later it will shift to compiler optimizations with performance benchmarks. Q. Is it OK to deal with performance later? A. … We keep performance in mind during making design decisions. Q. Is it possible to get good performance for a high-level language like LambdaCube 3D? A. … You can see LambdaCube 3D as an experiment achieving this goal.”
Overall, I am skeptical, considering design decisions made by Futhark for high performance. Futhark made innovations in monomorphisation and defunctionalisation and still needed somewhat big restrictions.
Futhark is probably better for high-performance, but writing shaders is still very error prone, so that LambdaCube would definitely solve a problem there. And LambdaCube works on WebGL while Futhark doesn’t. It’s good to see some innovative on shader programming, it’s a hot topic.