[schwoll]

For portability by far the vast majority will say C. In my experience the C compiler optimizer will do a lot with -O2 or -03 but it can't always infer correct SIMD optimization for some operations and on occasion you have to drop down into x86_64 assembly. The idea is do most things in C and use __asm__ to write custom assembly instructions. With #defines around the assembly for each processor you plan on supporting you get the benefit of both C optimizers and portability across different CPUs as well as any future updates to the compiler in the future. But the compiler writers will say to use intrinsics and extended assembly rather than raw assembly because when you write raw assembly your code becomes a black box to the compiler and it can't infer optimizations for your surrounding code that interfaces with the assembly. I think C with extended Asm is likely the most sane combination if you don't mind the slightly ugly syntax and the fact that there could be differences between compilers. That being said, C with compiler intrinsics seems to be a happy compromise for those that don't want to shift around registers and deal with the stack.

https://gcc.gnu.org/wiki/DontUseInlineAsm

https://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html

I don't use Rust so I can't comment on it but it also has compiler intrinsics + a memory safety model. It's compiler is really dog slow last time I used it so I hope that has improved but nobody is really killing C any time soon, even if there's enthusiasm for memory safety. Sooner or later you have to delve down into the depths of Narnia and you may as well get comfortable dealing with memory.

My likely favorite combination is Python + C (for the speed stuff) + Intrinsics (for the really speed stuff).