In the previous class, we noted that constraining the structure of the code you write is helpful for the creation of high-performance parallel algorithms. With general-purpose algorithms, the compiler can't give you that much help. Hailide constrains what you can do, since you can only write functions on arrays, but the compiler is able to take much larger role in optimizing and generating the low level code for you. With Hailide as inspiration, it's interesting to think about how to split up the responsibilities between the programmer and the system when performing computation on graphs.
In the previous class, we noted that constraining the structure of the code you write is helpful for the creation of high-performance parallel algorithms. With general-purpose algorithms, the compiler can't give you that much help. Hailide constrains what you can do, since you can only write functions on arrays, but the compiler is able to take much larger role in optimizing and generating the low level code for you. With Hailide as inspiration, it's interesting to think about how to split up the responsibilities between the programmer and the system when performing computation on graphs.