This makes me think of sequence operations in functional languages, where you can compose map's, reduces, flattens, etc. to elegantly perform a lot of tasks.
mario
Architects usually look at whether something composes with everything else. A primitive that does not compose with the given domain may not be apart of that domain.
o_o
By having primitives that compose, the system therefore doesn't need many primitives, as stated in #2, because the programmer will be able to use the simple and efficient primitives to construct more use cases for the system.
Firephinx
But with having primitives compose, what if we run into the problem where the system takes more time than if it had been specialized for the case? For example, restructuring loops for locality in quite a few of the lectures.
This makes me think of sequence operations in functional languages, where you can compose map's, reduces, flattens, etc. to elegantly perform a lot of tasks.
Architects usually look at whether something composes with everything else. A primitive that does not compose with the given domain may not be apart of that domain.
By having primitives that compose, the system therefore doesn't need many primitives, as stated in #2, because the programmer will be able to use the simple and efficient primitives to construct more use cases for the system.
But with having primitives compose, what if we run into the problem where the system takes more time than if it had been specialized for the case? For example, restructuring loops for locality in quite a few of the lectures.