[racket] Kudos: match and performance

On Feb 24, 2013, at 6:59 AM, Tim Brown wrote:

> Which is a bit annoying, since now I have to rethink how I implement solutions in racket, since the language now seems to be capable of behaving as if it's smarter than I am 

A couple of remarks: 

1. Compiling code is a bureaucratic task. Example: You need to keep track of all provably truths about the context of some piece of code to create good object code for it. There are usually tons of such truths. 

If software is good at anything, it is good a bureaucracy. You should therefore not be surprised that compilers are good at that kind of compilation task. Indeed, you should assume that the compiler is that good and program as if it were. 

2. Sadly, we all know that this isn't the case -- neither for C nor for languages with complex linguistic constructs that may impose a significant cost. 

That's what the Optimization Coach work is about. But we're only at the beginning and I am sure we have years of work ahead of us. 

-- Matthias