A couple of detailed, somewhat random thoughts:
-- preliminary style guide:
http://www.ccs.neu.edu/home/matthias/Style/style/
-- equip modules and functions with purpose statements
-- consider using (module+ test ...)
-- use define-inline, try the optimization coach
-- you might wish to put the math library into typed racket
I am sure others can suggest a few more detailed hints -- Matthias
On Jan 31, 2013, at 7:30 AM, Piotr Klibert wrote:
> I tried this a bit (this is why all the constants in bezier-math are
> floats) but I didn't notice any improvement. May be that I was using
> flonums inconsistently, I'll try once again.
> Best regards,
> Piotr Klibert
>
>
> 2013/1/31 Jens Axel Søgaard <jensa...@soegaard.net>:
>> 2013/1/31 Carl Eastlund <c...@ccs.neu.edu>:
>>> Piotr,
>>>
>>> This isn't good Racket code, it's *gorgeous*. That was my impression at
>>> first glance, and the more I look, the more I like it. Your Bezier library
>>> absolutely does not read like the code of a novice Racketeer. Few enough
>>> people regularly use a combination of objects, pattern matching, map, cut,
>>> macros, contracts, named let, multiple values, submodules... let alone using
>>> them all in readable and idiomatic fashion.
>>
>> I agree with this. Welcome to the club.
>>
>> I see that in the hunt for performance, you have used macros
>> as a form of "manual inlining". It might be worth to experiment
>> with fl+, fl* and friends, since they don't need to handle other
>> number types, they might give you some extra performance.
>>
>> http://docs.racket-lang.org/reference/flonums.html
>>
>> --
>> Jens Axel Søgaard
>
> ____________________
> Racket Users list:
> http://lists.racket-lang.org/users
____________________
Racket Users list:
http://lists.racket-lang.org/users
