freenode/#lisp - IRC Chatlog

1:08:35 no-defun-allowed fair enough

1:08:46 no-defun-allowed gyah, using make-array to make actual arrays doesn't speed it up

1:09:14 aeth no-defun-allowed: What are you trying to speed up?

1:09:14 no-defun-allowed there must be a compilation step then.

1:09:24 aeth This? https://gist.github.com/nodefunallowed/95b95469449124570f36bb05192b7ef8

1:10:57 no-defun-allowed yes

1:11:06 no-defun-allowed i'll typedef the functions then

1:11:11 aeth no-defun-allowed: (1) Because you have non-earmuffed globals with the name x, y, and w, you're using slow dynamic variables every time you have a local named x, y, or w

1:11:37 sjl at least the actual problem solving code is nice now that I have the gross iterate driver to do the hard work https://github.com/sjl/rosalind/blob/master/src/problems/gc.lisp#L38-L51

1:11:50 aeth no-defun-allowed: (2) Inline heavily! Disassemble it, and if its disassemble is small enough, inline it! It makes it a bit painful to redefine functions (you have to remember to redefine the whole file) but it can greatly optimize array-related code because it removes a ton of bounds checks

1:11:50 no-defun-allowed hmm true

1:11:56 no-defun-allowed whoops

1:13:14 aeth no-defun-allowed: (3) Make sure your arrays have a valid numeric :element-type (for this sort of thing you might want single-float, double-float, (complex single-float), or (complex double-float)) unless your as-matrix macro is taking care of that.

1:13:44 no-defun-allowed i'm not sure what petalisp does to the defuns but i'll try that

1:13:53 aeth Well you can just print type-of

1:13:54 no-defun-allowed there's a bit i've done that's not on the repo yet

1:14:10 no-defun-allowed alright

1:14:15 aeth If it's a T array, then you should be writing a(n inline) helper function that takes in an initial list and outputs the correct type

1:14:25 aeth it needs to be inline to avoid consing the initial-contents list

1:14:39 aeth (or a macro, but don't use a macro when an inline function will work)

1:17:41 no-defun-allowed hmm, sometimes petalisp passes REFERENCEs and APPLICATIONs to the functions

1:18:48 no-defun-allowed eh well, it's faster than doing it in python

1:19:22 aeth If correctly written, this sort of thing should be 30x to 50x faster than non-Numpy Python.

1:19:35 aeth At least ime porting algorithms from Python examples

1:22:22 aeth (of course the correct way to do this sort of thing in Python is to use Numpy)

1:24:05 on_ion julia

2:13:11 emaczen what does the #- reader macro do?

2:14:36 no-defun-allowed opposite of #+

2:16:37 no-defun-allowed #+ checks if the next keyword is in features, and #- checks that it isn't in there

2:17:04 no-defun-allowed eg #-sbcl "This is not my beautiful compiler!"

2:17:32 no-defun-allowed nothing is read if it's not SBCL and that string is read if it is

2:17:41 no-defun-allowed wait no that's the other way around

4:19:53 beach Good morning everyone!

4:51:12 no-defun-allowed morning beach

6:52:54 drdo_ ** NICK drdo

7:20:21 shrdlu68 Hello commoners!

8:11:48 splittist hello citizens

8:26:36 no-defun-allowed hi lispers

8:50:14 shka_ good morning

8:52:44 no-defun-allowed hi shka_

9:04:51 hhdave_ ** NICK hhdave

11:03:25 Bike_ ** NICK Bicyclidine

11:53:11 knobo how do I make cl-autowrap define macros like for example __USE_XOPEN?

11:53:45 knobo So that blocks inside #ifdef __USE_XOPEN2K is included

12:23:59 knobo I'm trying to use cl-autowrap with /usr/include/time.h but strptime is not generated

12:35:53 knobo I needed :sysincludes '("/usr/include/x86_64-linux-gnu")

12:42:24 jarindyk if I have a list such as '((foo . 1) (bar . 2)), is there a way to get "foo" from 1, "bar" from 2, and 1 from "foo", 2 from "bar"? or what would be the best way to do something like this? I could create two lists too but I'm a noob so I thought I'd ask

12:43:30 pfdietz You mean, search for the element whose cdr is some value, then get the car?

12:45:06 pfdietz clhs assoc

12:45:06 specbot http://www.lispworks.com/reference/HyperSpec/Body/f_assocc.htm

12:45:10 pfdietz clhs rassoc

12:45:10 specbot http://www.lispworks.com/reference/HyperSpec/Body/f_rassoc.htm

12:45:17 pfdietz clhs find

12:45:18 specbot http://www.lispworks.com/reference/HyperSpec/Body/f_find_.htm

12:45:19 jarindyk pfdietz: well I'm not sure if that's the right way to go about this, as the list is a constant, I'll just need a way to associate a string to an integer and be able to look up the key and value from value and key respectively

12:45:35 jarindyk if it makes any sense :p

12:45:54 pfdietz clhs 18

12:45:54 specbot Hash Tables: http://www.lispworks.com/reference/HyperSpec/Body/18_.htm

12:46:29 pfdietz Also, consider using symbols instead of strings, when you are not actually doing stringy stuff to them.

12:47:01 jarindyk yeah good idea

12:47:44 pfdietz Keyword symbols are useful where you might use enums in C (if you're not converting them to actual ints).

12:48:04 pfdietz :foo, :bar, etc.

12:49:22 pfdietz If this is all compile time constant stuff, then if they foo, bar are symbols you can use CASE and the like (can't do that with strings, since it does EQL comparison).

12:50:23 pfdietz To make it extensible, you could use a generic function, and use EQL method parameter specifiers.

12:50:36 pfdietz (defmethod f ((x (eql :foo))) 1)

12:50:37 dlowe alexandia:switch will take a function

12:50:58 pfdietz (defmethod f-inv ((y (eql 1))) :foo)

13:05:19 knobo yay! I have strptime working :)