freenode/#sicl - IRC Chatlog
Search
22:32:17
jcowan
though my ASR-33 did not understand them, which I think was a (mechanical) configuration problem.
4:08:02
beach
makomo: Thanks for you fix. I am slowly untabifying the files, by doing one from time to time. The tabs are a leftover from the default way that Emacs does it, and I only fairly recently stopped using them.
6:57:49
beach
makomo: If you feel like it, there are also a lot of instances where I signal a simple error, but I should define a specific condition, and define a condition reporter for it. That kind of stuff can be done one condition per day, so doesn't require a lot of uninterrupted time.
6:58:53
beach
In a directory, there is typically a conditions.lisp file and a condition-reporters-english.lisp file that contain some definitions already, so those files just have to be added to.
7:27:19
no-defun-allowed
i don't know why, haven't been doing much programming recently. maybe it's the weather, maybe i'm just procrastinating too hard
7:37:07
beach
Sure, but that's not enough. The paper has to be comprehensible to the referees. And it's a very narrow topic, so if I get the wrong referees, they won't understand anything.
7:37:32
beach
I tried to describe what I do to Didier Verna, and got a blank stare in return. And he is one of the knowledgeable ones.
7:39:19
no-defun-allowed
well, i can't really write anymore since i'd be in danger from a new anti-cryptography law passed here recently
7:41:44
no-defun-allowed
the cops can throw a warrant at me and say "backdoor your program" and i can't since it's distributed so i'd get a fine instead
7:41:47
beach
Are you saying that 1. It is now illegal to encrypt, and 2. When you write things, you do?
7:42:31
no-defun-allowed
i am saying it is illegal (as far as i know) now to publish software that uses cryptography that cannot be backdoored when requested by warrant
7:43:59
no-defun-allowed
and it's not likely i'd be able to make such a change since the software will be run on interconnected servers and additionally someone would notice the demanded changes and refuse to accept them
7:44:43
no-defun-allowed
it is very worrying here, our prime minister even said (paraphrasing) "the laws of encryption are not more powerful than the laws of australia"
7:45:52
no-defun-allowed
i may be able to get away with it since a) most of the crypto is ECDSA signing and b) i'm not a company but the media doesn't really cover either scenario since few readers would understand those
7:50:04
no-defun-allowed
there's about 150 pages of amendments in the bill, so i probably should see a lawyer or someone professional about this
7:52:35
makomo
beach: hah, i was just away, that is all. :-) well, it seems like a good way to start helping around, i'll see what i can do
7:53:14
makomo
first gotta finish some TODOs, along with that backquote wager we had though. the reason why i did these typos today is because i had them on my TODO list for so long
8:51:42
heisig
Speaking of contributing to SICL - I still intend to work on sequences. But other things are keeping me busy at the moment.
8:57:46
heisig
I am working on a library that introduces 'restricted-functions', such as cl:+ restricted to two arguments of type double-float and double-float.
8:58:40
heisig
Each restricted function has metadata, such as its cycle cost and, most importantly, whether it can be vectorized and what the maximal vector width is.
8:59:39
heisig
Petalisp IR is already in SSA form and has no inter-loop dependencies. So I can just bump the loop stride to the maximal vector width and vectorize the body of the innermost loop.
9:01:04
heisig
And I have to add a loop epilogue, and some other things. But that is the general idea.