|
FamDav posted:i'd like to think shaggs meant gathering that information through an actual run of the program, but maybe not Oh, in that case it's an even more ridiculous statement than I thought. BRB, just running production loads in my IDE.
|
#
?
Feb 21, 2013 21:04
|
|
|
|
| # ? Jun 11, 2024 02:24 |
|
|
FamDav posted:i'd like to think shaggs meant gathering that information through an actual run of the program, but maybe not yes.
|
|
#
?
Feb 21, 2013 21:23
|
|
|
Zombywuf posted:Oh, in that case it's an even more ridiculous statement than I thought. BRB, just running production loads in my IDE. lol if ur not profiling production loads in acceptance.
|
|
#
?
Feb 21, 2013 21:23
|
|
|
MononcQc posted:What your culture really says / Toxic lies about startups' 'culture' in silicon valley sent this to the company-wide 'neat poo poo' mailing list of the startup I work at got an 'interesting article' comment from the HR guy at lunch Shaggar posted:stuff about testing shaggar was right
|
|
#
?
Feb 21, 2013 21:28
|
|
|
Shaggar posted:lol if ur not profiling production loads in acceptance. Nah, I have the luxury of running testing directly on prod data sets. Data processing apps can be so much less stressful to develop. I run them on a cluster of machines though, not in my IDE. When I gently caress up my analysis though, test loads chew through hundreds of dollars of EC2 time.
|
|
#
?
Feb 21, 2013 21:32
|
|
|
Shaggar posted:lol if ur not profiling production loads in acceptance. lol if your poo poo is so unpopular or limited use you can even do that
|
|
#
?
Feb 21, 2013 21:47
|
|
|
PleasingFungus posted:sent this to the company-wide 'neat poo poo' mailing list of the startup I work at That article had a really good point: quote:Meetings are evil and we have them as little as possible. There are days I feel PTSDd from meetings where the objective seems to be to say as many words as possible.
|
|
#
?
Feb 21, 2013 21:49
|
|
|
quote:We are all makers who are focused on shipping. quote:We have a team of people who are responsible for organizing frequent employee social events, maintaining the office feel, and making sure work is a great place to hang out. We get served organic, vegan, farm-raised, nutritious lunches every day at work.
|
|
#
?
Feb 21, 2013 21:52
|
|
|
Cocoa Crispies posted:lol if your poo poo is so unpopular or limited use you can even do that yeah that's your excuse for not testing. sure.
|
|
#
?
Feb 21, 2013 21:52
|
|
|
its like an echo chamber an echo chamber of shitposting
|
|
#
?
Feb 21, 2013 21:55
|
|
|
Otto Skorzeny posted:thank you, don quixote. i think your arguments about analysis might have gone down better with some other measure than reliability. software in most cases is 100% reliable - it will do the wrong thing every time, in exactly the same way. i also think you already know this though 
|
|
#
?
Feb 21, 2013 21:57
|
|
|
Cocoa Crispies posted:lol if your poo poo is so unpopular or limited use you can even do that or you aren't in web development ork ork
|
|
#
?
Feb 21, 2013 22:01
|
|
|
ninjeff posted:i think your arguments about analysis might have gone down better with some other measure than reliability. software in most cases is 100% reliable - it will do the wrong thing every time, in exactly the same way. i also think you already know this though Says someone who's clearly never used more than one thread.
|
|
#
?
Feb 21, 2013 22:17
|
|
|
Zombywuf posted:Says someone who's clearly never used more than one thread. I wanted to ask about that. Are there any static analysis tools out there expressly for concurrency? That seems like it would be a difficult problem.
|
|
#
?
Feb 21, 2013 22:19
|
|
|
Cocoa Crispies posted:lol if your poo poo is so unpopular or limited use you can even do that lol if your department can afford machines as powerful as the customers use 
|
|
#
?
Feb 21, 2013 22:20
|
|
|
i want to do the kind of work where it is important enough to needs reliability calculations  or maybe i dont
|
|
#
?
Feb 21, 2013 22:25
|
|
|
Zombywuf posted:Says someone who's clearly never used more than one thread. ugh yeah that is actually a case where reliability is non-binary isn't it. my bad my point still holds for many many kinds of concurrency though, and it's telling that much of the point of many concurrency models is eliminating reliability as a factor. reliability still counts when you can't use one of those models for w/e reason though, you're right
|
|
#
?
Feb 21, 2013 22:36
|
|
|
Sweeper posted:i want to do the kind of work where it is important enough to needs reliability calculations if you aren't even personally auditing your work in some way then god help you mayne. also i wrote a boggle solver in go last night. they really need to think about string indexing and runes vs. bytes. why is it that I can iterate over the runes in a string s using range s, but in order to index to a specific rune I have to use []rune(s) and index into that array? edit: also they need to reconsider make/new/composite literal construction with maps, arrays/slices, and channels FamDav fucked around with this message at 22:57 on Feb 21, 2013 |
|
#
?
Feb 21, 2013 22:48
|
|
|
Engineering reliability of real world parts seems very different than software reliability. The closest analogy to components of varying failure profiles for software engineering are the fundamental algorithms, proven by mathematics. i.e. Quicksort has a speed profile where so and so percentage of data input is within acceptable performance. This percentage is based on empirical data gathered from use cases. Unless the input is coming from something derived from a hard science and not just ppl's input, doing high level analysis is like predicting whether people will like the new triple-quadrouple heartburn burger. Servers and harddrives do have multiple backups and cost calculations for the same reason real engineering will calculate the cost of building a bridge that won't collapse for 50 years. I've only done multi-threading in a past life, bioinformatics. Splitting data to run on multiple threads and sleeping until they are all done to rearrange in order so reliability was not an issue. Is there a real world example where threading would result in non-binary reliability? Is that just not being threadsafe for a few percent more performance? MeruFM fucked around with this message at 23:23 on Feb 21, 2013 |
|
#
?
Feb 21, 2013 23:19
|
|
|
yes people don't understand how to do safe concurrency and they make a mess of it. really the solution is to avoid it whenever possible.
|
|
#
?
Feb 21, 2013 23:23
|
|
|
Apparently LinkedIn is switching over to Play http://engineering.linkedin.com/play/play-framework-linkedin
|
|
#
?
Feb 21, 2013 23:26
|
|
|
im the super loving annoying animation on http://www.playframework.com/
|
|
#
?
Feb 21, 2013 23:28
|
|
|
MeruFM posted:Is there a real world example where threading would result in non-binary reliability? Is that just not being threadsafe for a few percent more performance?
|
|
#
?
Feb 21, 2013 23:28
|
|
|
FamDav posted:if you aren't even personally auditing your work in some way then god help you mayne. you can audit code in a way that doesn't involve stat
|
|
#
?
Feb 21, 2013 23:33
|
|
|
Sweeper posted:you can audit code in a way that doesn't involve stat yeah, but there are insights to be gained from statistical analysis even on personal projects.
|
|
#
?
Feb 21, 2013 23:57
|
|
|
FamDav posted:yeah, but there are insights to be gained from statistical analysis even on personal projects. go on
|
|
#
?
Feb 22, 2013 00:14
|
|
|
Sweeper posted:go on so since i've got boggle on the brain let's use that. i use a trie and i'm using that to search for words that start from each position on the board. everything runs alright but when I profile the code i notice that i'm spending an inordinate amount of time in calls to containsPrefix(trie) because i'm searching the whole prefix at each entry i visit. instead, i can just pass around subtries in my search and only check that the next character is in that subtrie. i've now reduced something that resulted in quadratic complexity down to something linear. probably not something you need to do if you just want to solve a single board, but if you want to search for optimal boggle boards then you want to be able to score them as quickly as possible
|
|
#
?
Feb 22, 2013 00:28
|
|
|
FamDav posted:so since i've got boggle on the brain let's use that. i use a trie and i'm using that to search for words that start from each position on the board. everything runs alright but when I profile the code i notice that i'm spending an inordinate amount of time in calls to containsPrefix(trie) because i'm searching the whole prefix at each entry i visit. instead, i can just pass around subtries in my search and only check that the next character is in that subtrie. i've now reduced something that resulted in quadratic complexity down to something linear. i was expecting more, not impressed please see me after class
|
|
#
?
Feb 22, 2013 00:33
|
|
|
Sweeper posted:i was expecting more, not impressed please see me after class lol Sweeper posted:i want to do the kind of work where it is important enough to needs reliability calculations have you even graduated yet
|
|
#
?
Feb 22, 2013 00:45
|
|
|
Shaggar posted:yes people don't understand how to do safe concurrency and they make a mess of it. really the solution is to avoid it whenever possible. yo, so this is totally true, and should be lasered onto the moon. But! if you are doing a non-trivial thing that needs threads, you need to do the math proof analysis thing to make sure you didn't gently caress something up. even if its informal. no amount of debugging/testing/logging can save you.
|
|
#
?
Feb 22, 2013 01:00
|
|
|
Bream posted:I wanted to ask about that. Are there any static analysis tools out there expressly for concurrency? That seems like it would be a difficult problem. for java this particular static checker ensures mutual exclusion is properly implemented. seems quite limited but I think it could be used to split UI-thread-only parts away from free-for-all-threads parts in an API and ensure it's followed clang looks to be implementing a much more expressive system for checking locking policies
|
|
#
?
Feb 22, 2013 04:03
|
|
|
threading is terrible shared mutable state is Satan's piss in a glass fixing someone else's "threading is easy guys, I got this" code is drinking Satan's piss from the glass
|
|
#
?
Feb 22, 2013 04:19
|
|
|
How about the c++11 thread features?
|
|
#
?
Feb 22, 2013 04:20
|
|
|
FamDav posted:lol no why would i go and do that
|
|
#
?
Feb 22, 2013 05:02
|
|
|
Nomnom Cookie posted:threading is terrible Man, shared mutable horrors are just a large subset of the problems with threading. It's a large subset that I've found Objective-C makes really really easy to get into...thanks NSMutableXXX.
|
|
#
?
Feb 22, 2013 05:56
|
|
|
Bream posted:How about the c++11 thread features? C++ is poo poo and threading is poo poo, what if we combine them
|
|
#
?
Feb 22, 2013 06:03
|
|
|
Shaggar posted:a theoretical model is always always wrong. always always?????????????
|
|
#
?
Feb 22, 2013 06:17
|
|
|
ultramiraculous posted:Man, shared mutable horrors are just a large subset of the problems with threading. It's a large subset that I've found Objective-C makes really really easy to get into...thanks NSMutableXXX. are you seriously blaming the library for your shortcomings ahahaha
|
|
#
?
Feb 22, 2013 06:19
|
|
|
pro tip: dont share NSMutable objects between threads
|
|
#
?
Feb 22, 2013 06:20
|
|
|
|
| # ? Jun 11, 2024 02:24 |
|
|
in fact theyre entirely non-safe for concurrency and if you use them you are guaranteed to get heisenbugs
|
|
#
?
Feb 22, 2013 06:21
|
|