| === nickmoeck_ is now known as nickmoeck | ||
| ripps | I have a dilema: intel i3 2100 or amd athlon x4. | 08:55 |
|---|---|---|
| ripps | sandybridge with 2 cores vs. athlon with 4. Sandybridge is suppose to have awesome performance, but does it trump an extra 2 cores | 08:56 |
| ripps | I've looked at some comparison benchmarks, but it they seem pretty comparable, except each one does better a certain operation than the other. | 09:05 |
| mikeputnam | ripps: intel i7! | 14:19 |
| KBme | well, i'd guess it depends on how parallelisable is what you'r doing | 14:22 |
| KBme | you're | 14:22 |
| KBme | if it can be parallelised easily more cores will help more | 14:22 |
| mikeputnam | so do "cores" present themselves as disticnt cpus to the os? | 14:23 |
| mikeputnam | s/disticnt/distinct/ | 14:24 |
| spikeb | yes | 14:24 |
| mikeputnam | ah | 14:24 |
| mikeputnam | so a dual processor with dual cores would appear to have 4 cpus | 14:25 |
| mikeputnam | got it | 14:25 |
| KBme | also intel Hyper Threading makes cores show up as two cores | 14:27 |
| mikeputnam | as a non-os programmer, i wonder how linux itself uses or does not use parallelism with regard to cpus | 14:27 |
| mikeputnam | KBme: i see. didn't realize that either | 14:28 |
| KBme | how it uses? | 14:28 |
| KBme | it has a scheduler to map OS level threads to cores | 14:28 |
| mikeputnam | so in my earlier example the 2 proc dual core may look like 8 cpus? | 14:28 |
| KBme | if it has hyper threading yes, each core shows up as two cpus | 14:29 |
| mikeputnam | crazy | 14:29 |
| KBme | what i understood about hyper threading is that if the process runs into a bottleneck it can be rescheduled and an other process can take over | 14:29 |
| KBme | but don't take my word on that one, we're getting on the murky water of CPU design | 14:30 |
| mikeputnam | heh | 14:30 |
| mikeputnam | so even if an application isn't designed for parallelism, it probably would benefit in someways with the os balancing it's own threads across cpus | 14:31 |
| KBme | well | 14:32 |
| KBme | the whole system still benefits for it | 14:32 |
| mikeputnam | right | 14:32 |
| KBme | your application won't | 14:32 |
| mikeputnam | so indirectly in my example themn | 14:32 |
| mikeputnam | *then | 14:32 |
| mikeputnam | ah | 14:32 |
| KBme | yeah | 14:32 |
| KBme | well, since it can fully use one core and the "other stuff" can be mapped to other cores | 14:33 |
| mikeputnam | so generally speaking without actively designing for parallelism, your app exists as a single thread (pid?) | 14:33 |
| KBme | it will get more cpu time than if the whole system is on the same core | 14:33 |
| mikeputnam | right | 14:33 |
| KBme | well, that depends too | 14:33 |
| KBme | you can thread your program and still not be parallel | 14:33 |
| mikeputnam | overall timeshare, but the os isn't going to magically separate out chunks of your app to leverage idle cpu cycles | 14:33 |
| KBme | look up for example the difference between concurrency and parallelism | 14:33 |
| KBme | no, it won't ☺ | 14:34 |
| KBme | though the haskell people are working on that :D | 14:34 |
| mikeputnam | it's interesting to think about concurrency and parallelism out of the context of computing | 14:35 |
| mikeputnam | like a manufacturing business for example | 14:35 |
| mikeputnam | get more orders -> get more people -> do more work | 14:35 |
| KBme | yeah | 14:36 |
| KBme | io bottleneck: an unfinished highway | 14:36 |
| mikeputnam | heh | 14:36 |
| KBme | mikeputnam, dhmn looks interesting :P | 14:37 |
| mikeputnam | or even open source software development. i believe human brains are basically single threaded. but get 15 people... :) | 14:37 |
| mikeputnam | KBme: it is. clever people doing fun and interesting stuff | 14:38 |
| mikeputnam | KBme: join the mailing list. lots of neat stuff goes by | 14:38 |
| KBme | mikeputnam, you are very wrong about the human brain | 14:38 |
| mikeputnam | yes i suppose. there's a lot going on: keep my breathing, seeing, hearing, fight or flight, etc etc | 14:39 |
| KBme | yep | 14:40 |
| KBme | also, if you're really interested, search for "Google Workshop on Quantum Biology" in youtube, pretty mind-boggling talks | 14:40 |
| mikeputnam | i guess i was referring to the voluntary thought->action process | 14:40 |
| KBme | that "counscious" part...hmm yeah, maybe | 14:40 |
| KBme | i wouldn't be sure, tho, even | 14:40 |
| mikeputnam | at least i know i can't write 3 programs in parallel | 14:41 |
| KBme | heh | 14:41 |
| KBme | mikeputnam, i'm in france for now ☺ | 14:41 |
| mikeputnam | neat. seek out "hackerspace"s near your locality. you'll be glad you did. | 14:43 |
| mikeputnam | http://hackerspaces.org/wiki/France | 14:43 |
| KBme | there seem to be non where I am.. | 14:44 |
| mikeputnam | :/ | 14:44 |
| mikeputnam | start one? | 14:45 |
| mikeputnam | :) | 14:45 |
| KBme | so, what is it? | 14:45 |
| KBme | hmmm | 14:45 |
| KBme | i guess i'll just rtfm ;) | 14:45 |
| KBme | mikeputnam, we're kind of starting one | 14:46 |
| mikeputnam | great! | 14:47 |
| mikeputnam | sort of a community workshop | 14:47 |
| KBme | for hackers | 14:47 |
| KBme | yeah | 14:47 |
| mikeputnam | like a health club membership, but for your brain | 14:47 |
| KBme | so what do you program, mikeputnam? | 14:48 |
| mikeputnam | computer people, metalworking people, woodworking people, crafty people, electronics people, art people, etc, etc | 14:48 |
| mikeputnam | i work on an Oracle/PeopleSoft ERP app | 14:48 |
| KBme | ☺ | 14:49 |
| KBme | ugg (oracle;) | 14:49 |
| mikeputnam | yeah | 14:49 |
| KBme | they do usually pay well tho | 14:49 |
| KBme | well, ogtta get back to hacking on my couchapp laterz | 14:50 |
| mikeputnam | enjoy | 14:51 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!