| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13
Comments:
<0> ...
<0> Because it'd be faster if it used both at 100%?
<1> there are compilers than can auto thread code during optimisation
<1> but its not that good
<0> And it wouldn't work with graphics
<1> true
<1> a friend of mine is working on a compiler like that for his thesis
<2> ok about graphics, but for content loading
<0> There's also compilers that vectorize stuff but nobody uses them
<0> Klaim: content loading is not significant, it doesn't take 3 months to load data
<1> who cares about loading
<1> in any case it will be HD bound
<2> MrAshes> yes but having it done by another thread don't stop your game, and that is a verry good thing
<0> Many games load levels while running
<0> And they don't need multithreading for that
<0> [00:26.35] <1> in any case it will be HD bound
<0> [00:19.55] <0> Loading the world is mostly harddisk bound
<0> I win
<0> Har har har
<2> yes but it's not as good
<1> ah true, didnt see that :)
<0> It's exactly the same, Klaim
<2> lol
<0> The game won't stop, in both cases
<2> nop, as i said, the game is not stopped, the rendering is not stoped
<1> unless DMA is disabled
<2> it will
<1> then it might stop
<2> look at something like morrowind
<2> it is int the same thread, it have to stop
<0> There are tons of games that don't stop when loading levels
<0> and they're not multithreaded
<3> lots of games are completely single threaded, Klaim
<3> in fact, most probably are
<0> It's obvious it makes no difference
<2> yes about all games
<0> 99.9% of games are singlethreaded
<2> but the game you speak about d'ont use this kind of loading
<0> They don't, and they don't stop when loading
<0> Just like Oblivion
<2> it loads about all the ressources on the level loading
<0> Having it multithreaded and about 5-10% on the other CPU is just the same as overclocking your CPU by 10% and using only one
<4> Ashe: So here's what I want to know: Is there no practical benefit to multithreading in games, or is it just a matter of coding style catching up with technology?
<0> Style
<0> Just in rendering, you need to have stuff in the right order and such
<0> If you render from 2 threads, it can be difficult
<0> Personally I use kind of special queues
<2> :p ok let me know a game that don't stop on loading and that don't load by levels :)
<1> Style + difficulty :)
<1> threading is hard
<0> And I get my 2 Xeons
<2> (on PC i mean)
<0> (+ 2 logical cores)
<0> 100% used
<0> Doesn't matter if it's a PC or not
<0> All consoles have a single CPU
<0> Except for xbox360
<3> Klaim what do you mean "stop on loading" ?
<0> He means the game stops and you get a "loading" screen
<2> yes it matters : on console, it will be parrallel processing in most cases
<0> Just all flight simulations load the level while playing
<0> No it won't be parallel processing on consoles
<3> most cases?
<0> Why would it be different than PCs
<0> Consoles have threads just like PCs
<3> only 1 console out there can support parallel processing, i think
<0> And a single CPU
<0> No, they all do
<2> rdragon> playing(graphic rendering)->loading(not rendering)->playing...
<0> Gamecube, PS2 and Xbox have threads
<3> threads != parallel processing
<0> And xbox360 is the only one with 3 cores + 3 logical cores
<2> MrAshes> hu
<4> Klaim: There is asynchronous I/O.
<0> rdragon: just SIMD is considered parallel processing
<1> without DMA it would stop :)
<3> the correct term is 'block' or 'wait', not 'stop'
<2> on the PS2, i heard ther is 4 processors : 1 for graphic processing, 1 for sound processing, those are like you say ok; but the to others are multi purpose as far as i know
<2> so how doi they exploit it?
<0> You heard wrong
<2> rdragon>ok
<0> I've been developping on PS2 for like 5 years
<2> ok so how is it?
<0> It's vector units
<0> And the sound processor is like having a sound card (duh)
<2> yes that is what i said
<0> Ok, so in your PC you have the chipset, the CPU, the sound card, the network card, they're all adapters
<0> That's 4 as well
<0> :/
<0> What you're saying is stupid
<0> So anyway you have only one CPU on PS2
<2> :/ ok back to multicore, there is two threads in my process, one is launched on one core, the other on another core, you say that is not exploiting multicore, so who say stupid?
<3> it's stupid if one thread ends up waiting a lot, or otherwise doesn't use up much of the 2nd core
<0> If it's doing almost nothing on the second one, that's not what I'd call exploiting, indeed
<2> yes i understand that, but if it is not?
<0> That's the case for Oblivion
<2> hmm ok
<1> load balancing = exploiting?
<2> so it's the difference between using and exploiting for you i guess
<3> a good exploitation of a multicore processor would have both processors doing a lot of work at the same time
<0> So both 100%
<3> both cores, whatever
<5> you want to keep each core/processor working as hard as possible
<5> much easier said than done in most applications :)
<0> And not 1 100% and the other one waiting for the harddisk
<0> I did it, it works perfect
<0> I have 15 threads running
<0> (when I have 2 pads plugged)
<5> well, games are not "most applications"
<1> MrAshe, that mustnt have been easy
<0> It was a pain, indeed, hehe
<5> more research into autovectorisation needed!
<0> So it's 7 threads per gamepad then 1 thread for the game
<0> Har har har
<3> MrAshe have you messed with the PS3 at all yet?
<0> Asriel: multithread is too much highlevel for a compiler
<2> yes i understand where is the problem, i know it but you know that it's not possible in a game to have really heavy processes being executed without synchronization
<5> MrAshe - Depends on the source language
<2> so what you say is that multicore exploitation is not possible?
<0> Klaim: I just said I did it and it works on 4 CPUs
<3> Klaim don't say it's not possible
<0> (with all used 100%)
<5> For C++, sure. But for some (mostly functional) languages, you can gain much easier paralell distribution that wasn't implicit in the code
<3> Klaim it just requires people to adjust and adapt to make the code exploit multiple cores
<0> And add a bunch of mutex/critical sections
<0> Every line
<0> Then it works :D
<1> hehe
<1> multithreading is a pain in the ***
<1> Asriel, use your phd time to solve some threading programming problems
<0> Like why I need 7 threads for a gamepad
<6> why do you need 7 threads?
<0> rdragon: I just tested some preview stuff (compiler/lib)
<5> I'd like too. Much extracuricular research is focusing on theoretical languages for distributed computations
<6> is it like 1 for the direction pad, then 1 per button or something?
<5> which is basically "multithreading", though all the grants coming in are for "networked computation"
<0> Noidea: I was just kidding, because I said my app is 15 threads with 2 pads (it's just 1 thread per pad)
<1> Asriel: Pi calculi powa
<7> Asriel is also doing Phd?
<1> he will be, he will be
<5> Yup. Working with a slightly newer version, Distributed Pi Calculus
<5> Not yet. Shortly though :)
<6> why isn't that stuff interrupt driven?
<1> Code Mobility?
<5> aye
<2> sorry, i'm back
<0> Noidea: what, pads?
<6> yeah
<1> Asriel got an exam on that soon :(
<0> Because DirectInput manages them
<5> I must arrange to see the research group again, actually. they sent me off with a load of stuff to read, and I'm about half way through
Return to
#c++
or
Go to some related
logs:
the file is ''corrupt arhive'' problem
overclock cpufreq
#c
adel skype
#linux
12ax heaven
ubuntu santa cruz digital out
#MissKitten
openvt remote session
www.mike.n.com
|
|