was just wondering if there's any chance that LFS supports multi CPU management in the future ? should the whole game be totally reprogramed ? or only some modification would be needed to make it compatible with multi CPU's ?
the point is , LFS will be more realistic with some more calculations i guess , maybe it would push even further the physics in term of realism..am i totally wrong here ?
The game would almost need to be completely re-written, but i'd really like this (Quad core power!), the problem is that while Scaview works away trying to busy my 4x 2.8 GHz Cores, the player with the old 1500 Single Core Sempron will get screwed.
LFS is a niche, it can't afford to alienate players with not-so-good PC's
TBH i'm not sure chucking CPU cycles at the physics engine will help, it's aparently all the "little things" it needs
Maybe it's possible to make it simple.
For exemple using one core for your car's physics and data sent to your GPU and using other cores to calculate other player's or AI's Physics.
It doesn't seem so complicated but maybe i'm utterly wrong here!
I do see advantages of LFS using a single ocore though. If I'm trying to record FRAPS, I can set LFS to use core 1 and FRAPS to use core 2. This means I get absolutely no slowdown when recording at normal speed. If I record at 0.125 speed and speed it up 8X, I can get nearly 500 solid FPS for a movie. Not bad.
However, I would really prefer having dual core support for LFS, maybe bring a new patch out that lets you choose whether to use a single core for the game or dual core, pleasing both worlds.
or maybee "virtualize" your bs and run lfs on that virtual machine.
Of corse, the virtualizer has to be SMT approved (simultaneus multithreading).
If theat can be done, we could let lfs be as it is, and scale it up to whatever machine availabe.
It *is* complicated because you have to decouple the graphics and physics engine. That is far from being a simple task, and while multicore support surely is the way of the future, I don't think LFS will implement it anytime soon, nor am I sure that it will benefit much from it. The physics calculations right now actually take up relatively little CPU time - the graphics stuff is what causes most stress on the CPU. You'd actually have to split up the GFX process to multiple cores to make 100% use of the available power.
PS: Jakg, I know you're proud of it, but could you please try to keep the "look, I have a quad core setup, woop woop" rate below 50% of your posts. Thanks.
yeah, as the quoted reply from scawen already says it prolly wont come into S2 anyways.
are there such weak dual cores that u need 2 cores to run the game properly?
no...WONDER my cousins centrino laptop is better than my Turion dual core....bastard.
+1....it really needs to be compatible, im only getting about 60 fps on single player, and about, 10? on full grid, maybe that could be a bit above 30, all the time, so i can race on the CTRA servers without a pile up
thanks
you should be able to download 2 different things, LFS single Core, LFS multi Core
Harjun, the bottleneck of most laptops isn't the CPU, its the GRAPHICS CARD. You have done nothing but screw up and damage your laptop trying to overclock it, and we both run 1.6 Ghz processors. Like I said, mine runs at 1.2 Ghz on LFS at any given time, but the fps swings up and down.
Not really. Scawen's a smart cookie; he's probably got a really nice codebase with good code seperation. Unlike some of the "dire pronouncements" made, its not a case of a complete code rewrite, but multithreaded applications (doesn't matter if its even one processor) does have additional considerations that take more work to support and then to rigorously test.
Most games tend to work on a single game loop that listens and processes OS messages, rendering the current graphics frame, does physics and AI calcuations, polls network, and polls controllers (not necessarily in that order). In fact the min. sleep in the LFS options is probably inside the game loop; essentially it means that the current thread (i.e. the LFS app) does not use up any processor time slices for X amout of time which allows other applications, the OS, etc. to process messages, etc. Anyways, a decent, high-level overview (with code) of a game loop can be found http://www.mvps.org/directx/ar ... writing_the_game_loop.htm. And in LFS case, you are also dealing with a constant rate physics heartbeat which adds a bit of a wrinkle to the game loop.
Back to the topic of multi-threaded support in games, here's a quote snippet from Tim Sweeny (of Unreal fame; he and John Carmack are very good at what they do... plenty of others could be mentioned in the same breath too) has to say:
If Epic Games finds this difficult, and they have a very good and large staff, hopefully it impresses on folks the non-trivial nature of the task with Scawen being pretty much the sole programmer on the LFS codebase.
Here's another article http://www.gamasutra.com/featu ... 060906/monkkonen_01.shtml that's a good discussion on the pro's and con's of several different types of multi-threaded approaches with regards to games (and actually any type of interactive application).
Well Epic (and 2k with Bioshock) did really good with the multi-threading, after playing Bioshock on a Dual-Core machine, it ran lovely and looked just as good (DX9). Other multi-threaded games I've played didn't do it all so well.
