RIFS-ES Linux Kernel Scheduler Released

[gamerk2]

Look at it this way: With lower latency, you guarantee that every thread will run as much as possible when not blocked. However, it will take a longer period of time for tasks to finish, because they spend less time running overall.

When you maximize throughput, you get things done a lot faster, but other background threads will spend a LOT more time waiting to run, increasing overall latency.

Windows is designed to try and maximize throughput for the highest priority tasks, will still allowing background threads to run within a "reasonable" period of time (via priority boosts as threads spend time waiting to run). RIFS tries to minimize latency, which is great for multimedia, but not as great when doing a single intensive task while a large collection of other tasks are also running.

Throughput vs latency is a major problem in computing, not just for scheduling. RAM/Cache access times vs total throughput, for instance.

[Paradox Uncreated]

I don`t know why you are talking about throughput vs latency. I have low latency desktop config, but the rest is more like a server config. 100hz, granularity in favor of "server" style configs.

This runs 0.8ms latency without clicks on a core 2 duo 2.5ghz. On what today is a modest computer. Increasing hz or granularity resolution, does not improve games. Infact this makes doom 3 run perfectly on linux. (no lost frames).

This is a "low jitter" config.

I just uploaded an Ubuntu kernel, for 12.04. Try it:

http://paradoxuncreated.com/tmp/linu...stom_amd64.deb
http://paradoxuncreated.com/tmp/linu...stom_amd64.deb

Michael if you are reading this, benchmark them. The best config for games. No reason to benchmark kernels not made for games, where frames are not reaching the screen.

Peace Be With You!

[crazycheese]

I just uploaded an Ubuntu kernel, for 12.04. Try it:

http://paradoxuncreated.com/tmp/linu...stom_amd64.deb
http://paradoxuncreated.com/tmp/linu...stom_amd64.deb

Paradox, would you please describe the changes you did to this kernel? You just merged RIFS?

[gamerk2]

No reason to benchmark kernels not made for games, where frames are not reaching the screen.

You make the somewhat silly assumption that 60FPS means that 60 unique frames will actually be drawn on screen. Sure, the GPU can spit out that much, but unless you have one frame being created every 16.2ms, you will see frames getting dropped.

Average frame latency is a FAR better benchmark then FPS is. Its theoretically possible, albeit realistically impossible, to have 60 FPS, yet have only a single frame drawn to the screen.

[crazycheese]

Its theoretically possible, albeit realistically impossible, to have 60 FPS, yet have only a single frame drawn to the screen.

If you have compiz installed and turned on on any good hardware, the card will draw 200fps+, yet only 30fps will be finally mapped on the screen resulting in huge lag.
You can track that easily: install or get distro with XFCE. Get compiz and compiz icon there. Get drivers and urban terror. In the later set fps limit to 125.
Now if you use xfwm4 and compiz, you will instantly see how fluid game is when using xfwm4 and how sluggish with compiz, yet both show 125fps+

[Paradox Uncreated]

Paradox, would you please describe the changes you did to this kernel? You just merged RIFS?

This is not RIFS. It is a low-jitter kernel, with standard code, although some hardwirings to CFS variables. http://paradoxuncreated.com/Blog/wordpress/?p=2268

Also lowest hz = lowest jitter, but too low hz, is not good. Currently using 100. If lower is possible, I will use that also.

Peace Be With You.

[Paradox Uncreated]

After some testing, I have now for 90hz timer to be optimal. Doom 3 is free of visible jitter. I will be updating the kernel, shortly.

Peace Be With You.

[gamerk2]

If you have compiz installed and turned on on any good hardware, the card will draw 200fps+, yet only 30fps will be finally mapped on the screen resulting in huge lag.
You can track that easily: install or get distro with XFCE. Get compiz and compiz icon there. Get drivers and urban terror. In the later set fps limit to 125.
Now if you use xfwm4 and compiz, you will instantly see how fluid game is when using xfwm4 and how sluggish with compiz, yet both show 125fps+

Average Frame Latency is a FAR better benchmark then FPS, because you can see, on average, how long its taking to create a frame. Even thats not perfect, but it gives you a better idea how many frames are actually being created, and NOT being sent to the screen.

After some testing, I have now for 90hz timer to be optimal. Doom 3 is free of visible jitter. I will be updating the kernel, shortly.

Peace Be With You.

Based on a sample size of one, correct? I'm sure someone else, with some different H/W setup, will appreciate all the testing you did on this.

[Paradox Uncreated]

Bitchy as usual gaymer2k? I think you just should be really happy, because nobody else has succeded with this. As for my setup it is modest, a dualcore 2.5ghz + nvidia gtx 280. That usually means that jitter will be less of a problem, for newer PCs. There could be deviations here and there, but that is probably in microsecond range. So it is probably a good average setting.

Anyone can ofcourse test it. Daemons in ubuntu seem to do little for jitter, and seems to be below 0.1ms difference (from minimal processes - in other words, you don`t have to run minimal as on windows, for optimal performance)

http://paradoxuncreated.com/Blog/wordpress/?p=2268

If there should be large deviations, I would consider that an oddity.

Peace Be With You.

[3766691]

If you have compiz installed and turned on on any good hardware, the card will draw 200fps+, yet only 30fps will be finally mapped on the screen resulting in huge lag.
You can track that easily: install or get distro with XFCE. Get compiz and compiz icon there. Get drivers and urban terror. In the later set fps limit to 125.
Now if you use xfwm4 and compiz, you will instantly see how fluid game is when using xfwm4 and how sluggish with compiz, yet both show 125fps+

An LCD screen flashes 60 times in a second(60 HZ). So the FPS can just reach 60 maximumly, but the GPU can draw 200 frames in a second.

Spitting shit that windows sucks. Yeah, that language probably goes along with thinking windows is "decent".

Why you choose to argue with me, is another issue you have too apparently. I have far superior knowledge, and consider you extremely ignorant. And ofcourse only the ignorant would defend windows, and think they could somehow refute what I say.

Now f*ck off.

I agree with you that Windows sucks. And I also agree that Windows sucks at everything. On file system, memory management, cpu scheduling etc, Windows sucks(especially file system).