Re:JS1k JavaScript Code Competition – Demos 2015
by madflame991 on Sunday January 24, 2016 @01:31PM
(#10963407)
Managed to close it twice, but it took a while. I don't understand how a single application can render a whole (multicore) system unresponsive. Isn't the OS supposed to take control from a long-running task once in a while? I've encountered dozens of other graphical demos in JS that cause the same thing. I'm running Chrome 48 on Linux.
Re:JS1k JavaScript Code Competition – Demos 2015
by rogerbinns on Sunday January 24, 2016 @02:59PM
(#10963775)
I did a bit more digging. Using Ubuntu 15.10 64 bit on a system with Intel HD4000 graphics. I was getting about one frame every 3 seconds and the system felt very unresponsive (eg takes a minute or two to kill the tab). But the mouse pointer moved.
Connecting in via ssh showed that the system was just fine and close to idle. It was the GPU where the issues are. I ran intel_gpu_top and what you see is that render busy is at 100% as is blitter busy (most of the time). Essentially the gpu is saturated. I do wonder if this is similar to bufferbloat, but with so much work pending for the gpu that other requests take a long time to run.
I ended up having to kill -9 chrome, and ended up with a hung system (ie even magic sysrq didn't work).
Re:JS1k JavaScript Code Competition – Demos 2015
by rogerbinns on Sunday January 24, 2016 @10:07PM
(#10965338)
Since another commentor mentioned it running fine on their mobile device, I decided to run it on my tablet which coincidentally also uses an Intel Atom cpu + gpu.
It chugged along at about one frame every two seconds for about 10 seconds, before an app not responding dialog appears. I reported it, and the tab was restarted. On the restart I get a bottom of Chrome message "Rats! WebGL hit a snag". It has a helpfully (not) grayed out "Learn More", Reload or Ignore. No matter what I did I couldn't get it working again, with the browser showing "WebGL not supported".
On my Nexus 5 it ran just fine - smoothly even although there is a lot of aliasing.
Re:JS1k JavaScript Code Competition – Demos 2015
by q3k on Sunday January 24, 2016 @06:51PM
(#10964648)
That is consistent with my theory as well. It would be awesome if someone more experienced in the Intel graphics stack would dig a bit more into what was happening...
Re:JS1k JavaScript Code Competition – Demos 2015
by rashkov on Sunday January 24, 2016 @01:33PM
(#10963418)
Works well for me on Linux Chrome 46.0.2490.86 (Official Build) (64-bit). Strangely though I've been having trouble with amazon.com locking up the tab. I've got a dedicated graphics card in this thing though, with a heap of ram, so that may be the difference.
Re:JS1k JavaScript Code Competition – Demos 2015
by kozak on Monday January 25, 2016 @07:10AM
(#10966718)
1k seems like too little. If we take classic 64k .exe demos as a "gold standard", then we need to define some limit that would allow to replicate that amount of stuff in JS. Should it be 16k, 32k, 64k? Is textual JS code more compact or is it less compact than legacy binary .exe files? That's not a straightforward question to answer.
Connecting in via ssh showed that the system was just fine and close to idle. It was the GPU where the issues are. I ran intel_gpu_top and what you see is that render busy is at 100% as is blitter busy (most of the time). Essentially the gpu is saturated. I do wonder if this is similar to bufferbloat, but with so much work pending for the gpu that other requests take a long time to run.
I ended up having to kill -9 chrome, and ended up with a hung system (ie even magic sysrq didn't work).
It chugged along at about one frame every two seconds for about 10 seconds, before an app not responding dialog appears. I reported it, and the tab was restarted. On the restart I get a bottom of Chrome message "Rats! WebGL hit a snag". It has a helpfully (not) grayed out "Learn More", Reload or Ignore. No matter what I did I couldn't get it working again, with the browser showing "WebGL not supported".
On my Nexus 5 it ran just fine - smoothly even although there is a lot of aliasing.
Maybe there should be a code competition where you get points for simplicity as wells as how impressive the result is!?
In sports there are no secrets, just different philosophies.