the heaviest code is the wii main thread...
which needs to be synced with the
wiimote
audio thread
gpu thread
...
there is no way to split the main thread into more threads
without losing sync... which is causing a crash
or how tryed...
to achieve a sync main thread with more helper threads
there is so much sync process power needed that nothing is achived
the heaviest parts like gpu and sound rendering are allready outsourced..
the other parts are just to small to create a thread for them
(08-12-2011, 03:00 PM)pajousek Wrote: [ -> ]Well, can't even PhysX/CUDA for counting lots of things [mainly precision floating points numbers] be implemented? Surely it would help increase performance a lot.
If it was that easy, it would've already been done by now.
Frankly said, noobs should refrain from doing those kind of suggestions if they have no idea since they end up talking nonsense in 99% of all cases.
(08-12-2011, 03:00 PM)pajousek Wrote: [ -> ]btw. does that OpenCL checkbox actually help or not? I'm starting to find these help tips on checkboxes kinda unreliable, so that's why I'm asking.
Why would the help tips be unreliable, I spend like hours on bringing them up to date recently?
It most likely won't give any speedup anymore.
Well, for example, emulator states that DX11 is slower than DX9; and when I was reading forums, I've seen few posts where people are telling that DX11 is faster for them. And from looking at last 100 commits, I haven't seen anything with tips updates, so I thought that they can be unreliable, as in outdated? But truth is, 100 commits can be a little, considering yesterday there was about 20 commits, all just about OpenGL.
What I didn't know is, that heaviest parts are allready outsourced and other things can't be outsourced without spending lots of CPU performance on keeping them in sync.
And about the OpenCL speedup; can I ask why it won't give speedups? Is it not implemented as good as it could be or the optimizations have to be done somewhere else? What's the biggest obstacle in making Dolphin faster anyway? Or it almost hit the limits and now we just need to get ultra fast CPUs?
Sorry for sounding like a n00b, I didn't look at the source code (although I'm probably going to) and don't have any experiences with emulation at all; only basic knowledge (and common sense).
....Comments*
dx11 for me personally is almost always a little better except in a select few games where 9 is much better, I've never used opengl so far.
I don't know about optimization, I'm sure it's "possible" to get things going faster, but we're not talking about an army of employees under EA working on the emulator for salary here, so it might be difficult and time consuming. But yes definitely at the processor comment, atleast. The latest intel stuff pretty much gets any and every game that actually works at 100% speed, and the next set of amd processors seemingly might do the same. This isn't the greatest solution, go out and spend a bunch of money, but that means in like 2 years nearly any processor you could even buy for a desktop would run this emulator really well, though a lot of us do fine already. With the resolution bump and other nice little graphical enhancements it's not too bad playing a 10+ year old game for me either, unlike trying it on my wii regularly..so atleast coming back to games later is much better than it could be.
(08-12-2011, 10:18 PM)pajousek Wrote: [ -> ]...
You shouldn't be using anything post-3.0 anyway, that's not supported. Just use 3.0, you'll get up-to-date tooltips/descriptions and a d3d9 backend which is faster than d3d11 for free
