"everyone should be happy" - really no. The new AX-HLE demostrate how shitty our zelda-HLE is. Before we are used to say it isn't possible with HLE, but you've prooved it. Now we really really need the zelda-HLE rewrite ;-)
Why Dolphin is getting rid of asynchronous audio processing
|
07-11-2013, 11:23 PM
(07-11-2013, 09:41 PM)degasus Wrote: Now we really really need the zelda-HLE rewrite ;-) Nothing would be more awesomer...lol
i5-8600K @ 4.5 GHz
GeForce GTX 1080 SC2 ASRock Z370 Taichi LGA 1151 CORSAIR Vengeance LPX 16GB 2400 DDR4 SAMSUNG 970 EVO M.2 500GB PCIe SSD Indigo Xtreme TIM Phanteks Pro ES614P Black Steel Case EVGA SuperNOVA 750W PSU Windows 10 64-bit 07-12-2013, 12:22 AM
But aren't there a lot more interrupts etc which need emulating meaning that it would end up almost as slow as LLE once everything was properly synchronised? I'm sure someone said that, and therefore it wouldn't be worth the effort.
OS: Windows 10 64 bit Professional
CPU: AMD Ryzen 5900X RAM: 48GB GPU: Radeon 7800 XT 07-12-2013, 12:29 AM
Yes, we are used to say it will be, but just try the new AX HLE. It is slower than the old HLE, but much much faster than LLE.
btw: DSP doesn't generate so many interrupts, AudioInterface (which is needed both for HLE+LLE) generates many more... 07-12-2013, 01:16 AM
@degasus: on Wii AX HLE sends about 2 interrupts every 3ms as far as I know. That's not a lot less than the AI. Zelda HLE uses a lot more interrupts than AX HLE for synchronization and communication, which is why I expect the worst.
But really, performance is not the issue with rewriting Zelda HLE to be synchronous. It's more an issue with rewriting the code as a kind-of-state-machine to handle the fact that we have to leave the function and reenter it later when a new interrupt comes. 07-13-2013, 07:44 AM
Just to be sure: there isn't any way of implementing timestretching without delay, right? In this case, how could we use timestretching with smaller delays than the current ones used by OpenAL (or by the timestretching branch)? Even setting the latency as low as my sound card support, the delay in sound effects are pretty high...
Avell A70 MOB: Core i7-11800H, GeForce RTX 3060, 32 GB DDR4-3200, Windows 11 (Insider Preview)
ASRock Z97M OC Formula: Pentium G3258, GeForce GT 440, 16 GB DDR3-1600, Windows 10 (22H2) 07-15-2013, 09:39 PM
After a little bit of testing I am not anymore saying Developers of Dolphin emulator gone crazy. But still waiting more multithreading...
07-15-2013, 10:47 PM
You feel free to wait for that. We'll phone you back when computers start working off magic, not physics.
A less patronising explanation: Most of dolphin's work is done emulating 2 chips, the CPU and GPU. This right off the bat gives us two main threads we can use. However, the type of instructions these chips carry out usually depends on the results from the previous instructions. If different bits of code were emulated on different threads, they'd spend there whole time waiting for each other to be ready, which would completely mitigate any speedup. There'd then be a tonne of work to make sure all the threads stayed in sync, and all this extra code would make things even slower. tl;dr it would be a tonne of work, and it would only slow down the emulator and break things.
OS: Windows 10 64 bit Professional
CPU: AMD Ryzen 5900X RAM: 48GB GPU: Radeon 7800 XT 07-16-2013, 02:47 AM
Also read https://dolphin-emu.org/docs/faq/#couldn...-go-faster about multithreading.
07-16-2013, 05:16 AM
Oh sorry. I long time ago read those but didn't remember.
For me emulator is running just fine... just only asked. That mad comment about devs was just dumping asynchronous processing. But later i think whole emulator plays now example zelda wind waker much better (stable) than ever so i should never challenge devs when i do know nothing about them. Thanks for very nice emulator. Keep doing mad things. (Example when Wii support came and this). |
« Next Oldest | Next Newest »
|
Users browsing this thread: 1 Guest(s)