Vulkan Support - What features would help?
#11
(12-29-2016, 12:28 AM)lightningterror Wrote: Hmm what about async compute? That could increase performance on amd cards even further.

On a side note the pcsx2 code is on github so you can download it and start playing with it from there Laugh

That's assuming pcsx2 uses compute shaders in the first place.

And pascal cards can benefit too, but I suspect nvidia's driver is already overlapping work like that in dx11 and opengl when possible.
#StopRNG
Reply

Sponsored links

#12
PCSX2 doesn't use compute, I think Gabest dabbled with it a while ago (there was an OpenCL renderer for a bit) but it was pretty slow. I think the problem is we don't have enough repeated mathematical operations which is the strong point for compute.
[Image: ref-sig-anim.gif]

Reply
#13
I'm afraid but you understand asynchronous compute in the wrong way. Due to GPU arch you have some penalties/overheads when you switch between normal and compute workload. Asynchronous compute is a way to reduce the switch overhead. For us it is useless. However programmable blending will be a huge improvement but I think it wasn't added to Vulkan v1 (otherwise most GPU won't be Vulkan compatible)

A basic Vulkan port won't be too complicated. However I'm afraid it will make the code more complex to maintain. Anyway a first target could be to use cross portable shader code (between Vulkan and openGL)

Ping me if I forget but I will profile you Zone Of Ender when I came back from my vacation.
Reply
#14
I will look into that. I have a four day weekend starting tomorrow and I plan to dive into everything head first and start learning. I will check to see if the programmable blending was added or not as it may have been in any updates they have made to Vulkan. I know you have been testing x64 viability gregory, would there be any benefit in PCSX2 being 64-bit with a specific 64-bit coding of Vulkan?
Reply
#15
Don't get ahead of yourself Tongue 64bit is a LONG way off
[Image: ref-sig-anim.gif]

Reply
#16
Hmm. Vulkan would be really interesting. I wish you best of luck Shinmegumi Smile . I only have a handful of games I can test in the future but it would be exciting nonetheless.
Reply
#17
I make no promises, but I look forward to learning this. I already have a copy of the current GitHub code and plan to look at all of it over the weekend and start decyphering what I can and learning more. Over the next four days I expect I will spend 40-50 hours learning and trying new things. I'm new to C++ and started learning and coding a month or so ago. I'm hoping that diving into something like this I could learn more about APIs like I have wanted to and bring something useful to a community that has the same love of the PS2 as I do.

--Edit--

@refraction yeah, I'm aware. I was just curious if there was any benefits in case I decided to code a 64-bit version of it for the future.
Reply
#18
If you're that new to coding, this is a huge leap in difficulty, especially considering the current state of the project.

If you want to learn how to do vulkan (which is possibly a little advanced for you without struggle at the moment) I suggest making your own application with it first.
[Image: ref-sig-anim.gif]

Reply
#19
Yeah, I noticed that when looking at the code and looking at the API tutorials I found and everything. I mean, I understand some of it based on my current knowledge. I know it will be a huge leap in difficulty, but honestly, I learn best when I'm challenged. Because I'm driven to find the answer and fix the problem. I'm sure it will take me a while to get proficient or even come close to your guys level. I do appreciate all the input and encouragement.
Reply
#20
Maybe this can help you..

https://www.phoronix.com/scan.php?page=n...nderer-Out
Reply




Users browsing this thread: 1 Guest(s)