[UNOFFICIAL] Ishiiruka-Dolphin Custom Version

[Qaazavaca Qaanic]

...wait didn't the developers say that tev-fixes-new was impossible on DX9? Then how have you implemented it?

[Tino]

xystus: supports everything 4.0.2 supports.
jimbo1qaz: it was less accurate than integer emulation and the other developers didn't want to support 2 shader generator paths, so they drop dx9 to make their work easy on the other back ends. About the impossible tag, is not impossible is just more work.

[Anti-Ultimate]

(05-29-2014, 11:41 PM)Tino Wrote: xystus: supports everything 4.0.2 supports.
jimbo1qaz: it was less accurate than integer emulation and the other developers didn't want to support 2 shader generator paths, so they drop dx9 to make their work easy on the other back ends. About the impossible tag, is not impossible is just more work.

I think that was kind of obvious.

Spoiler: (Show Spoiler)

Honestly, I'm pretty sure it was just to reduce the amount of DX9-removement butthurt. This would give them a reason to complain

[JMC47]

No? It's impossible for D3D9 to do integer math natively, it's as simple as that. Tino can implement whatever workarounds he wants for that fact. When it comes down to it, D3D9 will never be quite as accurate as actually doing the proper operation. neobrain would know better than anyone (except maybe Tino?) the troubles of getting the integer operations to work without integer emulation. tev_fixes_new was like the fourth interation I had seen of it (tev_fixes, tev_fixes2, I think there was a 3 at some point) after the first two had failed.

Maintaining multiple codepaths would be a nightmare as well; as D3D9 would have to get separate fixes, separate updates, and would likely produce different results than the other video backends in many situations. Imagine the issue reports, the extra work in adding features to videocommon, etc. To make matters worse, no one was working on the D3D9 backend for ages! It was only after the initial deprecation that any of the issues were starting to be addressed, and even then it was slow. It is nothing like how Tino is handling D3D9 final on his own.

I'll defend the impossible tag as it was merited: D3D9 couldn't support integers natively. Could it have supported them better through floating point math? Yes, obviously.

[purpasmart96]

I think the problem is that the devs are so focused emulation accuracy, they just drop anything that no longer fits their agenda, regardless if there is a 50 % speed drop, just to fix a few shaders.

[JMC47]

I think the problem is users who have no idea what they're talking about.

[purpasmart96]

(05-30-2014, 01:50 AM)JMC47 Wrote: I think the problem is users who have no idea what they're talking about.

Please, I've been in the emulator world for quite some time, just because they couple dozen posts doesn't mean they are always some dumb illiterately challenged person , this isn't a snes emulator where you can use some crazy hardware accuracies with no reproductions. Although I'm pretty sure you already knew that.

[StripTheSoul]

I think everything has pretty much been said about all this and everyone has had their legit reasons. Now things are the way they are, DX9 is branched out and can do its own little thing while master can continue on nice and lean. No point in arguing now what should have been.

[JMC47]

s/reproductions/repercussions

And why not strive for more accuracy? You may not have noticed, but for people without broken drivers, Dolphin has gotten much faster in the last 500 or so revisions.

[purpasmart96]

(05-30-2014, 02:02 AM)JMC47 Wrote: s/reproductions/repercussions

And why not strive for more accuracy? You may not have noticed, but for people without broken drivers, Dolphin has gotten much faster in the last 500 or so revisions.

I'm not saying you shouldn't strive for accuracy, I'm saying you shouldn't do it in away that makes high end computers run it like crap.