[UNOFFICIAL] Ishiiruka-Dolphin Custom Version

[JMC47]

You're getting overly defensive about this . How come everyone else can post problems that DX9 Final has but as soon as I comment I'm the bad guy? Honestly, I find that childish all things considered.

I actually did performance testing of DirectX9 Final vs OpenGL Master, and, honestly, there are ups and downs to it, but there was no clear winner. I actually expected D3D9 Final to be much faster due to the lower overhead. If it had won the performance test I did (and, I tested about 8 or 9 games) then maybe I'd be more forgiving for its flaws.

The lack of good performance testing on this site leads to a lot of snake oil and stupid choices. D3D9 final DOES have merit, just not in the ways you think. If you wanna point out any other games to add to my testing report, I'd greatly appreciate it honestly.

There's nothing wrong with supporting DX9 Final for now. I didn't mean any offense. It's fully possible that Tino could fix VC NES games; you guys are always posting problems it's having.

FWIW: My WIP Performance Test with DX9 Final Added: https://dl.dropboxusercontent.com/u/484730/Performancechart4.0.png

Rules: All tests were done at 1x IR; Fast Texture Cache and EFB2Texture except on INI overrides. I ran my Core i5 3570K at 3.8 GHz, my GTX 760 locked at performance clocks, using Windows 7 x64.

[gamax92]

Just wondering, what, if anything, does this DX9 Final do regarding redundant state changes? Can anything that galop1n changed be also imported to this?

@JMC47
That post is kinda written in the style that your saying DX9 is worthless, ofc its going to come off more harsh or offensive to others.

[StripTheSoul]

(05-18-2014, 01:19 PM)JMC47 Wrote: You're getting overly defensive about this . How come everyone else can post problems that DX9 Final has but as soon as I comment I'm the bad guy? Honestly, I find that childish all things considered.

I actually did performance testing of DirectX9 Final vs OpenGL Master, and, honestly, there are ups and downs to it, but there was no clear winner. I actually expected D3D9 Final to be much faster due to the lower overhead. If it had won the performance test I did (and, I tested about 8 or 9 games) then maybe I'd be more forgiving for its flaws.

The lack of good performance testing on this site leads to a lot of snake oil and stupid choices. D3D9 final DOES have merit, just not in the ways you think. If you wanna point out any other games to add to my testing report, I'd greatly appreciate it honestly.

There's nothing wrong with supporting DX9 Final for now. I didn't mean any offense. It's fully possible that Tino could fix VC NES games; you guys are always posting problems it's having.

FWIW: My WIP Performance Test with DX9 Final Added: https://dl.dropboxusercontent.com/u/484730/Performancechart4.0.png

Rules: All tests were done at 1x IR; Fast Texture Cache and EFB2Texture except on INI overrides. I ran my Core i5 3570K at 3.8 GHz, my GTX 760 locked at performance clocks, using Windows 7 x64.

I'm getting childishly defensive because people are being childishly offensive

Of course you can report issues. But your post wasn't just that. The way you said it feels stronger when considering your other comments on DX9. It wasn't a report like "here's something that can be improved or needs fixing" but "f**k this s**t, don't use it!"
But to be fair, my answer was a bit of an outburst that wasn't just directed to you and I apologize. I was just annoyed by all the comments, also by devs, which sounded just very disrespectful towards Tino without whom I'm sure Dolphin wouldn't be as famous an emulator as it is today. And after pushing out his DX9 backend (though I understand why it needed to be done!) continuing to kick it while it's down (let's add some drama ^^).

And you know, I DO get it. The devs put a shitload of work into Dolphin and they're disappointed when everybody would rather use an old build or something non-master. (and of course when there's shitty 12-year old users who just go "new audio is shit!")

But first of all, most people ARE using master (I guess) and they have strong PCs so they are having a swell time with it and you won't even find those people writing here on the forums... Secondly, though, there are some people (like myself) who have shitty machines but are actually using Dolphin to play (and in the end, what's a Dolphin without the urge to play, ey?). Now I think it's kind of arrogant to say "your PC sucks, Dolphin is too good for you, get a better job or something" (yeah, I'm going a little bit over the top here but the principle applies ). In the end it's simple: If you could and would force everyone to use recent master builds, Dolphin would plainly and simply lose (hundreds of?) thousands of users (actually I had pretty much not been using Dolphin over the last year and it was Tino's build that got me back on the train because I was thrilled I could actually play Xenoblade Chronicles now). If I can play a game with 95% speed which totally garbles up my sound all the time, I won't play it. If I can play that game fullspeed I'll be having some good times with Dolphin; even if that build I'm playing with has whatever bad accuracy as long as it's not messing with my experience with that game.
It's not about which build or backend or dev is better. It's simple pragmatics. In a few years, when hopefully I'll have the money to get a new machine, I'll probably use master builds. Until then, I'm thankful Tino is giving us "Dolphin for the poor", if you will

That was the principle side of things. Now, if it's actually faster is another point to argue/find out. And I'm just a humble layman user (yet with lots of years of experience with Dolphin (/emulators)). I'm not good with benchmarks and stuff. But I can see and feel, when I play the same scene with DX9 and OpenGL when FPS go down by 1-2 FPS or by 4-6 FPS (and that's a lot if you are barely making 30 FPS - it's acatually the difference between a game being enjoyable or not). However, I was only playing SMG and Xenoblade recently. I'll check out a few more games.

[StripTheSoul]

I've just tried SMG again, DX9 vs. OpenGL. Same settings and all. I'm just starting the game, standing in that spot where Mario starts on the observatory (I'm far into the game, so the place is quite crowded and pretty heavy performance-wise). With DX9 (build Ishiiruka.406a44b.x64, the most recent one has crashing issues) I get 45 to 48 FPS, with OpenGL (recent master build) I get 36 - 39 FPS. So I'm not sure what your benchmark is actually testing (or maybe I'm having something setup very wrong here - wouldn't know what, though).

EDIT: Kirby's Epic Yarn (NTSC): Fangora (first boss, pretty heavy on all backends): DX9: around 37 - 40 FPS, OpenGL: around 25 - 30 FPS.

Xenoblade Chronicles (PAL), in battles: DX9: except for shader cache hiccups (which are recently just as bad with OpenGL, though) fullspeed (30 FPS); with OpenGL speed is mostly between 26 and 29 FPS.

Mario Kart Wii (NTSC): In race mode menu (Grand Prix, Time Trial, etc.), DX9: 54 FPS, OGL: 48 FPS. In Race, track "Mario Circuit", at starting position (in last place, before everyone drives off): DX9: fullspeed; OGL: 56 FPS

A Boy And His Blob (NTSC): Unfortunately this game has been rather slow on Dolphin (it's not even a 3D game) for, like two years or so... Anyway, here both backends go pretty much head to head: DX9: 53-54; OGL: 53-55

Metroid Prime (NTSC), with EFB2Texture (also in ini): At the start, head just slightly tilted to the left. DX9: 38-39, OGL: 48-50 (now there's a big win for OGL for a change)

XG3 Extreme G Racing (NTSC): First track to play. Both backends with fullspeed (DX9 feels a little smoother but that might just be a feeling); OGL has some graphical issues which DX9 doesn't (flickering textures on those skyscrapers and the shield-reloading areas don't show right).

This might not be high science, but yeah, it does tell me as a user (with an old PC) that in most games DX9 is better for me.

[narcku]

Let me just say I agree with StripTheSoul.
I haven't posted much but I've been lurking the forum(well, not just the forum) for quite some time.
And I've seen the posts made by the devs when replying to these things(DX9, new HLE audio).
It's not a matter of whether you're correct/not(though I'm sure you are), it's you guys' attitude that's seemingly... offensive(dunno the right word for this, but the response certainly didn't seem friendly).

[dephined]

In my opinion Dx9 should be at first compared with Dx11...

[StripTheSoul]

(05-19-2014, 12:48 AM)dephined Wrote: In my opinion Dx9 should be at first compared with Dx11...

Does make sense. However, DX11 (or in recent revisions D3D) is not the one that is always said to be faster (or at least as fast as) DX9. That's why we're comparing those two here.

[JMC47]

Stripthesoul: I was one of the people who didn't mind D3D9 sticking around as long as it wasn't in the way. Tino disappeared after removing the deprecated tag, leaving none (or I think he fixed one or two; but introduced regressions) of the bugs fixed so it got deprecated again. I was a user looking from the outside in, so maybe there's more to the story, but the work I've seen, and improvements, to the D3D9 backend weren't coming when it was still in Dolphin master.

I was a bit snarky with the NES VC game comment? I don't know; I was just doing my performance testing for Dolphin and found it when playing with DX9 Final. I'm always looking for problems. Have I ever been rude about bugs in mainstream dolphin? Yeah! I definitely have. I thought it was funny, that something I cared about (NES VC games were something I really wanted working in Dolphin) and required a ton of work to get working, actually were improved in D3D9 Final vs Pre-Tev fixes new Master, but that just made it more agonizing. Compared to latest master, it's laughable, but it's a tev_fixes_new fix; the fact Tino got things showing at all surprised me.

And, you know about this glitch in XG3 Extreme racing, but never report a bug, never provide a fifolog? What, because it works in D3D9 it doesn't matter to you? That offends me more than anything, I see you constantly testing this branch, reporting bugs, testing again and again, but something outside of it? IGNORED. You're slowing down the forward progress of Dolphin; I've seen it before. It starts with a small number of users using a fork like this; which is fine. A developer (which I have all the respect in the world for, btw) does rogue efforts to improve something separate from the main project. He gains a small user base of people who get improvements, real or not (I'm going to assume real in this case, because I know D3D9 has advantages.) Those users then segment themselves from the main project, which has already happened for the most part. Then what do you do? Whenever anyone, regardless of their computer, has performance issues, you point them to this fork. Now suddenly Dolphin Master is losing testing, losing users, losing issue reports to a fork.

The psychological implications are even worse; because now if those people go back to Master, they're going to simply assume it's slower even if they have no concrete proof. They'll say it feels slower despite the same framerate, etc. And if they do use OpenGL/D3D11 in this branch? Worthless issue reports because they're using outdated versions of those backends. It's just, very little good can happen out of people using this fork religiously.

Dephined: How does comparing D3D9 to D3D11 make more sense than comparing it to OpenGL? That makes no sense whatsoever. What, because they're more similar in name and architecture? It's not apples to oranges, though, they're both playing the same games and are just as valid to test head to head. D3D is also faster in some games than OpenGL, I just didn't compare because I think people should first and foremost be using the most accurate backend in Dolphin. When D3D9 was in Dolphin, I also reported bugs featuring that as well, even though it wasn't my primary backend.

Narcku: Please don't compare DX9-Final to new-AX-HLE; you really don't know the mess you're getting into. I fully admit DX9-Final has merit for some users, and some games (and hell, I'm working on something that discusses why new-AX-HLE met controversy too .) Old HLE does not doing anything right, and you really shouldn't compare it unless you're very, very familiar with what happened. I understand the parallel you're trying to make, but it's not correct.

Tino: Sorry about the mess I caused in your thread. I have nothing against your or the work you're putting into the D3D9 backend.

[narcku]

See.. you're not even getting it right.
When did I ever even discuss/compare about these two?
I was merely saying that some of the responses the devs gave when it came to these topics came off as unfriendly.
And I already stressed out it's not about correctness, "Old HLE does nothing right", yeah sure.. thing is I wasn't talking about that.
I was talking about the way some devs responded to these, not all of course, just some that I've seen.
Maybe I just caught you guys posting at a bad day, who knows

[delroth]

Quote:It's not a matter of whether you're correct/not(though I'm sure you are), it's you guys' attitude that's seemingly... offensive(dunno the right word for this, but the response certainly didn't seem friendly).

Quote:I was merely saying that some of the responses the devs gave when it came to these topics came off as unfriendly.

You know what's offensive and unfriendly? Someone who has no idea about the technical challenges involved in emulation posting crap on the forums saying that developers are not doing their jobs properly and should be doing it in another way. You'll notice that this happened for both D3D9 removal and the synchronous audio changes.

Maybe, just maybe, if people who have never written a single line of code in their life stopped assuming they know better than the people who spent months (years?) improving Dolphin, things would go differently.