Fast Mipmaps was a hack, and it was eradicated so that the errors it was hiding could be found and fixed, instead of ignored. Getting rid of these bugs will generally make things better, so its healthier to get rid of them than to pretend that they aren't there.
(06-19-2012, 10:41 PM)etking Wrote: [ -> ]I do not want to use ICC builds but I have to, because the standard builds are just too slow. And don't forget that master is an unstable development branch too, as the removal of fast mipmaps shows.
How does the removal of an option make stuff "unstable"?
I think he means that the things it fixed don't work any more.
(06-19-2012, 10:08 PM)DefenderX Wrote: [ -> ] (06-19-2012, 06:03 AM)gjfklhg Wrote: [ -> ]eh? It works just fine for me.
If you apply https://dl.dropbox.com/u/11111638/0001-fix-xenoblade-hle-patch.patch and then the original patch with the changes to mixer.cpp and mixer.h (first two files in the patch) removed, does that work?
Nope, won't work. I tried it via command line, but i got the same result.
Did either patch work?
(06-19-2012, 10:41 PM)etking Wrote: [ -> ]You can compare whatever you want. ICC Build will always be much faster when you compare the same revisions ICC and non-ICC.
Bullshit. In fact, i just tested it yesterday, and the icc build got 74.3 fps, and the same build from the dolphin site got 74.6 fps. Same settings, same revision, and i even used a .dtm movie to ensure the emulation ran through exactly the same code both times. The only difference was one was an icc build, and one was not. The icc build have actually been slower every single time i've tested. That's with a 2500k @ 4.3 ghz. Others have also posted benchmarks showing similar results. Amazingly, i've never once seen a benchmark showing these 5 fps gains people keep talking about.
(06-20-2012, 02:45 AM)AnyOldName3 Wrote: [ -> ]I think he means that the things it fixed don't work any more.
I know what he meant, I was just trying to point out the ignorance of his statement.
I can get many additional FPS in Zelda SS and SMG for example, I am running dolphin at 3x internal resolution and 1080p DX11 16xAF, FPSLimit=Auto, HLE audio. Lock Threads to Cores and OpenMP texture decoder is enabled. My GPU is a comparatively slow, overclocked GT555M, my CPU a core I7 2670QM.
Method: I load a savegame and look at the min, max and average FPS displayed at a specific location while performing exactly the same actions with exactly the same settings in each build.
@neobrain: My understanding of an unstable branch is that everybody submits his changes or removes critical features without caring that other people are relying on them and cannot use the master branch anymore. This is the definition of ignorance by the way. A developer may not care about blue grass or missing textures but an advanced user does.
Quote:Method: I load a savegame and look at the min, max and average FPS displayed at a specific location while performing exactly the same actions with exactly the same settings in each build.
Aka, you eyeball the fps, and try to do roughly the same thing each time? That's not even a test.
(06-19-2012, 10:22 AM)AnyOldName3 Wrote: [ -> ]In theory it means that they are faster. In practice, it's probably barely noticeable, but the placebo effect gives a good impression.
I disagree.
Maybe it has no effect on most games because they aren't intensive but it does tip the scale if you are just barely out of reach of running it perfectly.
For example, The Last Story ran with a little stutter for me using the unoptimized builds.
As soon as I used Lectrode's builds, the stutter went away completely.
Of course now I use Defender's builds which work almost as good (sorry buddy lol. I still get a slight stutter with your builds on that game).
(06-20-2012, 02:31 PM)lionheart5656 Wrote: [ -> ] (06-19-2012, 10:22 AM)AnyOldName3 Wrote: [ -> ]In theory it means that they are faster. In practice, it's probably barely noticeable, but the placebo effect gives a good impression.
I disagree.
Disagree all you want, but it doesn't really mean anything. Every single benchmark that has been posted has shown them to be almost exactly the same. You would think if so many people are getting as large speedups as they claim, at least one of them would be able to post a benchmark proving it. Apparently not though.
You may want to read over
http://en.wikipedia.org/wiki/Confirmation_bias to see why any results you observe without measuring them are completely meaningless.
(06-20-2012, 03:35 PM)gjfklhg Wrote: [ -> ] (06-20-2012, 02:31 PM)lionheart5656 Wrote: [ -> ] (06-19-2012, 10:22 AM)AnyOldName3 Wrote: [ -> ]In theory it means that they are faster. In practice, it's probably barely noticeable, but the placebo effect gives a good impression.
I disagree.
Disagree all you want, but it doesn't really mean anything. Every single benchmark that has been posted has shown them to be almost exactly the same. You would think if so many people are getting as large speedups as they claim, at least one of them would be able to post a benchmark proving it. Apparently not though.
You may want to read over http://en.wikipedia.org/wiki/Confirmation_bias to see why any results you observe without measuring them are completely meaningless.
That might have made sense if I came into the emulator world believing a particular view point.
I came in fresh with no bias, tried the two, and saw a noticeable difference.
I don't know what the actual specs of it are and I frankly don't care. It made a clearly visible difference in the amount of stuttering and that's all that really matters.
From my experience, claiming it does nothing is untrue and misleading.