Dolphin, the GameCube and Wii emulator - Forums

Hi there,
I recently committed some code in the efb_scaling_fixes branch addressing various scaling issues in Dolphin. Most importantly, but not limited to, this fixes the issues mentioned in this thread.

However, I'm not quite sure it didn't break anything, so I'd like people to test around a bit with it. Nothing fancy, just run your favorite game with it and change around the various settings which directly or indirectly affect internal resolution (IR itself, antialiasing, XFB settings, EFB copies to RAM, etc...) before and/or while running the game. Report any reproducable oddities that arise and which cannot be observed in current master builds. All reports are welcome, thanks for testing!

Builds are available here: x64 x86

But I can't really understand what this branch should fix and I've never seen blurred graphic in Xenoblade Chronicles...

The visible part is a very subtle fix, but the architectural changes are nontrivial. I.e. Testing should be focused on making sure nothing broke. In particular it really shouldn't introduce any crashes, nor should it somehow break the behavior of internal resolution.

I will do some test, but should i use Fractional (Auto window size), i like to use 3X than fractional, did this change can leave the screen blurred or it can fix the blurred screen?

The last Story for example, the screen get blurred, on gameplay and on menu, after a while.

Edit: I didn't get any issues here

What matters is stress-testing, i.e. play around as much as possible with the settings. I don't need testers for actual fixes because I know the fix is correct. I just need people to verify that e.g. changing from 1xIR to 4xIR ingame really still works, etc.

I'll try to test this with a few games today.

*Wonders why NV doesn't test this*

Because he doesn't have a properly working computer ATM.

(10-22-2012, 12:41 AM)delroth Wrote: [ -> ]Because he doesn't have a properly working computer ATM.

True XD, I guess he'll try after he gets his new one set up :p

I still cannot completely understand the purpose of risking a regression for something that you cannot even see...anyway here's the first one that I've found:

ANOTHER CODE: R (RNOP01) now shows only 50% of the image (only the top half screen) with Dx9 and auto (window and multiple), 1.5x or 2.5x Internal Resolutions; the other Dx9 IR (1x, 2x, 3x and 4x) are instead ok (show full screen).
OpenGL is still ok with every IR (cannot test Dx11 because I have WinXP).