4.0-1288 and over causes "sparkles" in d3d

[KHg8m3r]

Intel basically dropped support for anything below HD 4600, as they didn't provide any WDDM 2.0 drivers for anything older than the Haswell 4600 when Windows 10 was released, and they have no plans on releasing drivers for the anymore.

[Craftyawesome]

(06-25-2016, 08:27 AM)TurboK Wrote: So Dolphin 5.0 was released earlier today (great!) and since it is a very stable version for all sorts of uses I wanted to update the version with the fix for old Intel HD drivers.

This version is based on the development build, and is already some commits ahead of 5.0 (note: this is for Windows!):
https://dl.dropboxusercontent.com/u/23618751/Dolphin4.0.1288_glitch/x64_Intel_TEV_5.0-12.zip

The source code for this can be found at
https://github.com/TurboK234/dolphin/tree/Intel_TEV
and the critical commit at
https://github.com/TurboK234/dolphin/commit/bce040e7035a32ab7c7ae1af13ff9e042fa5ee56

And if someone wants to keep their Dolphin up to date with the fix, I suggest that you
-create a Github account and fork Dolphin
-install Visual Studio 2015 with Git support (using the Dolphin wiki and other guides)
-clone (locally) your own Github fork as "origin" and set the official Dolphin master branch as "upstream"
-create a local "Intel_TEV" branch
-use VS2015 to edit the involved lines in PixelShaderGen.cpp (the commit linked above) (use copy-pasting, takes 5 minutes)
-save the changes, commit the changes locally
-optional: push your brand new fixed branch to your Github account ($ git push origin Intel_TEV)
-build your copy of Dolphin using VS2015
-when you want to update your local repository use $ git fetch upstream and $ git merge upstream/master and/or $ git rebase master
-when there eventually is a merge/rebase conflict either solve it manually or just delete the old fixed branch and create a new one (start from "-create a local..." line above)

Or at least these are the basic steps I keep following. There might be faster/easier ways, but with this (and guides that you find with Google) you'll get started.

Phire changed a lot of things the day after you posted that. I can't even figure out how to work with the commit.


Before I had used visual studio to edit the cpp and just injected it in an extracted dolphin master zip.

[TurboK]

(06-28-2016, 05:44 AM)Craftyawesome Wrote: Phire changed a lot of things the day after you posted that. I can't even figure out how to work with the commit.


Before I had used visual studio to edit the cpp and just injected it in an extracted dolphin master zip.

He really did  
This is a good example of a "merge conflict" that I referred to. You could fix it by hand, but I find it easier to just redo the changes to avoid the manual git resolving.
I took a quick look and the code lines are displaced from what they were, but the actual code is easily identifiable.

I don't want to mess with my previous build-source-commit links, but to illustrate the -present- changes needed in PixelShaderGen.cpp I made another branch. The commit of interest can be found at https://github.com/TurboK234/dolphin/commit/9a079c235777d3f6beda9d9916fde24d04cef382 .

(Qt2 interface seems to be broken ATM because of other things, one more reason not to rush into development builds right now, since there seems to be a lot of pending progress (and eventual regressions) happening)

[Craftyawesome]

(06-28-2016, 07:23 AM)TurboK Wrote: He really did  
This is a good example of a "merge conflict" that I referred to. You could fix it by hand, but I find it easier to just redo the changes to avoid the manual git resolving.
I took a quick look and the code lines are displaced from what they were, but the actual code is easily identifiable.

I don't want to mess with my previous build-source-commit links, but to illustrate the -present- changes needed in PixelShaderGen.cpp I made another branch. The commit of interest can be found at https://github.com/TurboK234/dolphin/commit/9a079c235777d3f6beda9d9916fde24d04cef382 .

(Qt2 interface seems to be broken ATM because of other things, one more reason not to rush into development builds right now, since there seems to be a lot of pending progress (and eventual regressions) happening)

Oh I was using control-f to find
WriteTevRegular(ShaderCode& out, const char* components, int bias, int op, int clamp, int shift);
and since that's now split between lines I couldn't find it lol.