D3D11: more smooth frametime, less FPS and NO errors on exit.
D3D12: frametime spikes, more FPS, ERRORS on exit:

UT2004 Build UT2004_Build_[2005-11-23_16.22]

OS: Windows NT 6.2 (Build: 9200)
CPU: GenuineIntel PentiumPro-class processor @ 3800 MHz with 2047MB RAM
Video: Intel(R) HD Graphics 530 (6)

General protection fault!

History: UD3DRenderDevice::Exit <- UWindowsClient::Destroy <- UObject::ConditionalDestroy <- (WindowsClient Package.WindowsClient) <- DispatchDestroy <- (15292: WindowsClient Package.WindowsClient Class WinDrv.WindowsClient) <- DispatchDestroys <- UObject::PurgeGarbage <- UObject::StaticExit <- appPreExit <- FMallocWindows::Free

Maybe in the next weeks I'll buy an i5 11500 (+B560 + 8 GB DDR4 3200 x2) with the UHD 750 Xe-based and test dgVoodoo2 extensively with it as a "baseline" system in this gaming GPUs dark times 😁
Someone with a Ryzen 5700G/5600G + DDR4 3600 would be really welcome.....

Indeed, D3D12 crashes at exit. Also, x64 D3D9 has problems with state blocks (crash). I fixed both.
Btw, for UT2004, fast vidmem access is very recommended. That game locks the frame buffer completely unnecessary at the end of each frame which drops the FPS to the range 80-90 by default.

i5 11500 (UHD 750) + B560 ordered today, will see how well Intel Xe(-LP) architecture performs with dgVoodoo2 😁

Dege wrote on 2021-10-04, 16:51:

Also, x64 D3D9 has problems with state blocks (crash). I fixed both.

Don't waste your time with the D3D9 renderer of UT 2004, it's (graphically) bugged by its own 😜

lowenz wrote on 2021-10-04, 17:15:

Dege wrote on 2021-10-04, 16:51:

Also, x64 D3D9 has problems with state blocks (crash). I fixed both.

Don't waste your time with the D3D9 renderer of UT 2004, it's (graphically) bugged by its own 😜

How are future generations supposed to know that if they can't run the buggy D3D9 UT2K4 renderer? 😁

It's not about UT2004... The x64 D3D9 SDK sample apps are the ones that're not working...

Just want to report that I got higher fps with dx12 than every before in the old Dx9 race games I use this wrapper for and it also looks great.
So another thumbs up for a great job.
Only thing is it doesn't like the blink video playing at game start but not an issue for me as I do not care about those.
It runs the video but a lot of stutter.

Halo 1.10 + Chimera (maximum quality, unlocked tick rate) can run decently @2560x1440.....on the UHD 530 of an old Skylake i3 6300+DDR4 2400 !

A new WIP:

=========================
WIP82:
=========================

- Fixing x64 D3D9 state block handling crash
- Fixing D3D12 crash on exit
- Fixing DDraw surface create parameter validation incompatibility + some changes for better performance (Rogue Spear Urban Operations)
- Option 'DirectXExt\NPatchTesselationLevel' is integrated into the CPL (GUI)
- Option 'DirectXExt\EnableSpecializedShaders' is added - this is intented to be a transitioning option, to be removed in a later version
- Option 'DirectX\KeepFilterIfPointSampled' is added
- Fixing a bug and doing some changes (optimizations) in the new vs FF DXBC generator
- Fixing a particular D3D8 lighting effect to be compatible with MS D3D8 (Blitzkrieg map editor)
- Enabling (supporting) mixed MSAA types of D3D9 render- and depthstencil targets (some scene demos)
- Fixing a bug in the ps.3.x DXBC generator
- A minor peformance optimization for handling ps FF specialized shaders

http://dege.fw.hu/temp/dgVoodooWIP82.zip
http://dege.fw.hu/temp/dgVoodooWIP82_dbg.zip

Chaos Theory + D3D12 = black intro videos + black menu
No issue with D3D11.

Intel UHD 530

Good news! No more frametime spikes in UT2004! It's nearly perfect now!

it is possible to add per pixel lightning option? its something implementd in Dolphin emulator of gamecube.
The lights fall only on triangle by default, perpixel makes more curved accurate light applying to triangle

It's already there: 'Apply Phong shading when possible'.

But, the thing is, it only affects the fixed function pipeline. Majority of D3D8/9 games use shaders and do their own lighting calculations (usually per pixel).
And the funny things is, majority of DX7 and older games feed D3D with T&L (transformed and lit) vertices, per pixel lighting cannot be applied for them.
So, this option isn't very useful in practice, I was thinking of simply removing it.

Edit: Phong cannot be combined with TruForm. TruForm with its quadratic interpolated normals and high tesselation level already gives results close to per pixel lighting (mentioned even in the ATI docs).

Went to test Diablo 2 again using Glide, it still has screen issues with both normal and NAPALM
also Bilinear is fubar and full of artifacts like D3D with anisotropic, which is still another issue.

it should be more like this
https://www.youtube.com/watch?v=bKeUHQ7jl2o

I hate when too much code changes. I found and fixed a bug that's been present from WIP80 (Splinter Cell 3). It affects only D3D9 with the D3D12 backend and I don't want to release a new WIP just because of this. I replaced D3D9 in WIP82 and reuploaded the zips.
Download it again if interested.

Same black screen 😐
Audio is working.

Here is Diablo 2 with filters on
Direct 3d

DDRAW

3dfx Glide, both Napalm and Normal

Same artifact issues, even with Median XL which brings the game to 1024x768

lowenz wrote on 2021-10-11, 20:12:

Same black screen 😐
Audio is working.

That bug was about 3D rendering, just look at Sam's dress how it's rendered.
The game itself however works fine for me on my Intel UHD 530 + D3D12 (1280x1024). The driver version is 27.20.100.9466, from 2021.03.30.

Wrong aspect ratio

Let it be there about phong shading when possible =)