Dege wrote on 2021-09-03, 08:25:

========================= WIP80.1 ========================= […]
Show full quote

=========================
WIP80.1
=========================

- Fixing a tesselation incompatibility

http://dege.fw.hu/temp/dgVoodooWIP80_1.zip
http://dege.fw.hu/temp/dgVoodooWIP80_1_dbg.zip

What's the tessellation incompatibility mentioned in the changelog?

That bug got fixed Re: ATI TruForm
discuss truform only on topic that i linked

RaVeN-05 wrote on 2021-07-25, 16:39:

...

Just want to say:

Please don't link directly to a download file, link to the page which contains the download you want to link to instead. So users even have the chance and advantage to find other useful things on the pages. Thanks in advance.

kind regards
soggi

Unreal Tournamet 2004 i have some completely black surfeces and some grass is black too, look screenshot atached

Is it Truformed? (Looks quite round-shaped.) If it is, how did you manage to achieve this? Last time I set the npatch-related elems in the UT2004 config file(s) but to no avail.

yes truformed , its messed in ini files at two files , ut2004.ini and default.ini in each file two times appears string of usenpatch and tesslevel.
Setting needed values everywhere makes it work.

2.75.1 D3D12 can't start Severance (crash @start), FL11 or FL12 on Intel GPU.
D3D11 no problem!

Will test WIP 80.1 now

lowenz wrote on 2021-09-19, 08:41:

Will test WIP 80.1 now

The same:

EXCEPTION_RECORD: (.exr -1)
ExceptionAddress: 6e5fe98c (d3d11on12!D3D12TranslationLayer::BatchedContext::ProcessBatchWork+0x00000042)
ExceptionCode: c0000005 (Access violation)
ExceptionFlags: 00000000
NumberParameters: 2
Parameter[0]: 00000000
Parameter[1]: 53b969ac
Attempt to read from address 53b969ac

PROCESS_NAME: Blade.exe

READ_ADDRESS: 53b969ac

ERROR_CODE: (NTSTATUS) 0xc0000005 - The instruction at 0x%p referenced memory at 0x%p. The memory could not be %s.

EXCEPTION_CODE_STR: c0000005

EXCEPTION_PARAMETER1: 00000000

EXCEPTION_PARAMETER2: 53b969ac

STACK_TEXT:
1504fef0 6e5ff18d 02780624 27d1f138 099f3ab8 d3d11on12!D3D12TranslationLayer::BatchedContext::ProcessBatchWork+0x42
1504ff24 6e5ff251 0278060c 27d1f174 6e5fce70 d3d11on12!D3D12TranslationLayer::BatchedContext::ProcessBatchImpl+0x2c
1504ff68 6e5fce7d 1504ff80 768afa29 099f3ab8 d3d11on12!D3D12TranslationLayer::BatchedContext::BatchThread+0x8f
1504ff70 768afa29 099f3ab8 768afa10 1504ffdc d3d11on12!<lambda_52be7ffa4176bc56d64c1be44fbd6b8b>::<lambda_invoker_stdcall>+0xd
1504ff80 77cf7a7e 099f3ab8 e7828362 00000000 kernel32!BaseThreadInitThunk+0x19
1504ffdc 77cf7a4e ffffffff 77d18a1b 00000000 ntdll!__RtlUserThreadStart+0x2f
1504ffec 00000000 6e5fce70 099f3ab8 00000000 ntdll!_RtlUserThreadStart+0x1b


SYMBOL_NAME: d3d11on12!D3D12TranslationLayer::BatchedContext::ProcessBatchWork+42

MODULE_NAME: d3d11on12

IMAGE_NAME: d3d11on12.dll

STACK_COMMAND: ~21s ; .ecxr ; kb

FAILURE_BUCKET_ID: INVALID_POINTER_READ_c0000005_d3d11on12.dll!D3D12TranslationLayer::BatchedContext::ProcessBatchWork

OS_VERSION: 10.0.19041.1

BUILDLAB_STR: vb_release

OSPLATFORM_TYPE: x86

OSNAME: Windows 10

IMAGE_VERSION: 6.2.19041.1081

FAILURE_ID_HASH: {72e6e7cf-f995-da4b-d586-be2b11fa0dc8}

Followup: MachineOwner

found a bit of an odd bug but i assume its specific to nvidia's multithreaded driver,

using DgVoodoo together with the sdb fix that applies singleprocaffinity to starlancer results in temporary high cpu hangs in nvwgf2um.dll:OpenAdapterQuery
without it multiple spawn with little cpu usage and no hang occurs.

Sorry for being silent, but I fooled myself into completely overhauling (replacing) the FF vertex pipeline implementation of D3D. It's basically ready and according to my preliminary testing, it yields much much better GPU performance than the old version, especially with TruForm.

Until it's not completely finished I won't do bugfixing/validating though. But after that, I'll return to bugfix mode.

Thanks man! Mad skills as always 😁

Hi Dege. Only wanted to thank you very much for your work. I've been trying to run Dungeon Keeper 2 on modern systems without constant crashes to desktop for years. I tested on a VMware Windows XP VM (crashed anyway), cnc-ddraw (crashes when entering the name when saving the game), but with DgVoodoo 2 it simply runs great. And using the 1080p resolution of my monitor, and those lovely fine scanlines, is a plus. Great job, very appreciated.

I'm trying to figure out the best way to play Valkyria Chronicles. So I did an image quality comparison.

dgVoodoo 1280x720 scaled to 3840x2160:

dgVoodoo 1280x720 resolution forced to 3840x2160:

Native 3840x2160

3480x2160 native obviously looks best, but the PC version's high resolution rendering is broken. Units don't evade properly.

dgVoodoo provides a nice boost in sharpness over 1280x720, but has a bit of a double image problem that gets worse over range.

1280x720 by itself is just a blurry mess.

Fun story, this port was praised on release for its arbitrary resolution and framerate support:
https://www.pcgamer.com/valkyria-chronicles-p … rantes-verdict/
https://www.ign.com/articles/2014/11/26/valky … icles-pc-review
Without going into details, high framerate support is also broken. This port needs to be ran at 1280x720@30Hz. I think I'm going to play through the PS4 port to see how it compares. But if you're stuck with the PC version, use dgVoodoo to force resolution.

=========================
WIP81
=========================

- Replacing the FF vertex pipeline in D3Ds with a new code generator to use specialized shaders
* This might be enabled only for N-patch tesselation in subsequent versions
- Adding a option DirectXExt\NPatchTesselationLevel to force N-patch tesselation for FF/vs.1.x shader pipeline
(not recommended, it's intended for experimenting/debugging purposes)
- Adding scaling mode 'Stretched, keep aspect ratio (CRT like)'
- Fixing a minor bug in DDraw with fast videomemory access

http://dege.fw.hu/temp/dgVoodooWIP81.zip
http://dege.fw.hu/temp/dgVoodooWIP81_dbg.zip

So, I finished the new FF vertex pipeline implementation, but I'm not sure if I enable it for other than TruForm vertex processing in the final version.
The GPU performance is better for plain vertex processing as well, but not at a level that worths factoring too many specialized shaders.
Since it's a new implementation, some games/things can be broken. I'll fix reported cases.

mgtroyas wrote on 2021-10-01, 09:28:

Hi Dege. Only wanted to thank you very much for your work. I've been trying to run Dungeon Keeper 2 on modern systems without constant crashes to desktop for years. I tested on a VMware Windows XP VM (crashed anyway), cnc-ddraw (crashes when entering the name when saving the game), but with DgVoodoo 2 it simply runs great. And using the 1080p resolution of my monitor, and those lovely fine scanlines, is a plus. Great job, very appreciated.

Thank you! 😎

lowenz wrote on 2021-09-19, 11:17:

The same: […]
Show full quote

lowenz wrote on 2021-09-19, 08:41:

Will test WIP 80.1 now

The same:

EXCEPTION_RECORD: (.exr -1)
ExceptionAddress: 6e5fe98c (d3d11on12!D3D12TranslationLayer::BatchedContext::ProcessBatchWork+0x00000042)
ExceptionCode: c0000005 (Access violation)
ExceptionFlags: 00000000
NumberParameters: 2
Parameter[0]: 00000000
Parameter[1]: 53b969ac
Attempt to read from address 53b969ac

PROCESS_NAME: Blade.exe

READ_ADDRESS: 53b969ac

ERROR_CODE: (NTSTATUS) 0xc0000005 - The instruction at 0x%p referenced memory at 0x%p. The memory could not be %s.

EXCEPTION_CODE_STR: c0000005

EXCEPTION_PARAMETER1: 00000000

EXCEPTION_PARAMETER2: 53b969ac

STACK_TEXT:
1504fef0 6e5ff18d 02780624 27d1f138 099f3ab8 d3d11on12!D3D12TranslationLayer::BatchedContext::ProcessBatchWork+0x42
1504ff24 6e5ff251 0278060c 27d1f174 6e5fce70 d3d11on12!D3D12TranslationLayer::BatchedContext::ProcessBatchImpl+0x2c
1504ff68 6e5fce7d 1504ff80 768afa29 099f3ab8 d3d11on12!D3D12TranslationLayer::BatchedContext::BatchThread+0x8f
1504ff70 768afa29 099f3ab8 768afa10 1504ffdc d3d11on12!<lambda_52be7ffa4176bc56d64c1be44fbd6b8b>::<lambda_invoker_stdcall>+0xd
1504ff80 77cf7a7e 099f3ab8 e7828362 00000000 kernel32!BaseThreadInitThunk+0x19
1504ffdc 77cf7a4e ffffffff 77d18a1b 00000000 ntdll!__RtlUserThreadStart+0x2f
1504ffec 00000000 6e5fce70 099f3ab8 00000000 ntdll!_RtlUserThreadStart+0x1b


SYMBOL_NAME: d3d11on12!D3D12TranslationLayer::BatchedContext::ProcessBatchWork+42

MODULE_NAME: d3d11on12

IMAGE_NAME: d3d11on12.dll

STACK_COMMAND: ~21s ; .ecxr ; kb

FAILURE_BUCKET_ID: INVALID_POINTER_READ_c0000005_d3d11on12.dll!D3D12TranslationLayer::BatchedContext::ProcessBatchWork

OS_VERSION: 10.0.19041.1

BUILDLAB_STR: vb_release

OSPLATFORM_TYPE: x86

OSNAME: Windows 10

IMAGE_VERSION: 6.2.19041.1081

FAILURE_ID_HASH: {72e6e7cf-f995-da4b-d586-be2b11fa0dc8}

Followup: MachineOwner

Close MSI Afterburner and it'll work. The crash is coming from d3d11on12 which is used by MSIA (it doesn't have a native D3D12 implementation) but I don't know why.
I wrote about it in the readme in the D3D12 section:

MSI Afterburner relies on D3D11on12 to render its own overlay when hooking D3D12. I ran into multiple cases when D3D11on12 crashed or hung at startup, for unknown reasons. First try a game with MSIA shut down if you have problem with launching it.

+14% in UT2004 on an Intel UHD 530 @1920x1440!

Hm, it's nice!
But, aren't you getting a lot of microlags or hiccups during the gameplay (typically with D3D12)?

Only when a new effect/shader (?) is loaded (compiled) the FIRST time!

1660 Ti (QMax)

vs

RX 570

See the pixel shaders test (Nature) difference!

Yes, interesting. You can try both with D3D11 and 12 to see if there is difference. With D3D11 the driver gets DXBC shader code, while DXIL with D3D12.