Issue running 32bit NV DX10 SDK Stencil Routed K-Buffer sample with new DXMT 0.6

[oscarbg]

I have DXMT 0.6 installed correctly on Wine Devel 10.9 gcenx build(https://github.com/Gcenx/macOS_Wine_builds/releases/tag/10.9)..
wanting to test simple DX10 app I download precompiled binary from:
https://developer.download.nvidia.com/SDK/10/Samples/StencilRoutedKBuffer.zip
on site:
https://developer.download.nvidia.com/SDK/10/direct3d/samples.html

app is 32 bit:
file StencilRoutedKBuffer.exe
StencilRoutedKBuffer.exe: PE32 executable (GUI) Intel 80386, for MS Windows

I get this error:

I run with:
wine --version
wine-10.9

wine dx10/Direct3D/Bin/StencilRoutedKBuffer.exe

I tried fixing by running:
winetricks d3dx9
winetricks d3dx10
and
winetricks d3dcompiler_43
and also copy d3dcompiler_43.dll to d3dcompiler_37.dll and d3dcompiler_39.dll in both system32 and syswow64 wine prefix folders as gcenx build wow64 enabled..

before as I was getting some d3dcompiler errors but now that error is now gone..

running with WINEDEBUG=+loaddll,+dll,+d3dcompiler,+d3d10 wine /Volumes/T9/dlss/dx10/Direct3D/Bin/StencilRoutedKBuffer.exe

can help?

thanks..

log:
part log shows seems to be where it shows unsing d3dcompiler_39 d3dcompiler_37 to compile shader:

WINEDEBUG=+loaddll,+dll,+d3dcompiler,+d3d10 wine /Volumes/T9/dlss/dx10/Direct3D/Bin/StencilRoutedKBuffer.exe
00e8:trace:loaddll:build_module Loaded L"C:\windows\system32\d3dcompiler_39.dll" at 79AA0000: native
00e8:trace:loaddll:build_module Loaded L"C:\windows\system32\opengl32.dll" at 79670000: builtin
00e8:trace:loaddll:build_module Loaded L"C:\windows\system32\wined3d.dll" at 797A0000: builtin
00e8:trace:loaddll:build_module Loaded L"C:\windows\system32\d3d10.dll" at 7A2F0000: builtin
00e8:err:winediag:wined3d_dll_init Using the Vulkan renderer.
00e8:trace:loaddll:build_module Loaded L"C:\windows\system32\d3d10_1.dll" at 79650000: builtin
00e8:trace:d3d10:d3d10_create_device adapter 00012F18, driver_type D3D10_DRIVER_TYPE_HARDWARE, swrast 00000000, flags 0, sdk_version 0x1d, device 0012FA14.
info: Maximum supported feature level: D3D_FEATURE_LEVEL_11_1
info: Using feature level D3D_FEATURE_LEVEL_10_0
00e8:trace:d3d10:d3d10_create_device Created ID3D10Device 00016138.
00e8:trace:d3d10:d3d10_create_device adapter 00000000, driver_type D3D10_DRIVER_TYPE_REFERENCE, swrast 00000000, flags 0, sdk_version 0x1d, device 0012FA14.
00e8:warn:d3d10:d3d10_create_device Failed to load refrast, returning E_FAIL.
00e8:trace:d3d10:d3d10_create_device adapter 01219A88, driver_type D3D10_DRIVER_TYPE_HARDWARE, swrast 00000000, flags 0, sdk_version 0x1d, device 0012F9D0.
info: Maximum supported feature level: D3D_FEATURE_LEVEL_11_1
info: Using feature level D3D_FEATURE_LEVEL_10_0
00e8:trace:d3d10:d3d10_create_device Created ID3D10Device 01219F90.
00e8:trace:d3d10:d3d10_create_device adapter 01213880, driver_type D3D10_DRIVER_TYPE_HARDWARE, swrast 00000000, flags 0, sdk_version 0x1d, device 0012F9D0.
info: Maximum supported feature level: D3D_FEATURE_LEVEL_11_1
info: Using feature level D3D_FEATURE_LEVEL_10_0
00e8:trace:d3d10:d3d10_create_device Created ID3D10Device 01212ED8.
00e8:trace:d3d10:d3d10_create_device1 adapter 01212E18, driver_type D3D10_DRIVER_TYPE_HARDWARE, swrast 00000000, flags 0, hw_level D3D10_FEATURE_LEVEL_10_1, sdk_version 32, device 0012FB40.
info: Maximum supported feature level: D3D_FEATURE_LEVEL_11_1
info: Using feature level D3D_FEATURE_LEVEL_10_1
00e8:trace:d3d10:d3d10_create_device1 Created device 012139A0.
00e8:trace:loaddll:build_module Loaded L"C:\windows\system32\D3DCompiler_37.dll" at 79430000: native
00e8:trace:loaddll:free_modref Unloaded module L"C:\windows\system32\D3DCompiler_37.dll" : native

[3Shain]

Have you checked the demo works on Windows? I can't make it work.

[3Shain]

Since you have the source code, can you take a look what exactly is causing this error https://github.com/microsoft/DXUT/blob/2b7f8caca999a79c279bb937195e208e4de6374c/Core/DXUT.cpp#L3471 ?

[oscarbg]

Have you checked the demo works on Windows? I can't make it work.

I had working some time (years) ago I believe..
you need "DirectX End-User Runtimes (June 2010)" for it to work.. https://www.microsoft.com/en-us/download/details.aspx?id=8109
have you installed it?
also what error are you getting?

[oscarbg]

Since you have the source code, can you take a look what exactly is causing this error https://github.com/microsoft/DXUT/blob/2b7f8caca999a79c279bb937195e208e4de6374c/Core/DXUT.cpp#L3471 ?

this is nice but since precompiled demo is compiled 17 years ago DXUT library code may have changed a lot and should look at DXUT code around april 2008..

[3Shain]

Have you checked the demo works on Windows? I can't make it work.

I had working some time (years) ago I believe.. you need "DirectX End-User Runtimes (June 2010)" for it to work.. https://www.microsoft.com/en-us/download/details.aspx?id=8109 have you installed it? also what error are you getting?

I'm getting the same error as yours even on Windows.

Since you have the source code, can you take a look what exactly is causing this error https://github.com/microsoft/DXUT/blob/2b7f8caca999a79c279bb937195e208e4de6374c/Core/DXUT.cpp#L3471 ?

this is nice but since precompiled demo is compiled 17 years ago DXUT library code may have changed a lot and should look at DXUT code around april 2008..

well yes the reference code is actually DXUT11 which is newer.

anyway I need to know which call is failing. if you can't modified the source, then at least I need an apitrace

[oscarbg]

just seen I picked a somewhat controversial sample to test as seen on as seen on:
https://developer.download.nvidia.com/SDK/10/direct3d/ReleaseNotes.htm
New in This 10.6 Release
2. Bug fix for StencilRoutedKBuffer: the previous version would create a DX 10.1 device, resulting in corruption errors. A modification to DXUTCore prevents the DX 10.1 device from being created for the StencilRoutedKBuffer project

and 10.6 release is from Version 10.6, January 2011, so maybe the exe isn't working after all on Windows even..

let me try on Windows..

[oscarbg]

@3Shain
demo is working on Windows without problems.. at least on my Nvidia GPU..
as said before I believe key is to have installed:
"DirectX End-User Runtimes (June 2010)"
see:

[oscarbg]

ok I attach apitrace:
https://drive.google.com/file/d/1_J2SrMuHGda9gCFDIZONup91BY0lTv_B/view?usp=sharing

capture and replay working OK!!
info:

C:\Users\oqui\Downloads\StencilRoutedKBuffer\apitrace-win32-x86\apitrace-latest-win32\apitrace-latest-win32\bin\apitrace trace -a dxgi -v -o dxmt.trace C:\Users\oqui\Downloads\StencilRoutedKBuffer\Direct3D\Bin\StencilRoutedKBuffer.exe
info: did not find C:\Users\oqui\Downloads\StencilRoutedKBuffer\apitrace-win32-x86\apitrace-latest-win32\apitrace-latest-win32\bin\..\wrappers\RelWithDebInfo\dxgitrace.dll
info: found C:\Users\oqui\Downloads\StencilRoutedKBuffer\apitrace-win32-x86\apitrace-latest-win32\apitrace-latest-win32\bin\..\lib\wrappers\dxgitrace.dll
inject -D C:\Users\oqui\Downloads\StencilRoutedKBuffer\apitrace-win32-x86\apitrace-latest-win32\apitrace-latest-win32\bin\..\lib\wrappers\dxgitrace.dll -- C:\Users\oqui\Downloads\StencilRoutedKBuffer\Direct3D\Bin\StencilRoutedKBuffer.exe

C:\Users\oqui\Downloads\StencilRoutedKBuffer\apitrace-win32-x86\apitrace-latest-win32\apitrace-latest-win32\bin\apitrace replay dxmt.trace

apitrace replay dxmt.trace
C:\Users\oqui\Downloads\StencilRoutedKBuffer\Direct3D\Bin\StencilRoutedKBuffer.exe
155 @0 IDXGIFactory7::EnumAdapters(this = 0x860200, Adapter = 1, ppAdapter = &0x3ac2c30) = S_OK
155: warning: ignoring non-default adapter 1
171 @0 IDXGIFactory7::EnumAdapters(this = 0x860200, Adapter = 1, ppAdapter = &0x39f31e8) = S_OK
171: warning: ignoring non-default adapter 1
206 @0 IDXGIFactory7::EnumAdapters(this = 0x860200, Adapter = 1, ppAdapter = &0x3ac2c30) = S_OK
206: warning: ignoring non-default adapter 1
Rendered 247 frames in 1.44285 secs, average of 171.189 fps

[oscarbg]

running the trace obtained on Windows (https://drive.google.com/file/d/1_J2SrMuHGda9gCFDIZONup91BY0lTv_B/view?usp=sharing) on Mac via qapitrace on a Wine with DXMT 0.6 correctly configured I get crash on first frame replay:

  1 0x00465789 retrace_ID3D10Device__ClearDepthStencilView+0x89(call=01005808) [C:\a\apitrace\apitrace\build\retrace\d3dretrace_dxgi.cpp:5811] in d3dretrace (0x0012fd2c)
  2 0x0048f409 retrace::Retracer::retrace+0xe9(call=00000000, this=0012FD98) [C:\a\apitrace\apitrace\retrace\retrace.cpp:159] in d3dretrace (0x0012fd54)
  3 0x004921fd contains(call=00000000) [C:\a\apitrace\apitrace\lib\..

also I see some error on console:

err:   CreateMTLTextureDescriptorInternal: sample count 8 is not supported.
-[_MTLDevice minimumLinearTextureAlignmentForPixelFormat:]:5213: failed assertion `Linear textures do not support depth/stencil pixel formats' 

EDIT: also seeing now:

err:   CreateMTLTextureDescriptorInternal: sample count 8 is not supported.
err:   CreateMTLTextureDescriptorInternal: sample count 8 is not supported.
err:   CreateMTLTextureDescriptorInternal: sample count 8 is not supported.
err:   No rendering attachment or uav is bounded

[3Shain]

I managed to make it render something with this patch but unfortunately it still looks incorrect

It's most likely caused by 8x sample count which is hardcoded by

#define MSAA_LEVEL 8

in the source of the demo. It can't really be workaround-ed as the shader is compiled to expect 8 samples (instead of querying GetDimensions()). So you may try to compile the demo with #define MSAA_LEVEL 4

[oscarbg]

@3Shain thanks for info..
compiled sample with:
#define MSAA_LEVEL 4
and did a trace..
renders mostry correctly..

can send executable compiled with "MSAA_LEVEL 4" if you want also..

link to trace:
https://drive.google.com/file/d/1ghsF0DPe9k_Tx67pyfuoXLaSjw-7YRyn/view?usp=sharing
is crashing also..
using qapitrace I see errors like:
-[_MTLDevice minimumLinearTextureAlignmentForPixelFormat:]:5213: failed assertion Linear textures do not support depth/stencil pixel formats'`
can you please give a look at any remaing issues? or if with your patch renders now correctly?

thanks..

[oscarbg]

just to confirm, part of the patch referenced below fixes:
-[_MTLDevice minimumLinearTextureAlignmentForPixelFormat:]:5213: failed assertion Linear textures do not support depth/stencil pixel formats'
` ?

if yes, it will be integrated on DXMT or is a "hack" and apps should behave correctly without it?

part of patch:

diff --git a/src/dxmt/dxmt_staging.hpp b/src/dxmt/dxmt_staging.hpp
index 2be0096..73db32c 100644
--- a/src/dxmt/dxmt_staging.hpp
+++ b/src/dxmt/dxmt_staging.hpp
@@ -65,9 +65,26 @@ private:
 
     ~StagingBuffer() {
 #ifdef __i386__
-    _aligned_free(info.memory.get());
+      if (info.memory.get())
+        _aligned_free(info.memory.get());
 #endif
     }
+
+    StagingBuffer() = default;
+    StagingBuffer(const StagingBuffer &) = delete;
+    StagingBuffer &operator=(const StagingBuffer &) = delete;
+    StagingBuffer(StagingBuffer &&move) {
+      info = move.info;
+      move.info.memory.set(nullptr);
+      allocation = std::move(move.allocation);
+    };
+    StagingBuffer &
+    operator=(StagingBuffer &&move) {
+      info = move.info;
+      move.info.memory.set(nullptr);
+      allocation = std::move(move.allocation);
+      return *this;
+    };
   };
   struct QueueEntry {
     uint64_t id;

[3Shain]

just to confirm, part of the patch referenced below fixes: -[_MTLDevice minimumLinearTextureAlignmentForPixelFormat:]:5213: failed assertion Linear textures do not support depth/stencil pixel formats' ` ?

if yes, it will be integrated on DXMT or is a "hack" and apps should behave correctly without it?

part of patch:

diff --git a/src/dxmt/dxmt_staging.hpp b/src/dxmt/dxmt_staging.hpp
index 2be0096..73db32c 100644
--- a/src/dxmt/dxmt_staging.hpp
+++ b/src/dxmt/dxmt_staging.hpp
@@ -65,9 +65,26 @@ private:
 
     ~StagingBuffer() {
 #ifdef __i386__
-    _aligned_free(info.memory.get());
+      if (info.memory.get())
+        _aligned_free(info.memory.get());
 #endif
     }
+
+    StagingBuffer() = default;
+    StagingBuffer(const StagingBuffer &) = delete;
+    StagingBuffer &operator=(const StagingBuffer &) = delete;
+    StagingBuffer(StagingBuffer &&move) {
+      info = move.info;
+      move.info.memory.set(nullptr);
+      allocation = std::move(move.allocation);
+    };
+    StagingBuffer &
+    operator=(StagingBuffer &&move) {
+      info = move.info;
+      move.info.memory.set(nullptr);
+      allocation = std::move(move.allocation);
+      return *this;
+    };
   };
   struct QueueEntry {
     uint64_t id;

It's on main now, you can just test the latest artifact or compile yourself.

And I don't get -[_MTLDevice minimumLinearTextureAlignmentForPixelFormat:]:5213: failed assertion Linear textures do not support depth/stencil pixel formats' ` when replaying your provided trace. What's the latest trace call id before you get this assertion?

[3Shain]

is there still anything to fix?