gpuopen-librariesandsdks / capsaicin Goto Github PK

core/src/math/pack.hlsl:481, the unpackNormal function fails to unpack normals with negative values packed with packNormal.

The current implementation of unpackNormal is:

/**
 * Convert 10bit snorms to normal vector.
 * @param packedValue Input snorm values to convert.
 * @returns Converted float values (range [-1,1]).
 */
float3 unpackNormal(uint packedValue)
{
    uint3 value = uint3(packedValue, packedValue >> 10,
        packedValue >> 20) & 0x3FFu.xxx;
    return float3(value) * (1.0f / 511.0f).xxx;
}

It fails to unpack negative values.

InterlockedExchange(g_ScreenProbes_ProbeSpawnBuffer[g_ScreenProbes_OverrideTileBuffer[index]], ScreenProbes_PackSeed(seed), index);

so as to match with the GI-1.0 technical report.

at capsaicin_internal.cpp: 665:

bool CapsaicinInternal::setDebugView(std::string_view const &name) noexcept
{
    auto debugView = std::find_if(
        debug_views_.cbegin(), debug_views_.cend(), [this](auto val) { return debug_view_ == val.first; });
    if (debugView == debug_views_.cend())
    {
        GFX_PRINTLN("Error: Requested invalid debug view: %s", name.data());
        return false;
    }
    debug_view_ = name;
    return true;
}

debug_view_ = name; assumes that name is a string view of a string whose lifetime expands across frames. When a string_view of a temporary string object is passed in (and it is "equal" to one of the strings within the vector debug_views), the function silently fails, debug_view sooner become meaningless and ultimately causes random behavior.
I'll suggest a fix about changing debug_view_ = name; to debug_view_ = *debugView;
Also, auto debugView = std::find_if(debug_views_.cbegin(), debug_views_.cend(), [this](auto val) { return debug_view_ == val.first; }); seems to not be doing correct validation.

In visibility_buffer.cpp, line 69-70:

    disocclusion_mask_program_ = gfxCreateProgram(
        gfx_, "render_techniques/visibility_buffer/disocclusion_mask", capsaicin.getShaderPath());

It potentially creats a shader program path src/core/src/render_techniques/visibility_buffer/disocclusion_mask, which is a string of length 66. It overflows the name property of class GfxProgram and cause the program to crash by chance.

At light_sampler_bounds.hlsl:282, by upgrading DXC to 2024.3.22 the inout decorator will not crash the compiler again. Thus the random numbers wont be reused in the reservoir generation.

Is this a normal phenomenon? Is there a way to eliminate obvious noise by adjusting parameters?

my gpu is 3080, use 38ms, the paper said 6900xt use 3ms.
I need set something?

OS: Windows 10
Visual Stuidio 2022

ninja: error: 'd3d12.lib', needed by 'lib/core.lib', missing and no known rule to make it

I noticed there's a ssgi_unroll_kernel_ option in the Render Options panel at the bottom of the scene viewer GUI:

This option is disabled by default, but enabling it improves performance on a GeForce RTX 4090:

Is there a reason it's not enabled by default? I can't spot any visual difference in quality, both in still scenes and when moving the camera.

Getting this when I'm running the scene viewer from Visual Studio in debug mode. In release mode I get a black screen and the app crashes after a while.

br,
Sami

Selecting Windows SDK version 10.0.19041.0 to target Windows 10.0.19045.
The C compiler identification is MSVC 19.29.30148.0
The CXX compiler identification is MSVC 19.29.30148.0
Detecting C compiler ABI info
Detecting C compiler ABI info - done
Check for working C compiler: C:/Program Files (x86)/Microsoft Visual Studio/2019/Community/VC/Tools/MSVC/14.29.30133/bin/Hostx64/x64/cl.exe - skipped
Detecting C compile features
Detecting C compile features - done
Detecting CXX compiler ABI info
Detecting CXX compiler ABI info - done
Check for working CXX compiler: C:/Program Files (x86)/Microsoft Visual Studio/2019/Community/VC/Tools/MSVC/14.29.30133/bin/Hostx64/x64/cl.exe - skipped
Detecting CXX compile features
Detecting CXX compile features - done
Found Bash: C:/Program Files/Git/bin/bash.exe
Looking for CL_VERSION_2_2
Looking for CL_VERSION_2_2 - not found
Looking for CL_VERSION_2_1
Looking for CL_VERSION_2_1 - not found
Looking for CL_VERSION_2_0
Looking for CL_VERSION_2_0 - not found
Looking for CL_VERSION_1_2
Looking for CL_VERSION_1_2 - not found
Looking for CL_VERSION_1_1
Looking for CL_VERSION_1_1 - not found
Looking for CL_VERSION_1_0
Looking for CL_VERSION_1_0 - not found
Could NOT find OpenCL (missing: OpenCL_LIBRARY OpenCL_INCLUDE_DIR)
Found Vulkan: C:/VulkanSDK/1.3.239.0/Lib/vulkan-1.lib
Found Perl: C:/Strawberry/perl/bin/perl.exe (found version "5.32.1")
AVX2 backend - ON
SSE4.1 backend - OFF
SSE2 backend - OFF
NEON backend - OFF
NONE backend - OFF
NATIVE backend - OFF
Decompressor - OFF
No invariance - OFF
Diagnostics - OFF
ASAN - OFF
Unit tests - OFF
Performing Test COMPILER_HAS_DEPRECATED_ATTR
Performing Test COMPILER_HAS_DEPRECATED_ATTR - Failed
Performing Test COMPILER_HAS_DEPRECATED
Performing Test COMPILER_HAS_DEPRECATED - Success
Configuring done
CMake Error at src/core/CMakeLists.txt:109 (add_custom_command):
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:core>

Expression did not evaluate to a known generator expression

CMake Error at src/core/CMakeLists.txt:109 (add_custom_command):
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:core>

Expression did not evaluate to a known generator expression

CMake Error at src/core/CMakeLists.txt:109 (add_custom_command):
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:core>

Expression did not evaluate to a known generator expression

CMake Error at src/core/CMakeLists.txt:109 (add_custom_command):
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:core>

Expression did not evaluate to a known generator expression

CMake Error at src/core/CMakeLists.txt:109 (add_custom_command):
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:core>

Expression did not evaluate to a known generator expression

CMake Error at src/core/CMakeLists.txt:109 (add_custom_command):
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:core>

Expression did not evaluate to a known generator expression

CMake Error at src/core/CMakeLists.txt:109 (add_custom_command):
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:core>

Expression did not evaluate to a known generator expression

CMake Error at src/core/CMakeLists.txt:109 (add_custom_command):
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:core>

Expression did not evaluate to a known generator expression

CMake Error:
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:scene_viewer>

Expression did not evaluate to a known generator expression

CMake Error:
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:scene_viewer>

Expression did not evaluate to a known generator expression

CMake Error:
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:scene_viewer>

Expression did not evaluate to a known generator expression

CMake Error:
Error evaluating generator expression:

$<TARGET_RUNTIME_DLLS:scene_viewer>

Expression did not evaluate to a known generator expression

Generating done

Full log

https://pastebin.com/CJT7Em9q

Widows 11 Pro, 10.0.22631 Build 22631
Cmake 3.28.3
Visual Studio 17 2022 up to date (17.9.3.)

More info in pastebin

OS: Windows 11
Compiler: Visual Studio 2022, CMake 3.26.3

I cloned this repository with --recursive and ran cmake -S ./ -B ./build -A x64 then cmake --build ./build --config RelWithDebInfo as instructed in getting_started.md. The scene_viewer.exe binary compiled successfully, but running it by double-clicking it spawns a window that closes immediately. Running --help shows a list of CLI arguments (including a functional --list-scenes), so the binary isn't broken.

I was able to resolve the issue by running build\bin\RelWithDebInfo\scene_viewer.exe from the cloned repository root. Ideally, scene_viewer.exe should automatically scan up to 3 folders upwards¹ so that you can double-click it and it'll just work 🙂

See this screenshot where I try to inspect the underside of the bed closely:

Could the near plane be moved closer to the camera? This issue isn't as noticeable in the Flying World scene, but it's quite noticeable in the other two.

Hi,

I'm getting a weird issue that looks almost like some kind of shadow map self-bias error on some materials. The scene I'm testing with is originally converted from UE, and it has been working fine with my other test projects.

Since it renders mostly OK, I assume there's something wrong in rendering certain material types? (I'm guessing reflections?)

Uploaded a video clip here:

https://drive.google.com/file/d/1CoWiPlcMJjRl91HPn2X--ldzToypptUN/view