arhix52/strelka

Realtime render for USD

Strelka

Path tracing render based on NVIDIA OptiX + NVIDIA MDL and Apple Metal

OpenUSD Hydra render delegate

Basis curves support

Project Dependencies

OpenUSD https://github.com/PixarAnimationStudios/OpenUSD

• Set evn var: USD_DIR=c:\work\USD_build

OptiX

• Set evn var: OPTIX_DIR=C:\work\OptiX SDK 8.0.0

Download MDL sdk (for example: mdl-sdk-367100.2992): https://developer.nvidia.com/nvidia-mdl-sdk-get-started

• unzip content to /external/mdl-sdk/

LLVM 12.0.1 (https://github.com/llvm/llvm-project/releases/tag/llvmorg-12.0.1) for MDL ptx code generator

• for win: https://github.com/llvm/llvm-project/releases/download/llvmorg-12.0.1/LLVM-12.0.1-win64.exe
• for linux: https://github.com/llvm/llvm-project/releases/download/llvmorg-12.0.1/clang+llvm-12.0.1-x86_64-linux-gnu-ubuntu-16.04.tar.xz
• install it to c:\work for example
• add to PATH: c:\work\LLVM\bin
• extract 2 header files files from external/clang12_patched to C:\work\LLVM\lib\clang\12.0.1\include

Strelka uses conan https://conan.io/

• install conan: pip install conan
• install ninja [https://ninja-build.org/] build system: sudo apt install ninja-build

detect conan profile: conan profile detect --force

1. conan install . --build=missing --settings=build_type=Debug
2. cd build
3. cmake .. -G "Visual Studio 17 2022" -DCMAKE_TOOLCHAIN_FILE=generators\conan_toolchain.cmake
4. cmake --build . --config Debug

On Mac/Linux:

1. conan install . -c tools.cmake.cmaketoolchain:generator=Ninja -c tools.system.package_manager:mode=install -c tools.system.package_manager:sudo=True --build=missing --settings=build_type=Debug
2. cd build/Debug
3. source ./generators/conanbuild.sh
4. cmake ../.. -DCMAKE_TOOLCHAIN_FILE=generators/conan_toolchain.cmake -DCMAKE_BUILD_TYPE=Debug
5. cmake --build .

Installation

Launch

Synopsis

Strelka -s <USD Scene path> [OPTION...] positional parameters

-s, --scene arg       scene path (default: "")
-i, --iteration arg  Iteration to capture (default: -1)
-h, --help            Print usage

To set log level use

`export SPDLOG_LEVEL=debug`

The available log levels are: trace, debug, info, warn, and err.

Example

./Strelka -s misc/coffeemaker.usdc -i 100

USD

USD env:
    export USD_DIR=/Users/<user>/work/usd_build/
    export PATH=/Users/<user>/work/usd_build/bin:$PATH
    export PYTHONPATH=/Users/<user>/work/usd_build/lib/python:$PYTHONPATH

Install plugin:
    cmake --install . --component HdStrelka

License

• USD plugin design and material translation code based on Pablo Gatling code: https://github.com/pablode/gatling

strelka's Issues

Port codebase on linux

• WSL build
• conan packages

Camera speed control

camera speed as number input for current scene in ui, default 1

Select light non uniformly

light selection pdf proportional to light power, area so on

Change default textures result in MDL init

Save render result to file

save render output buffer to file (png format)

Bug: curves phantom points rendering

ImGuI Integration

Enable ImGuI in render.

Stop criteria for PathTracing on reaching max samples count

Set Total Samples Per Pixel (SPP) using console arg and from UI - default 64
create setting spp for single path tracing subframe - default 1 (console arg + ui)
On reaching criteria stop PT and present accumulated image. Invalidate criteria on camera changes and setting changes.

Fix: Resizing window

Add Tangent calculation

calculate tangent vector on geometry loading

Tonemapping

Convert tonemapper kernels to cuda (reinhard, aces, filmic)
expose parameters to UI

Rect light sampling

Camera movement is broken

try to look using mouse on different scenes

Bug: light occlusion and offset in OptiX code path

Relative path for OptiX precompiled file

Textures in materials

Textures UV coordinates for curves

Depth of Field

https://forums.developer.nvidia.com/t/dof-lens-shader/78089/2

OptiX: add albedo outputs: ldr, hrd

output RGB image, LDR - after tonemapping values range: 0.0-1.0
HDR - 0.0 - 10000.0 no tone-mapping or gamma correction should be performed on HDR

OptiX: add compaction for Acceleration Structures

OptiX: add validation/performance mode

OptiX: add normal output

Remove GLM from cuda code

OptiX denoiser

https://raytracing-docs.nvidia.com/optix7/guide/index.html#ai_denoiser#nvidia-ai-denoiser