OptixCSP

OptixCSP is a modern, cross-platform (Windows and Linux) GPU‑accelerated ray‑tracing middleware library for modeling and simulating optical systems in engineering applications, such as Concentrated Solar Power (CSP). It leverages NVIDIA OptiX and can be used as a stand‑alone library or integrated into CSP workflows (e.g., SolTrace) to significantly accelerate ray‑tracing tasks.

OptixCSP focuses on accuracy, performance, and extensibility so you can:

• Rapidly simulate optical behavior in CSP systems: heliostats, receivers, sun modeling, optical properties, etc.
• Leverage GPU acceleration for complex scenes, large‑scale simulations, and dynamic/annual analyses.
• Integrate OptixCSP into existing CSP software as a high‑performance backend.
• Build standalone ray‑tracing applications using the OptixCSP core.

Prerequisites

Hardware & Drivers

• NVIDIA GPU, preferrably with RT cores, has been tested on RTX3050, RTX4070Ti and H100.
• NVIDIA driver, compatible with your target OptiX & CUDA toolkit.

Software

• CUDA Toolkit: 12.0 or newer.
• NVIDIA OptiX SDK: 8.x or newer.
• CMake: 3.18 or higher.
• C++ Compiler: C++17 or newer.

Note: The runtime (libnvoptix.so on Linux, nvoptix.dll on Windows) is provided by the NVIDIA driver. The OptiX SDK provides headers only. Make sure the optix version is supported by the driver on your machine.

Verified build/test configurations

• Windows: MS Visual Studio 2022, CUDA 12.8, OptiX 9.0
• Windows: MS Visual Studio 2022, CUDA 12.8, OptiX 8.1
• Windows: MS Visual Studio 2022, CUDA 12.3, OptiX 8.1
• Linux (redhat 8.0): gcc 11.2, CUDA 12.3, Optix 8.0
• Linux (redhat 8.0): gcc 12.1, CUDA 12.3, Optix 8.0
• Linux (ubuntu 22.04): gcc 11.4, CUDA 12.8, Optix 9.0

Quick checks (Linux)

Use these to confirm the system is ready before building:

# 1) GPU & driver
nvidia-smi
# Look for your GPU and the CUDA Version (max toolkit version supported by the driver)

# 2) OptiX runtime library from the driver, you should see libnvoptix.so.1
ldconfig -p | grep nvoptix

# 3) CUDA compiler (if installed on the node)
nvcc --version

Installing the OptiX SDK

Download the OptiX SDK from NVIDIA (link here). If your NVIDIA driver does not support the latest OptiX, please download the compatible version from the legacy download page. For Windows, follow the instructions; on Linux, run the provided script.

chmod +x OptiX-SDK-8.0.x-linux64.sh
./OptiX-SDK-8.0.x-linux64.sh --skip-license --prefix="$HOME/Optix8.0"

This creates a tree like:

$HOME/Optix8.0/
  include/
  SDK/

Building OptixCSP

Clone the repo,

git clone [email protected]:uwsbel/optixCSP.git
cd optixCSP

Linux

# configure cmake
mkdir -p build
cd build
cmake \
  -DCMAKE_BUILD_TYPE=Release \
  -DOptiX_INCLUDE="$HOME/Optix8.0/include" \
  -DOptiX_INSTALL_DIR="$HOME/Optix8.0/"
  ..

# build
cmake --build . -j

Notes: On kestrel, consider loading CUDA Toolkit and gcc first before running cmake: module load cuda/12.4 gcc/12.

Windows

• In CMake GUI, set OptiX_INCLUDE to the Optix SDK's include/ folder (e.g., C:/ProgramData/NVIDIA Corporation/OptiX SDK 8.1.0/include).
• Generate for Visual Studio 2022, then build the Release and Debug configuration.

Running the demos

After building, demos/executables are placed under your build output (e.g., build/bin/). Run them from the command line or your IDE.

Using OptixCSP as a library

You can link OptixCSP to third‑party software such as SolTrace.

1. Build and install (or just build) OptixCSP per the instructions above. The build exports a package config file OptixCSPConfig.cmake in its build tree (e.g., build/cmake/).

2. In your project’s CMakeLists.txt, configure dependencies and find OptixCSP:

cmake_minimum_required(VERSION 3.18)
project(MyApp CXX)

find_package(CUDAToolkit REQUIRED)  # Provides CUDA::cuda_driver
find_package(OptiX REQUIRED) # Provides optix headers

# Tell CMake where to find OptixCSP's package config
set(OptixCSP_DIR "/path/to/OptixCSP/build/cmake")
find_package(OptixCSP REQUIRED)

add_executable(MyApp main.cpp)

target_link_libraries(MyApp
  PUBLIC
    OptixCSP::OptixCSP_core
    CUDA::cuda_driver
)

# If you need headers from the OptiX SDK directly:
# target_include_directories(MyApp PRIVATE "$ENV{HOME}/optix-8.0/include")