This year, I did each day in a different programming model targeting GPU execution!
| Day | Programming Model | Run Command |
|---|---|---|
| 01 | CuTe | uv run python day01.py |
| 02 | cupy | uv run python day02.py |
| 03 | warp | uv run python day03.py |
| 04 | cudf | uv run --with "cudf-cu13=25.10.*" python day04.py |
| 05 | cuda.cccl | uv run python day05.py |
| 06 | numba-cuda | uv run python day06.py |
| 07 | C++ Standard Parallelism | ./build/day07 (after building) |
| 08 | CUDA with CCCL | ./build/day08 (after building) |
| 09 | OpenMP | ./build/day09 (after building) |
| 10 | CUDA | ./build/day10 (after building) |
| 11 | C++ std::execution | ./build/day11 (after building) |
| 12 | OpenACC | ./build/day12 (after building) |
I also did Day 04 in OpenACC and Day 12 in cudf.
uvwith Python >= 3.13 installed- CUDA >= 13.0
- NVIDIA HPC SDK >= 2025.09
- NVIDIA GPU with driver supporting CUDA 13
- CMake >= 4.0
Run/modify ./build_all.sh as you see fit:
./build_all.shAssuming you have all prerequisites in place and you already compiled the C++ solutions, simply invoke:
./run_all.sh