|
1 | 1 | # RTXpy |
2 | 2 |
|
3 | | -Ray tracing using CUDA accessible from Python. |
| 3 | +Ray tracing using CUDA, accessible from Python. |
4 | 4 |
|
5 | | -Requires CMake 3.10 or higher, either system CMake or can |
| 5 | +## Hardware requirements |
6 | 6 |
|
7 | | - pip install cmake |
| 7 | + * Nvidia Maxwell GPU or newer |
| 8 | + * Nvidia driver version: |
| 9 | + * 456.71 or newer for Windows |
| 10 | + * 455.28 or newer for Linux |
8 | 11 |
|
9 | | -To install |
| 12 | +## Installation |
10 | 13 |
|
11 | | - pip install -ve . |
| 14 | + pip install rtxpy |
12 | 15 |
|
13 | | -Optional runtime dependencies |
| 16 | +## Installation from source |
14 | 17 |
|
15 | | - cupy |
| 18 | +Requires CMake 3.10 or higher, either system CMake or |
16 | 19 |
|
17 | | -If you know the version of the CUDA toolkit that you have installed, which can |
18 | | -be obtained by running `nvcc --version`, you can install the appropriate `cupy` |
19 | | -wheel using |
| 20 | + pip install cmake |
20 | 21 |
|
21 | | - pip install cupy-cuda101 |
| 22 | +To install RTXpy from source use |
22 | 23 |
|
23 | | -which is for CUDA toolkit 10.1 |
| 24 | + pip install -ve . |
| 25 | + |
| 26 | +`cupy` is an optional runtime dependency. If you know the version of the CUDA |
| 27 | +toolkit you have installed, which can be obtained by running `nvcc --version`, |
| 28 | +you can install the appropriate `cupy` wheel. For example, for CUDA toolkit |
| 29 | +10.1 use |
| 30 | + |
| 31 | + pip install cupy-cuda101 |
24 | 32 |
|
25 | 33 | To run tests |
26 | 34 |
|
27 | 35 | pip install -ve .[tests] |
28 | 36 | pytest -v rtxpy/tests |
29 | | - |
30 | | -**Requires NVIDIA R456.71 driver or newer for Windows and 455.28 or newer for Linux.** |
31 | | -**Requires an NVidia Maxwell GPU or newer** |
|
0 commit comments