How to set the CMAKE_CUDA_ARCHITECTURES value? nvcc fatal : Option '--ptx (-ptx)' is not allowed when compiling for multiple GPU architectures about pbrt-v4 HOT 3 CLOSED

Try without the "sm_" prefix, i.e.

if(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
   set(CMAKE_CUDA_ARCHITECTURES 61)
endif(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)

Alternatively, if you are using CMake >= 3.20, you could define the environment variable before configuring the project with CMake CUDAARCHS=61, or passing -DCMAKE_CUDA_ARCHITURES=61 to the cmake command (if invoking on the command-line).

If you are using CMake 3.18, the best might be to call cmake_policy(SET CMP0104 NEW), so that the default behaviour is for CMAKE_CUDA_ARCHITURES to be initialised with the GPU architectures CUDA found in your computer.

Much Appreciated! It works now! Thank you sir!

from pbrt-v4.

My OS is UBUNTU 22.04.2, with cuda 12.2 and optix 7.7, and using a intel 6700k cpu and gtx 1070 gpu.

from pbrt-v4.

Try without the "sm_" prefix, i.e.

if(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
   set(CMAKE_CUDA_ARCHITECTURES 61)
endif(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)

Alternatively, if you are using CMake >= 3.20, you could define the environment variable before configuring the project with CMake CUDAARCHS=61, or passing -DCMAKE_CUDA_ARCHITURES=61 to the cmake command (if invoking on the command-line).

If you are using CMake 3.18, the best might be to call cmake_policy(SET CMP0104 NEW), so that the default behaviour is for CMAKE_CUDA_ARCHITURES to be initialised with the GPU architectures CUDA found in your computer.

from pbrt-v4.