Closed dacorvo closed 2 weeks ago
The __CUDA_ARCH__ preprocessor variable is not exported when compiling CUDA code for the host, but only for the device. This means that we need another preprocessor variable to decide whether we compile the AWQ kernels.
The __CUDA_ARCH__ preprocessor variable is not exported when compiling CUDA code for the host, but only for the device. This means that we need another preprocessor variable to decide whether we compile the AWQ kernels.