invalid configuration argument in top_k_gpu.cu at line 418

[minghust]

Hi,
When I run `run.sh', it shows "invalid configuration argument in top_k_gpu.cu at line 418".

my cuda version: cuda 9.0

Do you have any similar error? Thanks :)

[yuxianzhi]

413    if(shared_mem_usage < GPU_SHARED_MEM_THRESHOLD)
414        top_k_gpu_kernel3_2<<<1, threads, shared_mem_usage>>>(input, blocks, stride, k, output);
415    else
416        top_k_gpu_kernel3_2_orig<<<1, threads>>>(input, blocks, stride, k, output);
417    cudaError_t err = cudaGetLastError();
418    HANDLE_CUDA_ERROR(err);

You can print shared_mem_usage and threads to make sure that they are valid. (shared_mem_usage < ?KB and threads < 1024)

[minghust]

thx :)