Skip to content

Instantly share code, notes, and snippets.

View jsteube's full-sized avatar

Jens Steube jsteube

View GitHub Profile
@jsteube
jsteube / 3k.cu
Created February 6, 2020 10:17
NVCC issue POC
// I made this source code to demonstrate an issue which affects the compile process of CUDA kernels.
//
// - All GPU architectures are affected
// - Tested on CUDA SDK 10.x and 9.x
// - Tested on Ubuntu 18.04 LTS and Windows 10
__device__ void s (int *in, int *out);
__device__ void t (int *w0, int *w1, int *w2, int *w3, int *h);
// I created three artificial kernels. They all consist of 100% the same code.