GitHub - zkh2016/sgemm: Cuda-based matrix multiplication, compared with cuBLas performance. Refer to the https://github.com/NervanaSystems/maxas/wiki/SGEMM