//#define SH(_R, _C) sh[(_R)][(_C)] #define SH(_R, _C) sh[(_C)][(_R)] #define SVW(_R, _C) svw[(_R)][(_C)] //#define SVW(_R, _C) svw[(_C)][(_R)] KERNEL UpdateW_AD(cudafloat * W, cudafloat * H, cudafloat * V, cudafloat * WH, cudafloat * sumH, int n, int m, int r) { __shared__ cudafloat SH(32, 32); __shared__ cudafloat SVW(32, 32);