Try to use avx_hs() - slow and requires alignment 4096 bits < (l.size*l.siz...

AlexeyAB

2018-08-08 a284a7da8d1facbf984a22302665a2a50295a687

 src/dropout_layer_kernels.cu

@@ -1,8 +1,11 @@
#include "cuda_runtime.h"
#include "curand.h"
#include "cublas_v2.h"

extern "C" {
#include "dropout_layer.h"
#include "cuda.h"
#include "utils.h"
#include "params.h"
}

__global__ void yoloswag420blazeit360noscope(float *input, int size, float *rand, float prob, float scale)