Support yolov4

Signed-off-by: Micaela Verucchi <micaelaverucchi@gmail.com>

src/kernels/activation.mish.cu

@@ -0,0 +1,31 @@
+#include "kernels.h"
+#include <math.h>
+
+#define MISH_THRESHOLD 20
+
+__device__ float tanh_activate_kernel(float x){return (2/(1 + expf(-2*x)) - 1);}
+__device__ float softplus_kernel(float x, float threshold = 20) {
+    if (x > threshold) return x;                // too large
+    else if (x < -threshold) return expf(x);    // too small
+    return logf(expf(x) + 1);
+}
+
+// https://github.com/digantamisra98/Mish
+// https://github.com/AlexeyAB/darknet/blob/master/src/activation_kernels.cu
+__global__
+void activation_mish(dnnType *input, dnnType *output, int size) {
+    int i = (blockIdx.x + blockIdx.y*gridDim.x) * blockDim.x + threadIdx.x;
+    if (i < size) 
+        output[i] = input[i] * tanh_activate_kernel( softplus_kernel(input[i], MISH_THRESHOLD));    
+}
+
+/**
+    Mish activation function
+*/
+void activationMishForward(dnnType* srcData, dnnType* dstData, int size, cudaStream_t stream)
+{
+    int blocks = (size+255)/256;
+    int threads = 256;
+    
+    activation_mish<<<blocks, threads, 0, stream>>>(srcData, dstData, size);
+}

src/kernels/scaladd.cu

@@ -0,0 +1,16 @@
+#include "kernels.h"
+#include <math.h>
+
+__global__ void scal_add_kernel(dnnType* dstData, int size, float alpha, float beta, int inc)
+{
+    int i = (blockIdx.x + blockIdx.y*gridDim.x) * blockDim.x + threadIdx.x;
+    if (i < size) dstData[i*inc] = dstData[i*inc] * alpha + beta;
+}
+
+void scalAdd(dnnType* dstData, int size, float alpha, float beta, int inc, cudaStream_t stream)
+{
+    int blocks = (size+255)/256;
+    int threads = 256;
+    
+    scal_add_kernel<<<blocks, threads, 0, stream>>>(dstData, size, alpha, beta, inc);
+}