ximgproc: optimize add_mul using NEON intrinsics for ARM64

Author: pratham-mcw

modules/ximgproc/src/edgeaware_filters_common.cpp

@@ -60,6 +60,19 @@ inline bool CPU_SUPPORT_SSE1()
 }  // end
 #endif
 
+#if CV_NEON
+namespace
+{
+
+inline bool CPU_SUPPORT_NEON()
+{
+    static const bool is_supported = cv::checkHardwareSupport(CV_CPU_NEON);
+    return is_supported;
+}
+
+}  // end
+#endif
+
 namespace cv
 {
 namespace ximgproc
@@ -288,6 +301,20 @@ void add_mul(float *dst, float *src1, float *src2, int w)
             _mm_storeu_ps(dst + j, c);
         }
     }
+#elif CV_NEON
+    if (CPU_SUPPORT_NEON())
+    {
+        float32x4_t a, b, c;
+        for (; j < w - 3; j += 4)
+        {
+            a = vld1q_f32(src1 + j);
+            b = vld1q_f32(src2 + j);
+            b = vmulq_f32(b, a);
+            c = vld1q_f32(dst + j);
+            c = vaddq_f32(c, b);
+            vst1q_f32(dst + j, c);
+        }
+    }
 #endif
     for (; j < w; j++)
     {