Merge pull request #1898 from tompollok:3.4

Author: alalek

modules/optflow/src/dis_flow.cpp

@@ -41,8 +41,10 @@
 //M*/
 
 #include "precomp.hpp"
+#include "opencv2/core/hal/intrin.hpp"
 #include "opencl_kernels_optflow.hpp"
 
+
 using namespace std;
 #define EPS 0.001F
 #define INF 1E+10F
@@ -519,7 +521,7 @@ inline float processPatch(float &dst_dUx, float &dst_dUy, uchar *I0_ptr, uchar *
                           int I0_stride, int I1_stride, float w00, float w01, float w10, float w11, int patch_sz)
 {
     float SSD = 0.0f;
-#ifdef CV_SIMD128
+#if CV_SIMD128
     if (patch_sz == 8)
     {
         /* Variables to accumulate the sums */
@@ -571,7 +573,7 @@ inline float processPatch(float &dst_dUx, float &dst_dUy, uchar *I0_ptr, uchar *
                 dst_dUx += diff * I0x_ptr[i * I0_stride + j];
                 dst_dUy += diff * I0y_ptr[i * I0_stride + j];
             }
-#ifdef CV_SIMD128
+#if CV_SIMD128
     }
 #endif
     return SSD;
@@ -588,7 +590,7 @@ inline float processPatchMeanNorm(float &dst_dUx, float &dst_dUy, uchar *I0_ptr,
     float sum_I0x_mul = 0.0, sum_I0y_mul = 0.0;
     float n = (float)patch_sz * patch_sz;
 
-#ifdef CV_SIMD128
+#if CV_SIMD128
     if (patch_sz == 8)
     {
         /* Variables to accumulate the sums */
@@ -643,7 +645,7 @@ inline float processPatchMeanNorm(float &dst_dUx, float &dst_dUy, uchar *I0_ptr,
                 sum_I0x_mul += diff * I0x_ptr[i * I0_stride + j];
                 sum_I0y_mul += diff * I0y_ptr[i * I0_stride + j];
             }
-#ifdef CV_SIMD128
+#if CV_SIMD128
     }
 #endif
     dst_dUx = sum_I0x_mul - sum_diff * x_grad_sum / n;
@@ -656,7 +658,7 @@ inline float computeSSD(uchar *I0_ptr, uchar *I1_ptr, int I0_stride, int I1_stri
                         float w11, int patch_sz)
 {
     float SSD = 0.0f;
-#ifdef CV_SIMD128
+#if CV_SIMD128
     if (patch_sz == 8)
     {
         v_float32x4 SSD_vec = v_setall_f32(0);
@@ -681,7 +683,7 @@ inline float computeSSD(uchar *I0_ptr, uchar *I1_ptr, int I0_stride, int I1_stri
                        I0_ptr[i * I0_stride + j];
                 SSD += diff * diff;
             }
-#ifdef CV_SIMD128
+#if CV_SIMD128
     }
 #endif
     return SSD;
@@ -693,7 +695,7 @@ inline float computeSSDMeanNorm(uchar *I0_ptr, uchar *I1_ptr, int I0_stride, int
 {
     float sum_diff = 0.0f, sum_diff_sq = 0.0f;
     float n = (float)patch_sz * patch_sz;
-#ifdef CV_SIMD128
+#if CV_SIMD128
     if (patch_sz == 8)
     {
         v_float32x4 sum_diff_vec = v_setall_f32(0);
@@ -723,7 +725,7 @@ inline float computeSSDMeanNorm(uchar *I0_ptr, uchar *I1_ptr, int I0_stride, int
                 sum_diff += diff;
                 sum_diff_sq += diff * diff;
             }
-#ifdef CV_SIMD128
+#if CV_SIMD128
     }
 #endif
     return sum_diff_sq - sum_diff * sum_diff / n;

modules/optflow/src/variational_refinement.cpp

@@ -597,7 +597,7 @@ void VariationalRefinementImpl::ComputeDataTerm_ParBody::operator()(const Range
 #undef INIT_ROW_POINTERS
 
         int j = 0;
-#ifdef CV_SIMD128
+#if CV_SIMD128
         v_float32x4 zeta_vec = v_setall_f32(zeta_squared);
         v_float32x4 eps_vec = v_setall_f32(epsilon_squared);
         v_float32x4 delta_vec = v_setall_f32(delta2);
@@ -803,7 +803,7 @@ void VariationalRefinementImpl::ComputeSmoothnessTermHorPass_ParBody::operator()
     pA_v_next[j] += pWeight[j];
 
         int j = 0;
-#ifdef CV_SIMD128
+#if CV_SIMD128
         v_float32x4 alpha2_vec = v_setall_f32(alpha2);
         v_float32x4 eps_vec = v_setall_f32(epsilon_squared);
         v_float32x4 cW_u_vec, cW_v_vec;
@@ -913,7 +913,7 @@ void VariationalRefinementImpl::ComputeSmoothnessTermVertPass_ParBody::operator(
 #undef INIT_ROW_POINTERS
 
         int j = 0;
-#ifdef CV_SIMD128
+#if CV_SIMD128
         v_float32x4 pWeight_vec, uy_vec, vy_vec;
         for (; j < len - 3; j += 4)
         {
@@ -1015,7 +1015,7 @@ void VariationalRefinementImpl::RedBlackSOR_ParBody::operator()(const Range &ran
 #undef INIT_ROW_POINTERS
 
         j = 0;
-#ifdef CV_SIMD128
+#if CV_SIMD128
         v_float32x4 pW_prev_vec = v_setall_f32(pW_next[-1]);
         v_float32x4 pdu_prev_vec = v_setall_f32(pdu_next[-1]);
         v_float32x4 pdv_prev_vec = v_setall_f32(pdv_next[-1]);