diff --git a/src/UpperPHY/ChannelEstimation/arm_channel_estimation.cpp b/src/UpperPHY/ChannelEstimation/arm_channel_estimation.cpp
index e31b12fc85b19868b1dd4d916effc85dd089cef6..f4039fe705252aeff646c44ad9f6dc89eb6b3c78 100644
--- a/src/UpperPHY/ChannelEstimation/arm_channel_estimation.cpp
+++ b/src/UpperPHY/ChannelEstimation/arm_channel_estimation.cpp
@@ -361,26 +361,32 @@ lmmse_f32(armral_cmplx_f32_t *p_src_x, armral_cmplx_f32_t *p_src_y,
   return ARMRAL_SUCCESS;
 }
 
+// Interpolate values for the target symbol from pilot symbols
 void linear_interpolation_f32(armral_cmplx_f32_t *pilot_1,
                               armral_cmplx_f32_t *pilot_2,
                               armral_cmplx_f32_t *target_re, uint16_t rows,
                               uint16_t cols, uint16_t pilot_1_idx,
                               uint16_t pilot_2_idx, uint16_t target_idx) {
-  for (uint16_t i = 0; i < rows * cols; i++) {
-    float32_t slope_re =
-        (pilot_2[i].re - pilot_1[i].re) / (pilot_2_idx - pilot_1_idx);
-    float32_t slope_im =
-        (pilot_2[i].im - pilot_1[i].im) / (pilot_2_idx - pilot_1_idx);
-    target_re[i].re = pilot_1[i].re + slope_re * (target_idx - pilot_1_idx);
-    target_re[i].im = pilot_1[i].im + slope_im * (target_idx - pilot_1_idx);
+  float32_t m_denom = 1.0 / (pilot_2_idx - pilot_1_idx);
+  float32_t t_x = (target_idx - pilot_1_idx);
+  uint16_t idx = 0;
+
+  for (uint16_t i = (rows * cols); i > 0; i--) {
+    float32_t slope_re = (pilot_2[idx].re - pilot_1[idx].re) * m_denom;
+    float32_t slope_im = (pilot_2[idx].im - pilot_1[idx].im) * m_denom;
+    target_re[idx].re = pilot_1[idx].re + slope_re * t_x;
+    target_re[idx].im = pilot_1[idx].im + slope_im * t_x;
+    idx++;
   }
 }
 
 // If there is only one pilot symbol copy it to the rest of the REs
 void copy_pilot_f32(armral_cmplx_f32_t *pilot_re, armral_cmplx_f32_t *target_re,
                     uint16_t rows, uint16_t cols) {
-  for (uint16_t i = 0; i < rows * cols; i++) {
+  uint16_t i = 0;
+  for (uint16_t j = rows * cols; j > 0; j--) {
     target_re[i] = pilot_re[i];
+    i++;
   }
 }
 
@@ -393,11 +399,13 @@ void linear_interpolation_per_dimension_f32(
   uint16_t pilot_idx;
   uint32_t pilot1_re_idx = pilot_indices[0] * re_idx_stride + re_idx_offset;
   uint32_t pilot2_re_idx = pilot_indices[1] * re_idx_stride + re_idx_offset;
-  for (uint16_t idx = 0; idx < pilot_indices[0]; idx++) {
+  uint16_t idx = 0;
+  for (uint16_t j = pilot_indices[0]; j > 0; j--) {
     uint32_t re_idx = idx * re_idx_stride + re_idx_offset;
     linear_interpolation_f32(p_dst_h + pilot1_re_idx, p_dst_h + pilot2_re_idx,
                              p_dst_h + re_idx, rows, cols, pilot_indices[0],
                              pilot_indices[1], idx);
+    idx++;
   }
 
   // Interpolation between first and last pilot
@@ -405,8 +413,8 @@ void linear_interpolation_per_dimension_f32(
     pilot1_re_idx = pilot_indices[pilot_idx] * re_idx_stride + re_idx_offset;
     pilot2_re_idx =
         pilot_indices[pilot_idx + 1] * re_idx_stride + re_idx_offset;
-    for (uint16_t idx = pilot_indices[pilot_idx] + 1;
-         idx < pilot_indices[pilot_idx + 1]; idx++) {
+    for (idx = pilot_indices[pilot_idx] + 1; idx < pilot_indices[pilot_idx + 1];
+         idx++) {
       uint32_t re_idx = idx * re_idx_stride + re_idx_offset;
       linear_interpolation_f32(p_dst_h + pilot1_re_idx, p_dst_h + pilot2_re_idx,
                                p_dst_h + re_idx, rows, cols,
@@ -418,12 +426,15 @@ void linear_interpolation_per_dimension_f32(
   // Extrapolation after last pilots
   pilot1_re_idx = pilot_indices[pilot_idx - 1] * re_idx_stride + re_idx_offset;
   pilot2_re_idx = pilot_indices[pilot_idx] * re_idx_stride + re_idx_offset;
-  for (uint16_t idx = pilot_indices[pilot_idx] + 1; idx < max_idx; idx++) {
+  idx = pilot_indices[pilot_idx] + 1;
+  uint16_t loop_count = max_idx - (pilot_indices[pilot_idx] + 1);
+  for (uint16_t j = loop_count; j > 0; j--) {
     uint32_t re_idx = idx * re_idx_stride + re_idx_offset;
     linear_interpolation_f32(p_dst_h + pilot1_re_idx, p_dst_h + pilot2_re_idx,
                              p_dst_h + re_idx, rows, cols,
                              pilot_indices[pilot_idx - 1],
                              pilot_indices[pilot_idx], idx);
+    idx++;
   }
 }
 
@@ -436,15 +447,20 @@ void copy_pilot_per_dimension_f32(armral_cmplx_f32_t *p_dst_h,
   uint32_t pilot_re_idx = pilot_idx * re_idx_stride + re_idx_offset;
 
   // REs before the pilot
-  for (uint16_t idx = 0; idx < pilot_idx; idx++) {
+  uint16_t idx = 0;
+  for (uint16_t j = pilot_idx; j > 0; j--) {
     uint32_t re_idx = idx * re_idx_stride + re_idx_offset;
     copy_pilot_f32(p_dst_h + pilot_re_idx, p_dst_h + re_idx, rows, cols);
+    idx++;
   }
 
   // REs after the pilot
-  for (uint16_t idx = pilot_idx + 1; idx < max_idx; idx++) {
+  idx = pilot_idx + 1;
+  uint16_t loop_count1 = max_idx - pilot_idx - 1;
+  for (uint16_t j = loop_count1; j > 0; j--) {
     uint32_t re_idx = idx * re_idx_stride + re_idx_offset;
     copy_pilot_f32(p_dst_h + pilot_re_idx, p_dst_h + re_idx, rows, cols);
+    idx++;
   }
 }