diff --git a/kernels/volk/volk_8u_x4_conv_k7_r2_8u.h b/kernels/volk/volk_8u_x4_conv_k7_r2_8u.h
index 33ab9676..53b366c1 100644
--- a/kernels/volk/volk_8u_x4_conv_k7_r2_8u.h
+++ b/kernels/volk/volk_8u_x4_conv_k7_r2_8u.h
@@ -669,18 +669,18 @@ static inline void volk_8u_x4_conv_k7_r2_8u_neonspiral(unsigned char* Y,
         high_bits = vreinterpretq_u16_u8(vshrq_n_u8(d9, 7));
         paired16 = vreinterpretq_u32_u16(vsraq_n_u16(high_bits, high_bits, 6));
         paired32_new = vreinterpretq_u8_u32(vsraq_n_u32(paired16, paired16, 12));
-        s20 = ((unsigned int)vgetq_lane_u8(paired32_new, 0) << 17)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 4) << 25)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 8) << 1)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 12) << 9);
+        s20 = ((unsigned int)vgetq_lane_u8(paired32_new, 0) << 9)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 4) << 1)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 8) << 25)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 12) << 17);
 printf("%08x\n", s20);
         high_bits = vreinterpretq_u16_u8(vshrq_n_u8(d10, 7));
         paired16 = vreinterpretq_u32_u16(vsraq_n_u16(high_bits, high_bits, 6));
         paired32_new = vreinterpretq_u8_u32(vsraq_n_u32(paired16, paired16, 12));
-        s20 |= ((unsigned int)vgetq_lane_u8(paired32_new, 0) << 16)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 4) << 24)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 8) << 0)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 12) << 8);
+        s20 |= ((unsigned int)vgetq_lane_u8(paired32_new, 0) << 8)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 4) << 0)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 8) << 24)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 12) << 16);
 printf("%08x\n", s20);
         a91 = ((unsigned int*)dec);
         a92 = (4 * i9);
@@ -717,17 +717,17 @@ printf("%08x\n", s20);
         high_bits = vreinterpretq_u16_u8(vshrq_n_u8(d11, 7));
         paired16 = vreinterpretq_u32_u16(vsraq_n_u16(high_bits, high_bits, 6));
         paired32_new = vreinterpretq_u8_u32(vsraq_n_u32(paired16, paired16, 12));
-        s26 = ((unsigned int)vgetq_lane_u8(paired32_new, 0) << 17)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 4) << 25)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 8) << 1)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 12) << 9);
+        s26 = ((unsigned int)vgetq_lane_u8(paired32_new, 0) << 9)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 4) << 1)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 8) << 25)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 12) << 17);
         high_bits = vreinterpretq_u16_u8(vshrq_n_u8(d12, 7));
         paired16 = vreinterpretq_u32_u16(vsraq_n_u16(high_bits, high_bits, 6));
         paired32_new = vreinterpretq_u8_u32(vsraq_n_u32(paired16, paired16, 12));
-        s26 |= ((unsigned int)vgetq_lane_u8(paired32_new, 0) << 16)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 4) << 24)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 8) << 0)
-            | ((unsigned int)vgetq_lane_u8(paired32_new, 12) << 8);
+        s26 |= ((unsigned int)vgetq_lane_u8(paired32_new, 0) << 8)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 4) << 0)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 8) << 24)
+            | ((unsigned int)vgetq_lane_u8(paired32_new, 12) << 16);
         a110 = (a93 + 1);
         *(a110) = s26;
         s28 = vzip1q_u8(a108, a109);