Remove non-critical bound check for 8/16-bits operations

Author: junwha

src/layer/arm/shufflechannel_arm.cpp

@@ -374,7 +374,7 @@ int ShuffleChannel_arm::forward_bf16s_fp16s(const Mat& bottom_blob, Mat& top_blo
 
                 ptr1 += 4;
 
-                for (int i = 0; i < size - 1; i++)
+                for (int i = 0; i < size; i++)
                 {
                     uint16x4_t _p0 = vld1_u16(ptr0);
                     uint16x4_t _p1 = vld1_u16(ptr1);
@@ -388,21 +388,6 @@ int ShuffleChannel_arm::forward_bf16s_fp16s(const Mat& bottom_blob, Mat& top_blo
                     ptr1 += 8;
                     outptr0 += 8;
                 }
-
-                for (int i = 0; i < 8; i++)
-                {
-                    if (i % 2)
-                    {
-                        *outptr0 = *ptr1;
-                        ptr1 += 1;
-                    }
-                    else
-                    {
-                        *outptr0 = *ptr0;
-                        ptr0 += 1;
-                    }
-                    outptr0 += 1;
-                }
             }
 
             return 0;

src/layer/x86/shufflechannel_x86.cpp

@@ -116,7 +116,7 @@ int ShuffleChannel_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Opt
 
                 ptr1 += 8;
 
-                for (int i = 0; i < size - 1; i++)
+                for (int i = 0; i < size; i++)
                 {
                     __m256 _p0 = _mm256_loadu_ps(ptr0);
                     __m256 _p1 = _mm256_loadu_ps(ptr1);
@@ -134,21 +134,6 @@ int ShuffleChannel_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Opt
                     ptr1 += 16;
                     outptr += 16;
                 }
-
-                for (int i = 0; i < 16; i++)
-                {
-                    if (i % 2)
-                    {
-                        *outptr = *ptr1;
-                        ptr1 += 1;
-                    }
-                    else
-                    {
-                        *outptr = *ptr0;
-                        ptr0 += 1;
-                    }
-                    outptr += 1;
-                }
             }
 
             return 0;
@@ -387,7 +372,7 @@ int ShuffleChannel_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Opt
 
                 ptr1 += 4;
 
-                for (int i = 0; i < size - 1; i++)
+                for (int i = 0; i < size; i++)
                 {
                     __m128 _p0 = _mm_loadu_ps(ptr0);
                     __m128 _p1 = _mm_loadu_ps(ptr1);
@@ -402,21 +387,6 @@ int ShuffleChannel_x86::forward(const Mat& bottom_blob, Mat& top_blob, const Opt
                     ptr1 += 8;
                     outptr += 8;
                 }
-
-                for (int i = 0; i < 8; i++)
-                {
-                    if (i % 2)
-                    {
-                        *outptr = *ptr1;
-                        ptr1 += 1;
-                    }
-                    else
-                    {
-                        *outptr = *ptr0;
-                        ptr0 += 1;
-                    }
-                    outptr += 1;
-                }
             }
 
             return 0;