fix bus error in resize_bilinear_c2 on armv7

Author: nihui

src/mat_pixel_resize.cpp

@@ -392,7 +392,12 @@ void resize_bilinear_c2(const unsigned char* src, int srcw, int srch, unsigned c
                 int16x4_t _a0a1XX = vld1_s16(ialphap);
                 int16x4_t _a0a0a1a1 = vzip_s16(_a0a1XX, _a0a1XX).val[0];
                 uint8x8_t _S1 = uint8x8_t();
-                _S1 = vreinterpret_u8_u32(vld1_lane_u32((const unsigned int*)S1p, vreinterpret_u32_u8(_S1), 0));
+
+                _S1 = vld1_lane_u8(S1p, _S1, 0);
+                _S1 = vld1_lane_u8(S1p+1, _S1, 1);
+                _S1 = vld1_lane_u8(S1p+2, _S1, 2);
+                _S1 = vld1_lane_u8(S1p+3, _S1, 3);
+
                 int16x8_t _S116 = vreinterpretq_s16_u16(vmovl_u8(_S1));
                 int16x4_t _S1lowhigh = vget_low_s16(_S116);
                 int32x4_t _S1ma0a1 = vmull_s16(_S1lowhigh, _a0a0a1a1);
@@ -434,8 +439,17 @@ void resize_bilinear_c2(const unsigned char* src, int srcw, int srch, unsigned c
                 int16x4_t _a1 = vdup_n_s16(a1);
                 uint8x8_t _S0 = uint8x8_t();
                 uint8x8_t _S1 = uint8x8_t();
-                _S0 = vreinterpret_u8_u32(vld1_lane_u32((const unsigned int*)S0p, vreinterpret_u32_u8(_S0), 0));
-                _S1 = vreinterpret_u8_u32(vld1_lane_u32((const unsigned int*)S1p, vreinterpret_u32_u8(_S1), 0));
+
+                _S0 = vld1_lane_u8(S0p, _S0, 0);
+                _S0 = vld1_lane_u8(S0p+1, _S0, 1);
+                _S0 = vld1_lane_u8(S0p+2, _S0, 2);
+                _S0 = vld1_lane_u8(S0p+3, _S0, 3);
+
+                _S1 = vld1_lane_u8(S1p, _S1, 0);
+                _S1 = vld1_lane_u8(S1p+1, _S1, 1);
+                _S1 = vld1_lane_u8(S1p+2, _S1, 2);
+                _S1 = vld1_lane_u8(S1p+3, _S1, 3);
+
                 int16x8_t _S016 = vreinterpretq_s16_u16(vmovl_u8(_S0));
                 int16x8_t _S116 = vreinterpretq_s16_u16(vmovl_u8(_S1));
                 int16x4_t _S0lowhigh = vget_low_s16(_S016);