2337    __m128i row0, row1, row2, row3, row4, row5, row6, row7;  in stbi__idct_simd()  local
2395          __m128i sum04 = _mm_add_epi16(row0, row4); \  in stbi__idct_simd()
2396          __m128i dif04 = _mm_sub_epi16(row0, row4); \  in stbi__idct_simd()
2416          dct_bfly32o(row3,row4, x3,x4,bias,shift); \  in stbi__idct_simd()
2437    row4 = _mm_load_si128((const __m128i *) (data + 4*8));  in stbi__idct_simd()
2447       dct_interleave16(row0, row4);  in stbi__idct_simd()
2455       dct_interleave16(row4, row6);  in stbi__idct_simd()
2461       dct_interleave16(row4, row5);  in stbi__idct_simd()
2472       __m128i p2 = _mm_packus_epi16(row4, row5);  in stbi__idct_simd()
2517    int16x8_t row0, row1, row2, row3, row4, row5, row6, row7;  in stbi__idct_simd()  local
2570       int16x8_t sum04 = vaddq_s16(row0, row4); \  in stbi__idct_simd()
2571       int16x8_t dif04 = vsubq_s16(row0, row4); \  in stbi__idct_simd()
2600       dct_bfly32o(row3,row4, x3,x4,shiftop,shift); \  in stbi__idct_simd()
2608    row4 = vld1q_s16(data + 4*8);  in stbi__idct_simd()
2630       dct_trn16(row4, row5);  in stbi__idct_simd()
2636       dct_trn32(row4, row6);  in stbi__idct_simd()
2640       dct_trn64(row0, row4); // a0b0c0d0e0f0g0h0  in stbi__idct_simd()
2662       uint8x8_t p4 = vqrshrun_n_s16(row4, 1);  in stbi__idct_simd()