393 static WEBP_INLINE int16x8_t ConvertU8ToS16(uint32x2_t v) {  in ConvertU8ToS16()
400                                             const int16x8_t dst01,  in SaturateAndStore4x4()
401                                             const int16x8_t dst23) {  in SaturateAndStore4x4()
413 static WEBP_INLINE void Add4x4(const int16x8_t row01, const int16x8_t row23,  in Add4x4()
426     const int16x8_t dst01_s16 = ConvertU8ToS16(dst01);  in Add4x4()
427     const int16x8_t dst23_s16 = ConvertU8ToS16(dst23);  in Add4x4()
430     const int16x8_t out01 = vrsraq_n_s16(dst01_s16, row01, 3);  in Add4x4()
431     const int16x8_t out23 = vrsraq_n_s16(dst23_s16, row23, 3);  in Add4x4()
750   const int16x8_t kCst63 = vdupq_n_s16(63);  in ApplyFilter6()
756   const int16x8_t s1_lo = vmlal_s8(kCst63, kCst27, delta_lo);  // 63 + 27 * a  in ApplyFilter6()
757   const int16x8_t s1_hi = vmlal_s8(kCst63, kCst27, delta_hi);  // 63 + 27 * a  in ApplyFilter6()
758   const int16x8_t s2_lo = vmlal_s8(kCst63, kCst18, delta_lo);  // 63 + 18 * a  in ApplyFilter6()
759   const int16x8_t s2_hi = vmlal_s8(kCst63, kCst18, delta_hi);  // 63 + 18 * a  in ApplyFilter6()
760   const int16x8_t s3_lo = vmlal_s8(kCst63, kCst9, delta_lo);   // 63 + 9 * a  in ApplyFilter6()
761   const int16x8_t s3_hi = vmlal_s8(kCst63, kCst9, delta_hi);   // 63 + 9 * a  in ApplyFilter6()
990 static WEBP_INLINE void Transpose8x2(const int16x8_t in0, const int16x8_t in1,  in Transpose8x2()
1003   const int16x8_t B1 =  in TransformPass()
1007   const int16x8_t C0 = vsraq_n_s16(B1, vqdmulhq_n_s16(B1, kC1), 1);  in TransformPass()
1008   const int16x8_t C1 = vqdmulhq_n_s16(B1, kC2);  in TransformPass()
1017   const int16x8_t D0 = vcombine_s16(a, b);      // D0 = a | b  in TransformPass()
1018   const int16x8_t D1 = vcombine_s16(d, c);      // D1 = d | c  in TransformPass()
1019   const int16x8_t E0 = vqaddq_s16(D0, D1);      // a+d | b+c  in TransformPass()
1020   const int16x8_t E_tmp = vqsubq_s16(D0, D1);   // a-d | b-c  in TransformPass()
1021   const int16x8_t E1 = vcombine_s16(vget_high_s16(E_tmp), vget_low_s16(E_tmp));  in TransformPass()
1176   const int16x8_t DC = vdupq_n_s16(in[0]);  in TransformDC()
1255   const int16x8_t m0_m1 = vcombine_s16(vqadd_s16(B, d4), vqadd_s16(B, c4));  in TransformAC3()
1256   const int16x8_t m2_m3 = vcombine_s16(vqsub_s16(B, c4), vqsub_s16(B, d4));  in TransformAC3()