3152     __m128i c128, au, bu, sum;  in vrhaddq_s8()  local
3153     c128 = _mm_set1_epi8((int8_t)0x80); //-128  in vrhaddq_s8()
3154     au = _mm_sub_epi8(a, c128); //add 128  in vrhaddq_s8()
3155     bu = _mm_sub_epi8(b, c128); //add 128  in vrhaddq_s8()
3157     return _mm_add_epi8 (sum, c128); //sub 128  in vrhaddq_s8()
3815     __m128i b128, c128, res;  in vmla_s8()  local
3817     c128 = _MM_CVTEPI8_EPI16 (_pM128i(c)); // SSE 4.1 use low 64 bits  in vmla_s8()
3818     res = _mm_mullo_epi16 (c128, b128);  in vmla_s8()
3859     __m128i mask, b128, c128, res;  in vmla_u8()  local
3862     c128 = _MM_CVTEPU8_EPI16 (_pM128i(c)); // SSE 4.1 use low 64 bits  in vmla_u8()
3863     res = _mm_mullo_epi16 (c128, b128);  in vmla_u8()
4781     __m128i c128, au,bu;  in vhsubq_s8()  local
4782     c128 = _mm_set1_epi8((int8_t)128);  in vhsubq_s8()
4783     au = _mm_add_epi8( a, c128);  in vhsubq_s8()
4784     bu = _mm_add_epi8( b, c128);  in vhsubq_s8()
13196     __m128i c7, maskgt, sh, c128;  in vtbx1_u8()  local
13198     c128 = _pM128i(c);  in vtbx1_u8()
13199     maskgt = _mm_cmpgt_epi8(c128,c7);  in vtbx1_u8()
13201     sh = _mm_shuffle_epi8(_pM128i(b),c128);  in vtbx1_u8()
13217     __m128i c15, b01, maskgt15, sh, c128;  in vtbx2_u8()  local
13219     c128 = _pM128i(c);  in vtbx2_u8()
13220     maskgt15 = _mm_cmpgt_epi8(c128, c15);  in vtbx2_u8()
13223     sh =  _mm_shuffle_epi8(b01, c128);  in vtbx2_u8()
13240     __m128i c15, c23, maskgt15, maskgt23, sh0, sh1, b01, c128;  in vtbx3_u8()  local
13243     c128 = _pM128i(c);  in vtbx3_u8()
13244     maskgt15 = _mm_cmpgt_epi8(c128,c15);  in vtbx3_u8()
13245     maskgt23 = _mm_cmpgt_epi8(c128,c23);  in vtbx3_u8()
13248     sh0 =  _mm_shuffle_epi8(b01, c128);  in vtbx3_u8()
13249     sh1 =  _mm_shuffle_epi8(_pM128i(b.val[2]), c128); //for bi>15 bi is wrapped (bi-=15)  in vtbx3_u8()
13267     __m128i c15, c31, maskgt15, maskgt31, sh0, sh1, b01, b23, c128;  in vtbx4_u8()  local
13270     c128 = _pM128i(c);  in vtbx4_u8()
13271     maskgt15 = _mm_cmpgt_epi8(c128,c15);  in vtbx4_u8()
13272     maskgt31 = _mm_cmpgt_epi8(c128,c31);  in vtbx4_u8()
13277     sh0 =  _mm_shuffle_epi8(b01, c128);  in vtbx4_u8()
13278     sh1 =  _mm_shuffle_epi8(b23, c128); //for bi>15 bi is wrapped (bi-=15)  in vtbx4_u8()