124         const int8x8_t va0x1 = vld1_s8(a0); a0 += 8;  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()  local
151         vprod0x0 = vmlal_s8(vprod0x0, vb0x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
161         vprod0x1 = vmlal_s8(vprod0x1, vb1x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
171         vprod0x2 = vmlal_s8(vprod0x2, vb2x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
181         vprod0x3 = vmlal_s8(vprod0x3, vb3x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
191         vprod0x4 = vmlal_s8(vprod0x4, vb4x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
201         vprod0x5 = vmlal_s8(vprod0x5, vb5x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
211         vprod0x6 = vmlal_s8(vprod0x6, vb6x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
221         vprod0x7 = vmlal_s8(vprod0x7, vb7x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
231         vprod0x8 = vmlal_s8(vprod0x8, vb8x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
241         vprod0x9 = vmlal_s8(vprod0x9, vb9x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
251         vprod0x10 = vmlal_s8(vprod0x10, vb10x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
261         vprod0x11 = vmlal_s8(vprod0x11, vb11x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
271         vprod0x12 = vmlal_s8(vprod0x12, vb12x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
281         vprod0x13 = vmlal_s8(vprod0x13, vb13x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
291         vprod0x14 = vmlal_s8(vprod0x14, vb14x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
301         vprod0x15 = vmlal_s8(vprod0x15, vb15x1, va0x1);  in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()