144         float32x4_t va7 = vld1q_f32(a7); a7 += 4;  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()  local
157         vacc7x0123 = vmlaq_f32(vacc7x0123, va7, vb0123c0);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
165         vacc7x4567 = vmlaq_f32(vacc7x4567, va7, vb4567c0);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
174         va7 = vextq_f32(va7, va7, 1);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
186         vacc7x0123 = vmlaq_f32(vacc7x0123, va7, vb0123c1);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
194         vacc7x4567 = vmlaq_f32(vacc7x4567, va7, vb4567c1);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
203         va7 = vextq_f32(va7, va7, 1);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
215         vacc7x0123 = vmlaq_f32(vacc7x0123, va7, vb0123c2);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
223         vacc7x4567 = vmlaq_f32(vacc7x4567, va7, vb4567c2);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
232         va7 = vextq_f32(va7, va7, 1);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
244         vacc7x0123 = vmlaq_f32(vacc7x0123, va7, vb0123c3);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
252         vacc7x4567 = vmlaq_f32(vacc7x4567, va7, vb4567c3);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
267           const float32x4_t va7 = vld1q_dup_f32(a7); a7 += 1;  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()  local
279           vacc7x0123 = vmlaq_f32(vacc7x0123, va7, vb0123);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()
287           vacc7x4567 = vmlaq_f32(vacc7x4567, va7, vb4567);  in xnn_f32_igemm_minmax_ukernel_8x8s4__neon()