86       float32x4_t vacc0123p0 = vld1q_f32(w); w += 4;  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()  local
94       vacc0123p0 = vfmaq_f32(vacc0123p0, vi0x0123, vk0x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
101       vacc0123p0 = vfmaq_f32(vacc0123p0, vi1x0123, vk1x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
108       vacc0123p0 = vfmaq_f32(vacc0123p0, vi2x0123, vk2x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
115       vacc0123p0 = vfmaq_f32(vacc0123p0, vi3x0123, vk3x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
122       vacc0123p0 = vfmaq_f32(vacc0123p0, vi4x0123, vk4x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
129       vacc0123p0 = vfmaq_f32(vacc0123p0, vi5x0123, vk5x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
136       vacc0123p0 = vfmaq_f32(vacc0123p0, vi6x0123, vk6x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
143       vacc0123p0 = vfmaq_f32(vacc0123p0, vi7x0123, vk7x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
150       vacc0123p0 = vfmaq_f32(vacc0123p0, vi8x0123, vk8x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
154       float32x4_t vacc0123 = vmaxq_f32(vacc0123p0, vmin);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
163       float32x4_t vacc0123p0 = vld1q_f32(w); w += 4;  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()  local
168       vacc0123p0 = vfmaq_f32(vacc0123p0, vi0x0123, vk0x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
172       vacc0123p0 = vfmaq_f32(vacc0123p0, vi1x0123, vk1x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
176       vacc0123p0 = vfmaq_f32(vacc0123p0, vi2x0123, vk2x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
180       vacc0123p0 = vfmaq_f32(vacc0123p0, vi3x0123, vk3x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
184       vacc0123p0 = vfmaq_f32(vacc0123p0, vi4x0123, vk4x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
188       vacc0123p0 = vfmaq_f32(vacc0123p0, vi5x0123, vk5x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
192       vacc0123p0 = vfmaq_f32(vacc0123p0, vi6x0123, vk6x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
196       vacc0123p0 = vfmaq_f32(vacc0123p0, vi7x0123, vk7x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
200       vacc0123p0 = vfmaq_f32(vacc0123p0, vi8x0123, vk8x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
203       float32x4_t vacc0123 = vmaxq_f32(vacc0123p0, vmin);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
209       float32x4_t vacc0123p0 = vld1q_f32(w);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()  local
214       vacc0123p0 = vfmaq_f32(vacc0123p0, vi0x0123, vk0x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
218       vacc0123p0 = vfmaq_f32(vacc0123p0, vi1x0123, vk1x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
222       vacc0123p0 = vfmaq_f32(vacc0123p0, vi2x0123, vk2x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
226       vacc0123p0 = vfmaq_f32(vacc0123p0, vi3x0123, vk3x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
230       vacc0123p0 = vfmaq_f32(vacc0123p0, vi4x0123, vk4x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
234       vacc0123p0 = vfmaq_f32(vacc0123p0, vi5x0123, vk5x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
238       vacc0123p0 = vfmaq_f32(vacc0123p0, vi6x0123, vk6x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
242       vacc0123p0 = vfmaq_f32(vacc0123p0, vi7x0123, vk7x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
246       vacc0123p0 = vfmaq_f32(vacc0123p0, vi8x0123, vk8x0123);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()
249       float32x4_t vacc0123 = vmaxq_f32(vacc0123p0, vmin);  in xnn_f32_dwconv_minmax_ukernel_up8x9__neonfma()