87       float16x8_t vacc01234567p0 = vld1q_f16(w); w += 8;  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()  local
92       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi0x01234567, vk0x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
96       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi1x01234567, vk1x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
100       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi2x01234567, vk2x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
104       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi3x01234567, vk3x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
108       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi4x01234567, vk4x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
112       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi5x01234567, vk5x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
116       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
120       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi7x01234567, vk7x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
124       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi8x01234567, vk8x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
127       float16x8_t vacc01234567 = vmaxq_f16(vacc01234567p0, vmin);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
133       float16x8_t vacc01234567p0 = vld1q_f16(w); w += 8;  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()  local
138       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi0x01234567, vk0x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
142       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi1x01234567, vk1x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
146       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi2x01234567, vk2x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
150       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi3x01234567, vk3x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
154       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi4x01234567, vk4x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
158       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi5x01234567, vk5x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
162       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
166       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi7x01234567, vk7x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
170       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi8x01234567, vk8x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
173       float16x8_t vacc01234567 = vmaxq_f16(vacc01234567p0, vmin);  in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()