62       float16x8_t vacc01234567p0 = vld1q_f16(w); w += 8;  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()  local
70       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi0x01234567, vk0x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
77       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi1x01234567, vk1x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
84       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi2x01234567, vk2x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
91       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi3x01234567, vk3x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
95       float16x8_t vacc01234567 = vmaxq_f16(vacc01234567p0, vmin);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
104       float16x8_t vacc01234567p0 = vld1q_f16(w); w += 8;  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()  local
109       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi0x01234567, vk0x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
113       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi1x01234567, vk1x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
117       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi2x01234567, vk2x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
121       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi3x01234567, vk3x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
124       float16x8_t vacc01234567 = vmaxq_f16(vacc01234567p0, vmin);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
130       float16x8_t vacc01234567p0 = vld1q_f16(w);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()  local
135       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi0x01234567, vk0x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
139       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi1x01234567, vk1x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
143       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi2x01234567, vk2x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
147       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi3x01234567, vk3x01234567);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()
150       float16x8_t vacc01234567 = vmaxq_f16(vacc01234567p0, vmin);  in xnn_f16_dwconv_minmax_ukernel_up16x4__neonfp16arith()