167       float16x8_t vacc01234567p0 = vld1q_f16(w); w += 8;  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()  local
172       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi0x01234567, vk0x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
176       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi1x01234567, vk1x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
180       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi2x01234567, vk2x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
184       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi3x01234567, vk3x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
188       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi4x01234567, vk4x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
192       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi5x01234567, vk5x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
196       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
200       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi7x01234567, vk7x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
204       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi8x01234567, vk8x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
208       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi9x01234567, vk9x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
212       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi10x01234567, vk10x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
216       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi11x01234567, vk11x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
220       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi12x01234567, vk12x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
224       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi13x01234567, vk13x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
228       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
232       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi15x01234567, vk15x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
236       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
240       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi17x01234567, vk17x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
244       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi18x01234567, vk18x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
248       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi19x01234567, vk19x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
252       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi20x01234567, vk20x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
256       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi21x01234567, vk21x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
260       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi22x01234567, vk22x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
264       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi23x01234567, vk23x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
268       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi24x01234567, vk24x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
271       float16x8_t vacc01234567 = vmaxq_f16(vacc01234567p0, vmin);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
277       float16x8_t vacc01234567p0 = vld1q_f16(w); w += 8;  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()  local
282       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi0x01234567, vk0x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
286       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi1x01234567, vk1x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
290       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi2x01234567, vk2x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
294       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi3x01234567, vk3x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
298       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi4x01234567, vk4x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
302       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi5x01234567, vk5x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
306       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
310       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi7x01234567, vk7x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
314       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi8x01234567, vk8x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
318       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi9x01234567, vk9x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
322       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi10x01234567, vk10x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
326       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi11x01234567, vk11x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
330       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi12x01234567, vk12x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
334       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi13x01234567, vk13x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
338       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
342       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi15x01234567, vk15x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
346       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
350       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi17x01234567, vk17x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
354       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi18x01234567, vk18x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
358       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi19x01234567, vk19x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
362       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi20x01234567, vk20x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
366       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi21x01234567, vk21x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
370       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi22x01234567, vk22x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
374       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi23x01234567, vk23x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
378       vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi24x01234567, vk24x01234567);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
381       float16x8_t vacc01234567 = vmaxq_f16(vacc01234567p0, vmin);  in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()