90       int32x4_t vacc4567 = vld1q_s32(w); w = (const void*) ((uintptr_t) w + 4 * sizeof(int32_t));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()  local
105       vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi0x01234567), vget_high_s16(vk0x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
119       vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi1x01234567), vget_high_s16(vk1x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
133       vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi2x01234567), vget_high_s16(vk2x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
147       vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi3x01234567), vget_high_s16(vk3x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
161       vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi4x01234567), vget_high_s16(vk4x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
175       vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi5x01234567), vget_high_s16(vk5x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
189       vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
203       vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi7x01234567), vget_high_s16(vk7x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
217       vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi8x01234567), vget_high_s16(vk8x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
224       vacc4567 = vqrdmulhq_s32(vacc4567, vmultiplier);  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
231       vacc4567 = vsraq_n_s32(vacc4567, vbicq_s32(vacc4567, vzero_shift_mask), 31);  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
238       vacc4567 = vrshlq_s32(vacc4567, vright_shift);  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
245 …const int16x8_t vacc01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc0123), vacc4567), voutput…  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
252 …vacc01234567 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc0123), vqmovn_s32(vacc4567)), voutput_zero_p…  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
273         int32x4_t vacc4567 = vld1q_s32(w); w = (const void*) ((uintptr_t) w + 4 * sizeof(int32_t));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()  local
279         vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi0x01234567), vget_high_s16(vk0x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
284         vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi1x01234567), vget_high_s16(vk1x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
289         vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi2x01234567), vget_high_s16(vk2x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
294         vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi3x01234567), vget_high_s16(vk3x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
299         vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi4x01234567), vget_high_s16(vk4x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
304         vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi5x01234567), vget_high_s16(vk5x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
309         vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
314         vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi7x01234567), vget_high_s16(vk7x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
319         vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi8x01234567), vget_high_s16(vk8x01234567));  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
322         vacc4567 = vqrdmulhq_s32(vacc4567, vmultiplier);  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
325         vacc4567 = vsraq_n_s32(vacc4567, vbicq_s32(vacc4567, vzero_shift_mask), 31);  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
328         vacc4567 = vrshlq_s32(vacc4567, vright_shift);  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
331 …const int16x8_t vacc01234567 = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc0123), vacc4567), voutput…  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
335 …vacc01234567 = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc0123), vqmovn_s32(vacc4567)), voutput_zero_p…  in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()