401     vacc0x0123 = vsraq_n_s32(vacc0x0123, vbicq_s32(vacc0x0123, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
402     vacc0x4567 = vsraq_n_s32(vacc0x4567, vbicq_s32(vacc0x4567, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
403     vacc0x89AB = vsraq_n_s32(vacc0x89AB, vbicq_s32(vacc0x89AB, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
404     vacc0xCDEF = vsraq_n_s32(vacc0xCDEF, vbicq_s32(vacc0xCDEF, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
405     vacc1x0123 = vsraq_n_s32(vacc1x0123, vbicq_s32(vacc1x0123, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
406     vacc1x4567 = vsraq_n_s32(vacc1x4567, vbicq_s32(vacc1x4567, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
407     vacc1x89AB = vsraq_n_s32(vacc1x89AB, vbicq_s32(vacc1x89AB, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
408     vacc1xCDEF = vsraq_n_s32(vacc1xCDEF, vbicq_s32(vacc1xCDEF, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
409     vacc2x0123 = vsraq_n_s32(vacc2x0123, vbicq_s32(vacc2x0123, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
410     vacc2x4567 = vsraq_n_s32(vacc2x4567, vbicq_s32(vacc2x4567, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
411     vacc2x89AB = vsraq_n_s32(vacc2x89AB, vbicq_s32(vacc2x89AB, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
412     vacc2xCDEF = vsraq_n_s32(vacc2xCDEF, vbicq_s32(vacc2xCDEF, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
413     vacc3x0123 = vsraq_n_s32(vacc3x0123, vbicq_s32(vacc3x0123, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
414     vacc3x4567 = vsraq_n_s32(vacc3x4567, vbicq_s32(vacc3x4567, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
415     vacc3x89AB = vsraq_n_s32(vacc3x89AB, vbicq_s32(vacc3x89AB, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
416     vacc3xCDEF = vsraq_n_s32(vacc3xCDEF, vbicq_s32(vacc3xCDEF, vzero_shift_mask), 31);  in xnn_qs8_igemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()