76     int32x4_t vacc3x89AB = vacc0x89AB;  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()  local
221       vacc3x89AB = vpadalq_s16(vacc3x89AB, vprod3x89ABc0);  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
222       vacc3x89AB = vpadalq_s16(vacc3x89AB, vprod3x89ABc1);  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
223       vacc3x89AB = vpadalq_s16(vacc3x89AB, vprod3x89ABc2);  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
224       vacc3x89AB = vpadalq_s16(vacc3x89AB, vprod3x89ABc3);  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
277       vacc3x89AB = vpadalq_s16(vacc3x89AB, vprod3x89ABc0);  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
316         vacc3x89AB = vpadalq_s16(vacc3x89AB, vprod3x89ABc1);  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
355           vacc3x89AB = vpadalq_s16(vacc3x89AB, vprod3x89ABc2);  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
376     vacc3x89AB = vqrdmulhq_s32(vacc3x89AB, vmultiplier);  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
395     vacc3x89AB = vsraq_n_s32(vacc3x89AB, vbicq_s32(vacc3x89AB, vzero_shift_mask), 31);  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
412     vacc3x89AB = vrshlq_s32(vacc3x89AB, vright_shift);  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
424 …const int16x8_t vacc3x89ABCDEF = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc3x89AB), vacc3xCDEF), v…  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()
438 …const int16x8_t vacc3x89ABCDEF = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc3x89AB), vqmovn_s32(vacc3x…  in xnn_qs8_gemm_minmax_ukernel_4x16c2__neon_mull_padal_dup()