63     int32x4_t vacc1xCDEF = vacc0xCDEF;  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()  local
154       vacc1xCDEF = vpadalq_s16(vacc1xCDEF, vprod1xCDEFc0);  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()
155       vacc1xCDEF = vpadalq_s16(vacc1xCDEF, vprod1xCDEFc1);  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()
156       vacc1xCDEF = vpadalq_s16(vacc1xCDEF, vprod1xCDEFc2);  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()
157       vacc1xCDEF = vpadalq_s16(vacc1xCDEF, vprod1xCDEFc3);  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()
219       vacc1xCDEF = vpadalq_s16(vacc1xCDEF, vprod1xCDEFc0);  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()
250         vacc1xCDEF = vpadalq_s16(vacc1xCDEF, vprod1xCDEFc1);  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()
281           vacc1xCDEF = vpadalq_s16(vacc1xCDEF, vprod1xCDEFc2);  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()
301     vacc1xCDEF = vqrdmulhq_s32(vacc1xCDEF, vmultiplier);  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()
316     vacc1xCDEF = vsraq_n_s32(vacc1xCDEF, vbicq_s32(vacc1xCDEF, vzero_shift_mask), 31);  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()
329     vacc1xCDEF = vrshlq_s32(vacc1xCDEF, vright_shift);  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()
340 …const int16x8_t vacc1x89ABCDEF = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc1x89AB), vacc1xCDEF), v…  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()
351 …x89ABCDEF = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc1x89AB), vqmovn_s32(vacc1xCDEF)), voutput_zero_…  in xnn_qs8_gemm_minmax_ukernel_3x16c2__neon_mull_padal_dup()