132     const int32x2_t vpsum0x0 = vadd_s32(vget_low_s32(vacc0x0), vget_high_s32(vacc0x0));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
133     const int32x2_t vpsum0x1 = vadd_s32(vget_low_s32(vacc0x1), vget_high_s32(vacc0x1));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
134     const int32x2_t vpsum0x2 = vadd_s32(vget_low_s32(vacc0x2), vget_high_s32(vacc0x2));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
135     const int32x2_t vpsum0x3 = vadd_s32(vget_low_s32(vacc0x3), vget_high_s32(vacc0x3));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
139     const int32x2_t vpsum0x4 = vadd_s32(vget_low_s32(vacc0x4), vget_high_s32(vacc0x4));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
140     const int32x2_t vpsum0x5 = vadd_s32(vget_low_s32(vacc0x5), vget_high_s32(vacc0x5));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
141     const int32x2_t vpsum0x6 = vadd_s32(vget_low_s32(vacc0x6), vget_high_s32(vacc0x6));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
142     const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
146     const int32x2_t vpsum1x0 = vadd_s32(vget_low_s32(vacc1x0), vget_high_s32(vacc1x0));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
147     const int32x2_t vpsum1x1 = vadd_s32(vget_low_s32(vacc1x1), vget_high_s32(vacc1x1));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
148     const int32x2_t vpsum1x2 = vadd_s32(vget_low_s32(vacc1x2), vget_high_s32(vacc1x2));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
149     const int32x2_t vpsum1x3 = vadd_s32(vget_low_s32(vacc1x3), vget_high_s32(vacc1x3));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
153     const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
154     const int32x2_t vpsum1x5 = vadd_s32(vget_low_s32(vacc1x5), vget_high_s32(vacc1x5));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
155     const int32x2_t vpsum1x6 = vadd_s32(vget_low_s32(vacc1x6), vget_high_s32(vacc1x6));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
156     const int32x2_t vpsum1x7 = vadd_s32(vget_low_s32(vacc1x7), vget_high_s32(vacc1x7));  in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()