51       int8x8_t va0x0 = vld1_s8(a0); a0 += 8;  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()  local
70       int16x8_t vprod0x0123c0 = vmull_s8(vb0123c0x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
72       int16x8_t vprod0x4567c0 = vmull_s8(vb4567c0x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
74       int16x8_t vprod0x89ABc0 = vmull_s8(vb89ABc0x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
76       int16x8_t vprod0xCDEFc0 = vmull_s8(vbCDEFc0x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
78       va0x0 = vext_s8(va0x0, va0x0, 2);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
79       int16x8_t vprod0x0123c1 = vmull_s8(vb0123c1x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
81       int16x8_t vprod0x4567c1 = vmull_s8(vb4567c1x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
83       int16x8_t vprod0x89ABc1 = vmull_s8(vb89ABc1x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
85       int16x8_t vprod0xCDEFc1 = vmull_s8(vbCDEFc1x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
87       va0x0 = vext_s8(va0x0, va0x0, 2);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
88       int16x8_t vprod0x0123c2 = vmull_s8(vb0123c2x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
90       int16x8_t vprod0x4567c2 = vmull_s8(vb4567c2x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
92       int16x8_t vprod0x89ABc2 = vmull_s8(vb89ABc2x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
94       int16x8_t vprod0xCDEFc2 = vmull_s8(vbCDEFc2x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
96       va0x0 = vext_s8(va0x0, va0x0, 2);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
97       int16x8_t vprod0x0123c3 = vmull_s8(vb0123c3x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
99       int16x8_t vprod0x4567c3 = vmull_s8(vb4567c3x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
101       int16x8_t vprod0x89ABc3 = vmull_s8(vb89ABc3x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()
103       int16x8_t vprod0xCDEFc3 = vmull_s8(vbCDEFc3x0, va0x0);  in xnn_qs8_gemm_minmax_rndnu_ukernel_1x16c2s4__neon_mull()