57       int8x8_t va0x0 = vld1_s8(a0); a0 += 8;  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()  local
71       int16x8_t vprod0x0123c0 = vmull_s8(vb0123c0x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
78       int16x8_t vprod0x4567c0 = vmull_s8(vb4567c0x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
85       va0x0 = vext_s8(va0x0, va0x0, 2);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
89       int16x8_t vprod0x0123c1 = vmull_s8(vb0123c1x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
96       int16x8_t vprod0x4567c1 = vmull_s8(vb4567c1x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
103       va0x0 = vext_s8(va0x0, va0x0, 2);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
107       int16x8_t vprod0x0123c2 = vmull_s8(vb0123c2x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
114       int16x8_t vprod0x4567c2 = vmull_s8(vb4567c2x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
121       va0x0 = vext_s8(va0x0, va0x0, 2);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
125       int16x8_t vprod0x0123c3 = vmull_s8(vb0123c3x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
132       int16x8_t vprod0x4567c3 = vmull_s8(vb4567c3x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
143       int8x8_t va0x0 = vld1_s8(a0); a0 += 8;  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()  local
155       int16x8_t vprod0x0123c0 = vmull_s8(vb0123c0x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
159       int16x8_t vprod0x4567c0 = vmull_s8(vb4567c0x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
163       va0x0 = vext_s8(va0x0, va0x0, 2);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
165       int16x8_t vprod0x0123c1 = vmull_s8(vb0123c1x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
169       int16x8_t vprod0x4567c1 = vmull_s8(vb4567c1x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
173       va0x0 = vext_s8(va0x0, va0x0, 2);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
175       int16x8_t vprod0x0123c2 = vmull_s8(vb0123c2x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
179       int16x8_t vprod0x4567c2 = vmull_s8(vb4567c2x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
183       va0x0 = vext_s8(va0x0, va0x0, 2);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
185       int16x8_t vprod0x0123c3 = vmull_s8(vb0123c3x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
189       int16x8_t vprod0x4567c3 = vmull_s8(vb4567c3x0, va0x0);  in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()