84       float32x4_t vi1x0 = vld1q_f32(i1); i1 += 4;  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()  local
107         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk10c0x0123, vget_low_f32(vi1x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
109         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk10c0x4567, vget_low_f32(vi1x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
131         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk10c1x0123, vget_low_f32(vi1x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
133         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk10c1x4567, vget_low_f32(vi1x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
155         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk10c2x0123, vget_high_f32(vi1x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
157         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk10c2x4567, vget_high_f32(vi1x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
179         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk11c0x0123, vget_high_f32(vi1x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
181         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk11c0x4567, vget_high_f32(vi1x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
327         vi1x0 = vcombine_f32(vget_high_f32(vi1x1), vi1x2);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
396         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk10c0x0123, vget_low_f32(vi1x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
398         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk10c0x4567, vget_low_f32(vi1x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
420         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk10c1x0123, vget_low_f32(vi1x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
422         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk10c1x4567, vget_low_f32(vi1x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
444         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk10c2x0123, vget_high_f32(vi1x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
446         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk10c2x4567, vget_high_f32(vi1x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
468         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk11c0x0123, vget_high_f32(vi1x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
470         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk11c0x4567, vget_high_f32(vi1x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()