85       float32x4_t vi2x0 = vld1q_f32(i2); i2 += 4;  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()  local
100         vo1c0123 = vmlaq_lane_f32(vo1c0123, vk00c0x0123, vget_low_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
102         vo1c4567 = vmlaq_lane_f32(vo1c4567, vk00c0x4567, vget_low_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
115         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk20c0x0123, vget_low_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
117         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk20c0x4567, vget_low_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
124         vo1c0123 = vmlaq_lane_f32(vo1c0123, vk00c1x0123, vget_low_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
126         vo1c4567 = vmlaq_lane_f32(vo1c4567, vk00c1x4567, vget_low_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
139         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk20c1x0123, vget_low_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
141         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk20c1x4567, vget_low_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
148         vo1c0123 = vmlaq_lane_f32(vo1c0123, vk00c2x0123, vget_high_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
150         vo1c4567 = vmlaq_lane_f32(vo1c4567, vk00c2x4567, vget_high_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
163         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk20c2x0123, vget_high_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
165         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk20c2x4567, vget_high_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
172         vo1c0123 = vmlaq_lane_f32(vo1c0123, vk01c0x0123, vget_high_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
174         vo1c4567 = vmlaq_lane_f32(vo1c4567, vk01c0x4567, vget_high_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
187         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk21c0x0123, vget_high_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
189         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk21c0x4567, vget_high_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
328         vi2x0 = vcombine_f32(vget_high_f32(vi2x1), vi2x2);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
389         vo1c0123 = vmlaq_lane_f32(vo1c0123, vk00c0x0123, vget_low_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
391         vo1c4567 = vmlaq_lane_f32(vo1c4567, vk00c0x4567, vget_low_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
404         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk20c0x0123, vget_low_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
406         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk20c0x4567, vget_low_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
413         vo1c0123 = vmlaq_lane_f32(vo1c0123, vk00c1x0123, vget_low_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
415         vo1c4567 = vmlaq_lane_f32(vo1c4567, vk00c1x4567, vget_low_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
428         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk20c1x0123, vget_low_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
430         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk20c1x4567, vget_low_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
437         vo1c0123 = vmlaq_lane_f32(vo1c0123, vk00c2x0123, vget_high_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
439         vo1c4567 = vmlaq_lane_f32(vo1c4567, vk00c2x4567, vget_high_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
452         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk20c2x0123, vget_high_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
454         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk20c2x4567, vget_high_f32(vi2x0), 0);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
461         vo1c0123 = vmlaq_lane_f32(vo1c0123, vk01c0x0123, vget_high_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
463         vo1c4567 = vmlaq_lane_f32(vo1c4567, vk01c0x4567, vget_high_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
476         vo0c0123 = vmlaq_lane_f32(vo0c0123, vk21c0x0123, vget_high_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()
478         vo0c4567 = vmlaq_lane_f32(vo0c4567, vk21c0x4567, vget_high_f32(vi2x0), 1);  in xnn_f32_conv_hwc_ukernel_3x3s2p0p1c3x8__neon_2x1()