44     uint8x8_t vout = vld1_u8(x); x += 8;  in xnn_u8_clamp_ukernel__neon_x64()  local
45     vout = vmin_u8(vout, vget_low_u8(voutput_max));  in xnn_u8_clamp_ukernel__neon_x64()
46     vout = vmax_u8(vout, vget_low_u8(voutput_min));  in xnn_u8_clamp_ukernel__neon_x64()
47     vst1_u8(y, vout); y += 8;  in xnn_u8_clamp_ukernel__neon_x64()
50     uint8x8_t vout = vld1_u8(x);  in xnn_u8_clamp_ukernel__neon_x64()  local
51     vout = vmin_u8(vout, vget_low_u8(voutput_max));  in xnn_u8_clamp_ukernel__neon_x64()
52     vout = vmax_u8(vout, vget_low_u8(voutput_min));  in xnn_u8_clamp_ukernel__neon_x64()
55       vst1_lane_u32(__builtin_assume_aligned(y, 1), vreinterpret_u32_u8(vout), 0); y += 4;  in xnn_u8_clamp_ukernel__neon_x64()
56       vout = vext_u8(vout, vout, 4);  in xnn_u8_clamp_ukernel__neon_x64()
59       vst1_lane_u16(__builtin_assume_aligned(y, 1), vreinterpret_u16_u8(vout), 0); y += 2;  in xnn_u8_clamp_ukernel__neon_x64()
60       vout = vext_u8(vout, vout, 2);  in xnn_u8_clamp_ukernel__neon_x64()
63       vst1_lane_u8(y, vout, 0);  in xnn_u8_clamp_ukernel__neon_x64()