With　the　increase　of　vehicular　broadband　services,　network　slicing　is　regarded　as　a　promising　technology　to　meet　the　various　requirements.　In　the　Internet　of　Vehicles　(IoV)　enabled　by　network　slicing,　both　the　resource　allocation　and　task　scheduling　algorithms　have　received　extensive　attentions.　However,　most　of　the　existing　works　on　the　joint　optimization　of　inter-slice　resource　allocation　and　intra-slice　task　scheduling　ignore　the　power　consumption.　In　this　regard,　this　paper　further　considers　the　queuing　latency　and　the　power　consumption　in　the　uplink　transmission.　Three　sub　Markov　decision　processes　(MDP)　are　modeled　and　a　layered　deep　reinforcement　learning　(DRL)　based　algorithm　is　designed　for　the　joint　optimization.　The　simulation　consists　of　two　parts,　firstly　the　cumulative　reward　of　different　task　scheduling　algorithms　are　compared　and　analyzed,　then　the　performance　of　all　solutions　are　evaluated　under　different　task　densities.　The　results　show　that　the　proposed　algorithm　can　significantly　reduce　the　queuing　latency　while　the　power　consumption　is　comparable　to　that　of　second　best　algorithm.　©　2022　IEEE.