About rocketchip pipeline

[linkLiao]

why rocketchip pipeline detect WAW data-hazard at :

// stall for RAW/WAW hazards on CSRs, loads, AMOs, and mul/div in execute stage.
  val ex_cannot_bypass = ex_ctrl.csr =/= CSR.N || ex_ctrl.jalr || ex_ctrl.mem || ex_ctrl.mul || ex_ctrl.div || ex_ctrl.fp || ex_ctrl.rocc || ex_scie_pipelined
  val data_hazard_ex = ex_ctrl.wxd && checkHazards(hazard_targets, _ === ex_waddr)
  val fp_data_hazard_ex = ex_ctrl.wfd && checkHazards(fp_hazard_targets, _ === ex_waddr)
  val id_ex_hazard = ex_reg_valid && (data_hazard_ex && ex_cannot_bypass || fp_data_hazard_ex)

  // stall for RAW/WAW hazards on CSRs, LB/LH, and mul/div in memory stage.
  val mem_mem_cmd_bh =
    if (fastLoadWord) Bool(!fastLoadByte) && mem_reg_slow_bypass
    else Bool(true)
  val mem_cannot_bypass = mem_ctrl.csr =/= CSR.N || mem_ctrl.mem && mem_mem_cmd_bh || mem_ctrl.mul || mem_ctrl.div || mem_ctrl.fp || mem_ctrl.rocc
  val data_hazard_mem = mem_ctrl.wxd && checkHazards(hazard_targets, _ === mem_waddr)
  val fp_data_hazard_mem = mem_ctrl.wfd && checkHazards(fp_hazard_targets, _ === mem_waddr)
  val id_mem_hazard = mem_reg_valid && (data_hazard_mem && mem_cannot_bypass || fp_data_hazard_mem)
  id_load_use := mem_reg_valid && data_hazard_mem && mem_ctrl.mem

  // stall for RAW/WAW hazards on load/AMO misses and mul/div in writeback.
  val data_hazard_wb = wb_ctrl.wxd && checkHazards(hazard_targets, _ === wb_waddr)
  val fp_data_hazard_wb = wb_ctrl.wfd && checkHazards(fp_hazard_targets, _ === wb_waddr)
  val id_wb_hazard = wb_reg_valid && (data_hazard_wb && wb_set_sboard || fp_data_hazard_wb)

Is it require to check WAW data hazard ??

[aswaterman]

It's necessary because long-latency operations (memory-mapped I/O, division, etc.) can result in out-of-order writebacks.