xref: /XiangShan/src/main/scala/xiangshan/cache/dcache/mainpipe/MissQueue.scala (revision 51f9a957bed418a09ee0e36f827174ce8ef2ac8e)

/***************************************************************************************
* Copyright (c) 2020-2021 Institute of Computing Technology, Chinese Academy of Sciences
* Copyright (c) 2020-2021 Peng Cheng Laboratory
*
* XiangShan is licensed under Mulan PSL v2.
* You can use this software according to the terms and conditions of the Mulan PSL v2.
* You may obtain a copy of Mulan PSL v2 at:
*          http://license.coscl.org.cn/MulanPSL2
*
* THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
* EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
* MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
*
* See the Mulan PSL v2 for more details.
*
*
* Acknowledgement
*
* This implementation is inspired by several key papers:
* [1] David Kroft. "[Lockup-free instruction fetch/prefetch cache organization.]
* (https://dl.acm.org/doi/10.5555/800052.801868)" 8th Annual Symposium on Computer Architecture (ISCA). 1981.
***************************************************************************************/

package xiangshan.cache

import chisel3._
import chisel3.util._
import chisel3.experimental.dataview._
import coupledL2.VaddrKey
import coupledL2.IsKeywordKey
import difftest._
import freechips.rocketchip.tilelink.ClientStates._
import freechips.rocketchip.tilelink.MemoryOpCategories._
import freechips.rocketchip.tilelink.TLPermissions._
import freechips.rocketchip.tilelink.TLMessages._
import freechips.rocketchip.tilelink._
import huancun.{AliasKey, DirtyKey, PrefetchKey}
import org.chipsalliance.cde.config.Parameters
import utility._
import utils._
import xiangshan._
import xiangshan.mem.AddPipelineReg
import xiangshan.mem.prefetch._
import xiangshan.mem.trace._
import xiangshan.mem.LqPtr

class MissReqWoStoreData(implicit p: Parameters) extends DCacheBundle {
  val source = UInt(sourceTypeWidth.W)
  val pf_source = UInt(L1PfSourceBits.W)
  val cmd = UInt(M_SZ.W)
  val addr = UInt(PAddrBits.W)
  val vaddr = UInt(VAddrBits.W)
  val pc = UInt(VAddrBits.W)

  val lqIdx = new LqPtr
  // store
  val full_overwrite = Bool()

  // amo
  val word_idx = UInt(log2Up(blockWords).W)
  val amo_data   = UInt(QuadWordBits.W)
  val amo_mask   = UInt(QuadWordBytes.W)
  val amo_cmp    = UInt(QuadWordBits.W) // data to be compared in AMOCAS

  val req_coh = new ClientMetadata
  val id = UInt(reqIdWidth.W)

  // For now, miss queue entry req is actually valid when req.valid && !cancel
  // * req.valid is fast to generate
  // * cancel is slow to generate, it will not be used until the last moment
  //
  // cancel may come from the following sources:
  // 1. miss req blocked by writeback queue:
  //      a writeback req of the same address is in progress
  // 2. pmp check failed
  val cancel = Bool() // cancel is slow to generate, it will cancel missreq.valid

  // Req source decode
  // Note that req source is NOT cmd type
  // For instance, a req which isFromPrefetch may have R or W cmd
  def isFromLoad = source === LOAD_SOURCE.U
  def isFromStore = source === STORE_SOURCE.U
  def isFromAMO = source === AMO_SOURCE.U
  def isFromPrefetch = source >= DCACHE_PREFETCH_SOURCE.U
  def isPrefetchWrite = source === DCACHE_PREFETCH_SOURCE.U && cmd === MemoryOpConstants.M_PFW
  def isPrefetchRead = source === DCACHE_PREFETCH_SOURCE.U && cmd === MemoryOpConstants.M_PFR
  def hit = req_coh.isValid()
}

class MissReqStoreData(implicit p: Parameters) extends DCacheBundle {
  // store data and store mask will be written to miss queue entry
  // 1 cycle after req.fire() and meta write
  val store_data = UInt((cfg.blockBytes * 8).W)
  val store_mask = UInt(cfg.blockBytes.W)
}

class MissQueueRefillInfo(implicit p: Parameters) extends MissReqStoreData {
  // refill_info for mainpipe req awake
  val miss_param = UInt(TLPermissions.bdWidth.W)
  val miss_dirty = Bool()
  val error      = Bool()
}

class MissReq(implicit p: Parameters) extends MissReqWoStoreData {
  // store data and store mask will be written to miss queue entry
  // 1 cycle after req.fire() and meta write
  val store_data = UInt((cfg.blockBytes * 8).W)
  val store_mask = UInt(cfg.blockBytes.W)

  def toMissReqStoreData(): MissReqStoreData = {
    val out = Wire(new MissReqStoreData)
    out.store_data := store_data
    out.store_mask := store_mask
    out
  }

  def toMissReqWoStoreData(): MissReqWoStoreData = {
    this.viewAsSupertype(new MissReqWoStoreData)
  }
}

class MissResp(implicit p: Parameters) extends DCacheBundle {
  val id = UInt(log2Up(cfg.nMissEntries).W)
  // cache miss request is handled by miss queue, either merged or newly allocated
  val handled = Bool()
  // cache req missed, merged into one of miss queue entries
  // i.e. !miss_merged means this access is the first miss for this cacheline
  val merged = Bool()
}


/**
  * miss queue enq logic: enq is now splited into 2 cycles
  *  +---------------------------------------------------------------------+    pipeline reg  +-------------------------+
  *  +         s0: enq source arbiter, judge mshr alloc or merge           +     +-------+    + s1: real alloc or merge +
  *  +                      +-----+          primary_fire?       ->        +     | alloc |    +                         +
  *  + mainpipe  -> req0 -> |     |          secondary_fire?     ->        +     | merge |    +                         +
  *  + loadpipe0 -> req1 -> | arb | -> req                       ->        +  -> | req   | -> +                         +
  *  + loadpipe1 -> req2 -> |     |          mshr id             ->        +     | id    |    +                         +
  *  +                      +-----+                                        +     +-------+    +                         +
  *  +---------------------------------------------------------------------+                  +-------------------------+
  */

// a pipeline reg between MissReq and MissEntry
class MissReqPipeRegBundle(edge: TLEdgeOut)(implicit p: Parameters) extends DCacheBundle
 with HasCircularQueuePtrHelper
 {
  val req           = new MissReq
  // this request is about to merge to an existing mshr
  val merge         = Bool()
  // this request is about to allocate a new mshr
  val alloc         = Bool()
  val cancel        = Bool()
  val mshr_id       = UInt(log2Up(cfg.nMissEntries).W)

  def reg_valid(): Bool = {
    (merge || alloc)
  }

  def matched(new_req: MissReq): Bool = {
    val block_match = get_block(req.addr) === get_block(new_req.addr)
    block_match && reg_valid() && !(req.isFromPrefetch)
  }

  def prefetch_late_en(new_req: MissReqWoStoreData, new_req_valid: Bool): Bool = {
    val block_match = get_block(req.addr) === get_block(new_req.addr)
    new_req_valid && alloc && block_match && (req.isFromPrefetch) && !(new_req.isFromPrefetch)
  }

  def reject_req(new_req: MissReq): Bool = {
    val block_match = get_block(req.addr) === get_block(new_req.addr)
    val alias_match = is_alias_match(req.vaddr, new_req.vaddr)
    val merge_load = (req.isFromLoad || req.isFromStore || req.isFromPrefetch) && new_req.isFromLoad
    // store merge to a store is disabled, sbuffer should avoid this situation, as store to same address should preserver their program order to match memory model
    val merge_store = (req.isFromLoad || req.isFromPrefetch) && new_req.isFromStore

    val set_match = addr_to_dcache_set(req.vaddr) === addr_to_dcache_set(new_req.vaddr)

    Mux(
        alloc,
        block_match && (!alias_match || !(merge_load || merge_store)),
        false.B
      )
  }

  def merge_req(new_req: MissReq): Bool = {
    val block_match = get_block(req.addr) === get_block(new_req.addr)
    val alias_match = is_alias_match(req.vaddr, new_req.vaddr)
    val merge_load = (req.isFromLoad || req.isFromStore || req.isFromPrefetch) && new_req.isFromLoad
    // store merge to a store is disabled, sbuffer should avoid this situation, as store to same address should preserver their program order to match memory model
    val merge_store = (req.isFromLoad || req.isFromPrefetch) && new_req.isFromStore
    Mux(
        alloc,
        block_match && alias_match && (merge_load || merge_store),
        false.B
      )
  }

  def merge_isKeyword(new_req: MissReq): Bool = {
    val load_merge_load  = merge_req(new_req) && req.isFromLoad  && new_req.isFromLoad
    val store_merge_load = merge_req(new_req) && req.isFromStore && new_req.isFromLoad
    val load_merge_load_use_new_req_isKeyword = isAfter(req.lqIdx, new_req.lqIdx)
    val use_new_req_isKeyword = (load_merge_load && load_merge_load_use_new_req_isKeyword) || store_merge_load
    Mux (
      use_new_req_isKeyword,
        new_req.vaddr(5).asBool,
        req.vaddr(5).asBool
      )
  }

  def isKeyword(): Bool= {
    val alloc_isKeyword = Mux(
                           alloc,
                           Mux(
                            req.isFromLoad,
                            req.vaddr(5).asBool,
                            false.B),
                            false.B)
    Mux(
      merge_req(req),
      merge_isKeyword(req),
      alloc_isKeyword
    )
  }
  // send out acquire as soon as possible
  // if a new store miss req is about to merge into this pipe reg, don't send acquire now
  def can_send_acquire(valid: Bool, new_req: MissReq): Bool = {
    alloc && !(valid && merge_req(new_req) && new_req.isFromStore)
  }

  def get_acquire(l2_pf_store_only: Bool): TLBundleA = {
    val acquire = Wire(new TLBundleA(edge.bundle))
    val grow_param = req.req_coh.onAccess(req.cmd)._2
    val acquireBlock = edge.AcquireBlock(
      fromSource = mshr_id,
      toAddress = get_block_addr(req.addr),
      lgSize = (log2Up(cfg.blockBytes)).U,
      growPermissions = grow_param
    )._2
    val acquirePerm = edge.AcquirePerm(
      fromSource = mshr_id,
      toAddress = get_block_addr(req.addr),
      lgSize = (log2Up(cfg.blockBytes)).U,
      growPermissions = grow_param
    )._2
    acquire := Mux(req.full_overwrite, acquirePerm, acquireBlock)
    // resolve cache alias by L2
    acquire.user.lift(AliasKey).foreach(_ := req.vaddr(13, 12))
    // pass vaddr to l2
    acquire.user.lift(VaddrKey).foreach(_ := req.vaddr(VAddrBits - 1, blockOffBits))

    // miss req pipe reg pass keyword to L2, is priority
    acquire.echo.lift(IsKeywordKey).foreach(_ := isKeyword())

    // trigger prefetch
    acquire.user.lift(PrefetchKey).foreach(_ := Mux(l2_pf_store_only, req.isFromStore, true.B))
    // req source
    when(req.isFromLoad) {
      acquire.user.lift(ReqSourceKey).foreach(_ := MemReqSource.CPULoadData.id.U)
    }.elsewhen(req.isFromStore) {
      acquire.user.lift(ReqSourceKey).foreach(_ := MemReqSource.CPUStoreData.id.U)
    }.elsewhen(req.isFromAMO) {
      acquire.user.lift(ReqSourceKey).foreach(_ := MemReqSource.CPUAtomicData.id.U)
    }.otherwise {
      acquire.user.lift(ReqSourceKey).foreach(_ := MemReqSource.L1DataPrefetch.id.U)
    }

    acquire
  }

  def block_match(release_addr: UInt): Bool = {
    reg_valid() && get_block(req.addr) === get_block(release_addr)
  }
}

class CMOUnit(edge: TLEdgeOut)(implicit p: Parameters) extends DCacheModule {
  val io = IO(new Bundle() {
    val req = Flipped(DecoupledIO(new CMOReq))
    val req_chanA = DecoupledIO(new TLBundleA(edge.bundle))
    val resp_chanD = Flipped(DecoupledIO(new TLBundleD(edge.bundle)))
    val resp_to_lsq = DecoupledIO(new CMOResp)
  })

  val s_idle :: s_sreq :: s_wresp :: s_lsq_resp :: Nil = Enum(4)
  val state = RegInit(s_idle)
  val state_next = WireInit(state)
  val req = RegEnable(io.req.bits, io.req.fire)
  val nderr = RegInit(false.B)

  state := state_next

  switch (state) {
    is(s_idle) {
      when (io.req.fire) {
        state_next := s_sreq
        nderr := false.B
      }
    }
    is(s_sreq) {
      when (io.req_chanA.fire) {
        state_next := s_wresp
      }
    }
    is(s_wresp) {
      when (io.resp_chanD.fire) {
        state_next := s_lsq_resp
        nderr := io.resp_chanD.bits.denied || io.resp_chanD.bits.corrupt
      }
    }
    is(s_lsq_resp) {
      when (io.resp_to_lsq.fire) {
        state_next := s_idle
      }
    }
  }

  io.req.ready := state === s_idle

  io.req_chanA.valid := state === s_sreq
  io.req_chanA.bits := edge.CacheBlockOperation(
    fromSource = (cfg.nMissEntries + 1).U,
    toAddress = req.address,
    lgSize = (log2Up(cfg.blockBytes)).U,
    opcode = req.opcode
  )._2

  io.resp_chanD.ready := state === s_wresp

  io.resp_to_lsq.valid := state === s_lsq_resp
  io.resp_to_lsq.bits.address := req.address
  io.resp_to_lsq.bits.nderr   := nderr

  assert(!(state =/= s_idle && io.req.valid))
  assert(!(state =/= s_wresp && io.resp_chanD.valid))
}

class MissEntry(edge: TLEdgeOut, reqNum: Int)(implicit p: Parameters) extends DCacheModule
  with HasCircularQueuePtrHelper
 {
  val io = IO(new Bundle() {
    val hartId = Input(UInt(hartIdLen.W))
    // MSHR ID
    val id = Input(UInt(log2Up(cfg.nMissEntries).W))
    // client requests
    // MSHR update request, MSHR state and addr will be updated when req.fire
    val req = Flipped(ValidIO(new MissReqWoStoreData))
    val wbq_block_miss_req = Input(Bool())
    // pipeline reg
    val miss_req_pipe_reg = Input(new MissReqPipeRegBundle(edge))
    // allocate this entry for new req
    val primary_valid = Input(Bool())
    // this entry is free and can be allocated to new reqs
    val primary_ready = Output(Bool())
    // this entry is busy, but it can merge the new req
    val secondary_ready = Output(Bool())
    // this entry is busy and it can not merge the new req
    val secondary_reject = Output(Bool())
    // way selected for replacing, used to support plru update
    // bus
    val mem_acquire = DecoupledIO(new TLBundleA(edge.bundle))
    val mem_grant = Flipped(DecoupledIO(new TLBundleD(edge.bundle)))
    val mem_finish = DecoupledIO(new TLBundleE(edge.bundle))

    val queryME = Vec(reqNum, Flipped(new DCacheMEQueryIOBundle))

    // send refill info to load queue, useless now
    val refill_to_ldq = ValidIO(new Refill)

    // replace pipe
    val l2_hint = Input(Valid(new L2ToL1Hint())) // Hint from L2 Cache

    // main pipe: amo miss
    val main_pipe_req = DecoupledIO(new MainPipeReq)
    val main_pipe_resp = Input(Bool())
    val main_pipe_refill_resp = Input(Bool())
    val main_pipe_replay = Input(Bool())

    // for main pipe s2
    val refill_info = ValidIO(new MissQueueRefillInfo)

    val block_addr = ValidIO(UInt(PAddrBits.W))

    val req_addr = ValidIO(UInt(PAddrBits.W))

    val req_handled_by_this_entry = Output(Bool())

    val forwardInfo = Output(new MissEntryForwardIO)
    val l2_pf_store_only = Input(Bool())

    // whether the pipeline reg has send out an acquire
    val acquire_fired_by_pipe_reg = Input(Bool())
    val memSetPattenDetected = Input(Bool())

    val perf_pending_prefetch = Output(Bool())
    val perf_pending_normal   = Output(Bool())

    val rob_head_query = new DCacheBundle {
      val vaddr = Input(UInt(VAddrBits.W))
      val query_valid = Input(Bool())

      val resp = Output(Bool())

      def hit(e_vaddr: UInt): Bool = {
        require(e_vaddr.getWidth == VAddrBits)
        query_valid && vaddr(VAddrBits - 1, DCacheLineOffset) === e_vaddr(VAddrBits - 1, DCacheLineOffset)
      }
    }

    val latency_monitor = new DCacheBundle {
      val load_miss_refilling  = Output(Bool())
      val store_miss_refilling = Output(Bool())
      val amo_miss_refilling   = Output(Bool())
      val pf_miss_refilling    = Output(Bool())
    }

    val prefetch_info = new DCacheBundle {
      val late_prefetch = Output(Bool())
    }
    val nMaxPrefetchEntry = Input(UInt(64.W))
    val matched = Output(Bool())
    val l1Miss = Output(Bool())
  })

  assert(!RegNext(io.primary_valid && !io.primary_ready))

  val req = Reg(new MissReqWoStoreData)
  val req_primary_fire = Reg(new MissReqWoStoreData) // for perf use
  val req_store_mask = Reg(UInt(cfg.blockBytes.W))
  val req_valid = RegInit(false.B)
  val set = addr_to_dcache_set(req.vaddr)
  // initial keyword
  val isKeyword = RegInit(false.B)

  val miss_req_pipe_reg_bits = io.miss_req_pipe_reg.req

  val input_req_is_prefetch = isPrefetch(miss_req_pipe_reg_bits.cmd)

  val s_acquire = RegInit(true.B)
  val s_grantack = RegInit(true.B)
  val s_mainpipe_req = RegInit(true.B)

  val w_grantfirst = RegInit(true.B)
  val w_grantlast = RegInit(true.B)
  val w_mainpipe_resp = RegInit(true.B)
  val w_refill_resp = RegInit(true.B)
  val w_l2hint = RegInit(true.B)

  val mainpipe_req_fired = RegInit(true.B)

  val release_entry = s_grantack && w_mainpipe_resp && w_refill_resp

  val acquire_not_sent = !s_acquire && !io.mem_acquire.ready
  val data_not_refilled = !w_grantfirst

  val error = RegInit(false.B)
  val prefetch = RegInit(false.B)
  val access = RegInit(false.B)

  val should_refill_data_reg =  Reg(Bool())
  val should_refill_data = WireInit(should_refill_data_reg)

  val should_replace = RegInit(false.B)

  val full_overwrite = Reg(Bool())

  val (_, _, refill_done, refill_count) = edge.count(io.mem_grant)
  val grant_param = Reg(UInt(TLPermissions.bdWidth.W))

  // refill data with store data, this reg will be used to store:
  // 1. store data (if needed), before l2 refill data
  // 2. store data and l2 refill data merged result (i.e. new cacheline taht will be write to data array)
  val refill_and_store_data = Reg(Vec(blockRows, UInt(rowBits.W)))
  // raw data refilled to l1 by l2
  val refill_data_raw = Reg(Vec(blockBytes/beatBytes, UInt(beatBits.W)))

  // allocate current miss queue entry for a miss req
  val primary_fire = WireInit(io.req.valid && io.primary_ready && io.primary_valid && !io.req.bits.cancel && !io.wbq_block_miss_req)
  val primary_accept = WireInit(io.req.valid && io.primary_ready && io.primary_valid && !io.req.bits.cancel)
  // merge miss req to current miss queue entry
  val secondary_fire = WireInit(io.req.valid && io.secondary_ready && !io.req.bits.cancel && !io.wbq_block_miss_req)
  val secondary_accept = WireInit(io.req.valid && io.secondary_ready && !io.req.bits.cancel)

  val req_handled_by_this_entry = primary_accept || secondary_accept

  // for perf use
  val secondary_fired = RegInit(false.B)

  io.perf_pending_prefetch := req_valid && prefetch && !secondary_fired
  io.perf_pending_normal   := req_valid && (!prefetch || secondary_fired)

  io.rob_head_query.resp   := io.rob_head_query.hit(req.vaddr) && req_valid

  io.req_handled_by_this_entry := req_handled_by_this_entry

  when (release_entry && req_valid) {
    req_valid := false.B
  }

  when (io.miss_req_pipe_reg.alloc && !io.miss_req_pipe_reg.cancel) {
    assert(RegNext(primary_fire), "after 1 cycle of primary_fire, entry will be allocated")
    req_valid := true.B

    req := miss_req_pipe_reg_bits.toMissReqWoStoreData()
    req_primary_fire := miss_req_pipe_reg_bits.toMissReqWoStoreData()
    req.addr := get_block_addr(miss_req_pipe_reg_bits.addr)
    //only  load miss need keyword
    isKeyword := Mux(miss_req_pipe_reg_bits.isFromLoad, miss_req_pipe_reg_bits.vaddr(5).asBool,false.B)

    s_acquire := io.acquire_fired_by_pipe_reg
    s_grantack := false.B
    s_mainpipe_req := false.B

    w_grantfirst := false.B
    w_grantlast := false.B
    w_l2hint := false.B
    mainpipe_req_fired := false.B

    when(miss_req_pipe_reg_bits.isFromStore) {
      req_store_mask := miss_req_pipe_reg_bits.store_mask
      for (i <- 0 until blockRows) {
        refill_and_store_data(i) := miss_req_pipe_reg_bits.store_data(rowBits * (i + 1) - 1, rowBits * i)
      }
    }
    full_overwrite := miss_req_pipe_reg_bits.isFromStore && miss_req_pipe_reg_bits.full_overwrite

    when (!miss_req_pipe_reg_bits.isFromAMO) {
      w_refill_resp := false.B
    }

    when (miss_req_pipe_reg_bits.isFromAMO) {
      w_mainpipe_resp := false.B
    }

    should_refill_data_reg := miss_req_pipe_reg_bits.isFromLoad
    error := false.B
    prefetch := input_req_is_prefetch && !io.miss_req_pipe_reg.prefetch_late_en(io.req.bits, io.req.valid)
    access := false.B
    secondary_fired := false.B
  }

  when (io.miss_req_pipe_reg.merge && !io.miss_req_pipe_reg.cancel) {
    assert(RegNext(secondary_fire) || RegNext(RegNext(primary_fire)), "after 1 cycle of secondary_fire or 2 cycle of primary_fire, entry will be merged")
    assert(miss_req_pipe_reg_bits.req_coh.state <= req.req_coh.state || (prefetch && !access))
    assert(!(miss_req_pipe_reg_bits.isFromAMO || req.isFromAMO))
    // use the most uptodate meta
    req.req_coh := miss_req_pipe_reg_bits.req_coh

    isKeyword := Mux(
      before_req_sent_can_merge(miss_req_pipe_reg_bits),
      before_req_sent_merge_iskeyword(miss_req_pipe_reg_bits),
      isKeyword)
    assert(!miss_req_pipe_reg_bits.isFromPrefetch, "can not merge a prefetch req, late prefetch should always be ignored!")

    when (miss_req_pipe_reg_bits.isFromStore) {
      req := miss_req_pipe_reg_bits
      req.addr := get_block_addr(miss_req_pipe_reg_bits.addr)
      req_store_mask := miss_req_pipe_reg_bits.store_mask
      for (i <- 0 until blockRows) {
        refill_and_store_data(i) := miss_req_pipe_reg_bits.store_data(rowBits * (i + 1) - 1, rowBits * i)
      }
      full_overwrite := miss_req_pipe_reg_bits.isFromStore && miss_req_pipe_reg_bits.full_overwrite
      assert(is_alias_match(req.vaddr, miss_req_pipe_reg_bits.vaddr), "alias bits should be the same when merging store")
    }

    should_refill_data := should_refill_data_reg || miss_req_pipe_reg_bits.isFromLoad
    should_refill_data_reg := should_refill_data
    when (!input_req_is_prefetch) {
      access := true.B // when merge non-prefetch req, set access bit
    }
    secondary_fired := true.B
  }

  when (io.mem_acquire.fire) {
    s_acquire := true.B
  }

  // merge data refilled by l2 and store data, update miss queue entry, gen refill_req
  val new_data = Wire(Vec(blockRows, UInt(rowBits.W)))
  val new_mask = Wire(Vec(blockRows, UInt(rowBytes.W)))
  // merge refilled data and store data (if needed)
  def mergePutData(old_data: UInt, new_data: UInt, wmask: UInt): UInt = {
    val full_wmask = FillInterleaved(8, wmask)
    (~full_wmask & old_data | full_wmask & new_data)
  }
  for (i <- 0 until blockRows) {
    // new_data(i) := req.store_data(rowBits * (i + 1) - 1, rowBits * i)
    new_data(i) := refill_and_store_data(i)
    // we only need to merge data for Store
    new_mask(i) := Mux(req.isFromStore, req_store_mask(rowBytes * (i + 1) - 1, rowBytes * i), 0.U)
  }

  val hasData = RegInit(true.B)
  val isDirty = RegInit(false.B)
  when (io.mem_grant.fire) {
    w_grantfirst := true.B
    grant_param := io.mem_grant.bits.param
    when (edge.hasData(io.mem_grant.bits)) {
      // GrantData
      when (isKeyword) {
       for (i <- 0 until beatRows) {
         val idx = ((refill_count << log2Floor(beatRows)) + i.U) ^ 4.U
         val grant_row = io.mem_grant.bits.data(rowBits * (i + 1) - 1, rowBits * i)
         refill_and_store_data(idx) := mergePutData(grant_row, new_data(idx), new_mask(idx))
        }
      }
      .otherwise{
       for (i <- 0 until beatRows) {
         val idx = (refill_count << log2Floor(beatRows)) + i.U
         val grant_row = io.mem_grant.bits.data(rowBits * (i + 1) - 1, rowBits * i)
         refill_and_store_data(idx) := mergePutData(grant_row, new_data(idx), new_mask(idx))
        }
      }
      w_grantlast := w_grantlast || refill_done
      hasData := true.B
    }.otherwise {
      // Grant
      assert(full_overwrite)
      for (i <- 0 until blockRows) {
        refill_and_store_data(i) := new_data(i)
      }
      w_grantlast := true.B
      hasData := false.B
    }

    error := io.mem_grant.bits.denied || io.mem_grant.bits.corrupt || error

    refill_data_raw(refill_count ^ isKeyword) := io.mem_grant.bits.data
    isDirty := io.mem_grant.bits.echo.lift(DirtyKey).getOrElse(false.B)
  }

  when (io.mem_finish.fire) {
    s_grantack := true.B
  }

  when (io.main_pipe_req.fire) {
    s_mainpipe_req := true.B
    mainpipe_req_fired := true.B
  }

  when (io.main_pipe_replay) {
    s_mainpipe_req := false.B
  }

  when (io.main_pipe_resp) {
    w_mainpipe_resp := true.B
  }

  when(io.main_pipe_refill_resp) {
    w_refill_resp := true.B
  }

  when (io.l2_hint.valid) {
    w_l2hint := true.B
  }

  def before_req_sent_can_merge(new_req: MissReqWoStoreData): Bool = {
    // acquire_not_sent && (new_req.isFromLoad || new_req.isFromStore)

    // Since most acquire requests have been issued from pipe_reg,
    // the number of such merge situations is currently small,
    // So dont Merge anything for better timing.
    false.B
  }

  def before_data_refill_can_merge(new_req: MissReqWoStoreData): Bool = {
    data_not_refilled && new_req.isFromLoad
  }

  // Note that late prefetch will be ignored

  def should_merge(new_req: MissReqWoStoreData): Bool = {
    val block_match = get_block(req.addr) === get_block(new_req.addr)
    val alias_match = is_alias_match(req.vaddr, new_req.vaddr)
    block_match && alias_match &&
    (
      before_req_sent_can_merge(new_req) ||
      before_data_refill_can_merge(new_req)
    )
  }

  def before_req_sent_merge_iskeyword(new_req: MissReqWoStoreData): Bool = {
    val need_check_isKeyword = acquire_not_sent && req.isFromLoad && new_req.isFromLoad && should_merge(new_req)
    val use_new_req_isKeyword = isAfter(req.lqIdx, new_req.lqIdx)
    Mux(
      need_check_isKeyword,
      Mux(
        use_new_req_isKeyword,
        new_req.vaddr(5).asBool,
        req.vaddr(5).asBool
      ),
      isKeyword
      )
  }

  // store can be merged before io.mem_acquire.fire
  // store can not be merged the cycle that io.mem_acquire.fire
  // load can be merged before io.mem_grant.fire
  //
  // TODO: merge store if possible? mem_acquire may need to be re-issued,
  // but sbuffer entry can be freed
  def should_reject(new_req: MissReqWoStoreData): Bool = {
    val block_match = get_block(req.addr) === get_block(new_req.addr)
    val set_match = set === addr_to_dcache_set(new_req.vaddr)
    val alias_match = is_alias_match(req.vaddr, new_req.vaddr)

    req_valid && Mux(
        block_match,
        (!before_req_sent_can_merge(new_req) && !before_data_refill_can_merge(new_req)) || !alias_match,
        false.B
      )
  }

  // req_valid will be updated 1 cycle after primary_fire, so next cycle, this entry cannot accept a new req
  when(GatedValidRegNext(io.id >= ((cfg.nMissEntries).U - io.nMaxPrefetchEntry))) {
    // can accept prefetch req
    io.primary_ready := !req_valid && !GatedValidRegNext(primary_fire)
  }.otherwise {
    // cannot accept prefetch req except when a memset patten is detected
    io.primary_ready := !req_valid && (!io.req.bits.isFromPrefetch || io.memSetPattenDetected) && !GatedValidRegNext(primary_fire)
  }
  io.secondary_ready := should_merge(io.req.bits)
  io.secondary_reject := should_reject(io.req.bits)

  // generate primary_ready & secondary_(ready | reject) for each miss request
  for (i <- 0 until reqNum) {
    when(GatedValidRegNext(io.id >= ((cfg.nMissEntries).U - io.nMaxPrefetchEntry))) {
      io.queryME(i).primary_ready := !req_valid && !GatedValidRegNext(primary_fire)
    }.otherwise {
      io.queryME(i).primary_ready := !req_valid && !GatedValidRegNext(primary_fire) &&
                                    (!io.queryME(i).req.bits.isFromPrefetch || io.memSetPattenDetected)
    }
    io.queryME(i).secondary_ready  := should_merge(io.queryME(i).req.bits)
    io.queryME(i).secondary_reject := should_reject(io.queryME(i).req.bits)
  }

  // should not allocate, merge or reject at the same time
  assert(RegNext(PopCount(Seq(io.primary_ready, io.secondary_ready, io.secondary_reject)) <= 1.U || !io.req.valid))

  val refill_data_splited = WireInit(VecInit(Seq.tabulate(cfg.blockBytes * 8 / l1BusDataWidth)(i => {
    val data = refill_and_store_data.asUInt
    data((i + 1) * l1BusDataWidth - 1, i * l1BusDataWidth)
  })))
  // when granted data is all ready, wakeup lq's miss load
  val refill_to_ldq_en = !w_grantlast && io.mem_grant.fire
  io.refill_to_ldq.valid := GatedValidRegNext(refill_to_ldq_en)
  io.refill_to_ldq.bits.addr := RegEnable(req.addr + ((refill_count ^ isKeyword) << refillOffBits), refill_to_ldq_en)
  io.refill_to_ldq.bits.data := refill_data_splited(RegEnable(refill_count ^ isKeyword, refill_to_ldq_en))
  io.refill_to_ldq.bits.error := RegEnable(io.mem_grant.bits.corrupt || io.mem_grant.bits.denied, refill_to_ldq_en)
  io.refill_to_ldq.bits.refill_done := RegEnable(refill_done && io.mem_grant.fire, refill_to_ldq_en)
  io.refill_to_ldq.bits.hasdata := hasData
  io.refill_to_ldq.bits.data_raw := refill_data_raw.asUInt
  io.refill_to_ldq.bits.id := io.id

  // if the entry has a pending merge req, wait for it
  // Note: now, only wait for store, because store may acquire T
  io.mem_acquire.valid := !s_acquire && !(io.miss_req_pipe_reg.merge && !io.miss_req_pipe_reg.cancel && miss_req_pipe_reg_bits.isFromStore)
  val grow_param = req.req_coh.onAccess(req.cmd)._2
  val acquireBlock = edge.AcquireBlock(
    fromSource = io.id,
    toAddress = req.addr,
    lgSize = (log2Up(cfg.blockBytes)).U,
    growPermissions = grow_param
  )._2
  val acquirePerm = edge.AcquirePerm(
    fromSource = io.id,
    toAddress = req.addr,
    lgSize = (log2Up(cfg.blockBytes)).U,
    growPermissions = grow_param
  )._2
  io.mem_acquire.bits := Mux(full_overwrite, acquirePerm, acquireBlock)
  // resolve cache alias by L2
  io.mem_acquire.bits.user.lift(AliasKey).foreach( _ := req.vaddr(13, 12))
  // pass vaddr to l2
  io.mem_acquire.bits.user.lift(VaddrKey).foreach( _ := req.vaddr(VAddrBits-1, blockOffBits))
  // pass keyword to L2
  io.mem_acquire.bits.echo.lift(IsKeywordKey).foreach(_ := isKeyword)
  // trigger prefetch
  io.mem_acquire.bits.user.lift(PrefetchKey).foreach(_ := Mux(io.l2_pf_store_only, req.isFromStore, true.B))
  // req source
  when(prefetch && !secondary_fired) {
    io.mem_acquire.bits.user.lift(ReqSourceKey).foreach(_ := MemReqSource.L1DataPrefetch.id.U)
  }.otherwise {
    when(req.isFromStore) {
      io.mem_acquire.bits.user.lift(ReqSourceKey).foreach(_ := MemReqSource.CPUStoreData.id.U)
    }.elsewhen(req.isFromLoad) {
      io.mem_acquire.bits.user.lift(ReqSourceKey).foreach(_ := MemReqSource.CPULoadData.id.U)
    }.elsewhen(req.isFromAMO) {
      io.mem_acquire.bits.user.lift(ReqSourceKey).foreach(_ := MemReqSource.CPUAtomicData.id.U)
    }.otherwise {
      io.mem_acquire.bits.user.lift(ReqSourceKey).foreach(_ := MemReqSource.L1DataPrefetch.id.U)
    }
  }
  require(nSets <= 256)

  // io.mem_grant.ready := !w_grantlast && s_acquire
  io.mem_grant.ready := true.B
  assert(!(io.mem_grant.valid && !(!w_grantlast && s_acquire)), "dcache should always be ready for mem_grant now")

  val grantack = RegEnable(edge.GrantAck(io.mem_grant.bits), io.mem_grant.fire)
  assert(RegNext(!io.mem_grant.fire || edge.isRequest(io.mem_grant.bits)))
  io.mem_finish.valid := !s_grantack && w_grantfirst
  io.mem_finish.bits := grantack

  // Send mainpipe_req when receive hint from L2 or receive data without hint
  io.main_pipe_req.valid := !s_mainpipe_req && (w_l2hint || w_grantlast)
  io.main_pipe_req.bits := DontCare
  io.main_pipe_req.bits.miss := true.B
  io.main_pipe_req.bits.miss_id := io.id
  io.main_pipe_req.bits.probe := false.B
  io.main_pipe_req.bits.source := req.source
  io.main_pipe_req.bits.cmd := req.cmd
  io.main_pipe_req.bits.vaddr := req.vaddr
  io.main_pipe_req.bits.addr := req.addr
  io.main_pipe_req.bits.word_idx := req.word_idx
  io.main_pipe_req.bits.amo_data := req.amo_data
  io.main_pipe_req.bits.amo_mask := req.amo_mask
  io.main_pipe_req.bits.amo_cmp  := req.amo_cmp
  io.main_pipe_req.bits.id := req.id
  io.main_pipe_req.bits.pf_source := req.pf_source
  io.main_pipe_req.bits.access := access

  io.block_addr.valid := req_valid && w_grantlast
  io.block_addr.bits := req.addr

  io.req_addr.valid := req_valid
  io.req_addr.bits := req.addr

  io.refill_info.valid := req_valid && w_grantlast
  io.refill_info.bits.store_data := refill_and_store_data.asUInt
  io.refill_info.bits.store_mask := ~0.U(blockBytes.W)
  io.refill_info.bits.miss_param := grant_param
  io.refill_info.bits.miss_dirty := isDirty
  io.refill_info.bits.error      := error

  XSPerfAccumulate("miss_refill_mainpipe_req", io.main_pipe_req.fire)
  XSPerfAccumulate("miss_refill_without_hint", io.main_pipe_req.fire && !mainpipe_req_fired && !w_l2hint)
  XSPerfAccumulate("miss_refill_replay", io.main_pipe_replay)

  val w_grantfirst_forward_info = Mux(isKeyword, w_grantlast, w_grantfirst)
  val w_grantlast_forward_info = Mux(isKeyword, w_grantfirst, w_grantlast)
  io.forwardInfo.inflight := req_valid
  io.forwardInfo.paddr := req.addr
  io.forwardInfo.raw_data := refill_and_store_data
  io.forwardInfo.firstbeat_valid := w_grantfirst_forward_info
  io.forwardInfo.lastbeat_valid := w_grantlast_forward_info
  io.forwardInfo.corrupt := error

  io.matched := req_valid && (get_block(req.addr) === get_block(io.req.bits.addr)) && !prefetch
  io.prefetch_info.late_prefetch := io.req.valid && !(io.req.bits.isFromPrefetch) && req_valid && (get_block(req.addr) === get_block(io.req.bits.addr)) && prefetch

  when(io.prefetch_info.late_prefetch) {
    prefetch := false.B
  }

  io.l1Miss := req_valid
  // refill latency monitor
  val start_counting = GatedValidRegNext(io.mem_acquire.fire) || (GatedValidRegNextN(primary_fire, 2) && s_acquire)
  io.latency_monitor.load_miss_refilling  := req_valid && req_primary_fire.isFromLoad     && BoolStopWatch(start_counting, io.mem_grant.fire && !refill_done, true, true)
  io.latency_monitor.store_miss_refilling := req_valid && req_primary_fire.isFromStore    && BoolStopWatch(start_counting, io.mem_grant.fire && !refill_done, true, true)
  io.latency_monitor.amo_miss_refilling   := req_valid && req_primary_fire.isFromAMO      && BoolStopWatch(start_counting, io.mem_grant.fire && !refill_done, true, true)
  io.latency_monitor.pf_miss_refilling    := req_valid && req_primary_fire.isFromPrefetch && BoolStopWatch(start_counting, io.mem_grant.fire && !refill_done, true, true)

  XSPerfAccumulate("miss_req_primary", primary_fire)
  XSPerfAccumulate("miss_req_merged", secondary_fire)
  XSPerfAccumulate("load_miss_penalty_to_use",
    should_refill_data &&
      BoolStopWatch(primary_fire, io.refill_to_ldq.valid, true)
  )
  XSPerfAccumulate("penalty_between_grantlast_and_release",
    BoolStopWatch(!RegNext(w_grantlast) && w_grantlast, release_entry, true)
  )
  XSPerfAccumulate("main_pipe_penalty", BoolStopWatch(io.main_pipe_req.fire, io.main_pipe_resp))
  XSPerfAccumulate("penalty_blocked_by_channel_A", io.mem_acquire.valid && !io.mem_acquire.ready)
  XSPerfAccumulate("penalty_waiting_for_channel_D", s_acquire && !w_grantlast && !io.mem_grant.valid)
  XSPerfAccumulate("penalty_waiting_for_channel_E", io.mem_finish.valid && !io.mem_finish.ready)
  XSPerfAccumulate("prefetch_req_primary", primary_fire && io.req.bits.source === DCACHE_PREFETCH_SOURCE.U)
  XSPerfAccumulate("prefetch_req_merged", secondary_fire && io.req.bits.source === DCACHE_PREFETCH_SOURCE.U)
  XSPerfAccumulate("can_not_send_acquire_because_of_merging_store", !s_acquire && io.miss_req_pipe_reg.merge && io.miss_req_pipe_reg.cancel && miss_req_pipe_reg_bits.isFromStore)

  val (mshr_penalty_sample, mshr_penalty) = TransactionLatencyCounter(GatedValidRegNextN(primary_fire, 2), release_entry)
  XSPerfHistogram("miss_penalty", mshr_penalty, mshr_penalty_sample, 0, 20, 1, true, true)
  XSPerfHistogram("miss_penalty", mshr_penalty, mshr_penalty_sample, 20, 100, 10, true, false)

  val load_miss_begin = primary_fire && io.req.bits.isFromLoad
  val refill_finished = GatedValidRegNext(!w_grantlast && refill_done) && should_refill_data
  val (load_miss_penalty_sample, load_miss_penalty) = TransactionLatencyCounter(load_miss_begin, refill_finished) // not real refill finish time
  XSPerfHistogram("load_miss_penalty_to_use", load_miss_penalty, load_miss_penalty_sample, 0, 20, 1, true, true)
  XSPerfHistogram("load_miss_penalty_to_use", load_miss_penalty, load_miss_penalty_sample, 20, 100, 10, true, false)
  XSPerfHistogram("load_miss_penalty_to_use", load_miss_penalty, load_miss_penalty_sample, 100, 400, 30, true, false)

  val (a_to_d_penalty_sample, a_to_d_penalty) = TransactionLatencyCounter(start_counting, GatedValidRegNext(io.mem_grant.fire && refill_done))
  XSPerfHistogram("a_to_d_penalty", a_to_d_penalty, a_to_d_penalty_sample, 0, 20, 1, true, true)
  XSPerfHistogram("a_to_d_penalty", a_to_d_penalty, a_to_d_penalty_sample, 20, 100, 10, true, false)
}

class MissQueue(edge: TLEdgeOut, reqNum: Int)(implicit p: Parameters) extends DCacheModule
  with HasPerfEvents
  {
  val io = IO(new Bundle {
    val hartId = Input(UInt(hartIdLen.W))
    val req = Flipped(DecoupledIO(new MissReq))
    val resp = Output(new MissResp)
    val refill_to_ldq = ValidIO(new Refill)

    // cmo req
    val cmo_req = Flipped(DecoupledIO(new CMOReq))
    val cmo_resp = DecoupledIO(new CMOResp)

    val queryMQ = Vec(reqNum, Flipped(new DCacheMQQueryIOBundle))

    val mem_acquire = DecoupledIO(new TLBundleA(edge.bundle))
    val mem_grant = Flipped(DecoupledIO(new TLBundleD(edge.bundle)))
    val mem_finish = DecoupledIO(new TLBundleE(edge.bundle))

    val l2_hint = Input(Valid(new L2ToL1Hint())) // Hint from L2 Cache

    val main_pipe_req = DecoupledIO(new MainPipeReq)
    val main_pipe_resp = Flipped(ValidIO(new MainPipeResp))

    val mainpipe_info = Input(new MainPipeInfoToMQ)
    val refill_info = ValidIO(new MissQueueRefillInfo)

    // block probe
    val probe_addr = Input(UInt(PAddrBits.W))
    val probe_block = Output(Bool())

    // block replace when release an addr valid in mshr
    val replace_addr = Flipped(ValidIO(UInt(PAddrBits.W)))
    val replace_block = Output(Bool())

    // req blocked by wbq
    val wbq_block_miss_req = Input(Bool())

    val full = Output(Bool())

    // forward missqueue
    val forward = Vec(LoadPipelineWidth, new LduToMissqueueForwardIO)
    val l2_pf_store_only = Input(Bool())

    val memSetPattenDetected = Output(Bool())
    val lqEmpty = Input(Bool())

    val prefetch_info = new Bundle {
      val naive = new Bundle {
        val late_miss_prefetch = Output(Bool())
      }

      val fdp = new Bundle {
        val late_miss_prefetch = Output(Bool())
        val prefetch_monitor_cnt = Output(Bool())
        val total_prefetch = Output(Bool())
      }
    }

    val debugTopDown = new DCacheTopDownIO
    val l1Miss = Output(Bool())
  })

  // 128KBL1: FIXME: provide vaddr for l2

  val entries = Seq.fill(cfg.nMissEntries)(Module(new MissEntry(edge, reqNum)))
  val cmo_unit = Module(new CMOUnit(edge))

  val miss_req_pipe_reg = RegInit(0.U.asTypeOf(new MissReqPipeRegBundle(edge)))
  val acquire_from_pipereg = Wire(chiselTypeOf(io.mem_acquire))

  val primary_ready_vec = entries.map(_.io.primary_ready)
  val secondary_ready_vec = entries.map(_.io.secondary_ready)
  val secondary_reject_vec = entries.map(_.io.secondary_reject)
  val probe_block_vec = entries.map { case e => e.io.block_addr.valid && e.io.block_addr.bits === io.probe_addr }

  val merge = ParallelORR(Cat(secondary_ready_vec ++ Seq(miss_req_pipe_reg.merge_req(io.req.bits))))
  val reject = ParallelORR(Cat(secondary_reject_vec ++ Seq(miss_req_pipe_reg.reject_req(io.req.bits))))
  val alloc = !reject && !merge && ParallelORR(Cat(primary_ready_vec))
  val accept = alloc || merge

  // generate req_ready for each miss request for better timing
  for (i <- 0 until reqNum) {
    val _primary_ready_vec = entries.map(_.io.queryME(i).primary_ready)
    val _secondary_ready_vec = entries.map(_.io.queryME(i).secondary_ready)
    val _secondary_reject_vec = entries.map(_.io.queryME(i).secondary_reject)
    val _merge = ParallelORR(Cat(_secondary_ready_vec ++ Seq(miss_req_pipe_reg.merge_req(io.queryMQ(i).req.bits))))
    val _reject = ParallelORR(Cat(_secondary_reject_vec ++ Seq(miss_req_pipe_reg.reject_req(io.queryMQ(i).req.bits))))
    val _alloc = !_reject && !_merge && ParallelORR(Cat(_primary_ready_vec))
    val _accept = _alloc || _merge

    io.queryMQ(i).ready := _accept
  }

  val req_mshr_handled_vec = entries.map(_.io.req_handled_by_this_entry)
  // merged to pipeline reg
  val req_pipeline_reg_handled = miss_req_pipe_reg.merge_req(io.req.bits) && io.req.valid
  assert(PopCount(Seq(req_pipeline_reg_handled, VecInit(req_mshr_handled_vec).asUInt.orR)) <= 1.U, "miss req will either go to mshr or pipeline reg")
  assert(PopCount(req_mshr_handled_vec) <= 1.U, "Only one mshr can handle a req")
  io.resp.id := Mux(!req_pipeline_reg_handled, OHToUInt(req_mshr_handled_vec), miss_req_pipe_reg.mshr_id)
  io.resp.handled := Cat(req_mshr_handled_vec).orR || req_pipeline_reg_handled
  io.resp.merged := merge

  /*  MissQueue enq logic is now splitted into 2 cycles
   *
   */
  when(io.req.valid){
    miss_req_pipe_reg.req     := io.req.bits
  }
  // miss_req_pipe_reg.req     := io.req.bits
  miss_req_pipe_reg.alloc   := alloc && io.req.valid && !io.req.bits.cancel && !io.wbq_block_miss_req
  miss_req_pipe_reg.merge   := merge && io.req.valid && !io.req.bits.cancel && !io.wbq_block_miss_req
  miss_req_pipe_reg.cancel  := io.wbq_block_miss_req
  miss_req_pipe_reg.mshr_id := io.resp.id

  assert(PopCount(Seq(alloc && io.req.valid, merge && io.req.valid)) <= 1.U, "allocate and merge a mshr in same cycle!")

  val source_except_load_cnt = RegInit(0.U(10.W))
  when(VecInit(req_mshr_handled_vec).asUInt.orR || req_pipeline_reg_handled) {
    when(io.req.bits.isFromLoad) {
      source_except_load_cnt := 0.U
    }.otherwise {
      when(io.req.bits.isFromStore) {
        source_except_load_cnt := source_except_load_cnt + 1.U
      }
    }
  }
  val Threshold = 8
  val memSetPattenDetected = GatedValidRegNext((source_except_load_cnt >= Threshold.U) && io.lqEmpty)

  io.memSetPattenDetected := memSetPattenDetected

  val forwardInfo_vec = VecInit(entries.map(_.io.forwardInfo))
  (0 until LoadPipelineWidth).map(i => {
    val id = io.forward(i).mshrid
    val req_valid = io.forward(i).valid
    val paddr = io.forward(i).paddr

    val (forward_mshr, forwardData) = forwardInfo_vec(id).forward(req_valid, paddr)
    io.forward(i).forward_result_valid := forwardInfo_vec(id).check(req_valid, paddr)
    io.forward(i).forward_mshr := forward_mshr
    io.forward(i).forwardData := forwardData
    io.forward(i).corrupt := RegNext(forwardInfo_vec(id).corrupt)
  })

  assert(RegNext(PopCount(secondary_ready_vec) <= 1.U || !io.req.valid))
//  assert(RegNext(PopCount(secondary_reject_vec) <= 1.U))
  // It is possible that one mshr wants to merge a req, while another mshr wants to reject it.
  // That is, a coming req has the same paddr as that of mshr_0 (merge),
  // while it has the same set and the same way as mshr_1 (reject).
  // In this situation, the coming req should be merged by mshr_0
//  assert(RegNext(PopCount(Seq(merge, reject)) <= 1.U))

  def select_valid_one[T <: Bundle](
    in: Seq[DecoupledIO[T]],
    out: DecoupledIO[T],
    name: Option[String] = None): Unit = {

    if (name.nonEmpty) { out.suggestName(s"${name.get}_select") }
    out.valid := Cat(in.map(_.valid)).orR
    out.bits := ParallelMux(in.map(_.valid) zip in.map(_.bits))
    in.map(_.ready := out.ready)
    assert(!RegNext(out.valid && PopCount(Cat(in.map(_.valid))) > 1.U))
  }

  io.mem_grant.ready := false.B

  val nMaxPrefetchEntry = Constantin.createRecord(s"nMaxPrefetchEntry${p(XSCoreParamsKey).HartId}", initValue = 14)
  entries.zipWithIndex.foreach {
    case (e, i) =>
      val former_primary_ready = if(i == 0)
        false.B
      else
        Cat((0 until i).map(j => entries(j).io.primary_ready)).orR

      e.io.hartId := io.hartId
      e.io.id := i.U
      e.io.l2_pf_store_only := io.l2_pf_store_only
      e.io.req.valid := io.req.valid
      e.io.wbq_block_miss_req := io.wbq_block_miss_req
      e.io.primary_valid := io.req.valid &&
        !merge &&
        !reject &&
        !former_primary_ready &&
        e.io.primary_ready
      e.io.req.bits := io.req.bits.toMissReqWoStoreData()

      e.io.mem_grant.valid := false.B
      e.io.mem_grant.bits := DontCare
      when (io.mem_grant.bits.source === i.U) {
        e.io.mem_grant <> io.mem_grant
      }

      when(miss_req_pipe_reg.reg_valid() && miss_req_pipe_reg.mshr_id === i.U) {
        e.io.miss_req_pipe_reg := miss_req_pipe_reg
      }.otherwise {
        e.io.miss_req_pipe_reg       := DontCare
        e.io.miss_req_pipe_reg.merge := false.B
        e.io.miss_req_pipe_reg.alloc := false.B
      }

      e.io.acquire_fired_by_pipe_reg := acquire_from_pipereg.fire

      e.io.main_pipe_resp := io.main_pipe_resp.valid && io.main_pipe_resp.bits.ack_miss_queue && io.main_pipe_resp.bits.miss_id === i.U
      e.io.main_pipe_replay := io.mainpipe_info.s2_valid && io.mainpipe_info.s2_replay_to_mq && io.mainpipe_info.s2_miss_id === i.U
      e.io.main_pipe_refill_resp := io.mainpipe_info.s3_valid && io.mainpipe_info.s3_refill_resp && io.mainpipe_info.s3_miss_id === i.U

      e.io.memSetPattenDetected := memSetPattenDetected
      e.io.nMaxPrefetchEntry := nMaxPrefetchEntry

      e.io.main_pipe_req.ready := io.main_pipe_req.ready

      for (j <- 0 until reqNum) {
        e.io.queryME(j).req.valid := io.queryMQ(j).req.valid
        e.io.queryME(j).req.bits  := io.queryMQ(j).req.bits.toMissReqWoStoreData()
      }

      when(io.l2_hint.bits.sourceId === i.U) {
        e.io.l2_hint <> io.l2_hint
      } .otherwise {
        e.io.l2_hint.valid := false.B
        e.io.l2_hint.bits := DontCare
      }
  }

  cmo_unit.io.req <> io.cmo_req
  io.cmo_resp <> cmo_unit.io.resp_to_lsq
  when (io.mem_grant.valid && io.mem_grant.bits.opcode === TLMessages.CBOAck) {
    cmo_unit.io.resp_chanD <> io.mem_grant
  } .otherwise {
    cmo_unit.io.resp_chanD.valid := false.B
    cmo_unit.io.resp_chanD.bits := DontCare
  }

  io.req.ready := accept
  io.refill_to_ldq.valid := Cat(entries.map(_.io.refill_to_ldq.valid)).orR
  io.refill_to_ldq.bits := ParallelMux(entries.map(_.io.refill_to_ldq.valid) zip entries.map(_.io.refill_to_ldq.bits))

  io.refill_info.valid := VecInit(entries.zipWithIndex.map{ case(e,i) => e.io.refill_info.valid && io.mainpipe_info.s2_valid && io.mainpipe_info.s2_miss_id === i.U}).asUInt.orR
  io.refill_info.bits := Mux1H(entries.zipWithIndex.map{ case(e,i) => (io.mainpipe_info.s2_miss_id === i.U) -> e.io.refill_info.bits })

  acquire_from_pipereg.valid := miss_req_pipe_reg.can_send_acquire(io.req.valid, io.req.bits)
  acquire_from_pipereg.bits := miss_req_pipe_reg.get_acquire(io.l2_pf_store_only)

  XSPerfAccumulate("acquire_fire_from_pipereg", acquire_from_pipereg.fire)
  XSPerfAccumulate("pipereg_valid", miss_req_pipe_reg.reg_valid())

  val acquire_sources = Seq(cmo_unit.io.req_chanA, acquire_from_pipereg) ++ entries.map(_.io.mem_acquire)
  TLArbiter.lowest(edge, io.mem_acquire, acquire_sources:_*)
  TLArbiter.lowest(edge, io.mem_finish, entries.map(_.io.mem_finish):_*)

  // amo's main pipe req out
  fastArbiter(entries.map(_.io.main_pipe_req), io.main_pipe_req, Some("main_pipe_req"))

  io.probe_block := Cat(probe_block_vec).orR

  io.replace_block := io.replace_addr.valid && Cat(entries.map(e => e.io.req_addr.valid && e.io.req_addr.bits === io.replace_addr.bits) ++ Seq(miss_req_pipe_reg.block_match(io.replace_addr.bits))).orR

  io.full := ~Cat(entries.map(_.io.primary_ready)).andR

  // prefetch related
  io.prefetch_info.naive.late_miss_prefetch := io.req.valid && io.req.bits.isPrefetchRead && (miss_req_pipe_reg.matched(io.req.bits) || Cat(entries.map(_.io.matched)).orR)

  io.prefetch_info.fdp.late_miss_prefetch := (miss_req_pipe_reg.prefetch_late_en(io.req.bits.toMissReqWoStoreData(), io.req.valid) || Cat(entries.map(_.io.prefetch_info.late_prefetch)).orR)
  io.prefetch_info.fdp.prefetch_monitor_cnt := io.main_pipe_req.fire
  io.prefetch_info.fdp.total_prefetch := alloc && io.req.valid && !io.req.bits.cancel && isFromL1Prefetch(io.req.bits.pf_source)

  // L1MissTrace Chisel DB
  val debug_miss_trace = Wire(new L1MissTrace)
  debug_miss_trace.vaddr := io.req.bits.vaddr
  debug_miss_trace.paddr := io.req.bits.addr
  debug_miss_trace.source := io.req.bits.source
  debug_miss_trace.pc := io.req.bits.pc

  val isWriteL1MissQMissTable = Constantin.createRecord(s"isWriteL1MissQMissTable${p(XSCoreParamsKey).HartId}")
  val table = ChiselDB.createTable(s"L1MissQMissTrace_hart${p(XSCoreParamsKey).HartId}", new L1MissTrace)
  table.log(debug_miss_trace, isWriteL1MissQMissTable.orR && io.req.valid && !io.req.bits.cancel && alloc, "MissQueue", clock, reset)

  // Difftest
  if (env.EnableDifftest) {
    val difftest = DifftestModule(new DiffRefillEvent, dontCare = true)
    difftest.coreid := io.hartId
    difftest.index := 1.U
    difftest.valid := io.refill_to_ldq.valid && io.refill_to_ldq.bits.hasdata && io.refill_to_ldq.bits.refill_done
    difftest.addr := io.refill_to_ldq.bits.addr
    difftest.data := io.refill_to_ldq.bits.data_raw.asTypeOf(difftest.data)
    difftest.idtfr := DontCare
  }

  // Perf count
  XSPerfAccumulate("miss_req", io.req.fire && !io.req.bits.cancel)
  XSPerfAccumulate("miss_req_allocate", io.req.fire && !io.req.bits.cancel && alloc)
  XSPerfAccumulate("miss_req_load_allocate", io.req.fire && !io.req.bits.cancel && alloc && io.req.bits.isFromLoad)
  XSPerfAccumulate("miss_req_store_allocate", io.req.fire && !io.req.bits.cancel && alloc && io.req.bits.isFromStore)
  XSPerfAccumulate("miss_req_amo_allocate", io.req.fire && !io.req.bits.cancel && alloc && io.req.bits.isFromAMO)
  XSPerfAccumulate("miss_req_prefetch_allocate", io.req.fire && !io.req.bits.cancel && alloc && io.req.bits.isFromPrefetch)
  XSPerfAccumulate("miss_req_merge_load", io.req.fire && !io.req.bits.cancel && merge && io.req.bits.isFromLoad)
  XSPerfAccumulate("miss_req_reject_load", io.req.valid && !io.req.bits.cancel && reject && io.req.bits.isFromLoad)
  XSPerfAccumulate("probe_blocked_by_miss", io.probe_block)
  XSPerfAccumulate("prefetch_primary_fire", io.req.fire && !io.req.bits.cancel && alloc && io.req.bits.isFromPrefetch)
  XSPerfAccumulate("prefetch_secondary_fire", io.req.fire && !io.req.bits.cancel && merge && io.req.bits.isFromPrefetch)
  XSPerfAccumulate("memSetPattenDetected", memSetPattenDetected)
  val max_inflight = RegInit(0.U((log2Up(cfg.nMissEntries) + 1).W))
  val num_valids = PopCount(~Cat(primary_ready_vec).asUInt)
  when (num_valids > max_inflight) {
    max_inflight := num_valids
  }
  // max inflight (average) = max_inflight_total / cycle cnt
  XSPerfAccumulate("max_inflight", max_inflight)
  QueuePerf(cfg.nMissEntries, num_valids, num_valids === cfg.nMissEntries.U)
  io.full := num_valids === cfg.nMissEntries.U
  io.l1Miss := RegNext(Cat(entries.map(_.io.l1Miss)).orR)
  XSPerfHistogram("num_valids", num_valids, true.B, 0, cfg.nMissEntries, 1)

  XSPerfHistogram("L1DMLP_CPUData", PopCount(VecInit(entries.map(_.io.perf_pending_normal)).asUInt), true.B, 0, cfg.nMissEntries, 1)
  XSPerfHistogram("L1DMLP_Prefetch", PopCount(VecInit(entries.map(_.io.perf_pending_prefetch)).asUInt), true.B, 0, cfg.nMissEntries, 1)
  XSPerfHistogram("L1DMLP_Total", num_valids, true.B, 0, cfg.nMissEntries, 1)

  XSPerfAccumulate("miss_load_refill_latency", PopCount(entries.map(_.io.latency_monitor.load_miss_refilling)))
  XSPerfAccumulate("miss_store_refill_latency", PopCount(entries.map(_.io.latency_monitor.store_miss_refilling)))
  XSPerfAccumulate("miss_amo_refill_latency", PopCount(entries.map(_.io.latency_monitor.amo_miss_refilling)))
  XSPerfAccumulate("miss_pf_refill_latency", PopCount(entries.map(_.io.latency_monitor.pf_miss_refilling)))

  val rob_head_miss_in_dcache = VecInit(entries.map(_.io.rob_head_query.resp)).asUInt.orR

  entries.foreach {
    case e => {
      e.io.rob_head_query.query_valid := io.debugTopDown.robHeadVaddr.valid
      e.io.rob_head_query.vaddr := io.debugTopDown.robHeadVaddr.bits
    }
  }

  io.debugTopDown.robHeadMissInDCache := rob_head_miss_in_dcache

  val perfValidCount = RegNext(PopCount(entries.map(entry => (!entry.io.primary_ready))))
  val perfEvents = Seq(
    ("dcache_missq_req      ", io.req.fire),
    ("dcache_missq_1_4_valid", (perfValidCount < (cfg.nMissEntries.U/4.U))),
    ("dcache_missq_2_4_valid", (perfValidCount > (cfg.nMissEntries.U/4.U)) & (perfValidCount <= (cfg.nMissEntries.U/2.U))),
    ("dcache_missq_3_4_valid", (perfValidCount > (cfg.nMissEntries.U/2.U)) & (perfValidCount <= (cfg.nMissEntries.U*3.U/4.U))),
    ("dcache_missq_4_4_valid", (perfValidCount > (cfg.nMissEntries.U*3.U/4.U))),
  )
  generatePerfEvent()
}