updates

src/main/scala/gemmini/Configs.scala

@@ -4,7 +4,7 @@ import chisel3._
 import org.chipsalliance.cde.config.{Config, Parameters}
 import freechips.rocketchip.diplomacy.LazyModule
 import freechips.rocketchip.subsystem._
-import freechips.rocketchip.tile.{BuildRoCC, OpcodeSet, XLen}
+import freechips.rocketchip.tile.{BuildRoCC, OpcodeSet}
 import freechips.rocketchip.rocket._
 import freechips.rocketchip.tile._
 import freechips.rocketchip.system._

src/main/scala/gemmini/Controller.scala

@@ -33,7 +33,7 @@ class Gemmini[T <: Data : Arithmetic, U <: Data, V <: Data](val config: GemminiA
     System.exit(1)
   }
 
-  val xLen = p(XLen)
+  val xLen = p(TileKey).core.xLen
   val spad = LazyModule(new Scratchpad(config))
 
   override lazy val module = new GemminiModule(this)
@@ -415,7 +415,7 @@ class GemminiModule[T <: Data: Arithmetic, U <: Data, V <: Data]
 
   // Debugging signals
   val pipeline_stall_counter = RegInit(0.U(32.W))
-  when (io.cmd.fire()) {
+  when (io.cmd.fire) {
     pipeline_stall_counter := 0.U
   }.elsewhen(io.busy) {
     pipeline_stall_counter := pipeline_stall_counter + 1.U

src/main/scala/gemmini/DMA.scala

@@ -3,7 +3,6 @@ package gemmini
 
 import chisel3._
 import chisel3.util._
-import chisel3.experimental.DataMirror
 
 import org.chipsalliance.cde.config.Parameters
 import freechips.rocketchip.diplomacy.{IdRange, LazyModule, LazyModuleImp}
@@ -211,7 +210,7 @@ class StreamReaderCore[T <: Data, U <: Data, V <: Data](config: GemminiArrayConf
     )._2
 
     class TLBundleAWithInfo extends Bundle {
-      val tl_a = DataMirror.internal.chiselTypeClone[TLBundleA](tl.a.bits)
+      val tl_a = tl.a.bits.cloneType
       val vaddr = Output(UInt(vaddrBits.W))
       val status = Output(new MStatus)
     }
@@ -499,7 +498,7 @@ class StreamWriter[T <: Data: Arithmetic](nXacts: Int, beatBits: Int, maxBytes:
     )._2
 
     class TLBundleAWithInfo extends Bundle {
-      val tl_a = DataMirror.internal.chiselTypeClone[TLBundleA](tl.a.bits)
+      val tl_a = tl.a.bits.cloneType
       val vaddr = Output(UInt(vaddrBits.W))
       val status = Output(new MStatus)
     }

src/main/scala/gemmini/LoopConv.scala

@@ -1026,7 +1026,7 @@ class LoopConvSt(block_size: Int, coreMaxAddrBits: Int, large_iterator_bitwidth:
   io.loop_id := req.loop_id
 
   command_p.io.in.valid := state =/= idle && !skip && io.ex_completed
-  command_p.io.in.bits.cmd := MuxLookup(state.asUInt, mvout_cmd, Seq(
+  command_p.io.in.bits.cmd := MuxLookup(state.asUInt, mvout_cmd) (Seq(
     pre_pool_config.asUInt -> pre_pool_config_cmd,
     pool.asUInt -> pool_cmd,
     post_pool_config.asUInt -> post_pool_config_cmd)

src/main/scala/gemmini/LoopMatadd.scala

@@ -511,7 +511,7 @@ class LoopMataddStC(block_size: Int, coreMaxAddrBits: Int, iterator_bitwidth: In
 
   when (req.dram_addr === 0.U) {
     state := idle
-  }.elsewhen (io.cmd.fire() && state === st) {
+  }.elsewhen (io.cmd.fire && state === st) {
     // The order here is k, j, i
     val next_i = floorAdd(i, 1.U, req.max_i)
     val next_j = floorAdd(j, max_blocks, req.max_j(b), next_i === 0.U)

src/main/scala/gemmini/LoopMatmul.scala

@@ -606,7 +606,7 @@ class LoopMatmulStC(block_size: Int, coreMaxAddrBits: Int, iterator_bitwidth: In
 
   when (req.dram_addr === 0.U) {
     state := idle
-  }.elsewhen (io.cmd.fire() && state === st) {
+  }.elsewhen (io.cmd.fire && state === st) {
     // The order here is k, j, i
     val next_i = floorAdd(i, 1.U, req.max_i)
     val next_j = floorAdd(j, max_blocks, req.max_j, next_i === 0.U)

src/main/scala/gemmini/NormCmd.scala

@@ -3,7 +3,6 @@ package gemmini
 
 import chisel3._
 import chisel3.util._
-import chisel3.experimental.ChiselEnum
 
 object NormCmd extends ChiselEnum {
   val RESET, SUM, MEAN, VARIANCE, INV_STDDEV, MAX, SUM_EXP, INV_SUM_EXP = Value

src/main/scala/gemmini/Normalizer.scala

@@ -2,7 +2,6 @@
 package gemmini
 
 import chisel3._
-import chisel3.experimental.ChiselEnum
 import chisel3.util._
 import gemmini.AccumulatorScale.iexp
 import hardfloat.{DivSqrtRecFN_small, INToRecFN, MulRecFN, consts, fNFromRecFN, recFNFromFN}
@@ -348,7 +347,7 @@ class Normalizer[T <: Data, U <: Data](max_len: Int, num_reduce_lanes: Int, num_
     lanes.io.ins.bits.stats_id := in_lanes_stats_id
     lanes.io.ins.bits.iexp_const := iexp_const
 
-    when (lanes.io.ins.fire()) {
+    when (lanes.io.ins.fire) {
       stat.elems_left := stat.elems_left - len
     }
   }
@@ -359,7 +358,7 @@ class Normalizer[T <: Data, U <: Data](max_len: Int, num_reduce_lanes: Int, num_
 
     val stat = stats(out_lanes_stats_id)
 
-    when (lanes.io.out.fire()) {
+    when (lanes.io.out.fire) {
       stat.sum := stat.sum + lanes.io.out.bits.result
     }
   }
@@ -379,7 +378,7 @@ class Normalizer[T <: Data, U <: Data](max_len: Int, num_reduce_lanes: Int, num_
     max_lanes.io.ins.bits.len := len
     max_lanes.io.ins.bits.stats_id := max_in_lanes_stats_id
 
-    when (max_lanes.io.ins.fire()) {
+    when (max_lanes.io.ins.fire) {
       stat.elems_left := stat.elems_left - len
     }
   }
@@ -390,7 +389,7 @@ class Normalizer[T <: Data, U <: Data](max_len: Int, num_reduce_lanes: Int, num_
 
     val stat = stats(max_out_lanes_stats_id)
 
-    when (max_lanes.io.out.fire()) {
+    when (max_lanes.io.out.fire) {
       val new_max = Mux(max_lanes.io.out.bits.result > stat.running_max, max_lanes.io.out.bits.result, stat.running_max)
       stat.running_max := new_max
       stat.max := new_max
@@ -645,13 +644,13 @@ class Normalizer[T <: Data, U <: Data](max_len: Int, num_reduce_lanes: Int, num_
       next_state := idle
       done := DontCare
     }.elsewhen(state === output) {
-      next_state := Mux(io.out.fire() && out_stats_id === id.U, idle, state)
-      done := io.out.fire() && out_stats_id === id.U
+      next_state := Mux(io.out.fire && out_stats_id === id.U, idle, state)
+      done := io.out.fire && out_stats_id === id.U
     }.elsewhen(state === get_max) {
       val is_last_lane_input = stat.vec_groups_left === 0.U ||
         (stat.vec_groups_left === 1.U &&
           max_lanes.io.ins.bits.stats_id === id.U &&
-          max_lanes.io.ins.fire())
+          max_lanes.io.ins.fire)
 
       next_state := Mux(
         is_last_lane_input,
@@ -667,7 +666,7 @@ class Normalizer[T <: Data, U <: Data](max_len: Int, num_reduce_lanes: Int, num_
       val is_last_lane_input = stat.vec_groups_left === 0.U ||
         (stat.vec_groups_left === 1.U &&
           lanes.io.ins.bits.stats_id === id.U &&
-          lanes.io.ins.fire())
+          lanes.io.ins.fire)
 
       next_state := Mux(
         is_last_lane_input,
@@ -688,51 +687,51 @@ class Normalizer[T <: Data, U <: Data](max_len: Int, num_reduce_lanes: Int, num_
 
       done := is_last_lane_input && cmd =/= NormCmd.MEAN && cmd =/= NormCmd.INV_STDDEV && cmd =/= NormCmd.INV_SUM_EXP
     }.elsewhen(state === get_mean || state === get_variance) {
-      next_state := Mux(divider_in.fire() && sum_to_divide_id === id.U, state.next, state)
+      next_state := Mux(divider_in.fire && sum_to_divide_id === id.U, state.next, state)
       done := false.B
     }.elsewhen(state === waiting_for_mean) {
-      next_state := Mux(divider_out.fire(), idle, state)
-      done := divider_out.fire()
+      next_state := Mux(divider_out.fire, idle, state)
+      done := divider_out.fire
     }.elsewhen(state === waiting_for_variance) {
-      next_state := Mux(divider_out.fire(), get_stddev, state)
+      next_state := Mux(divider_out.fire, get_stddev, state)
       done := false.B
     }.elsewhen(state === get_stddev) {
-      next_state := Mux(sqrt_in.fire() && variance_to_sqrt_id === id.U, state.next, state)
+      next_state := Mux(sqrt_in.fire && variance_to_sqrt_id === id.U, state.next, state)
       done := false.B
     }.elsewhen(state === waiting_for_stddev) {
-      next_state := Mux(sqrt_out.fire(), state.next, state)
+      next_state := Mux(sqrt_out.fire, state.next, state)
       done := false.B
     }.elsewhen(state === get_inv_stddev) {
-      next_state := Mux(reciprocal_in.fire() && stddev_to_inv_id === id.U, state.next, state)
+      next_state := Mux(reciprocal_in.fire && stddev_to_inv_id === id.U, state.next, state)
       done := false.B
     }.elsewhen(state === waiting_for_inv_stddev) {
-      next_state := Mux(reciprocal_out.fire(), state.next, state)
+      next_state := Mux(reciprocal_out.fire, state.next, state)
       done := false.B
     }.elsewhen(state === get_scaled_inv_stddev) {
-      next_state := Mux(inv_stddev_scale_mul_pipe.io.ins.fire() && inv_stddev_to_scale_id === id.U, state.next, state)
+      next_state := Mux(inv_stddev_scale_mul_pipe.io.ins.fire && inv_stddev_to_scale_id === id.U, state.next, state)
       done := false.B
     }.elsewhen(state === waiting_for_scaled_inv_stddev) {
-      next_state := Mux(inv_stddev_scale_mul_pipe.io.out.fire(), idle, state)
-      done := inv_stddev_scale_mul_pipe.io.out.fire()
+      next_state := Mux(inv_stddev_scale_mul_pipe.io.out.fire, idle, state)
+      done := inv_stddev_scale_mul_pipe.io.out.fire
     }.elsewhen(state === get_inv_sum_exp) {
-      next_state := Mux(exp_divider_in.fire() && sum_exp_to_inv_id === id.U, state.next, state)
+      next_state := Mux(exp_divider_in.fire && sum_exp_to_inv_id === id.U, state.next, state)
       done := false.B
     }.elsewhen(state === waiting_for_inv_sum_exp) {
-      next_state := Mux(exp_divider_out.fire(), state.next, state)
+      next_state := Mux(exp_divider_out.fire, state.next, state)
       done := false.B
     }.elsewhen(state === get_scaled_inv_sum_exp) {
-      next_state := Mux(inv_sum_exp_scale_mul_pipe.io.ins.fire() && inv_sum_exp_to_scale_id === id.U, state.next, state)
+      next_state := Mux(inv_sum_exp_scale_mul_pipe.io.ins.fire && inv_sum_exp_to_scale_id === id.U, state.next, state)
       done := false.B
     }.elsewhen(state === waiting_for_scaled_inv_sum_exp) {
-      next_state := Mux(inv_sum_exp_scale_mul_pipe.io.out.fire(), idle, state)
-      done := inv_sum_exp_scale_mul_pipe.io.out.fire()
+      next_state := Mux(inv_sum_exp_scale_mul_pipe.io.out.fire, idle, state)
+      done := inv_sum_exp_scale_mul_pipe.io.out.fire
     }.otherwise {
       assert(false.B, "invalid state in Normalizer")
       next_state := DontCare
       done := DontCare
     }
 
-    when (io.in.fire() && in_stats_id === id.U) {
+    when (io.in.fire && in_stats_id === id.U) {
       next_state := Mux(io.in.bits.cmd === NormCmd.RESET, output,
         Mux(io.in.bits.cmd === NormCmd.MAX, get_max, get_sum))
     }
@@ -747,7 +746,7 @@ class Normalizer[T <: Data, U <: Data](max_len: Int, num_reduce_lanes: Int, num_
         (state === get_mean && next_state =/= get_mean) ||
         (state === get_variance && next_state =/= get_variance)
 
-    val is_input = io.in.fire() && in_stats_id === id.U
+    val is_input = io.in.fire && in_stats_id === id.U
 
     when (is_input) {
       stat.req := io.in.bits

src/main/scala/gemmini/ReservationStation.scala

@@ -19,7 +19,7 @@ class ReservationStationIssue[T <: Data](cmd_t: T, id_width: Int) extends Bundle
   val cmd = Output(cmd_t.cloneType)
   val rob_id = Output(UInt(id_width.W))
 
-  def fire(dummy: Int=0) = valid && ready
+  def fire = valid && ready
 }
 
 // TODO we don't need to store the full command in here. We should be able to release the command directly into the relevant controller and only store the associated metadata in the ROB. This would reduce the size considerably
@@ -191,7 +191,7 @@ class ReservationStation[T <: Data : Arithmetic, U <: Data, V <: Data](config: G
   val new_entry_oh = new_allocs_oh_ld ++ new_allocs_oh_ex ++ new_allocs_oh_st
   new_entry_oh.foreach(_ := false.B)
 
-  val alloc_fire = io.alloc.fire()
+  val alloc_fire = io.alloc.fire
 
   io.alloc.ready := false.B
   when (io.alloc.valid) {
@@ -424,7 +424,7 @@ class ReservationStation[T <: Data : Arithmetic, U <: Data, V <: Data](config: G
     val from_matmul_fsm = entries_type(issue_id).bits.cmd.from_matmul_fsm
     val from_matadd_fsm = entries_type(issue_id).bits.cmd.from_matadd_fsm
 
-    when (io.fire()) {
+    when (io.fire) {
       entries_type.zipWithIndex.foreach { case (e, i) =>
         when (issue_sel(i)) {
           e.bits.issued := true.B
@@ -538,7 +538,7 @@ class ReservationStation[T <: Data : Arithmetic, U <: Data, V <: Data](config: G
 
   val cycles_since_issue = RegInit(0.U(16.W))
 
-  when (io.issue.ld.fire() || io.issue.st.fire() || io.issue.ex.fire() || !io.busy || io.completed.fire) {
+  when (io.issue.ld.fire || io.issue.st.fire || io.issue.ex.fire || !io.busy || io.completed.fire) {
     cycles_since_issue := 0.U
   }.elsewhen(io.busy) {
     cycles_since_issue := cycles_since_issue + 1.U

src/main/scala/gemmini/Scratchpad.scala

@@ -597,7 +597,7 @@ class Scratchpad[T <: Data, U <: Data, V <: Data](config: GemminiArrayConfig[T,
     acc_scale_unit.io.in.valid := acc_norm_unit_out.valid && acc_waiting_to_be_scaled
     acc_scale_unit.io.in.bits  := acc_norm_unit_out.bits
 
-    when (acc_scale_unit.io.in.fire()) {
+    when (acc_scale_unit.io.in.fire) {
       write_issue_q.io.enq <> write_scale_q.io.deq
     }
 

src/main/scala/gemmini/TransposePreloadUnroller.scala

@@ -2,7 +2,6 @@ package gemmini
 
 import chisel3._
 import chisel3.util._
-import chisel3.experimental.ChiselEnum
 import org.chipsalliance.cde.config.Parameters
 import Util._
 import midas.targetutils.PerfCounter