diff --git a/src/main/scala/radiance/memory/DoubleOutXbar.scala b/src/main/scala/radiance/memory/DoubleOutXbar.scala
index 20c5eec..67a9158 100644
--- a/src/main/scala/radiance/memory/DoubleOutXbar.scala
+++ b/src/main/scala/radiance/memory/DoubleOutXbar.scala
@@ -2,7 +2,7 @@ package radiance.memory
 
 import chisel3._
 import chisel3.util._
-import freechips.rocketchip.diplomacy.{AddressSet, TransferSizes, IdRange}
+import freechips.rocketchip.diplomacy.{AddressSet, TransferSizes, IdRange, BufferParams}
 import freechips.rocketchip.tilelink._
 import freechips.rocketchip.util.BundleField
 import org.chipsalliance.cde.config.Parameters
@@ -77,23 +77,29 @@ class DuplicatorNode(override val name: String = "dup")
     sourceEnq.bits := nodeIn.a.bits.source
 
     val idQueue = Queue(sourceEnq, entries = 4, pipe = false, flow = false)
+
     val srcMatch = nodeOuts.map(_.d.bits.source(inSourceWidth - 1, 0) === idQueue.bits)
-    idQueue.ready := nodeIn.d.ready && srcMatch.reduce(_ || _)
+    idQueue.ready := nodeIn.d.fire
 
     assert(sourceEnq.fire === nodeIn.a.fire)
     assert(idQueue.fire === nodeIn.d.fire)
 
-    (nodeOuts zip srcMatch).foreach { case (o, m) =>
+    (nodeOuts lazyZip srcMatch lazyZip Seq(0, inSourceEnd)).foreach { case (o, m, p) =>
       o.a.bits := nodeIn.a.bits
-      o.a.bits.source := nodeIn.a.bits.source | inSourceEnd.U
+      o.a.bits.source := nodeIn.a.bits.source | p.U
       o.a.valid := nodeIn.a.valid
-      nodeIn.d.bits := o.d.bits
-      nodeIn.d.bits.source := o.d.bits.source(inSourceWidth - 1, 0)
-      nodeIn.d.valid := o.d.valid
       o.d.ready := nodeIn.d.ready && m
     }
-    assert(!(nodeOuts.head.a.ready && nodeOuts.last.a.ready) || !nodeIn.a.valid, "double output fire")
+    nodeIn.d.bits := MuxCase(DontCare, (nodeOuts zip srcMatch).map { case (o, m) =>
+      m -> o.d.bits
+    })
+    nodeIn.d.bits.source := MuxCase(DontCare, (nodeOuts zip srcMatch).map { case (o, m) =>
+      m -> o.d.bits.source(inSourceWidth - 1, 0)
+    })
+    nodeIn.d.valid := (nodeOuts zip srcMatch).map { case (o, m) => o.d.valid && m }.reduce(_ || _)
     nodeIn.a.ready := nodeOuts.map(_.a.ready).reduce(_ || _) && sourceEnq.ready
+
+    assert(!(nodeOuts.head.a.ready && nodeOuts.last.a.ready) || !nodeIn.a.valid, "double output fire")
   }
 }
 
@@ -105,24 +111,30 @@ object DuplicatorNode {
 
 class DoubleOutXbar(clients: Seq[TLNode], override val name: String = "2o_xbar")
                    (implicit p: Parameters) extends LazyModule {
-  val xbar0 = TLXbar(TLArbiter.lowestIndexFirst)
-  val xbar1 = TLXbar(TLArbiter.lowestIndexFirst)
+  val xbar0 = TLXbar(TLArbiter.lowestIndexFirst, Some("double_out_xbar0"))
+  val xbar1 = TLXbar(TLArbiter.lowestIndexFirst, Some("double_out_xbar1"))
 
   implicit val disableMonitors: Boolean = false
 
+  val bufGen = () => TLBuffer(ace = BufferParams(0), bd = BufferParams(2, flow = false, pipe = false))
   val dupedIds = clients.map(connectOne(_, DuplicatorNode.apply)).map { c =>
     val id0 = connectOne(c, TLIdentityNode.apply)
     val id1 = connectOne(c, TLIdentityNode.apply)
-    xbar0 := id0
-    xbar1 := id1
+    xbar0 := connectOne(id0, bufGen)
+    xbar1 := connectOne(id1, bufGen)
     Seq(id0, id1)
   }.transpose
 
   lazy val module = new LazyModuleImp(this) {
     val id0InReadys = VecInit(dupedIds.head.map(_.in.head._1.a.ready)).asUInt
     val id1InValids = VecInit(dupedIds.last.map(_.in.head._1.a.valid)).asUInt
-    (dupedIds.last.map(_.out.head._1.a.valid) zip (id1InValids & (~id0InReadys).asUInt).asBools)
+    val id1OutValids = dupedIds.last.map(_.out.head._1.a.valid)
+    val id1InReadys = dupedIds.last.map(_.in.head._1.a.ready)
+    val id1OutReadys = VecInit(dupedIds.last.map(_.out.head._1.a.ready)).asUInt
+    (id1OutValids zip (id1InValids & (~id0InReadys).asUInt).asBools)
       .foreach { case (o, i) => o := i }
+    (id1InReadys zip (id1OutReadys & (~id0InReadys).asUInt).asBools)
+      .foreach { case (i, o) => i := o }
   }
 }
 
diff --git a/src/main/scala/radiance/memory/RWSplitterNode.scala b/src/main/scala/radiance/memory/RWSplitterNode.scala
index 5464b72..1cfceba 100644
--- a/src/main/scala/radiance/memory/RWSplitterNode.scala
+++ b/src/main/scala/radiance/memory/RWSplitterNode.scala
@@ -16,6 +16,7 @@ class RWSplitterNode(visibility: Option[AddressSet], override val name: String =
   // splits & arbitrates them into one client node per type of operation;
   // there will be N incoming edges, two outgoing edges, with two N:1 muxes;
   // it keeps the read and write channels fully separate to allow parallel processing.
+  suggestName(name)
   val node = TLNexusNode(
     clientFn = { seq =>
       val in_mapping = TLXbar.mapInputIds(seq)
@@ -155,6 +156,11 @@ object RWSplitterNode {
     LazyModule(new RWSplitterNode(None, name = valName.value)).node
   }
 
+  def apply(name: String)
+           (implicit p: Parameters, valName: ValName, sourceInfo: SourceInfo): TLNexusNode = {
+    LazyModule(new RWSplitterNode(None, name = name)).node
+  }
+
   def apply(visibility: AddressSet)
            (implicit p: Parameters, valName: ValName, sourceInfo: SourceInfo): TLNexusNode = {
     apply(visibility, valName.value)
diff --git a/src/main/scala/radiance/memory/XbarWithExtPolicy.scala b/src/main/scala/radiance/memory/XbarWithExtPolicy.scala
index 071c106..58131e1 100644
--- a/src/main/scala/radiance/memory/XbarWithExtPolicy.scala
+++ b/src/main/scala/radiance/memory/XbarWithExtPolicy.scala
@@ -53,10 +53,12 @@ object XbarWithExtPolicy {
 
 object XbarWithExtPolicyNoFallback {
   def apply(nameSuffix: Option[String] = None)
-           (implicit p: Parameters): (XbarWithExtPolicy, TLIdentityNode) = {
+           (implicit p: Parameters): (XbarWithExtPolicy, TLIdentityNode, TLIdentityNode) = {
     val inIdNode = TLIdentityNode()
+    val outIdNode = TLIdentityNode()
     val xbar = LazyModule(new XbarWithExtPolicy(nameSuffix, false))
     xbar.node :=* inIdNode
-    (xbar, inIdNode)
+    outIdNode :=* xbar.node
+    (xbar, inIdNode, outIdNode)
   }
 }
diff --git a/src/main/scala/radiance/tile/VirgoSharedMemComponents.scala b/src/main/scala/radiance/tile/VirgoSharedMemComponents.scala
index 8cc0072..ae143eb 100644
--- a/src/main/scala/radiance/tile/VirgoSharedMemComponents.scala
+++ b/src/main/scala/radiance/tile/VirgoSharedMemComponents.scala
@@ -108,7 +108,8 @@ class VirgoSharedMemComponents(
     // tensor core read nodes
     val tcDistNodes = Seq.fill(smemBanks)(tcNodeFanouts.map(connectOne(_, () => DistributorNode(smemWidth, wordSize))))
     val tcNodes = tcDistNodes.map { tcBank =>
-      Seq.fill(smemSubbanks)(tcBank.map(connectOne(_, () => TLBuffer(BufferParams(2, false, false)))).map(connectXbarName(_, Some("tc_dist_fanout"))))
+      Seq.fill(smemSubbanks)(tcBank.map(connectOne(_,
+        () => TLBuffer(BufferParams(2, false, false)))).map(connectXbarName(_, Some("tc_dist_fanout"))))
     } // (banks, subbanks, tc client)
 
     val unalignedRWNodes: ArrayBuffer[ArrayBuffer[TLNexusNode]] = // mutable for readability
@@ -195,7 +196,7 @@ class VirgoSharedMemComponents(
             val laneSerialXbar = laneSerialXbars.get(rw)(lid)
             laneSerialXbar._1.policySlaveNode := coreSerialPolicy.get(rw)(lid)
             coresRW.foreach(laneSerialXbar._2 := _)
-            connectXbarName(connectOne(laneSerialXbar._1.node, TLEphemeralNode.apply), Some(s"lane_${lid}_serial_out"))
+            connectXbarName(connectOne(laneSerialXbar._3, TLEphemeralNode.apply), Some(s"lane_${lid}_serial_out"))
           }
         }
         case NotSerialized => Seq.fill(2)(unalignedRWNodes.toSeq.flatten.map(connectXbar.apply))
@@ -216,7 +217,7 @@ class VirgoSharedMemComponents(
 
       (uniformRNodes, uniformWNodes, nonuniformRNodes, nonuniformWNodes)
     } else {
-      val splitterNodes = radianceSmemFanout.map { connectOne(_, RWSplitterNode.apply) }
+      val splitterNodes = radianceSmemFanout.map { connectOne(_, () => RWSplitterNode("rad_fanout_splitter")) }
       // these nodes access an entire line simultaneously
       val uniformRNodes: Seq[Seq[Seq[TLNexusNode]]] = spadReadNodes
       val uniformWNodes: Seq[Seq[Seq[TLNexusNode]]] = (spadWriteNodes zip spadSpWriteNodes).map { case (wb, wsb) =>
@@ -259,7 +260,7 @@ class VirgoSharedMemComponentsImp[T <: VirgoSharedMemComponents]
     (xbarsRW zip policiesRW).foreach { case (xbars, policies) =>
       // for each lane, if any core is valid
       val coreValids = xbars.map(_._2.in.map(_._1)).transpose.map { core => VecInit(core.map(_.a.valid)).asUInt.orR }
-      val select = xbars.map(_._2.out.map(_._1)).transpose.map { core => VecInit(core.map(_.a.ready)).asUInt.orR }
+      val select = xbars.map(_._3.in.map(_._1)).transpose.map { core => VecInit(core.map(_.a.fire)).asUInt.orR }
       val coreSelect = TLArbiter.roundRobin(outer.numCores, VecInit(coreValids).asUInt, VecInit(select).asUInt.orR)
       // TODO: roll this into XbarWithExtPolicy
       xbars.foreach { lane =>