tensor: Fix metadata of C req; fix dequeue / req gen timing

src/main/scala/radiance/core/TensorCoreDecoupled.scala

@@ -252,15 +252,15 @@ class TensorCoreDecoupled(
   // C access from regfile
   //
 
-  // since regfile is fixed-latency, respC valid should be determined at the
+  // regfile is fast; don't need a deep response queue
-  // request sending side.
-  val respCValid = RegInit(false.B)
-
-  // regfile latency is 1 cycle; don't need a deep response queue
   val respQueueCDepth = 2
   val respQueueC = Module(new Queue(
     new Bundle {
-      val tag = new TensorMemTag
+      val tag = new Bundle {
+        val warp = UInt(numWarpBits.W)
+        val set = UInt(setBits.W)
+        val step = UInt(stepBits.W)
+      }
       val data = UInt(io.respC.widthOption.get.W)
     },
     respQueueCDepth
@@ -276,8 +276,25 @@ class TensorCoreDecoupled(
                                (respQueueC.entries - 1).U,
                                (respQueueC.entries - 2).U)
   val genReqC = (state === AccessorState.access) && hasSpace && !doneReqC
-  // 1-cycle delay
+
-  respCValid := genReqC
+  // set/step state of the C accumulator value that will be latched ath the
+  // next cycle.
+  val setAccessC = RegInit(0.U(setBits.W))
+  val stepAccessC = RegInit(0.U(stepBits.W))
+  val substepAccessC = RegInit(0.U(1.W))
+  val nextStepAccessC = genReqC && (substepAccessC === 1.U)
+  when (genReqC) {
+    substepAccessC := substepAccessC + 1.U
+  }
+  dontTouch(stepAccessC)
+  dontTouch(substepAccessC)
+  dontTouch(nextStepAccessC)
+
+  // give 1-cycle delay to sync valid/metadata with C regfile response
+  val respCValid = RegNext(genReqC)
+  val warpAccessCDelayed = RegNext(warpAccess)
+  val setAccessCDelayed = RegNext(setAccessC)
+  val stepAccessCDelayed = RegNext(stepAccessC)
 
   // note rd is independent to sets
   def rdGen(step: UInt, substep: UInt): UInt = {
@@ -286,25 +303,14 @@ class TensorCoreDecoupled(
     (step << 1/*2 substeps*/) + substep
   }
   io.reqC.valid := genReqC
-  io.reqC.bits := 5.U // FIXME
+  io.reqC.bits := rdGen(stepAccessC, substepAccessC)
-
-  // set/index state of the C accumulator value that will be latched ath the
-  // next cycle.
-  val stateC = RegInit(stateInit)
-  when (genReqC) {
-    when (stateC.index === lastIndex.U) {
-      stateC.set := stateC.set + 1.U
-    }
-    stateC.index := stateC.index + 1.U
-  }
-  dontTouch(stateC)
 
   // queue the regfile response to buffers
   // these strictly belong to the execute stage
   respQueueC.io.enq.valid := respCValid
-  respQueueC.io.enq.bits.tag.warp := warpAccess
+  respQueueC.io.enq.bits.tag.warp := warpAccessCDelayed
-  respQueueC.io.enq.bits.tag.set := stateC.set
+  respQueueC.io.enq.bits.tag.set := setAccessCDelayed
-  respQueueC.io.enq.bits.tag.index := stateC.index
+  respQueueC.io.enq.bits.tag.step := stepAccessCDelayed
   respQueueC.io.enq.bits.data := io.respC
 
   // serialize every two C responses into one full 4x4 C tile
@@ -315,7 +321,8 @@ class TensorCoreDecoupled(
   fullC.io.enq.bits := respQueueC.io.deq.bits.data
   respQueueC.io.deq.ready := fullC.io.enq.ready
   val fullCTag = Module(new Queue(
-    new TensorMemTag, entries = 1, pipe = true
+    chiselTypeOf(respQueueC.io.deq.bits.tag),
+    entries = 1, pipe = true
   ))
   fullCTag.io.enq.valid := respQueueC.io.deq.valid
   fullCTag.io.enq.bits := respQueueC.io.deq.bits.tag
@@ -323,10 +330,10 @@ class TensorCoreDecoupled(
   // finalize state when everything has been accessed
   val lastReqA = (stateA.set === lastSet.U) && (stateA.index === lastIndex.U)
   val lastReqB = (stateB.set === lastSet.U) && (stateB.index === lastIndex.U)
-  val lastReqC = (stateC.set === lastSet.U) && (stateC.index === lastIndex.U)
+  val lastReqC = (setAccessC === lastSet.U) && (stepAccessC === lastStep.U)
   when (lastReqA && io.reqA.fire) { doneReqA := true.B }
   when (lastReqB && io.reqB.fire) { doneReqB := true.B }
-  when (lastReqC && genReqC) { doneReqC := true.B }
+  when (lastReqC && nextStepAccessC) { doneReqC := true.B }
   when (state === AccessorState.finish) {
     doneReqA := false.B
     doneReqB := false.B
@@ -335,8 +342,9 @@ class TensorCoreDecoupled(
     stateA.index := 0.U
     stateB.set := 0.U
     stateB.index := 0.U
-    stateC.set := 0.U
+    setAccessC := 0.U
-    stateC.index := 0.U
+    stepAccessC := 0.U
+    substepAccessC := 0.U
   }
 
   allReqsDone := doneReqA && doneReqB && doneReqC
@@ -432,7 +440,7 @@ class TensorCoreDecoupled(
 
   val fullCBuf = Module(new Queue(
     new Bundle {
-      val tag = new TensorMemTag
+      val tag = chiselTypeOf(fullCTag.io.deq.bits)
       val data = chiselTypeOf(fullC.io.deq.bits)
     }, entries = 1, pipe = true
   ))
@@ -496,9 +504,9 @@ class TensorCoreDecoupled(
     ((stepCompute & shouldDequeueBMask) === shouldDequeueBMask) &&
     (substepCompute === 1.U)
   fullBBuf.io.deq.ready := dpuFire && shouldDequeueB
-
+  // C should be dequeued everytime a 4x4 tile is done computing
-  // C deq should be synced with B deq
+  val shouldDequeueC = (substepCompute === 1.U)
-  fullCBuf.io.deq.ready := dpuFire && shouldDequeueB
+  fullCBuf.io.deq.ready := dpuFire && shouldDequeueC
 
   dontTouch(respQueueA)
   dontTouch(respQueueB)
@@ -639,6 +647,7 @@ class TensorCoreDecoupled(
       }
     }
   }
+  sequenceSetStep(setAccessC, stepAccessC, nextStepAccessC)
   sequenceSetStep(setCompute, stepCompute, nextStepCompute)
 }
 
@@ -739,7 +748,7 @@ class TensorCoreDecoupledTLImp(outer: TensorCoreDecoupledTL)
   tensor.io.respC := 42.U // FIXME bogus
 
   tensor.io.initiate.valid := io.start
-  tensor.io.initiate.bits.wid := 0.U // FIXME bogus
+  tensor.io.initiate.bits.wid := 3.U // bogus, static value
   tensor.io.writeback.ready := true.B
 
   io.finished := tensor.io.writeback.valid && tensor.io.writeback.bits.last