From fb9905545cd12b83d9a1fe4eb99519436a318aa3 Mon Sep 17 00:00:00 2001
From: Valery Pykhtin <Valery.Pykhtin@amd.com>
Date: Mon, 27 Mar 2017 18:22:39 +0000
Subject: [PATCH] [AMDGPU] SISched: Detect dependency types between blocks

Patch by Axel Davy (axel.davy@normalesup.org)

Differential revision: https://reviews.llvm.org/D30153

llvm-svn: 298872
---
 llvm/lib/Target/AMDGPU/SIMachineScheduler.cpp | 52 +++++++++++++++------------
 llvm/lib/Target/AMDGPU/SIMachineScheduler.h   | 13 +++++--
 2 files changed, 39 insertions(+), 26 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/SIMachineScheduler.cpp b/llvm/lib/Target/AMDGPU/SIMachineScheduler.cpp
index f779053..ce4e458 100644
--- a/llvm/lib/Target/AMDGPU/SIMachineScheduler.cpp
+++ b/llvm/lib/Target/AMDGPU/SIMachineScheduler.cpp
@@ -539,21 +539,30 @@ void SIScheduleBlock::addPred(SIScheduleBlock *Pred) {
   Preds.push_back(Pred);
 
   assert(none_of(Succs,
-                 [=](SIScheduleBlock *S) { return PredID == S->getID(); }) &&
+                 [=](std::pair<SIScheduleBlock*,
+                     SIScheduleBlockLinkKind> S) {
+                   return PredID == S.first->getID();
+                    }) &&
          "Loop in the Block Graph!");
 }
 
-void SIScheduleBlock::addSucc(SIScheduleBlock *Succ) {
+void SIScheduleBlock::addSucc(SIScheduleBlock *Succ,
+                              SIScheduleBlockLinkKind Kind) {
   unsigned SuccID = Succ->getID();
 
   // Check if not already predecessor.
-  for (SIScheduleBlock* S : Succs) {
-    if (SuccID == S->getID())
+  for (std::pair<SIScheduleBlock*, SIScheduleBlockLinkKind> &S : Succs) {
+    if (SuccID == S.first->getID()) {
+      if (S.second == SIScheduleBlockLinkKind::NoData &&
+          Kind == SIScheduleBlockLinkKind::Data)
+        S.second = Kind;
       return;
+    }
   }
   if (Succ->isHighLatencyBlock())
     ++NumHighLatencySuccessors;
-  Succs.push_back(Succ);
+  Succs.push_back(std::make_pair(Succ, Kind));
+
   assert(none_of(Preds,
                  [=](SIScheduleBlock *P) { return SuccID == P->getID(); }) &&
          "Loop in the Block Graph!");
@@ -573,8 +582,10 @@ void SIScheduleBlock::printDebug(bool full) {
   }
 
   dbgs() << "\nSuccessors:\n";
-  for (SIScheduleBlock* S : Succs) {
-    S->printDebug(false);
+  for (std::pair<SIScheduleBlock*, SIScheduleBlockLinkKind> S : Succs) {
+    if (S.second == SIScheduleBlockLinkKind::Data)
+      dbgs() << "(Data Dep) ";
+    S.first->printDebug(false);
   }
 
   if (Scheduled) {
@@ -1096,7 +1107,8 @@ void SIScheduleBlockCreator::createBlocksForVariant(SISchedulerBlockCreatorVaria
       if (SuccDep.isWeak() || Succ->NodeNum >= DAGSize)
         continue;
       if (Node2CurrentBlock[Succ->NodeNum] != SUID)
-        CurrentBlocks[SUID]->addSucc(CurrentBlocks[Node2CurrentBlock[Succ->NodeNum]]);
+        CurrentBlocks[SUID]->addSucc(CurrentBlocks[Node2CurrentBlock[Succ->NodeNum]],
+                                     SuccDep.isCtrl() ? NoData : Data);
     }
     for (SDep& PredDep : SU->Preds) {
       SUnit *Pred = PredDep.getSUnit();
@@ -1290,10 +1302,8 @@ void SIScheduleBlockCreator::fillStats() {
       Block->Height = 0;
     else {
       unsigned Height = 0;
-      for (SIScheduleBlock *Succ : Block->getSuccs()) {
-        if (Height < Succ->Height + 1)
-          Height = Succ->Height + 1;
-      }
+      for (const auto &Succ : Block->getSuccs())
+        Height = std::min(Height, Succ.first->Height + 1);
       Block->Height = Height;
     }
   }
@@ -1574,17 +1584,13 @@ void SIScheduleBlockScheduler::decreaseLiveRegs(SIScheduleBlock *Block,
 }
 
 void SIScheduleBlockScheduler::releaseBlockSuccs(SIScheduleBlock *Parent) {
-  for (SIScheduleBlock* Block : Parent->getSuccs()) {
-    --BlockNumPredsLeft[Block->getID()];
-    if (BlockNumPredsLeft[Block->getID()] == 0) {
-      ReadyBlocks.push_back(Block);
-    }
-    // TODO: Improve check. When the dependency between the high latency
-    // instructions and the instructions of the other blocks are WAR or WAW
-    // there will be no wait triggered. We would like these cases to not
-    // update LastPosHighLatencyParentScheduled.
-    if (Parent->isHighLatencyBlock())
-      LastPosHighLatencyParentScheduled[Block->getID()] = NumBlockScheduled;
+  for (const auto &Block : Parent->getSuccs()) {
+    if (--BlockNumPredsLeft[Block.first->getID()] == 0)
+      ReadyBlocks.push_back(Block.first);
+
+    if (Parent->isHighLatencyBlock() &&
+        Block.second == SIScheduleBlockLinkKind::Data)
+      LastPosHighLatencyParentScheduled[Block.first->getID()] = NumBlockScheduled;
   }
 }
 
diff --git a/llvm/lib/Target/AMDGPU/SIMachineScheduler.h b/llvm/lib/Target/AMDGPU/SIMachineScheduler.h
index 6978b83..085ffcd 100644
--- a/llvm/lib/Target/AMDGPU/SIMachineScheduler.h
+++ b/llvm/lib/Target/AMDGPU/SIMachineScheduler.h
@@ -54,6 +54,11 @@ struct SISchedulerCandidate {
 class SIScheduleDAGMI;
 class SIScheduleBlockCreator;
 
+enum SIScheduleBlockLinkKind {
+  NoData,
+  Data
+};
+
 class SIScheduleBlock {
   SIScheduleDAGMI *DAG;
   SIScheduleBlockCreator *BC;
@@ -92,7 +97,8 @@ class SIScheduleBlock {
   unsigned ID;
 
   std::vector<SIScheduleBlock*> Preds;  // All blocks predecessors.
-  std::vector<SIScheduleBlock*> Succs;  // All blocks successors.
+  // All blocks successors, and the kind of link
+  std::vector<std::pair<SIScheduleBlock*, SIScheduleBlockLinkKind>> Succs;
   unsigned NumHighLatencySuccessors = 0;
 
 public:
@@ -112,10 +118,11 @@ public:
 
   // Add block pred, which has instruction predecessor of SU.
   void addPred(SIScheduleBlock *Pred);
-  void addSucc(SIScheduleBlock *Succ);
+  void addSucc(SIScheduleBlock *Succ, SIScheduleBlockLinkKind Kind);
 
   const std::vector<SIScheduleBlock*>& getPreds() const { return Preds; }
-  const std::vector<SIScheduleBlock*>& getSuccs() const { return Succs; }
+  ArrayRef<std::pair<SIScheduleBlock*, SIScheduleBlockLinkKind>>
+    getSuccs() const { return Succs; }
 
   unsigned Height;  // Maximum topdown path length to block without outputs
   unsigned Depth;   // Maximum bottomup path length to block without inputs
-- 
2.7.4