commit b7322e8ac7a99b059dc7c3930436cebe72b53b8e
author Ikhlas Ajbar <iajbar@codeaurora.org> Tue Apr 03 03:39:43 2018 +0000
committer Ikhlas Ajbar <iajbar@codeaurora.org> Tue Apr 03 03:39:43 2018 +0000
tree abfa9e7c3351221fbd99c7ebd8076a03e7fe870c
parent 35a75cae92a0ed65966fa934fc087a7eb1f9ca26

peel loops with runtime small trip counts

For Hexagon, peeling loops with small runtime trip count is beneficial for our
benchmarks. We set PeelCount in HexagonTargetInfo.cpp and we use PeelCount set
by the target for computing the desired peel count.

Differential Revision: https://reviews.llvm.org/D44880

llvm-svn: 329042

llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp

diff --git a/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp b/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp
index 3290b59..786f8eb 100644
--- a/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp
@@ -44,6 +44,13 @@
 void HexagonTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
                                              TTI::UnrollingPreferences &UP) {
   UP.Runtime = UP.Partial = true;
+  // Only try to peel innermost loops with small runtime trip counts.
+  if (L && L->empty() &&
+      SE.getSmallConstantTripCount(L) == 0 &&
+      SE.getSmallConstantMaxTripCount(L) > 0 &&
+      SE.getSmallConstantMaxTripCount(L) <= 5) {
+    UP.PeelCount = 2;
+  }
 }
 
 bool HexagonTTIImpl::shouldFavorPostInc() const {

llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp

diff --git a/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp b/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
index a1b25a2..ad454d7 100644
--- a/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
@@ -978,6 +978,9 @@
   if (UP.Threshold == 0 && (!UP.Partial || UP.PartialThreshold == 0))
     return LoopUnrollResult::Unmodified;
 
+  if (UP.PeelCount && !canPeel(L))
+    UP.PeelCount = 0;
+
   SmallPtrSet<const Value *, 32> EphValues;
   CodeMetrics::collectEphemeralValues(L, &AC, EphValues);
 

llvm/lib/Transforms/Utils/LoopUnrollPeel.cpp

diff --git a/llvm/lib/Transforms/Utils/LoopUnrollPeel.cpp b/llvm/lib/Transforms/Utils/LoopUnrollPeel.cpp
index 5f465f3..f47744d 100644
--- a/llvm/lib/Transforms/Utils/LoopUnrollPeel.cpp
+++ b/llvm/lib/Transforms/Utils/LoopUnrollPeel.cpp
@@ -69,7 +69,7 @@
     std::numeric_limits<unsigned>::max();
 
 // Check whether we are capable of peeling this loop.
-static bool canPeel(Loop *L) {
+bool llvm::canPeel(Loop *L) {
   // Make sure the loop is in simplified form
   if (!L->isLoopSimplifyForm())
     return false;
@@ -221,6 +221,9 @@
                             TargetTransformInfo::UnrollingPreferences &UP,
                             unsigned &TripCount, ScalarEvolution &SE) {
   assert(LoopSize > 0 && "Zero loop size is not allowed!");
+  // Save the UP.PeelCount value set by the target in
+  // TTI.getUnrollingPreferences or by the flag -unroll-peel-count.
+  unsigned TargetPeelCount = UP.PeelCount;
   UP.PeelCount = 0;
   if (!canPeel(L))
     return;
@@ -240,7 +243,9 @@
     SmallDenseMap<PHINode *, unsigned> IterationsToInvariance;
     // Now go through all Phis to calculate their the number of iterations they
     // need to become invariants.
-    unsigned DesiredPeelCount = 0;
+    // Start the max computation with the UP.PeelCount value set by the target
+    // in TTI.getUnrollingPreferences or by the flag -unroll-peel-count.
+    unsigned DesiredPeelCount = TargetPeelCount;
     BasicBlock *BackEdge = L->getLoopLatch();
     assert(BackEdge && "Loop is not in simplified form?");
     for (auto BI = L->getHeader()->begin(); isa<PHINode>(&*BI); ++BI) {