Use profile summary to disable peeling for huge working sets

Summary:
Detect when the working set size of a profiled application is huge,
by comparing the number of counts required to reach the hot percentile
in the profile summary to a large threshold*.

When the working set size is determined to be huge, disable peeling
to avoid bloating the working set further.

*Note that the selected threshold (15K) is significantly larger than the
largest working set value in SPEC cpu2006 (which is gcc at around 11K).

Reviewers: davidxl

Subscribers: mehdi_amini, mzolotukhin, eraman, llvm-commits

Differential Revision: https://reviews.llvm.org/D36288

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@310005 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/llvm/Analysis/ProfileSummaryInfo.h b/include/llvm/Analysis/ProfileSummaryInfo.h
index 6aaabe1d1889084bb291a513cb9b4ce333e249b7..d788665fc10a31188eb03193d02150695837fa70 100644 (file)
--- a/include/llvm/Analysis/ProfileSummaryInfo.h
+++ b/include/llvm/Analysis/ProfileSummaryInfo.h
@@ -49,6 +49,10 @@ private:
   void computeThresholds();
   // Count thresholds to answer isHotCount and isColdCount queries.
   Optional<uint64_t> HotCountThreshold, ColdCountThreshold;
+  // True if the working set size of the code is considered huge,
+  // because the number of profile counts required to reach the hot
+  // percentile is above a huge threshold.
+  Optional<bool> HasHugeWorkingSetSize;
 
 public:
   ProfileSummaryInfo(Module &M) : M(M) {}
@@ -84,6 +88,8 @@ public:
   /// Returns the profile count for \p CallInst.
   Optional<uint64_t> getProfileCount(const Instruction *CallInst,
                                      BlockFrequencyInfo *BFI);
+  /// Returns true if the working set size of the code is considered huge.
+  bool hasHugeWorkingSetSize();
   /// \brief Returns true if \p F has hot function entry.
   bool isFunctionEntryHot(const Function *F);
   /// Returns true if \p F has hot function entry or hot call edge.

diff --git a/lib/Analysis/ProfileSummaryInfo.cpp b/lib/Analysis/ProfileSummaryInfo.cpp
index 3c2e04de4a5411d736e608a648a4164e06abfd6f..48bd54f80510e45815a9622f7320c5a46472f7c5 100644 (file)
--- a/lib/Analysis/ProfileSummaryInfo.cpp
+++ b/lib/Analysis/ProfileSummaryInfo.cpp
@@ -44,10 +44,16 @@ static cl::opt<bool> AccurateSampleProfile(
     cl::desc("If the sample profile is accurate, we will mark all un-sampled "
              "callsite as cold. Otherwise, treat un-sampled callsites as if "
              "we have no profile."));
+static cl::opt<unsigned> ProfileSummaryHugeWorkingSetSizeThreshold(
+    "profile-summary-huge-working-set-size-threshold", cl::Hidden,
+    cl::init(15000), cl::ZeroOrMore,
+    cl::desc("The code working set size is considered huge if the number of"
+             " blocks required to reach the -profile-summary-cutoff-hot"
+             " percentile exceeds this count."));
 
-// Find the minimum count to reach a desired percentile of counts.
-static uint64_t getMinCountForPercentile(SummaryEntryVector &DS,
-                                         uint64_t Percentile) {
+// Find the summary entry for a desired percentile of counts.
+static const ProfileSummaryEntry &getEntryForPercentile(SummaryEntryVector &DS,
+                                                        uint64_t Percentile) {
   auto Compare = [](const ProfileSummaryEntry &Entry, uint64_t Percentile) {
     return Entry.Cutoff < Percentile;
   };
@@ -56,7 +62,7 @@ static uint64_t getMinCountForPercentile(SummaryEntryVector &DS,
   // detailed summary.
   if (It == DS.end())
     report_fatal_error("Desired percentile exceeds the maximum cutoff");
-  return It->MinCount;
+  return *It;
 }
 
 // The profile summary metadata may be attached either by the frontend or by
@@ -169,10 +175,20 @@ void ProfileSummaryInfo::computeThresholds() {
   if (!computeSummary())
     return;
   auto &DetailedSummary = Summary->getDetailedSummary();
-  HotCountThreshold =
-      getMinCountForPercentile(DetailedSummary, ProfileSummaryCutoffHot);
-  ColdCountThreshold =
-      getMinCountForPercentile(DetailedSummary, ProfileSummaryCutoffCold);
+  auto &HotEntry =
+      getEntryForPercentile(DetailedSummary, ProfileSummaryCutoffHot);
+  HotCountThreshold = HotEntry.MinCount;
+  auto &ColdEntry =
+      getEntryForPercentile(DetailedSummary, ProfileSummaryCutoffCold);
+  ColdCountThreshold = ColdEntry.MinCount;
+  HasHugeWorkingSetSize =
+      HotEntry.NumCounts > ProfileSummaryHugeWorkingSetSizeThreshold;
+}
+
+bool ProfileSummaryInfo::hasHugeWorkingSetSize() {
+  if (!HasHugeWorkingSetSize)
+    computeThresholds();
+  return HasHugeWorkingSetSize && HasHugeWorkingSetSize.getValue();
 }
 
 bool ProfileSummaryInfo::isHotCount(uint64_t C) {

diff --git a/lib/Transforms/Scalar/LoopUnrollPass.cpp b/lib/Transforms/Scalar/LoopUnrollPass.cpp
index cd0e432869755da32fa60ee69a4bef370e7640cb..b692c08c9cc941c555544f35f1c6ec21a8989709 100644 (file)
--- a/lib/Transforms/Scalar/LoopUnrollPass.cpp
+++ b/lib/Transforms/Scalar/LoopUnrollPass.cpp
@@ -21,6 +21,7 @@
 #include "llvm/Analysis/LoopPass.h"
 #include "llvm/Analysis/LoopUnrollAnalyzer.h"
 #include "llvm/Analysis/OptimizationDiagnosticInfo.h"
+#include "llvm/Analysis/ProfileSummaryInfo.h"
 #include "llvm/Analysis/ScalarEvolution.h"
 #include "llvm/Analysis/ScalarEvolutionExpressions.h"
 #include "llvm/IR/DataLayout.h"
@@ -1252,6 +1253,11 @@ PreservedAnalyses LoopUnrollPass::run(Function &F,
   auto &AC = AM.getResult<AssumptionAnalysis>(F);
   auto &ORE = AM.getResult<OptimizationRemarkEmitterAnalysis>(F);
 
+  const ModuleAnalysisManager &MAM =
+      AM.getResult<ModuleAnalysisManagerFunctionProxy>(F).getManager();
+  ProfileSummaryInfo *PSI =
+      MAM.getCachedResult<ProfileSummaryAnalysis>(*F.getParent());
+
   bool Changed = false;
 
   // The unroller requires loops to be in simplified form, and also needs LCSSA.
@@ -1280,12 +1286,18 @@ PreservedAnalyses LoopUnrollPass::run(Function &F,
     // states we support: partial and full (or "simple") unrolling. However, to
     // enable these things we actually pass "None" in for the optional to avoid
     // providing an explicit choice.
-    Optional<bool> AllowPartialParam, RuntimeParam, UpperBoundParam;
-    bool CurChanged = tryToUnrollLoop(
-        &L, DT, &LI, SE, TTI, AC, ORE,
-        /*PreserveLCSSA*/ true, OptLevel, /*Count*/ None,
-        /*Threshold*/ None, AllowPartialParam, RuntimeParam, UpperBoundParam,
-        /*AllowPeeling*/ None);
+    Optional<bool> AllowPartialParam, RuntimeParam, UpperBoundParam,
+        AllowPeeling;
+    // Check if the profile summary indicates that the profiled application
+    // has a huge working set size, in which case we disable peeling to avoid
+    // bloating it further.
+    if (PSI && PSI->hasHugeWorkingSetSize())
+      AllowPeeling = false;
+    bool CurChanged =
+        tryToUnrollLoop(&L, DT, &LI, SE, TTI, AC, ORE,
+                        /*PreserveLCSSA*/ true, OptLevel, /*Count*/ None,
+                        /*Threshold*/ None, AllowPartialParam, RuntimeParam,
+                        UpperBoundParam, AllowPeeling);
     Changed |= CurChanged;
 
     // The parent must not be damaged by unrolling!

diff --git a/test/Other/new-pm-defaults.ll b/test/Other/new-pm-defaults.ll
index 4136cf5da30114e2ff738a9963e4522bdd965a27..cb16b048cf2c1d45618ecb33a72e043db23554e3 100644 (file)
--- a/test/Other/new-pm-defaults.ll
+++ b/test/Other/new-pm-defaults.ll
@@ -198,6 +198,7 @@
 ; CHECK-O-NEXT: Running pass: SimplifyCFGPass
 ; CHECK-O-NEXT: Running pass: InstCombinePass
 ; CHECK-O-NEXT: Running pass: LoopUnrollPass
+; CHECK-O-NEXT: Running analysis: OuterAnalysisManagerProxy
 ; CHECK-O-NEXT: Running pass: InstCombinePass
 ; CHECK-O-NEXT: Running pass: RequireAnalysisPass<{{.*}}OptimizationRemarkEmitterAnalysis
 ; CHECK-O-NEXT: Running pass: FunctionToLoopPassAdaptor<{{.*}}LICMPass

diff --git a/test/Other/new-pm-thinlto-defaults.ll b/test/Other/new-pm-thinlto-defaults.ll
index 07603f2c2d339c9f73b039f8101923173ef4555b..5a280491e8490507e4ab0ccca8aeae94b5458251 100644 (file)
--- a/test/Other/new-pm-thinlto-defaults.ll
+++ b/test/Other/new-pm-thinlto-defaults.ll
@@ -185,6 +185,7 @@
 ; CHECK-POSTLINK-O-NEXT: Running pass: SimplifyCFGPass
 ; CHECK-POSTLINK-O-NEXT: Running pass: InstCombinePass
 ; CHECK-POSTLINK-O-NEXT: Running pass: LoopUnrollPass
+; CHECK-POSTLINK-O-NEXT: Running analysis: OuterAnalysisManagerProxy
 ; CHECK-POSTLINK-O-NEXT: Running pass: InstCombinePass
 ; CHECK-POSTLINK-O-NEXT: Running pass: RequireAnalysisPass<{{.*}}OptimizationRemarkEmitterAnalysis
 ; CHECK-POSTLINK-O-NEXT: Running pass: FunctionToLoopPassAdaptor<{{.*}}LICMPass

diff --git a/test/Transforms/LoopUnroll/peel-loop-pgo.ll b/test/Transforms/LoopUnroll/peel-loop-pgo.ll
index ee5846c6325839ef0104bf2a123a695df018cc19..361a2ca71c322bb50e88a9edfdb09a0c62c0be29 100644 (file)
--- a/test/Transforms/LoopUnroll/peel-loop-pgo.ll
+++ b/test/Transforms/LoopUnroll/peel-loop-pgo.ll
@@ -1,5 +1,8 @@
 ; RUN: opt < %s -S -debug-only=loop-unroll -loop-unroll 2>&1 | FileCheck %s
-; RUN: opt < %s -S -debug-only=loop-unroll -passes='require<opt-remark-emit>,unroll' 2>&1 | FileCheck %s
+; RUN: opt < %s -S -debug-only=loop-unroll -passes='require<profile-summary>,function(require<opt-remark-emit>,unroll)' 2>&1 | FileCheck %s
+; Confirm that peeling is disabled if the number of counts required to reach
+; the hot percentile is above the threshold.
+; RUN: opt < %s -S -profile-summary-huge-working-set-size-threshold=9 -debug-only=loop-unroll -passes='require<profile-summary>,function(require<opt-remark-emit>,unroll)' 2>&1 | FileCheck %s --check-prefix=NOPEEL
 ; REQUIRES: asserts
 
 ; Make sure we use the profile information correctly to peel-off 3 iterations
@@ -11,19 +14,19 @@
 ; CHECK-NOT: PEELING
 
 ; Confirm that no peeling occurs when we are performing full unrolling.
-; RUN: opt < %s -S -debug-only=loop-unroll -passes='require<opt-remark-emit>,loop(unroll-full)' 2>&1 | FileCheck %s --check-prefix=FULLUNROLL
-; FULLUNROLL-NOT: PEELING
+; RUN: opt < %s -S -debug-only=loop-unroll -passes='require<opt-remark-emit>,loop(unroll-full)' 2>&1 | FileCheck %s --check-prefix=NOPEEL
+; NOPEEL-NOT: PEELING
 
 ; CHECK-LABEL: @basic
-; CHECK: br i1 %{{.*}}, label %[[NEXT0:.*]], label %for.cond.for.end_crit_edge, !prof !1
+; CHECK: br i1 %{{.*}}, label %[[NEXT0:.*]], label %for.cond.for.end_crit_edge, !prof !15
 ; CHECK: [[NEXT0]]:
-; CHECK: br i1 %{{.*}}, label %[[NEXT1:.*]], label %for.cond.for.end_crit_edge, !prof !2
+; CHECK: br i1 %{{.*}}, label %[[NEXT1:.*]], label %for.cond.for.end_crit_edge, !prof !16
 ; CHECK: [[NEXT1]]:
-; CHECK: br i1 %{{.*}}, label %[[NEXT2:.*]], label %for.cond.for.end_crit_edge, !prof !3
+; CHECK: br i1 %{{.*}}, label %[[NEXT2:.*]], label %for.cond.for.end_crit_edge, !prof !17
 ; CHECK: [[NEXT2]]:
-; CHECK: br i1 %{{.*}}, label %for.body, label %{{.*}}, !prof !4
+; CHECK: br i1 %{{.*}}, label %for.body, label %{{.*}}, !prof !18
 
-define void @basic(i32* %p, i32 %k) #0 !prof !0 {
+define void @basic(i32* %p, i32 %k) #0 !prof !15 {
 entry:
   %cmp3 = icmp slt i32 0, %k
   br i1 %cmp3, label %for.body.lr.ph, label %for.end
@@ -38,7 +41,7 @@ for.body:                                         ; preds = %for.body.lr.ph, %fo
   store i32 %i.05, i32* %p.addr.04, align 4
   %inc = add nsw i32 %i.05, 1
   %cmp = icmp slt i32 %inc, %k
-  br i1 %cmp, label %for.body, label %for.cond.for.end_crit_edge, !prof !1
+  br i1 %cmp, label %for.body, label %for.cond.for.end_crit_edge, !prof !16
 
 for.cond.for.end_crit_edge:                       ; preds = %for.body
   br label %for.end
@@ -54,7 +57,7 @@ for.end:                                          ; preds = %for.cond.for.end_cr
 ; CHECK: for.body:
 ; CHECK-NOT: br
 ; CHECK: br i1 %cmp, label %for.body, label %for.cond.for.end_crit_edge
-define void @optsize(i32* %p, i32 %k) #1 !prof !0 {
+define void @optsize(i32* %p, i32 %k) #1 !prof !15 {
 entry:
   %cmp3 = icmp slt i32 0, %k
   br i1 %cmp3, label %for.body.lr.ph, label %for.end
@@ -69,7 +72,7 @@ for.body:                                         ; preds = %for.body.lr.ph, %fo
   store i32 %i.05, i32* %p.addr.04, align 4
   %inc = add nsw i32 %i.05, 1
   %cmp = icmp slt i32 %inc, %k
-  br i1 %cmp, label %for.body, label %for.cond.for.end_crit_edge, !prof !1
+  br i1 %cmp, label %for.body, label %for.cond.for.end_crit_edge, !prof !16
 
 for.cond.for.end_crit_edge:                       ; preds = %for.body
   br label %for.end
@@ -81,11 +84,27 @@ for.end:                                          ; preds = %for.cond.for.end_cr
 attributes #0 = { nounwind }
 attributes #1 = { nounwind optsize }
 
-!0 = !{!"function_entry_count", i64 1}
-!1 = !{!"branch_weights", i32 3001, i32 1001}
-
-;CHECK: !1 = !{!"branch_weights", i32 900, i32 101}
-;CHECK: !2 = !{!"branch_weights", i32 540, i32 360}
-;CHECK: !3 = !{!"branch_weights", i32 162, i32 378}
-;CHECK: !4 = !{!"branch_weights", i32 1399, i32 162}
+!llvm.module.flags = !{!1}
+
+!1 = !{i32 1, !"ProfileSummary", !2}
+!2 = !{!3, !4, !5, !6, !7, !8, !9, !10}
+!3 = !{!"ProfileFormat", !"InstrProf"}
+!4 = !{!"TotalCount", i64 10}
+!5 = !{!"MaxCount", i64 3}
+!6 = !{!"MaxInternalCount", i64 1}
+!7 = !{!"MaxFunctionCount", i64 3}
+!8 = !{!"NumCounts", i64 2}
+!9 = !{!"NumFunctions", i64 2}
+!10 = !{!"DetailedSummary", !11}
+!11 = !{!12, !13, !14}
+!12 = !{i32 10000, i64 3, i32 2}
+!13 = !{i32 999000, i64 1, i32 10}
+!14 = !{i32 999999, i64 1, i32 10}
+!15 = !{!"function_entry_count", i64 1}
+!16 = !{!"branch_weights", i32 3001, i32 1001}
+
+;CHECK: !15 = !{!"branch_weights", i32 900, i32 101}
+;CHECK: !16 = !{!"branch_weights", i32 540, i32 360}
+;CHECK: !17 = !{!"branch_weights", i32 162, i32 378}
+;CHECK: !18 = !{!"branch_weights", i32 1399, i32 162}