[llvm-exegesis] Output the unscaled value as well as the scaled one.

Summary: See PR38936 for context.

Reviewers: gchatelet

Subscribers: tschuett, llvm-commits

Differential Revision: https://reviews.llvm.org/D52500

llvm-svn: 343081

13 files changed, 70 insertions, 49 deletions

diff --git a/llvm/tools/llvm-exegesis/lib/Analysis.cpp b/llvm/tools/llvm-exegesis/lib/Analysis.cpp
index bb511808096..54dc4be374b 100644
--- a/llvm/tools/llvm-exegesis/lib/Analysis.cpp
+++ b/llvm/tools/llvm-exegesis/lib/Analysis.cpp
@@ -139,7 +139,7 @@ void Analysis::printInstructionRowCsv(const size_t PointId,
 #endif
   for (const auto &Measurement : Point.Measurements) {
     OS << kCsvSep;
-    writeMeasurementValue<kEscapeCsv>(OS, Measurement.Value);
+    writeMeasurementValue<kEscapeCsv>(OS, Measurement.PerInstructionValue);
   }
   OS << "\n";
 }
@@ -410,14 +410,14 @@ bool Analysis::SchedClassCluster::measurementsMatch(
       return false;
     }
     // Find the latency.
-    SchedClassPoint[0].Value = 0.0;
+    SchedClassPoint[0].PerInstructionValue = 0.0;
     for (unsigned I = 0; I < SC.SCDesc->NumWriteLatencyEntries; ++I) {
       const llvm::MCWriteLatencyEntry *const WLE =
           STI.getWriteLatencyEntry(SC.SCDesc, I);
-      SchedClassPoint[0].Value =
-          std::max<double>(SchedClassPoint[0].Value, WLE->Cycles);
+      SchedClassPoint[0].PerInstructionValue =
+          std::max<double>(SchedClassPoint[0].PerInstructionValue, WLE->Cycles);
     }
-    ClusterCenterPoint[0].Value = Representative[0].avg();
+    ClusterCenterPoint[0].PerInstructionValue = Representative[0].avg();
   } else if (Mode == InstructionBenchmark::Uops) {
     for (int I = 0, E = Representative.size(); I < E; ++I) {
       // Find the pressure on ProcResIdx `Key`.
@@ -433,11 +433,11 @@ bool Analysis::SchedClassCluster::measurementsMatch(
                        [ProcResIdx](const std::pair<uint16_t, float> &WPR) {
                          return WPR.first == ProcResIdx;
                        });
-      SchedClassPoint[I].Value =
+      SchedClassPoint[I].PerInstructionValue =
           ProcResPressureIt == SC.IdealizedProcResPressure.end()
               ? 0.0
               : ProcResPressureIt->second;
-      ClusterCenterPoint[I].Value = Representative[I].avg();
+      ClusterCenterPoint[I].PerInstructionValue = Representative[I].avg();
     }
   } else {
     llvm::errs() << "unimplemented measurement matching for mode " << Mode
diff --git a/llvm/tools/llvm-exegesis/lib/Analysis.h b/llvm/tools/llvm-exegesis/lib/Analysis.h
index 3c33cafc5fb..1e6294b8770 100644
--- a/llvm/tools/llvm-exegesis/lib/Analysis.h
+++ b/llvm/tools/llvm-exegesis/lib/Analysis.h
@@ -69,7 +69,7 @@ private:
     const std::vector<size_t> &getPointIds() const { return PointIds; }
 
     // Return the cluster centroid.
-    const std::vector<BenchmarkMeasureStats> &getRepresentative() const {
+    const std::vector<PerInstructionStats> &getRepresentative() const {
       return Representative;
     }
 
@@ -85,7 +85,7 @@ private:
     InstructionBenchmarkClustering::ClusterId ClusterId;
     std::vector<size_t> PointIds;
     // Measurement stats for the points in the SchedClassCluster.
-    std::vector<BenchmarkMeasureStats> Representative;
+    std::vector<PerInstructionStats> Representative;
   };
 
   void printInstructionRowCsv(size_t PointId, llvm::raw_ostream &OS) const;
diff --git a/llvm/tools/llvm-exegesis/lib/BenchmarkResult.cpp b/llvm/tools/llvm-exegesis/lib/BenchmarkResult.cpp
index 9a41114a853..1748c53479c 100644
--- a/llvm/tools/llvm-exegesis/lib/BenchmarkResult.cpp
+++ b/llvm/tools/llvm-exegesis/lib/BenchmarkResult.cpp
@@ -164,9 +164,10 @@ template <> struct SequenceElementTraits<exegesis::BenchmarkMeasure> {
 // e.g. { "key": "the key", "value": 0123 }
 template <> struct MappingTraits<exegesis::BenchmarkMeasure> {
   static void mapping(IO &Io, exegesis::BenchmarkMeasure &Obj) {
-    Io.mapRequired("key", Obj.Key);
-    Io.mapRequired("value", Obj.Value);
     Io.mapOptional("debug_string", Obj.DebugString);
+    Io.mapRequired("value", Obj.PerInstructionValue);
+    Io.mapOptional("per_snippet_value", Obj.PerSnippetValue);
+    Io.mapRequired("key", Obj.Key);
   }
   static const bool flow = true;
 };
@@ -345,14 +346,14 @@ llvm::Error InstructionBenchmark::writeYaml(const LLVMState &State,
   return llvm::Error::success();
 }
 
-void BenchmarkMeasureStats::push(const BenchmarkMeasure &BM) {
+void PerInstructionStats::push(const BenchmarkMeasure &BM) {
   if (Key.empty())
     Key = BM.Key;
   assert(Key == BM.Key);
   ++NumValues;
-  SumValues += BM.Value;
-  MaxValue = std::max(MaxValue, BM.Value);
-  MinValue = std::min(MinValue, BM.Value);
+  SumValues += BM.PerInstructionValue;
+  MaxValue = std::max(MaxValue, BM.PerInstructionValue);
+  MinValue = std::min(MinValue, BM.PerInstructionValue);
 }
 
 } // namespace exegesis
diff --git a/llvm/tools/llvm-exegesis/lib/BenchmarkResult.h b/llvm/tools/llvm-exegesis/lib/BenchmarkResult.h
index 62653cd7604..e4e29ef1523 100644
--- a/llvm/tools/llvm-exegesis/lib/BenchmarkResult.h
+++ b/llvm/tools/llvm-exegesis/lib/BenchmarkResult.h
@@ -42,7 +42,12 @@ struct InstructionBenchmarkKey {
 
 struct BenchmarkMeasure {
   std::string Key;
-  double Value;
+  // This is the per-instruction value, i.e. measured quantity scaled per
+  // instruction.
+  double PerInstructionValue;
+  // This is the per-snippet value, i.e. measured quantity for one repetition of
+  // the whole snippet.
+  double PerSnippetValue;
   std::string DebugString;
 };
 
@@ -81,7 +86,7 @@ struct InstructionBenchmark {
 // Utilities to work with Benchmark measures.
 
 // A class that measures stats over benchmark measures.
-class BenchmarkMeasureStats {
+class PerInstructionStats {
 public:
   void push(const BenchmarkMeasure &BM);
 
diff --git a/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.cpp b/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.cpp
index efda689ce6b..7addb0acd7e 100644
--- a/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.cpp
+++ b/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.cpp
@@ -88,7 +88,15 @@ BenchmarkRunner::runConfiguration(const BenchmarkCode &BC,
                << *ObjectFilePath << "\n";
   const ExecutableFunction EF(State.createTargetMachine(),
                               getObjectFromFile(*ObjectFilePath));
-  InstrBenchmark.Measurements = runMeasurements(EF, *Scratch, NumRepetitions);
+  InstrBenchmark.Measurements = runMeasurements(EF, *Scratch);
+  assert(InstrBenchmark.NumRepetitions > 0 && "invalid NumRepetitions");
+  for (BenchmarkMeasure &BM : InstrBenchmark.Measurements) {
+    // Scale the measurements by instruction.
+    BM.PerInstructionValue /= InstrBenchmark.NumRepetitions;
+    // Scale the measurements by snippet.
+    BM.PerSnippetValue *= static_cast<double>(BC.Instructions.size()) /
+                   InstrBenchmark.NumRepetitions;
+  }
 
   return InstrBenchmark;
 }
diff --git a/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.h b/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.h
index 07ab23c336b..e5b567f2463 100644
--- a/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.h
+++ b/llvm/tools/llvm-exegesis/lib/BenchmarkRunner.h
@@ -69,8 +69,8 @@ protected:
 
 private:
   virtual std::vector<BenchmarkMeasure>
-  runMeasurements(const ExecutableFunction &EF, ScratchSpace &Scratch,
-                  const unsigned NumRepetitions) const = 0;
+  runMeasurements(const ExecutableFunction &EF,
+                  ScratchSpace &Scratch) const = 0;
 
   llvm::Expected<std::string>
   writeObjectFile(const BenchmarkCode &Configuration,
diff --git a/llvm/tools/llvm-exegesis/lib/Clustering.cpp b/llvm/tools/llvm-exegesis/lib/Clustering.cpp
index a966c722a78..b63afec945f 100644
--- a/llvm/tools/llvm-exegesis/lib/Clustering.cpp
+++ b/llvm/tools/llvm-exegesis/lib/Clustering.cpp
@@ -53,7 +53,7 @@ bool InstructionBenchmarkClustering::isNeighbour(
     const std::vector<BenchmarkMeasure> &Q) const {
   double DistanceSquared = 0.0;
   for (size_t I = 0, E = P.size(); I < E; ++I) {
-    const auto Diff = P[I].Value - Q[I].Value;
+    const auto Diff = P[I].PerInstructionValue - Q[I].PerInstructionValue;
     DistanceSquared += Diff * Diff;
   }
   return DistanceSquared <= EpsilonSquared_;
diff --git a/llvm/tools/llvm-exegesis/lib/Latency.cpp b/llvm/tools/llvm-exegesis/lib/Latency.cpp
index d09cdc597fd..3c402c4700c 100644
--- a/llvm/tools/llvm-exegesis/lib/Latency.cpp
+++ b/llvm/tools/llvm-exegesis/lib/Latency.cpp
@@ -109,8 +109,7 @@ LatencyBenchmarkRunner::~LatencyBenchmarkRunner() = default;
 
 std::vector<BenchmarkMeasure>
 LatencyBenchmarkRunner::runMeasurements(const ExecutableFunction &Function,
-                                        ScratchSpace &Scratch,
-                                        const unsigned NumRepetitions) const {
+                                        ScratchSpace &Scratch) const {
   // Cycle measurements include some overhead from the kernel. Repeat the
   // measure several times and take the minimum value.
   constexpr const int NumMeasurements = 30;
@@ -131,7 +130,8 @@ LatencyBenchmarkRunner::runMeasurements(const ExecutableFunction &Function,
     if (Value < MinLatency)
       MinLatency = Value;
   }
-  return {{"latency", static_cast<double>(MinLatency) / NumRepetitions, ""}};
+  return {{"latency", static_cast<double>(MinLatency),
+           static_cast<double>(MinLatency), ""}};
 }
 
 } // namespace exegesis
diff --git a/llvm/tools/llvm-exegesis/lib/Latency.h b/llvm/tools/llvm-exegesis/lib/Latency.h
index 5b9be005c1f..3a72d026d9f 100644
--- a/llvm/tools/llvm-exegesis/lib/Latency.h
+++ b/llvm/tools/llvm-exegesis/lib/Latency.h
@@ -44,8 +44,8 @@ public:
 
 private:
   std::vector<BenchmarkMeasure>
-  runMeasurements(const ExecutableFunction &EF, ScratchSpace &Scratch,
-                  const unsigned NumRepetitions) const override;
+  runMeasurements(const ExecutableFunction &EF,
+                  ScratchSpace &Scratch) const override;
 
   virtual const char *getCounterName() const;
 };
diff --git a/llvm/tools/llvm-exegesis/lib/Uops.cpp b/llvm/tools/llvm-exegesis/lib/Uops.cpp
index beb9fa8acf5..dbecbfe5441 100644
--- a/llvm/tools/llvm-exegesis/lib/Uops.cpp
+++ b/llvm/tools/llvm-exegesis/lib/Uops.cpp
@@ -252,8 +252,7 @@ UopsSnippetGenerator::generateCodeTemplate(unsigned Opcode) const {
 
 std::vector<BenchmarkMeasure>
 UopsBenchmarkRunner::runMeasurements(const ExecutableFunction &Function,
-                                     ScratchSpace &Scratch,
-                                     const unsigned NumRepetitions) const {
+                                     ScratchSpace &Scratch) const {
   const auto &SchedModel = State.getSubtargetInfo().getSchedModel();
 
   std::vector<BenchmarkMeasure> Result;
@@ -281,7 +280,8 @@ UopsBenchmarkRunner::runMeasurements(const ExecutableFunction &Function,
       CounterValue += Counter.read();
     }
     Result.push_back({llvm::itostr(ProcResIdx),
-                      static_cast<double>(CounterValue) / NumRepetitions,
+                      static_cast<double>(CounterValue),
+                      static_cast<double>(CounterValue),
                       SchedModel.getProcResource(ProcResIdx)->Name});
   }
   return Result;
diff --git a/llvm/tools/llvm-exegesis/lib/Uops.h b/llvm/tools/llvm-exegesis/lib/Uops.h
index 160397affed..aff583b4dcd 100644
--- a/llvm/tools/llvm-exegesis/lib/Uops.h
+++ b/llvm/tools/llvm-exegesis/lib/Uops.h
@@ -71,8 +71,8 @@ public:
 
 private:
   std::vector<BenchmarkMeasure>
-  runMeasurements(const ExecutableFunction &EF, ScratchSpace &Scratch,
-                  const unsigned NumRepetitions) const override;
+  runMeasurements(const ExecutableFunction &EF,
+                  ScratchSpace &Scratch) const override;
 };
 
 } // namespace exegesis
diff --git a/llvm/unittests/tools/llvm-exegesis/ClusteringTest.cpp b/llvm/unittests/tools/llvm-exegesis/ClusteringTest.cpp
index ecdb7d2a274..5d583a08951 100644
--- a/llvm/unittests/tools/llvm-exegesis/ClusteringTest.cpp
+++ b/llvm/unittests/tools/llvm-exegesis/ClusteringTest.cpp
@@ -26,13 +26,18 @@ TEST(ClusteringTest, Clusters3D) {
   std::vector<InstructionBenchmark> Points(6);
 
   // Cluster around (x=0, y=1, z=2): points {0, 3}.
-  Points[0].Measurements = {{"x", 0.01, ""}, {"y", 1.02, ""}, {"z", 1.98, "A"}};
-  Points[3].Measurements = {{"x", -0.01, ""}, {"y", 1.02, ""}, {"z", 1.98, ""}};
+  Points[0].Measurements = {
+      {"x", 0.01, 0.0, ""}, {"y", 1.02, 0.0, ""}, {"z", 1.98, 0.0, "A"}};
+  Points[3].Measurements = {
+      {"x", -0.01, 0.0, ""}, {"y", 1.02, 0.0, ""}, {"z", 1.98, 0.0, ""}};
   // Cluster around (x=1, y=1, z=2): points {1, 4}.
-  Points[1].Measurements = {{"x", 1.01, ""}, {"y", 1.02, ""}, {"z", 1.98, ""}};
-  Points[4].Measurements = {{"x", 0.99, ""}, {"y", 1.02, ""}, {"z", 1.98, ""}};
+  Points[1].Measurements = {
+      {"x", 1.01, 0.0, ""}, {"y", 1.02, 0.0, ""}, {"z", 1.98, 0.0, ""}};
+  Points[4].Measurements = {
+      {"x", 0.99, 0.0, ""}, {"y", 1.02, 0.0, ""}, {"z", 1.98, 0.0, ""}};
   // Cluster around (x=0, y=0, z=0): points {5}, marked as noise.
-  Points[5].Measurements = {{"x", 0.0, ""}, {"y", 0.01, ""}, {"z", -0.02, ""}};
+  Points[5].Measurements = {
+      {"x", 0.0, 0.0, ""}, {"y", 0.01, 0.0, ""}, {"z", -0.02, 0.0, ""}};
   // Error cluster: points {2}
   Points[2].Error = "oops";
 
@@ -64,8 +69,9 @@ TEST(ClusteringTest, Clusters3D) {
 
 TEST(ClusteringTest, Clusters3D_InvalidSize) {
   std::vector<InstructionBenchmark> Points(6);
-  Points[0].Measurements = {{"x", 0.01, ""}, {"y", 1.02, ""}, {"z", 1.98, ""}};
-  Points[1].Measurements = {{"y", 1.02, ""}, {"z", 1.98, ""}};
+  Points[0].Measurements = {
+      {"x", 0.01, 0.0, ""}, {"y", 1.02, 0.0, ""}, {"z", 1.98, 0.0, ""}};
+  Points[1].Measurements = {{"y", 1.02, 0.0, ""}, {"z", 1.98, 0.0, ""}};
   auto Error =
       InstructionBenchmarkClustering::create(Points, 2, 0.25).takeError();
   ASSERT_TRUE((bool)Error);
@@ -74,8 +80,8 @@ TEST(ClusteringTest, Clusters3D_InvalidSize) {
 
 TEST(ClusteringTest, Clusters3D_InvalidOrder) {
   std::vector<InstructionBenchmark> Points(6);
-  Points[0].Measurements = {{"x", 0.01, ""}, {"y", 1.02, ""}};
-  Points[1].Measurements = {{"y", 1.02, ""}, {"x", 1.98, ""}};
+  Points[0].Measurements = {{"x", 0.01, 0.0, ""}, {"y", 1.02, 0.0, ""}};
+  Points[1].Measurements = {{"y", 1.02, 0.0, ""}, {"x", 1.98, 0.0, ""}};
   auto Error =
       InstructionBenchmarkClustering::create(Points, 2, 0.25).takeError();
   ASSERT_TRUE((bool)Error);
diff --git a/llvm/unittests/tools/llvm-exegesis/X86/BenchmarkResultTest.cpp b/llvm/unittests/tools/llvm-exegesis/X86/BenchmarkResultTest.cpp
index bdac559276b..67e8d2c5bc9 100644
--- a/llvm/unittests/tools/llvm-exegesis/X86/BenchmarkResultTest.cpp
+++ b/llvm/unittests/tools/llvm-exegesis/X86/BenchmarkResultTest.cpp
@@ -28,7 +28,8 @@ using ::testing::Property;
 namespace exegesis {
 
 bool operator==(const BenchmarkMeasure &A, const BenchmarkMeasure &B) {
-  return std::tie(A.Key, A.Value) == std::tie(B.Key, B.Value);
+  return std::tie(A.Key, A.PerInstructionValue, A.PerSnippetValue) ==
+         std::tie(B.Key, B.PerInstructionValue, B.PerSnippetValue);
 }
 
 static std::string Dump(const llvm::MCInst &McInst) {
@@ -75,8 +76,8 @@ TEST(BenchmarkResultTest, WriteToAndReadFromDisk) {
   ToDisk.CpuName = "cpu_name";
   ToDisk.LLVMTriple = "llvm_triple";
   ToDisk.NumRepetitions = 1;
-  ToDisk.Measurements.push_back(BenchmarkMeasure{"a", 1, "debug a"});
-  ToDisk.Measurements.push_back(BenchmarkMeasure{"b", 2, ""});
+  ToDisk.Measurements.push_back(BenchmarkMeasure{"a", 1, 1, "debug a"});
+  ToDisk.Measurements.push_back(BenchmarkMeasure{"b", 2, 2, ""});
   ToDisk.Error = "error";
   ToDisk.Info = "info";
 
@@ -123,12 +124,12 @@ TEST(BenchmarkResultTest, WriteToAndReadFromDisk) {
   }
 }
 
-TEST(BenchmarkResultTest, BenchmarkMeasureStats) {
-  BenchmarkMeasureStats Stats;
-  Stats.push(BenchmarkMeasure{"a", 0.5, "debug a"});
-  Stats.push(BenchmarkMeasure{"a", 1.5, "debug a"});
-  Stats.push(BenchmarkMeasure{"a", -1.0, "debug a"});
-  Stats.push(BenchmarkMeasure{"a", 0.0, "debug a"});
+TEST(BenchmarkResultTest, PerInstructionStats) {
+  PerInstructionStats Stats;
+  Stats.push(BenchmarkMeasure{"a", 0.5, 0.0, "debug a"});
+  Stats.push(BenchmarkMeasure{"a", 1.5, 0.0, "debug a"});
+  Stats.push(BenchmarkMeasure{"a", -1.0, 0.0, "debug a"});
+  Stats.push(BenchmarkMeasure{"a", 0.0, 0.0, "debug a"});
   EXPECT_EQ(Stats.min(), -1.0);
   EXPECT_EQ(Stats.max(), 1.5);
   EXPECT_EQ(Stats.avg(), 0.25); // (0.5+1.5-1.0+0.0) / 4