1 //===-- Clustering.cpp ------------------------------------------*- C++ -*-===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 
9 #include "Clustering.h"
10 #include "Error.h"
11 #include "llvm/ADT/SetVector.h"
12 #include "llvm/ADT/SmallSet.h"
13 #include "llvm/ADT/SmallVector.h"
14 #include <algorithm>
15 #include <string>
16 #include <vector>
17 #include <deque>
18 
19 namespace llvm {
20 namespace exegesis {
21 
22 // The clustering problem has the following characteristics:
23 //  (A) - Low dimension (dimensions are typically proc resource units,
24 //    typically < 10).
25 //  (B) - Number of points : ~thousands (points are measurements of an MCInst)
26 //  (C) - Number of clusters: ~tens.
27 //  (D) - The number of clusters is not known /a priory/.
28 //  (E) - The amount of noise is relatively small.
29 // The problem is rather small. In terms of algorithms, (D) disqualifies
30 // k-means and makes algorithms such as DBSCAN[1] or OPTICS[2] more applicable.
31 //
32 // We've used DBSCAN here because it's simple to implement. This is a pretty
33 // straightforward and inefficient implementation of the pseudocode in [2].
34 //
35 // [1] https://en.wikipedia.org/wiki/DBSCAN
36 // [2] https://en.wikipedia.org/wiki/OPTICS_algorithm
37 
38 // Finds the points at distance less than sqrt(EpsilonSquared) of Q (not
39 // including Q).
rangeQuery(const size_t Q,std::vector<size_t> & Neighbors) const40 void InstructionBenchmarkClustering::rangeQuery(
41     const size_t Q, std::vector<size_t> &Neighbors) const {
42   Neighbors.clear();
43   Neighbors.reserve(Points_.size() - 1); // The Q itself isn't a neighbor.
44   const auto &QMeasurements = Points_[Q].Measurements;
45   for (size_t P = 0, NumPoints = Points_.size(); P < NumPoints; ++P) {
46     if (P == Q)
47       continue;
48     const auto &PMeasurements = Points_[P].Measurements;
49     if (PMeasurements.empty()) // Error point.
50       continue;
51     if (isNeighbour(PMeasurements, QMeasurements,
52                     AnalysisClusteringEpsilonSquared_)) {
53       Neighbors.push_back(P);
54     }
55   }
56 }
57 
58 // Given a set of points, checks that all the points are neighbours
59 // up to AnalysisClusteringEpsilon. This is O(2*N).
areAllNeighbours(ArrayRef<size_t> Pts) const60 bool InstructionBenchmarkClustering::areAllNeighbours(
61     ArrayRef<size_t> Pts) const {
62   // First, get the centroid of this group of points. This is O(N).
63   SchedClassClusterCentroid G;
64   for_each(Pts, [this, &G](size_t P) {
65     assert(P < Points_.size());
66     ArrayRef<BenchmarkMeasure> Measurements = Points_[P].Measurements;
67     if (Measurements.empty()) // Error point.
68       return;
69     G.addPoint(Measurements);
70   });
71   const std::vector<BenchmarkMeasure> Centroid = G.getAsPoint();
72 
73   // Since we will be comparing with the centroid, we need to halve the epsilon.
74   double AnalysisClusteringEpsilonHalvedSquared =
75       AnalysisClusteringEpsilonSquared_ / 4.0;
76 
77   // And now check that every point is a neighbour of the centroid. Also O(N).
78   return all_of(
79       Pts, [this, &Centroid, AnalysisClusteringEpsilonHalvedSquared](size_t P) {
80         assert(P < Points_.size());
81         const auto &PMeasurements = Points_[P].Measurements;
82         if (PMeasurements.empty()) // Error point.
83           return true;             // Pretend that error point is a neighbour.
84         return isNeighbour(PMeasurements, Centroid,
85                            AnalysisClusteringEpsilonHalvedSquared);
86       });
87 }
88 
InstructionBenchmarkClustering(const std::vector<InstructionBenchmark> & Points,const double AnalysisClusteringEpsilonSquared)89 InstructionBenchmarkClustering::InstructionBenchmarkClustering(
90     const std::vector<InstructionBenchmark> &Points,
91     const double AnalysisClusteringEpsilonSquared)
92     : Points_(Points),
93       AnalysisClusteringEpsilonSquared_(AnalysisClusteringEpsilonSquared),
94       NoiseCluster_(ClusterId::noise()), ErrorCluster_(ClusterId::error()) {}
95 
validateAndSetup()96 Error InstructionBenchmarkClustering::validateAndSetup() {
97   ClusterIdForPoint_.resize(Points_.size());
98   // Mark erroneous measurements out.
99   // All points must have the same number of dimensions, in the same order.
100   const std::vector<BenchmarkMeasure> *LastMeasurement = nullptr;
101   for (size_t P = 0, NumPoints = Points_.size(); P < NumPoints; ++P) {
102     const auto &Point = Points_[P];
103     if (!Point.Error.empty()) {
104       ClusterIdForPoint_[P] = ClusterId::error();
105       ErrorCluster_.PointIndices.push_back(P);
106       continue;
107     }
108     const auto *CurMeasurement = &Point.Measurements;
109     if (LastMeasurement) {
110       if (LastMeasurement->size() != CurMeasurement->size()) {
111         return make_error<ClusteringError>(
112             "inconsistent measurement dimensions");
113       }
114       for (size_t I = 0, E = LastMeasurement->size(); I < E; ++I) {
115         if (LastMeasurement->at(I).Key != CurMeasurement->at(I).Key) {
116           return make_error<ClusteringError>(
117               "inconsistent measurement dimensions keys");
118         }
119       }
120     }
121     LastMeasurement = CurMeasurement;
122   }
123   if (LastMeasurement) {
124     NumDimensions_ = LastMeasurement->size();
125   }
126   return Error::success();
127 }
128 
clusterizeDbScan(const size_t MinPts)129 void InstructionBenchmarkClustering::clusterizeDbScan(const size_t MinPts) {
130   std::vector<size_t> Neighbors; // Persistent buffer to avoid allocs.
131   for (size_t P = 0, NumPoints = Points_.size(); P < NumPoints; ++P) {
132     if (!ClusterIdForPoint_[P].isUndef())
133       continue; // Previously processed in inner loop.
134     rangeQuery(P, Neighbors);
135     if (Neighbors.size() + 1 < MinPts) { // Density check.
136       // The region around P is not dense enough to create a new cluster, mark
137       // as noise for now.
138       ClusterIdForPoint_[P] = ClusterId::noise();
139       continue;
140     }
141 
142     // Create a new cluster, add P.
143     Clusters_.emplace_back(ClusterId::makeValid(Clusters_.size()));
144     Cluster &CurrentCluster = Clusters_.back();
145     ClusterIdForPoint_[P] = CurrentCluster.Id; /* Label initial point */
146     CurrentCluster.PointIndices.push_back(P);
147 
148     // Process P's neighbors.
149     SetVector<size_t, std::deque<size_t>> ToProcess;
150     ToProcess.insert(Neighbors.begin(), Neighbors.end());
151     while (!ToProcess.empty()) {
152       // Retrieve a point from the set.
153       const size_t Q = *ToProcess.begin();
154       ToProcess.erase(ToProcess.begin());
155 
156       if (ClusterIdForPoint_[Q].isNoise()) {
157         // Change noise point to border point.
158         ClusterIdForPoint_[Q] = CurrentCluster.Id;
159         CurrentCluster.PointIndices.push_back(Q);
160         continue;
161       }
162       if (!ClusterIdForPoint_[Q].isUndef()) {
163         continue; // Previously processed.
164       }
165       // Add Q to the current custer.
166       ClusterIdForPoint_[Q] = CurrentCluster.Id;
167       CurrentCluster.PointIndices.push_back(Q);
168       // And extend to the neighbors of Q if the region is dense enough.
169       rangeQuery(Q, Neighbors);
170       if (Neighbors.size() + 1 >= MinPts) {
171         ToProcess.insert(Neighbors.begin(), Neighbors.end());
172       }
173     }
174   }
175   // assert(Neighbors.capacity() == (Points_.size() - 1));
176   // ^ True, but it is not quaranteed to be true in all the cases.
177 
178   // Add noisy points to noise cluster.
179   for (size_t P = 0, NumPoints = Points_.size(); P < NumPoints; ++P) {
180     if (ClusterIdForPoint_[P].isNoise()) {
181       NoiseCluster_.PointIndices.push_back(P);
182     }
183   }
184 }
185 
clusterizeNaive(unsigned NumOpcodes)186 void InstructionBenchmarkClustering::clusterizeNaive(unsigned NumOpcodes) {
187   // Given an instruction Opcode, which are the benchmarks of this instruction?
188   std::vector<SmallVector<size_t, 1>> OpcodeToPoints;
189   OpcodeToPoints.resize(NumOpcodes);
190   size_t NumOpcodesSeen = 0;
191   for (size_t P = 0, NumPoints = Points_.size(); P < NumPoints; ++P) {
192     const InstructionBenchmark &Point = Points_[P];
193     const unsigned Opcode = Point.keyInstruction().getOpcode();
194     assert(Opcode < NumOpcodes && "NumOpcodes is incorrect (too small)");
195     SmallVectorImpl<size_t> &PointsOfOpcode = OpcodeToPoints[Opcode];
196     if (PointsOfOpcode.empty()) // If we previously have not seen any points of
197       ++NumOpcodesSeen; // this opcode, then naturally this is the new opcode.
198     PointsOfOpcode.emplace_back(P);
199   }
200   assert(OpcodeToPoints.size() == NumOpcodes && "sanity check");
201   assert(NumOpcodesSeen <= NumOpcodes &&
202          "can't see more opcodes than there are total opcodes");
203   assert(NumOpcodesSeen <= Points_.size() &&
204          "can't see more opcodes than there are total points");
205 
206   Clusters_.reserve(NumOpcodesSeen); // One cluster per opcode.
207   for (ArrayRef<size_t> PointsOfOpcode :
208        make_filter_range(OpcodeToPoints, [](ArrayRef<size_t> PointsOfOpcode) {
209          return !PointsOfOpcode.empty(); // Ignore opcodes with no points.
210        })) {
211     // Create a new cluster.
212     Clusters_.emplace_back(ClusterId::makeValid(
213         Clusters_.size(), /*IsUnstable=*/!areAllNeighbours(PointsOfOpcode)));
214     Cluster &CurrentCluster = Clusters_.back();
215     // Mark points as belonging to the new cluster.
216     for_each(PointsOfOpcode, [this, &CurrentCluster](size_t P) {
217       ClusterIdForPoint_[P] = CurrentCluster.Id;
218     });
219     // And add all the points of this opcode to the new cluster.
220     CurrentCluster.PointIndices.reserve(PointsOfOpcode.size());
221     CurrentCluster.PointIndices.assign(PointsOfOpcode.begin(),
222                                        PointsOfOpcode.end());
223     assert(CurrentCluster.PointIndices.size() == PointsOfOpcode.size());
224   }
225   assert(Clusters_.size() == NumOpcodesSeen);
226 }
227 
228 // Given an instruction Opcode, we can make benchmarks (measurements) of the
229 // instruction characteristics/performance. Then, to facilitate further analysis
230 // we group the benchmarks with *similar* characteristics into clusters.
231 // Now, this is all not entirely deterministic. Some instructions have variable
232 // characteristics, depending on their arguments. And thus, if we do several
233 // benchmarks of the same instruction Opcode, we may end up with *different*
234 // performance characteristics measurements. And when we then do clustering,
235 // these several benchmarks of the same instruction Opcode may end up being
236 // clustered into *different* clusters. This is not great for further analysis.
237 // We shall find every opcode with benchmarks not in just one cluster, and move
238 // *all* the benchmarks of said Opcode into one new unstable cluster per Opcode.
stabilize(unsigned NumOpcodes)239 void InstructionBenchmarkClustering::stabilize(unsigned NumOpcodes) {
240   // Given an instruction Opcode and Config, in which clusters do benchmarks of
241   // this instruction lie? Normally, they all should be in the same cluster.
242   struct OpcodeAndConfig {
243     explicit OpcodeAndConfig(const InstructionBenchmark &IB)
244         : Opcode(IB.keyInstruction().getOpcode()), Config(&IB.Key.Config) {}
245     unsigned Opcode;
246     const std::string *Config;
247 
248     auto Tie() const -> auto { return std::tie(Opcode, *Config); }
249 
250     bool operator<(const OpcodeAndConfig &O) const { return Tie() < O.Tie(); }
251     bool operator!=(const OpcodeAndConfig &O) const { return Tie() != O.Tie(); }
252   };
253   std::map<OpcodeAndConfig, SmallSet<ClusterId, 1>> OpcodeConfigToClusterIDs;
254   // Populate OpcodeConfigToClusterIDs and UnstableOpcodes data structures.
255   assert(ClusterIdForPoint_.size() == Points_.size() && "size mismatch");
256   for (auto Point : zip(Points_, ClusterIdForPoint_)) {
257     const ClusterId &ClusterIdOfPoint = std::get<1>(Point);
258     if (!ClusterIdOfPoint.isValid())
259       continue; // Only process fully valid clusters.
260     const OpcodeAndConfig Key(std::get<0>(Point));
261     SmallSet<ClusterId, 1> &ClusterIDsOfOpcode = OpcodeConfigToClusterIDs[Key];
262     ClusterIDsOfOpcode.insert(ClusterIdOfPoint);
263   }
264 
265   for (const auto &OpcodeConfigToClusterID : OpcodeConfigToClusterIDs) {
266     const SmallSet<ClusterId, 1> &ClusterIDs = OpcodeConfigToClusterID.second;
267     const OpcodeAndConfig &Key = OpcodeConfigToClusterID.first;
268     // We only care about unstable instructions.
269     if (ClusterIDs.size() < 2)
270       continue;
271 
272     // Create a new unstable cluster, one per Opcode.
273     Clusters_.emplace_back(ClusterId::makeValidUnstable(Clusters_.size()));
274     Cluster &UnstableCluster = Clusters_.back();
275     // We will find *at least* one point in each of these clusters.
276     UnstableCluster.PointIndices.reserve(ClusterIDs.size());
277 
278     // Go through every cluster which we recorded as containing benchmarks
279     // of this UnstableOpcode. NOTE: we only recorded valid clusters.
280     for (const ClusterId &CID : ClusterIDs) {
281       assert(CID.isValid() &&
282              "We only recorded valid clusters, not noise/error clusters.");
283       Cluster &OldCluster = Clusters_[CID.getId()]; // Valid clusters storage.
284       // Within each cluster, go through each point, and either move it to the
285       // new unstable cluster, or 'keep' it.
286       // In this case, we'll reshuffle OldCluster.PointIndices vector
287       // so that all the points that are *not* for UnstableOpcode are first,
288       // and the rest of the points is for the UnstableOpcode.
289       const auto it = std::stable_partition(
290           OldCluster.PointIndices.begin(), OldCluster.PointIndices.end(),
291           [this, &Key](size_t P) {
292             return OpcodeAndConfig(Points_[P]) != Key;
293           });
294       assert(std::distance(it, OldCluster.PointIndices.end()) > 0 &&
295              "Should have found at least one bad point");
296       // Mark to-be-moved points as belonging to the new cluster.
297       std::for_each(it, OldCluster.PointIndices.end(),
298                     [this, &UnstableCluster](size_t P) {
299                       ClusterIdForPoint_[P] = UnstableCluster.Id;
300                     });
301       // Actually append to-be-moved points to the new cluster.
302       UnstableCluster.PointIndices.insert(UnstableCluster.PointIndices.end(),
303                                           it, OldCluster.PointIndices.end());
304       // And finally, remove "to-be-moved" points form the old cluster.
305       OldCluster.PointIndices.erase(it, OldCluster.PointIndices.end());
306       // Now, the old cluster may end up being empty, but let's just keep it
307       // in whatever state it ended up. Purging empty clusters isn't worth it.
308     };
309     assert(UnstableCluster.PointIndices.size() > 1 &&
310            "New unstable cluster should end up with more than one point.");
311     assert(UnstableCluster.PointIndices.size() >= ClusterIDs.size() &&
312            "New unstable cluster should end up with no less points than there "
313            "was clusters");
314   }
315 }
316 
create(const std::vector<InstructionBenchmark> & Points,const ModeE Mode,const size_t DbscanMinPts,const double AnalysisClusteringEpsilon,Optional<unsigned> NumOpcodes)317 Expected<InstructionBenchmarkClustering> InstructionBenchmarkClustering::create(
318     const std::vector<InstructionBenchmark> &Points, const ModeE Mode,
319     const size_t DbscanMinPts, const double AnalysisClusteringEpsilon,
320     Optional<unsigned> NumOpcodes) {
321   InstructionBenchmarkClustering Clustering(
322       Points, AnalysisClusteringEpsilon * AnalysisClusteringEpsilon);
323   if (auto Error = Clustering.validateAndSetup()) {
324     return std::move(Error);
325   }
326   if (Clustering.ErrorCluster_.PointIndices.size() == Points.size()) {
327     return Clustering; // Nothing to cluster.
328   }
329 
330   if (Mode == ModeE::Dbscan) {
331     Clustering.clusterizeDbScan(DbscanMinPts);
332 
333     if (NumOpcodes.hasValue())
334       Clustering.stabilize(NumOpcodes.getValue());
335   } else /*if(Mode == ModeE::Naive)*/ {
336     if (!NumOpcodes.hasValue())
337       return make_error<Failure>(
338           "'naive' clustering mode requires opcode count to be specified");
339     Clustering.clusterizeNaive(NumOpcodes.getValue());
340   }
341 
342   return Clustering;
343 }
344 
addPoint(ArrayRef<BenchmarkMeasure> Point)345 void SchedClassClusterCentroid::addPoint(ArrayRef<BenchmarkMeasure> Point) {
346   if (Representative.empty())
347     Representative.resize(Point.size());
348   assert(Representative.size() == Point.size() &&
349          "All points should have identical dimensions.");
350 
351   for (auto I : zip(Representative, Point))
352     std::get<0>(I).push(std::get<1>(I));
353 }
354 
getAsPoint() const355 std::vector<BenchmarkMeasure> SchedClassClusterCentroid::getAsPoint() const {
356   std::vector<BenchmarkMeasure> ClusterCenterPoint(Representative.size());
357   for (auto I : zip(ClusterCenterPoint, Representative))
358     std::get<0>(I).PerInstructionValue = std::get<1>(I).avg();
359   return ClusterCenterPoint;
360 }
361 
validate(InstructionBenchmark::ModeE Mode) const362 bool SchedClassClusterCentroid::validate(
363     InstructionBenchmark::ModeE Mode) const {
364   size_t NumMeasurements = Representative.size();
365   switch (Mode) {
366   case InstructionBenchmark::Latency:
367     if (NumMeasurements != 1) {
368       errs()
369           << "invalid number of measurements in latency mode: expected 1, got "
370           << NumMeasurements << "\n";
371       return false;
372     }
373     break;
374   case InstructionBenchmark::Uops:
375     // Can have many measurements.
376     break;
377   case InstructionBenchmark::InverseThroughput:
378     if (NumMeasurements != 1) {
379       errs() << "invalid number of measurements in inverse throughput "
380                 "mode: expected 1, got "
381              << NumMeasurements << "\n";
382       return false;
383     }
384     break;
385   default:
386     llvm_unreachable("unimplemented measurement matching mode");
387     return false;
388   }
389 
390   return true; // All good.
391 }
392 
393 } // namespace exegesis
394 } // namespace llvm
395