Switch SmallSetVector to use DenseSet when it overflows its inline space.

Summary:
SetVector already used DenseSet, but SmallSetVector used std::set.  This
leads to surprising performance differences.  Moreover, it means that
the set of key types accepted by SetVector and SmallSetVector are
quite different!

In order to make this change, we had to convert some callsites that used
SmallSetVector<std::string, N> to use SmallSetVector<CachedHashString, N>
instead.

Reviewers: timshen

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D25648

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@284887 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/llvm/ADT/SetVector.h b/include/llvm/ADT/SetVector.h
index 455f8a89e107c3caa6aa57ea224c234dc5fba630..c6bc9bfb546b4574d49cc079aea04683fe5f446d 100644 (file)
--- a/include/llvm/ADT/SetVector.h
+++ b/include/llvm/ADT/SetVector.h
@@ -282,7 +282,8 @@ private:
 /// \brief A SetVector that performs no allocations if smaller than
 /// a certain size.
 template <typename T, unsigned N>
-class SmallSetVector : public SetVector<T, SmallVector<T, N>, SmallSet<T, N> > {
+class SmallSetVector
+    : public SetVector<T, SmallVector<T, N>, SmallDenseSet<T, N>> {
 public:
   SmallSetVector() {}
 

diff --git a/utils/TableGen/AsmMatcherEmitter.cpp b/utils/TableGen/AsmMatcherEmitter.cpp
index fb715364b7143dec586249f65d3c8859c2f861b6..5dbd5ec9bb041c7427bba67ecbf0da6b9c4cae86 100644 (file)
--- a/utils/TableGen/AsmMatcherEmitter.cpp
+++ b/utils/TableGen/AsmMatcherEmitter.cpp
@@ -97,6 +97,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "CodeGenTarget.h"
+#include "llvm/ADT/CachedHashString.h"
 #include "llvm/ADT/PointerUnion.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallPtrSet.h"
@@ -1822,10 +1823,11 @@ void MatchableInfo::buildAliasResultOperands() {
   }
 }
 
-static unsigned getConverterOperandID(const std::string &Name,
-                                      SmallSetVector<std::string, 16> &Table,
-                                      bool &IsNew) {
-  IsNew = Table.insert(Name);
+static unsigned
+getConverterOperandID(const std::string &Name,
+                      SmallSetVector<CachedHashString, 16> &Table,
+                      bool &IsNew) {
+  IsNew = Table.insert(CachedHashString(Name));
 
   unsigned ID = IsNew ? Table.size() - 1 : find(Table, Name) - Table.begin();
 
@@ -1838,8 +1840,8 @@ static void emitConvertFuncs(CodeGenTarget &Target, StringRef ClassName,
                              std::vector<std::unique_ptr<MatchableInfo>> &Infos,
                              bool HasMnemonicFirst, bool HasOptionalOperands,
                              raw_ostream &OS) {
-  SmallSetVector<std::string, 16> OperandConversionKinds;
-  SmallSetVector<std::string, 16> InstructionConversionKinds;
+  SmallSetVector<CachedHashString, 16> OperandConversionKinds;
+  SmallSetVector<CachedHashString, 16> InstructionConversionKinds;
   std::vector<std::vector<uint8_t> > ConversionTable;
   size_t MaxRowLength = 2; // minimum is custom converter plus terminator.
 
@@ -1911,9 +1913,9 @@ static void emitConvertFuncs(CodeGenTarget &Target, StringRef ClassName,
 
   // Pre-populate the operand conversion kinds with the standard always
   // available entries.
-  OperandConversionKinds.insert("CVT_Done");
-  OperandConversionKinds.insert("CVT_Reg");
-  OperandConversionKinds.insert("CVT_Tied");
+  OperandConversionKinds.insert(CachedHashString("CVT_Done"));
+  OperandConversionKinds.insert(CachedHashString("CVT_Reg"));
+  OperandConversionKinds.insert(CachedHashString("CVT_Tied"));
   enum { CVT_Done, CVT_Reg, CVT_Tied };
 
   for (auto &II : Infos) {
@@ -1925,13 +1927,13 @@ static void emitConvertFuncs(CodeGenTarget &Target, StringRef ClassName,
       II->ConversionFnKind = Signature;
 
       // Check if we have already generated this signature.
-      if (!InstructionConversionKinds.insert(Signature))
+      if (!InstructionConversionKinds.insert(CachedHashString(Signature)))
         continue;
 
       // Remember this converter for the kind enum.
       unsigned KindID = OperandConversionKinds.size();
-      OperandConversionKinds.insert("CVT_" +
-                                    getEnumNameForToken(AsmMatchConverter));
+      OperandConversionKinds.insert(
+          CachedHashString("CVT_" + getEnumNameForToken(AsmMatchConverter)));
 
       // Add the converter row for this instruction.
       ConversionTable.emplace_back();
@@ -2113,7 +2115,7 @@ static void emitConvertFuncs(CodeGenTarget &Target, StringRef ClassName,
 
     // Save the signature. If we already have it, don't add a new row
     // to the table.
-    if (!InstructionConversionKinds.insert(Signature))
+    if (!InstructionConversionKinds.insert(CachedHashString(Signature)))
       continue;
 
     // Add the row to the table.
@@ -2130,14 +2132,14 @@ static void emitConvertFuncs(CodeGenTarget &Target, StringRef ClassName,
 
   // Output the operand conversion kind enum.
   OS << "enum OperatorConversionKind {\n";
-  for (const std::string &Converter : OperandConversionKinds)
+  for (const auto &Converter : OperandConversionKinds)
     OS << "  " << Converter << ",\n";
   OS << "  CVT_NUM_CONVERTERS\n";
   OS << "};\n\n";
 
   // Output the instruction conversion kind enum.
   OS << "enum InstructionConversionKind {\n";
-  for (const std::string &Signature : InstructionConversionKinds)
+  for (const auto &Signature : InstructionConversionKinds)
     OS << "  " << Signature << ",\n";
   OS << "  CVT_NUM_SIGNATURES\n";
   OS << "};\n\n";

diff --git a/utils/TableGen/FixedLenDecoderEmitter.cpp b/utils/TableGen/FixedLenDecoderEmitter.cpp
index d0c4e0fd2f9960e171e3fe9489ebfcaef2702057..f2d54e6fe5afb8aac3d7cef5d1eaa144c0de5c6c 100644 (file)
--- a/utils/TableGen/FixedLenDecoderEmitter.cpp
+++ b/utils/TableGen/FixedLenDecoderEmitter.cpp
@@ -14,6 +14,7 @@
 
 #include "CodeGenTarget.h"
 #include "llvm/ADT/APInt.h"
+#include "llvm/ADT/CachedHashString.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/ADT/StringRef.h"
@@ -66,8 +67,8 @@ typedef std::vector<uint8_t> DecoderTable;
 typedef uint32_t DecoderFixup;
 typedef std::vector<DecoderFixup> FixupList;
 typedef std::vector<FixupList> FixupScopeList;
-typedef SmallSetVector<std::string, 16> PredicateSet;
-typedef SmallSetVector<std::string, 16> DecoderSet;
+typedef SmallSetVector<CachedHashString, 16> PredicateSet;
+typedef SmallSetVector<CachedHashString, 16> DecoderSet;
 struct DecoderTableInfo {
   DecoderTable Table;
   FixupScopeList FixupStack;
@@ -1106,7 +1107,7 @@ unsigned FilterChooser::getDecoderIndex(DecoderSet &Decoders,
   // overkill for now, though.
 
   // Make sure the predicate is in the table.
-  Decoders.insert(StringRef(Decoder));
+  Decoders.insert(CachedHashString(Decoder));
   // Now figure out the index for when we write out the table.
   DecoderSet::const_iterator P = find(Decoders, Decoder.str());
   return (unsigned)(P - Decoders.begin());
@@ -1179,9 +1180,9 @@ unsigned FilterChooser::getPredicateIndex(DecoderTableInfo &TableInfo,
   // overkill for now, though.
 
   // Make sure the predicate is in the table.
-  TableInfo.Predicates.insert(Predicate.str());
+  TableInfo.Predicates.insert(CachedHashString(Predicate));
   // Now figure out the index for when we write out the table.
-  PredicateSet::const_iterator P = find(TableInfo.Predicates, Predicate.str());
+  PredicateSet::const_iterator P = find(TableInfo.Predicates, Predicate);
   return (unsigned)(P - TableInfo.Predicates.begin());
 }