Sync to upstream/release/593 (#1024)

- Updated Roblox copyright to 2023
- Floor division operator `//` (implements #832)
- Autocomplete now shows `end` within `do` blocks
- Restore BraceType when using `Lexer::lookahead` (fixes #1019)

# New typechecker

- Subtyping tests between metatables and tables
- Subtyping tests between string singletons and tables
- Subtyping tests for class types

# Native codegen

- Fixed macOS test failure (wrong spill restore offset)
- Fixed clobbering of non-volatile xmm registers on Windows
- Fixed wrong storage location of SSA reg spills
- Implemented A64 support for add/sub extended
- Eliminated zextReg from A64 lowering
- Remove identical table slot lookups
- Propagate values from predecessor into the linear block
- Disabled reuse slot optimization
- Keep `LuaNode::val` check for nil when optimizing `CHECK_SLOT_MATCH`
- Implemented IR translation of `table.insert` builtin
- Fixed mmap error handling on macOS/Linux

# Tooling

- Used `|` as a column separator instead of `+` in `bench.py`
- Added a `table.sort` micro-benchmark
- Switched `libprotobuf-mutator` to a less problematic version

Analysis/include/Luau/Metamethods.h

@@ -19,6 +19,7 @@ static const std::unordered_map<AstExprBinary::Op, const char*> kBinaryOpMetamet
     {AstExprBinary::Op::Sub, "__sub"},
     {AstExprBinary::Op::Mul, "__mul"},
     {AstExprBinary::Op::Div, "__div"},
+    {AstExprBinary::Op::FloorDiv, "__idiv"},
     {AstExprBinary::Op::Pow, "__pow"},
     {AstExprBinary::Op::Mod, "__mod"},
     {AstExprBinary::Op::Concat, "__concat"},

Analysis/include/Luau/Subtyping.h

@@ -2,6 +2,7 @@
 #pragma once
 
 #include "Luau/Type.h"
+#include "Luau/TypePack.h"
 #include "Luau/UnifierSharedState.h"
 
 #include <vector>
@@ -14,8 +15,10 @@ template<typename A, typename B>
 struct TryPair;
 struct InternalErrorReporter;
 
+class TypeIds;
 class Normalizer;
 struct NormalizedType;
+struct NormalizedClassType;
 
 struct SubtypingResult
 {
@@ -30,6 +33,8 @@ struct SubtypingResult
     void andAlso(const SubtypingResult& other);
     void orElse(const SubtypingResult& other);
 
+    // Only negates the `isSubtype`.
+    static SubtypingResult negate(const SubtypingResult& result);
     static SubtypingResult all(const std::vector<SubtypingResult>& results);
     static SubtypingResult any(const std::vector<SubtypingResult>& results);
 };
@@ -63,7 +68,7 @@ struct Subtyping
     DenseHashMap<TypePackId, TypePackId> mappedGenericPacks{nullptr};
 
     using SeenSet = std::unordered_set<std::pair<TypeId, TypeId>, TypeIdPairHash>;
-    
+
     SeenSet seenTypes;
 
     // TODO cache
@@ -88,8 +93,19 @@ private:
     SubtypingResult isSubtype_(const SingletonType* subSingleton, const PrimitiveType* superPrim);
     SubtypingResult isSubtype_(const SingletonType* subSingleton, const SingletonType* superSingleton);
     SubtypingResult isSubtype_(const TableType* subTable, const TableType* superTable);
+    SubtypingResult isSubtype_(const MetatableType* subMt, const MetatableType* superMt);
+    SubtypingResult isSubtype_(const MetatableType* subMt, const TableType* superTable);
+    SubtypingResult isSubtype_(const ClassType* subClass, const ClassType* superClass);
+    SubtypingResult isSubtype_(const ClassType* subClass, const TableType* superTable); // Actually a class <: shape.
     SubtypingResult isSubtype_(const FunctionType* subFunction, const FunctionType* superFunction);
+    SubtypingResult isSubtype_(const PrimitiveType* subPrim, const TableType* superTable);
+    SubtypingResult isSubtype_(const SingletonType* subSingleton, const TableType* superTable);
+
     SubtypingResult isSubtype_(const NormalizedType* subNorm, const NormalizedType* superNorm);
+    SubtypingResult isSubtype_(const NormalizedClassType& subClass, const NormalizedClassType& superClass, const TypeIds& superTables);
+    SubtypingResult isSubtype_(const TypeIds& subTypes, const TypeIds& superTypes);
+
+    SubtypingResult isSubtype_(const VariadicTypePack* subVariadic, const VariadicTypePack* superVariadic);
 
     bool bindGeneric(TypeId subTp, TypeId superTp);
     bool bindGeneric(TypePackId subTp, TypePackId superTp);

Analysis/src/AstJsonEncoder.cpp

@@ -8,6 +8,8 @@
 
 #include <math.h>
 
+LUAU_FASTFLAG(LuauFloorDivision)
+
 namespace Luau
 {
 
@@ -514,6 +516,9 @@ struct AstJsonEncoder : public AstVisitor
             return writeString("Mul");
         case AstExprBinary::Div:
             return writeString("Div");
+        case AstExprBinary::FloorDiv:
+            LUAU_ASSERT(FFlag::LuauFloorDivision);
+            return writeString("FloorDiv");
         case AstExprBinary::Mod:
             return writeString("Mod");
         case AstExprBinary::Pow:
@@ -536,6 +541,8 @@ struct AstJsonEncoder : public AstVisitor
             return writeString("And");
         case AstExprBinary::Or:
             return writeString("Or");
+        default:
+            LUAU_ASSERT(!"Unknown Op");
         }
     }
 

Analysis/src/Autocomplete.cpp

@@ -15,6 +15,7 @@
 LUAU_FASTFLAG(DebugLuauReadWriteProperties)
 LUAU_FASTFLAGVARIABLE(LuauAnonymousAutofilled1, false);
 LUAU_FASTFLAGVARIABLE(LuauAutocompleteLastTypecheck, false)
+LUAU_FASTFLAGVARIABLE(LuauAutocompleteDoEnd, false)
 LUAU_FASTFLAGVARIABLE(LuauAutocompleteStringLiteralBounds, false);
 
 static const std::unordered_set<std::string> kStatementStartingKeywords = {
@@ -1089,14 +1090,19 @@ static AutocompleteEntryMap autocompleteStatement(
     {
         if (AstStatForIn* statForIn = (*it)->as<AstStatForIn>(); statForIn && !statForIn->hasEnd)
             result.emplace("end", AutocompleteEntry{AutocompleteEntryKind::Keyword});
-        if (AstStatFor* statFor = (*it)->as<AstStatFor>(); statFor && !statFor->hasEnd)
+        else if (AstStatFor* statFor = (*it)->as<AstStatFor>(); statFor && !statFor->hasEnd)
             result.emplace("end", AutocompleteEntry{AutocompleteEntryKind::Keyword});
-        if (AstStatIf* statIf = (*it)->as<AstStatIf>(); statIf && !statIf->hasEnd)
+        else if (AstStatIf* statIf = (*it)->as<AstStatIf>(); statIf && !statIf->hasEnd)
             result.emplace("end", AutocompleteEntry{AutocompleteEntryKind::Keyword});
-        if (AstStatWhile* statWhile = (*it)->as<AstStatWhile>(); statWhile && !statWhile->hasEnd)
+        else if (AstStatWhile* statWhile = (*it)->as<AstStatWhile>(); statWhile && !statWhile->hasEnd)
             result.emplace("end", AutocompleteEntry{AutocompleteEntryKind::Keyword});
-        if (AstExprFunction* exprFunction = (*it)->as<AstExprFunction>(); exprFunction && !exprFunction->hasEnd)
+        else if (AstExprFunction* exprFunction = (*it)->as<AstExprFunction>(); exprFunction && !exprFunction->hasEnd)
             result.emplace("end", AutocompleteEntry{AutocompleteEntryKind::Keyword});
+        if (FFlag::LuauAutocompleteDoEnd)
+        {
+            if (AstStatBlock* exprBlock = (*it)->as<AstStatBlock>(); exprBlock && !exprBlock->hasEnd)
+                result.emplace("end", AutocompleteEntry{AutocompleteEntryKind::Keyword});
+        }
     }
 
     if (ancestry.size() >= 2)

Analysis/src/ConstraintGraphBuilder.cpp

@@ -24,6 +24,7 @@ LUAU_FASTINT(LuauCheckRecursionLimit);
 LUAU_FASTFLAG(DebugLuauLogSolverToJson);
 LUAU_FASTFLAG(DebugLuauMagicTypes);
 LUAU_FASTFLAG(LuauParseDeclareClassIndexer);
+LUAU_FASTFLAG(LuauFloorDivision);
 
 namespace Luau
 {
@@ -1170,7 +1171,8 @@ static bool isMetamethod(const Name& name)
 {
     return name == "__index" || name == "__newindex" || name == "__call" || name == "__concat" || name == "__unm" || name == "__add" ||
            name == "__sub" || name == "__mul" || name == "__div" || name == "__mod" || name == "__pow" || name == "__tostring" ||
-           name == "__metatable" || name == "__eq" || name == "__lt" || name == "__le" || name == "__mode" || name == "__iter" || name == "__len";
+           name == "__metatable" || name == "__eq" || name == "__lt" || name == "__le" || name == "__mode" || name == "__iter" || name == "__len" ||
+           (FFlag::LuauFloorDivision && name == "__idiv");
 }
 
 ControlFlow ConstraintGraphBuilder::visit(const ScopePtr& scope, AstStatDeclareClass* declaredClass)

Analysis/src/ConstraintSolver.cpp

@@ -22,6 +22,7 @@
 #include "Luau/VisitType.h"
 
 LUAU_FASTFLAGVARIABLE(DebugLuauLogSolver, false);
+LUAU_FASTFLAG(LuauFloorDivision);
 
 namespace Luau
 {
@@ -719,6 +720,8 @@ bool ConstraintSolver::tryDispatch(const BinaryConstraint& c, NotNull<const Cons
     // Metatables go first, even if there is primitive behavior.
     if (auto it = kBinaryOpMetamethods.find(c.op); it != kBinaryOpMetamethods.end())
     {
+        LUAU_ASSERT(FFlag::LuauFloorDivision || c.op != AstExprBinary::Op::FloorDiv);
+
         // Metatables are not the same. The metamethod will not be invoked.
         if ((c.op == AstExprBinary::Op::CompareEq || c.op == AstExprBinary::Op::CompareNe) &&
             getMetatable(leftType, builtinTypes) != getMetatable(rightType, builtinTypes))
@@ -806,9 +809,12 @@ bool ConstraintSolver::tryDispatch(const BinaryConstraint& c, NotNull<const Cons
     case AstExprBinary::Op::Sub:
     case AstExprBinary::Op::Mul:
     case AstExprBinary::Op::Div:
+    case AstExprBinary::Op::FloorDiv:
     case AstExprBinary::Op::Pow:
     case AstExprBinary::Op::Mod:
     {
+        LUAU_ASSERT(FFlag::LuauFloorDivision || c.op != AstExprBinary::Op::FloorDiv);
+
         const NormalizedType* normLeftTy = normalizer->normalize(leftType);
         if (hasTypeInIntersection<FreeType>(leftType) && force)
             asMutable(leftType)->ty.emplace<BoundType>(anyPresent ? builtinTypes->anyType : builtinTypes->numberType);

Analysis/src/Error.cpp

@@ -12,7 +12,6 @@
 #include <stdexcept>
 #include <type_traits>
 
-LUAU_FASTFLAGVARIABLE(LuauIndentTypeMismatch, false)
 LUAU_FASTINTVARIABLE(LuauIndentTypeMismatchMaxTypeLength, 10)
 
 static std::string wrongNumberOfArgsString(
@@ -94,31 +93,18 @@ struct ErrorConverter
                     {
                         std::string givenModuleName = fileResolver->getHumanReadableModuleName(*givenDefinitionModule);
                         std::string wantedModuleName = fileResolver->getHumanReadableModuleName(*wantedDefinitionModule);
-                        if (FFlag::LuauIndentTypeMismatch)
-                            result = constructErrorMessage(givenTypeName, wantedTypeName, givenModuleName, wantedModuleName);
-                        else
-                            result = "Type '" + givenTypeName + "' from '" + givenModuleName + "' could not be converted into '" + wantedTypeName +
-                                     "' from '" + wantedModuleName + "'";
+                        result = constructErrorMessage(givenTypeName, wantedTypeName, givenModuleName, wantedModuleName);
                     }
                     else
                     {
-                        if (FFlag::LuauIndentTypeMismatch)
-                            result = constructErrorMessage(givenTypeName, wantedTypeName, *givenDefinitionModule, *wantedDefinitionModule);
-                        else
-                            result = "Type '" + givenTypeName + "' from '" + *givenDefinitionModule + "' could not be converted into '" +
-                                     wantedTypeName + "' from '" + *wantedDefinitionModule + "'";
+                        result = constructErrorMessage(givenTypeName, wantedTypeName, *givenDefinitionModule, *wantedDefinitionModule);
                     }
                 }
             }
         }
 
         if (result.empty())
-        {
-            if (FFlag::LuauIndentTypeMismatch)
-                result = constructErrorMessage(givenTypeName, wantedTypeName, std::nullopt, std::nullopt);
-            else
-                result = "Type '" + givenTypeName + "' could not be converted into '" + wantedTypeName + "'";
-        }
+            result = constructErrorMessage(givenTypeName, wantedTypeName, std::nullopt, std::nullopt);
 
 
         if (tm.error)
@@ -126,7 +112,7 @@ struct ErrorConverter
             result += "\ncaused by:\n  ";
 
             if (!tm.reason.empty())
-                result += tm.reason + (FFlag::LuauIndentTypeMismatch ? " \n" : " ");
+                result += tm.reason + " \n";
 
             result += Luau::toString(*tm.error, TypeErrorToStringOptions{fileResolver});
         }

Analysis/src/Subtyping.cpp

@@ -58,6 +58,15 @@ void SubtypingResult::orElse(const SubtypingResult& other)
     normalizationTooComplex |= other.normalizationTooComplex;
 }
 
+SubtypingResult SubtypingResult::negate(const SubtypingResult& result)
+{
+    return SubtypingResult{
+        !result.isSubtype,
+        result.isErrorSuppressing,
+        result.normalizationTooComplex,
+    };
+}
+
 SubtypingResult SubtypingResult::all(const std::vector<SubtypingResult>& results)
 {
     SubtypingResult acc{true, false};
@@ -137,10 +146,10 @@ SubtypingResult Subtyping::isSubtype_(TypeId subTy, TypeId superTy)
 
     SeenSetPopper ssp{&seenTypes, typePair};
 
-    if (auto superUnion = get<UnionType>(superTy))
-        return isSubtype_(subTy, superUnion);
-    else if (auto subUnion = get<UnionType>(subTy))
+    if (auto subUnion = get<UnionType>(subTy))
         return isSubtype_(subUnion, superTy);
+    else if (auto superUnion = get<UnionType>(superTy))
+        return isSubtype_(subTy, superUnion);
     else if (auto superIntersection = get<IntersectionType>(superTy))
         return isSubtype_(subTy, superIntersection);
     else if (auto subIntersection = get<IntersectionType>(subTy))
@@ -196,6 +205,18 @@ SubtypingResult Subtyping::isSubtype_(TypeId subTy, TypeId superTy)
         return isSubtype_(p);
     else if (auto p = get2<TableType, TableType>(subTy, superTy))
         return isSubtype_(p);
+    else if (auto p = get2<MetatableType, MetatableType>(subTy, superTy))
+        return isSubtype_(p);
+    else if (auto p = get2<MetatableType, TableType>(subTy, superTy))
+        return isSubtype_(p);
+    else if (auto p = get2<ClassType, ClassType>(subTy, superTy))
+        return isSubtype_(p);
+    else if (auto p = get2<ClassType, TableType>(subTy, superTy))
+        return isSubtype_(p);
+    else if (auto p = get2<PrimitiveType, TableType>(subTy, superTy))
+        return isSubtype_(p);
+    else if (auto p = get2<SingletonType, TableType>(subTy, superTy))
+        return isSubtype_(p);
 
     return {false};
 }
@@ -323,7 +344,7 @@ SubtypingResult Subtyping::isSubtype_(TypePackId subTp, TypePackId superTp)
     {
         if (auto p = get2<VariadicTypePack, VariadicTypePack>(*subTail, *superTail))
         {
-            results.push_back(isSubtype_(p.first->ty, p.second->ty));
+            results.push_back(isSubtype_(p));
         }
         else if (auto p = get2<GenericTypePack, GenericTypePack>(*subTail, *superTail))
         {
@@ -472,7 +493,6 @@ SubtypingResult Subtyping::isSubtype_(TypeId subTy, const IntersectionType* supe
 
 SubtypingResult Subtyping::isSubtype_(const IntersectionType* subIntersection, TypeId superTy)
 {
-    // TODO: Semantic subtyping here.
     // As per TAPL: A & B <: T iff A <: T || B <: T
     std::vector<SubtypingResult> subtypings;
     for (TypeId ty : subIntersection)
@@ -520,6 +540,59 @@ SubtypingResult Subtyping::isSubtype_(const TableType* subTable, const TableType
     return result;
 }
 
+SubtypingResult Subtyping::isSubtype_(const MetatableType* subMt, const MetatableType* superMt)
+{
+    return SubtypingResult::all({
+        isSubtype_(subMt->table, superMt->table),
+        isSubtype_(subMt->metatable, superMt->metatable),
+    });
+}
+
+SubtypingResult Subtyping::isSubtype_(const MetatableType* subMt, const TableType* superTable)
+{
+    if (auto subTable = get<TableType>(subMt->table)) {
+        // Metatables cannot erase properties from the table they're attached to, so
+        // the subtyping rule for this is just if the table component is a subtype
+        // of the supertype table.
+        //
+        // There's a flaw here in that if the __index metamethod contributes a new
+        // field that would satisfy the subtyping relationship, we'll erronously say
+        // that the metatable isn't a subtype of the table, even though they have
+        // compatible properties/shapes. We'll revisit this later when we have a
+        // better understanding of how important this is.
+        return isSubtype_(subTable, superTable);
+    }
+    else
+    {
+        // TODO: This may be a case we actually hit?
+        return {false};
+    }
+}
+
+SubtypingResult Subtyping::isSubtype_(const ClassType* subClass, const ClassType* superClass)
+{
+    return {isSubclass(subClass, superClass)};
+}
+
+SubtypingResult Subtyping::isSubtype_(const ClassType* subClass, const TableType* superTable)
+{
+    SubtypingResult result{true};
+
+    for (const auto& [name, prop]: superTable->props)
+    {
+        if (auto classProp = lookupClassProp(subClass, name))
+        {
+            // Table properties are invariant
+            result.andAlso(isSubtype_(classProp->type(), prop.type()));
+            result.andAlso(isSubtype_(prop.type(), classProp->type()));
+        }
+        else
+            return SubtypingResult{false};
+    }
+
+    return result;
+}
+
 SubtypingResult Subtyping::isSubtype_(const FunctionType* subFunction, const FunctionType* superFunction)
 {
     SubtypingResult result;
@@ -533,6 +606,47 @@ SubtypingResult Subtyping::isSubtype_(const FunctionType* subFunction, const Fun
     return result;
 }
 
+SubtypingResult Subtyping::isSubtype_(const PrimitiveType* subPrim, const TableType* superTable)
+{
+    SubtypingResult result{false};
+    if (subPrim->type == PrimitiveType::String)
+    {
+        if (auto metatable = getMetatable(builtinTypes->stringType, builtinTypes))
+        {
+            if (auto mttv = get<TableType>(follow(metatable)))
+            {
+                if (auto it = mttv->props.find("__index"); it != mttv->props.end())
+                {
+                    if (auto stringTable = get<TableType>(it->second.type()))
+                        result.orElse(isSubtype_(stringTable, superTable));
+                }
+            }
+        }
+    }
+
+    return result;
+}
+
+SubtypingResult Subtyping::isSubtype_(const SingletonType* subSingleton, const TableType* superTable)
+{
+    SubtypingResult result{false};
+    if (auto stringleton = get<StringSingleton>(subSingleton))
+    {
+        if (auto metatable = getMetatable(builtinTypes->stringType, builtinTypes))
+        {
+            if (auto mttv = get<TableType>(follow(metatable)))
+            {
+                if (auto it = mttv->props.find("__index"); it != mttv->props.end())
+                {
+                    if (auto stringTable = get<TableType>(it->second.type()))
+                        result.orElse(isSubtype_(stringTable, superTable));
+                }
+            }
+        }
+    }
+    return result;
+}
+
 SubtypingResult Subtyping::isSubtype_(const NormalizedType* subNorm, const NormalizedType* superNorm)
 {
     if (!subNorm || !superNorm)
@@ -540,15 +654,14 @@ SubtypingResult Subtyping::isSubtype_(const NormalizedType* subNorm, const Norma
 
     SubtypingResult result = isSubtype_(subNorm->tops, superNorm->tops);
     result.andAlso(isSubtype_(subNorm->booleans, superNorm->booleans));
-    // isSubtype_(subNorm->classes, superNorm->classes);
-    // isSubtype_(subNorm->classes, superNorm->tables);
+    result.andAlso(isSubtype_(subNorm->classes, superNorm->classes, superNorm->tables));
     result.andAlso(isSubtype_(subNorm->errors, superNorm->errors));
     result.andAlso(isSubtype_(subNorm->nils, superNorm->nils));
     result.andAlso(isSubtype_(subNorm->numbers, superNorm->numbers));
     result.isSubtype &= Luau::isSubtype(subNorm->strings, superNorm->strings);
     // isSubtype_(subNorm->strings, superNorm->tables);
     result.andAlso(isSubtype_(subNorm->threads, superNorm->threads));
-    // isSubtype_(subNorm->tables, superNorm->tables);
+    result.andAlso(isSubtype_(subNorm->tables, superNorm->tables));
     // isSubtype_(subNorm->tables, superNorm->strings);
     // isSubtype_(subNorm->tables, superNorm->classes);
     // isSubtype_(subNorm->functions, superNorm->functions);
@@ -557,6 +670,58 @@ SubtypingResult Subtyping::isSubtype_(const NormalizedType* subNorm, const Norma
     return result;
 }
 
+SubtypingResult Subtyping::isSubtype_(const NormalizedClassType& subClass, const NormalizedClassType& superClass, const TypeIds& superTables)
+{
+    for (const auto& [subClassTy, _] : subClass.classes)
+    {
+        SubtypingResult result;
+
+        for (const auto& [superClassTy, superNegations] : superClass.classes)
+        {
+            result.orElse(isSubtype_(subClassTy, superClassTy));
+            if (!result.isSubtype)
+                continue;
+
+            for (TypeId negation : superNegations)
+            {
+                result.andAlso(SubtypingResult::negate(isSubtype_(subClassTy, negation)));
+                if (result.isSubtype)
+                    break;
+            }
+        }
+
+        if (result.isSubtype)
+            continue;
+
+        for (TypeId superTableTy : superTables)
+            result.orElse(isSubtype_(subClassTy, superTableTy));
+
+        if (!result.isSubtype)
+            return result;
+    }
+
+    return {true};
+}
+
+SubtypingResult Subtyping::isSubtype_(const TypeIds& subTypes, const TypeIds& superTypes)
+{
+    std::vector<SubtypingResult> results;
+
+    for (TypeId subTy : subTypes)
+    {
+        results.emplace_back();
+        for (TypeId superTy : superTypes)
+            results.back().orElse(isSubtype_(subTy, superTy));
+    }
+
+    return SubtypingResult::all(results);
+}
+
+SubtypingResult Subtyping::isSubtype_(const VariadicTypePack* subVariadic, const VariadicTypePack* superVariadic)
+{
+    return isSubtype_(subVariadic->ty, superVariadic->ty);
+}
+
 bool Subtyping::bindGeneric(TypeId subTy, TypeId superTy)
 {
     if (variance == Variance::Covariant)

Analysis/src/Transpiler.cpp

@@ -10,6 +10,8 @@
 #include <limits>
 #include <math.h>
 
+LUAU_FASTFLAG(LuauFloorDivision)
+
 namespace
 {
 bool isIdentifierStartChar(char c)
@@ -467,10 +469,13 @@ struct Printer
             case AstExprBinary::Sub:
             case AstExprBinary::Mul:
             case AstExprBinary::Div:
+            case AstExprBinary::FloorDiv:
             case AstExprBinary::Mod:
             case AstExprBinary::Pow:
             case AstExprBinary::CompareLt:
             case AstExprBinary::CompareGt:
+                LUAU_ASSERT(FFlag::LuauFloorDivision || a->op != AstExprBinary::FloorDiv);
+
                 writer.maybeSpace(a->right->location.begin, 2);
                 writer.symbol(toString(a->op));
                 break;
@@ -487,6 +492,8 @@ struct Printer
                 writer.maybeSpace(a->right->location.begin, 4);
                 writer.keyword(toString(a->op));
                 break;
+            default:
+                LUAU_ASSERT(!"Unknown Op");
             }
 
             visualize(*a->right);
@@ -753,6 +760,12 @@ struct Printer
                 writer.maybeSpace(a->value->location.begin, 2);
                 writer.symbol("/=");
                 break;
+            case AstExprBinary::FloorDiv:
+                LUAU_ASSERT(FFlag::LuauFloorDivision);
+
+                writer.maybeSpace(a->value->location.begin, 2);
+                writer.symbol("//=");
+                break;
             case AstExprBinary::Mod:
                 writer.maybeSpace(a->value->location.begin, 2);
                 writer.symbol("%=");

Analysis/src/TypeChecker2.cpp

@@ -23,6 +23,7 @@
 #include <algorithm>
 
 LUAU_FASTFLAG(DebugLuauMagicTypes)
+LUAU_FASTFLAG(LuauFloorDivision);
 
 namespace Luau
 {
@@ -1817,6 +1818,8 @@ struct TypeChecker2
         bool typesHaveIntersection = normalizer.isIntersectionInhabited(leftType, rightType);
         if (auto it = kBinaryOpMetamethods.find(expr->op); it != kBinaryOpMetamethods.end())
         {
+            LUAU_ASSERT(FFlag::LuauFloorDivision || expr->op != AstExprBinary::Op::FloorDiv);
+
             std::optional<TypeId> leftMt = getMetatable(leftType, builtinTypes);
             std::optional<TypeId> rightMt = getMetatable(rightType, builtinTypes);
             bool matches = leftMt == rightMt;
@@ -2002,8 +2005,11 @@ struct TypeChecker2
         case AstExprBinary::Op::Sub:
         case AstExprBinary::Op::Mul:
         case AstExprBinary::Op::Div:
+        case AstExprBinary::Op::FloorDiv:
         case AstExprBinary::Op::Pow:
         case AstExprBinary::Op::Mod:
+            LUAU_ASSERT(FFlag::LuauFloorDivision || expr->op != AstExprBinary::Op::FloorDiv);
+
             reportErrors(tryUnify(scope, expr->left->location, leftType, builtinTypes->numberType));
             reportErrors(tryUnify(scope, expr->right->location, rightType, builtinTypes->numberType));
 

Analysis/src/TypeInfer.cpp

@@ -40,6 +40,7 @@ LUAU_FASTFLAG(LuauOccursIsntAlwaysFailure)
 LUAU_FASTFLAGVARIABLE(LuauTinyControlFlowAnalysis, false)
 LUAU_FASTFLAGVARIABLE(LuauAlwaysCommitInferencesOfFunctionCalls, false)
 LUAU_FASTFLAG(LuauParseDeclareClassIndexer)
+LUAU_FASTFLAG(LuauFloorDivision);
 
 namespace Luau
 {
@@ -200,7 +201,8 @@ static bool isMetamethod(const Name& name)
 {
     return name == "__index" || name == "__newindex" || name == "__call" || name == "__concat" || name == "__unm" || name == "__add" ||
            name == "__sub" || name == "__mul" || name == "__div" || name == "__mod" || name == "__pow" || name == "__tostring" ||
-           name == "__metatable" || name == "__eq" || name == "__lt" || name == "__le" || name == "__mode" || name == "__iter" || name == "__len";
+           name == "__metatable" || name == "__eq" || name == "__lt" || name == "__le" || name == "__mode" || name == "__iter" || name == "__len" ||
+           (FFlag::LuauFloorDivision && name == "__idiv");
 }
 
 size_t HashBoolNamePair::operator()(const std::pair<bool, Name>& pair) const
@@ -2571,6 +2573,9 @@ std::string opToMetaTableEntry(const AstExprBinary::Op& op)
         return "__mul";
     case AstExprBinary::Div:
         return "__div";
+    case AstExprBinary::FloorDiv:
+        LUAU_ASSERT(FFlag::LuauFloorDivision);
+        return "__idiv";
     case AstExprBinary::Mod:
         return "__mod";
     case AstExprBinary::Pow:
@@ -3064,8 +3069,11 @@ TypeId TypeChecker::checkBinaryOperation(
     case AstExprBinary::Sub:
     case AstExprBinary::Mul:
     case AstExprBinary::Div:
+    case AstExprBinary::FloorDiv:
     case AstExprBinary::Mod:
     case AstExprBinary::Pow:
+        LUAU_ASSERT(FFlag::LuauFloorDivision || expr.op != AstExprBinary::FloorDiv);
+
         reportErrors(tryUnify(lhsType, numberType, scope, expr.left->location));
         reportErrors(tryUnify(rhsType, numberType, scope, expr.right->location));
         return numberType;

Ast/include/Luau/Ast.h

@@ -457,6 +457,7 @@ public:
         Sub,
         Mul,
         Div,
+        FloorDiv,
         Mod,
         Pow,
         Concat,
@@ -467,7 +468,9 @@ public:
         CompareGt,
         CompareGe,
         And,
-        Or
+        Or,
+
+        Op__Count
     };
 
     AstExprBinary(const Location& location, Op op, AstExpr* left, AstExpr* right);
@@ -531,11 +534,12 @@ class AstStatBlock : public AstStat
 public:
     LUAU_RTTI(AstStatBlock)
 
-    AstStatBlock(const Location& location, const AstArray<AstStat*>& body);
+    AstStatBlock(const Location& location, const AstArray<AstStat*>& body, bool hasEnd=true);
 
     void visit(AstVisitor* visitor) override;
 
     AstArray<AstStat*> body;
+    bool hasEnd = false;
 };
 
 class AstStatIf : public AstStat

Ast/include/Luau/Lexer.h

@@ -62,6 +62,7 @@ struct Lexeme
         Dot3,
         SkinnyArrow,
         DoubleColon,
+        FloorDiv,
 
         InterpStringBegin,
         InterpStringMid,
@@ -73,6 +74,7 @@ struct Lexeme
         SubAssign,
         MulAssign,
         DivAssign,
+        FloorDivAssign,
         ModAssign,
         PowAssign,
         ConcatAssign,

Ast/src/Ast.cpp

@@ -3,6 +3,8 @@
 
 #include "Luau/Common.h"
 
+LUAU_FASTFLAG(LuauFloorDivision)
+
 namespace Luau
 {
 
@@ -279,6 +281,9 @@ std::string toString(AstExprBinary::Op op)
         return "*";
     case AstExprBinary::Div:
         return "/";
+    case AstExprBinary::FloorDiv:
+        LUAU_ASSERT(FFlag::LuauFloorDivision);
+        return "//";
     case AstExprBinary::Mod:
         return "%";
     case AstExprBinary::Pow:
@@ -375,9 +380,10 @@ void AstExprError::visit(AstVisitor* visitor)
     }
 }
 
-AstStatBlock::AstStatBlock(const Location& location, const AstArray<AstStat*>& body)
+AstStatBlock::AstStatBlock(const Location& location, const AstArray<AstStat*>& body, bool hasEnd)
     : AstStat(ClassIndex(), location)
     , body(body)
+    , hasEnd(hasEnd)
 {
 }
 

Ast/src/Lexer.cpp

@@ -6,7 +6,9 @@
 
 #include <limits.h>
 
+LUAU_FASTFLAGVARIABLE(LuauFloorDivision, false)
 LUAU_FASTFLAGVARIABLE(LuauLexerConsumeFast, false)
+LUAU_FASTFLAGVARIABLE(LuauLexerLookaheadRemembersBraceType, false)
 
 namespace Luau
 {
@@ -138,6 +140,9 @@ std::string Lexeme::toString() const
     case DoubleColon:
         return "'::'";
 
+    case FloorDiv:
+        return FFlag::LuauFloorDivision ? "'//'" : "<unknown>";
+
     case AddAssign:
         return "'+='";
 
@@ -150,6 +155,9 @@ std::string Lexeme::toString() const
     case DivAssign:
         return "'/='";
 
+    case FloorDivAssign:
+        return FFlag::LuauFloorDivision ? "'//='" : "<unknown>";
+
     case ModAssign:
         return "'%='";
 
@@ -402,6 +410,8 @@ Lexeme Lexer::lookahead()
     unsigned int currentLineOffset = lineOffset;
     Lexeme currentLexeme = lexeme;
     Location currentPrevLocation = prevLocation;
+    size_t currentBraceStackSize = braceStack.size();
+    BraceType currentBraceType = braceStack.empty() ? BraceType::Normal : braceStack.back();
 
     Lexeme result = next();
 
@@ -410,6 +420,13 @@ Lexeme Lexer::lookahead()
     lineOffset = currentLineOffset;
     lexeme = currentLexeme;
     prevLocation = currentPrevLocation;
+    if (FFlag::LuauLexerLookaheadRemembersBraceType)
+    {
+        if (braceStack.size() < currentBraceStackSize)
+            braceStack.push_back(currentBraceType);
+        else if (braceStack.size() > currentBraceStackSize)
+            braceStack.pop_back();
+    }
 
     return result;
 }
@@ -901,15 +918,46 @@ Lexeme Lexer::readNext()
             return Lexeme(Location(start, 1), '+');
 
     case '/':
-        consume();
-
-        if (peekch() == '=')
+    {
+        if (FFlag::LuauFloorDivision)
         {
             consume();
-            return Lexeme(Location(start, 2), Lexeme::DivAssign);
+
+            char ch = peekch();
+
+            if (ch == '=')
+            {
+                consume();
+                return Lexeme(Location(start, 2), Lexeme::DivAssign);
+            }
+            else if (ch == '/')
+            {
+                consume();
+
+                if (peekch() == '=')
+                {
+                    consume();
+                    return Lexeme(Location(start, 3), Lexeme::FloorDivAssign);
+                }
+                else
+                    return Lexeme(Location(start, 2), Lexeme::FloorDiv);
+            }
+            else
+                return Lexeme(Location(start, 1), '/');
         }
         else
-            return Lexeme(Location(start, 1), '/');
+        {
+            consume();
+
+            if (peekch() == '=')
+            {
+                consume();
+                return Lexeme(Location(start, 2), Lexeme::DivAssign);
+            }
+            else
+                return Lexeme(Location(start, 1), '/');
+        }
+    }
 
     case '*':
         consume();

Ast/src/Parser.cpp

@@ -14,6 +14,7 @@
 LUAU_FASTINTVARIABLE(LuauRecursionLimit, 1000)
 LUAU_FASTINTVARIABLE(LuauParseErrorLimit, 100)
 LUAU_FASTFLAGVARIABLE(LuauParseDeclareClassIndexer, false)
+LUAU_FASTFLAG(LuauFloorDivision)
 
 namespace Luau
 {
@@ -460,11 +461,11 @@ AstStat* Parser::parseDo()
     Lexeme matchDo = lexer.current();
     nextLexeme(); // do
 
-    AstStat* body = parseBlock();
+    AstStatBlock* body = parseBlock();
 
     body->location.begin = start.begin;
 
-    expectMatchEndAndConsume(Lexeme::ReservedEnd, matchDo);
+    body->hasEnd = expectMatchEndAndConsume(Lexeme::ReservedEnd, matchDo);
 
     return body;
 }
@@ -1766,6 +1767,12 @@ std::optional<AstExprBinary::Op> Parser::parseBinaryOp(const Lexeme& l)
         return AstExprBinary::Mul;
     else if (l.type == '/')
         return AstExprBinary::Div;
+    else if (l.type == Lexeme::FloorDiv)
+    {
+        LUAU_ASSERT(FFlag::LuauFloorDivision);
+
+        return AstExprBinary::FloorDiv;
+    }
     else if (l.type == '%')
         return AstExprBinary::Mod;
     else if (l.type == '^')
@@ -1802,6 +1809,12 @@ std::optional<AstExprBinary::Op> Parser::parseCompoundOp(const Lexeme& l)
         return AstExprBinary::Mul;
     else if (l.type == Lexeme::DivAssign)
         return AstExprBinary::Div;
+    else if (l.type == Lexeme::FloorDivAssign)
+    {
+        LUAU_ASSERT(FFlag::LuauFloorDivision);
+
+        return AstExprBinary::FloorDiv;
+    }
     else if (l.type == Lexeme::ModAssign)
         return AstExprBinary::Mod;
     else if (l.type == Lexeme::PowAssign)
@@ -1872,12 +1885,13 @@ std::optional<AstExprBinary::Op> Parser::checkBinaryConfusables(const BinaryOpPr
 AstExpr* Parser::parseExpr(unsigned int limit)
 {
     static const BinaryOpPriority binaryPriority[] = {
-        {6, 6}, {6, 6}, {7, 7}, {7, 7}, {7, 7}, // `+' `-' `*' `/' `%'
-        {10, 9}, {5, 4},                        // power and concat (right associative)
-        {3, 3}, {3, 3},                         // equality and inequality
-        {3, 3}, {3, 3}, {3, 3}, {3, 3},         // order
-        {2, 2}, {1, 1}                          // logical (and/or)
+        {6, 6}, {6, 6}, {7, 7}, {7, 7}, {7, 7}, {7, 7}, // `+' `-' `*' `/' `//' `%'
+        {10, 9}, {5, 4},                                // power and concat (right associative)
+        {3, 3}, {3, 3},                                 // equality and inequality
+        {3, 3}, {3, 3}, {3, 3}, {3, 3},                 // order
+        {2, 2}, {1, 1}                                  // logical (and/or)
     };
+    static_assert(sizeof(binaryPriority) / sizeof(binaryPriority[0]) == size_t(AstExprBinary::Op__Count), "binaryPriority needs an entry per op");
 
     unsigned int recursionCounterOld = recursionCounter;
 

CodeGen/include/Luau/AssemblyBuilderA64.h

@@ -221,6 +221,7 @@ private:
     void placeFMOV(const char* name, RegisterA64 dst, double src, uint32_t op);
     void placeBM(const char* name, RegisterA64 dst, RegisterA64 src1, uint32_t src2, uint8_t op);
     void placeBFM(const char* name, RegisterA64 dst, RegisterA64 src1, int src2, uint8_t op, int immr, int imms);
+    void placeER(const char* name, RegisterA64 dst, RegisterA64 src1, RegisterA64 src2, uint8_t op, int shift);
 
     void place(uint32_t word);
 

CodeGen/include/Luau/IrData.h

@@ -67,6 +67,7 @@ enum class IrCmd : uint8_t
     // Get pointer (LuaNode) to table node element at the active cached slot index
     // A: pointer (Table)
     // B: unsigned int (pcpos)
+    // C: Kn
     GET_SLOT_NODE_ADDR,
 
     // Get pointer (LuaNode) to table node element at the main position of the specified key hash
@@ -132,6 +133,7 @@ enum class IrCmd : uint8_t
     SUB_NUM,
     MUL_NUM,
     DIV_NUM,
+    IDIV_NUM,
     MOD_NUM,
 
     // Get the minimum/maximum of two numbers
@@ -253,6 +255,11 @@ enum class IrCmd : uint8_t
     // A: pointer (Table)
     DUP_TABLE,
 
+    // Insert an integer key into a table
+    // A: pointer (Table)
+    // B: int (key)
+    TABLE_SETNUM,
+
     // Try to convert a double number into a table index (int) or jump if it's not an integer
     // A: double
     // B: block
@@ -411,6 +418,12 @@ enum class IrCmd : uint8_t
     // When undef is specified instead of a block, execution is aborted on check failure
     CHECK_NODE_NO_NEXT,
 
+    // Guard against table node with 'nil' value
+    // A: pointer (LuaNode)
+    // B: block/vmexit/undef
+    // When undef is specified instead of a block, execution is aborted on check failure
+    CHECK_NODE_VALUE,
+
     // Special operations
 
     // Check interrupt handler
@@ -832,6 +845,8 @@ struct IrBlock
     uint32_t finish = ~0u;
 
     uint32_t sortkey = ~0u;
+    uint32_t chainkey = 0;
+    uint32_t expectedNextBlock = ~0u;
 
     Label label;
 };
@@ -993,23 +1008,26 @@ struct IrFunction
         valueRestoreOps[instIdx] = location;
     }
 
-    IrOp findRestoreOp(uint32_t instIdx) const
+    IrOp findRestoreOp(uint32_t instIdx, bool limitToCurrentBlock) const
     {
         if (instIdx >= valueRestoreOps.size())
             return {};
 
         const IrBlock& block = blocks[validRestoreOpBlockIdx];
 
-        // Values can only reference restore operands in the current block
-        if (instIdx < block.start || instIdx > block.finish)
-            return {};
+        // When spilled, values can only reference restore operands in the current block
+        if (limitToCurrentBlock)
+        {
+            if (instIdx < block.start || instIdx > block.finish)
+                return {};
+        }
 
         return valueRestoreOps[instIdx];
     }
 
-    IrOp findRestoreOp(const IrInst& inst) const
+    IrOp findRestoreOp(const IrInst& inst, bool limitToCurrentBlock) const
     {
-        return findRestoreOp(getInstIndex(inst));
+        return findRestoreOp(getInstIndex(inst), limitToCurrentBlock);
     }
 };
 

CodeGen/include/Luau/IrUtils.h

@@ -128,6 +128,7 @@ inline bool isNonTerminatingJump(IrCmd cmd)
     case IrCmd::CHECK_ARRAY_SIZE:
     case IrCmd::CHECK_SLOT_MATCH:
     case IrCmd::CHECK_NODE_NO_NEXT:
+    case IrCmd::CHECK_NODE_VALUE:
         return true;
     default:
         break;
@@ -156,6 +157,7 @@ inline bool hasResult(IrCmd cmd)
     case IrCmd::SUB_NUM:
     case IrCmd::MUL_NUM:
     case IrCmd::DIV_NUM:
+    case IrCmd::IDIV_NUM:
     case IrCmd::MOD_NUM:
     case IrCmd::MIN_NUM:
     case IrCmd::MAX_NUM:
@@ -168,6 +170,7 @@ inline bool hasResult(IrCmd cmd)
     case IrCmd::NOT_ANY:
     case IrCmd::CMP_ANY:
     case IrCmd::TABLE_LEN:
+    case IrCmd::TABLE_SETNUM:
     case IrCmd::STRING_LEN:
     case IrCmd::NEW_TABLE:
     case IrCmd::DUP_TABLE:

CodeGen/include/Luau/RegisterA64.h

@@ -47,18 +47,6 @@ constexpr RegisterA64 castReg(KindA64 kind, RegisterA64 reg)
     return RegisterA64{kind, reg.index};
 }
 
-// This is equivalent to castReg(KindA64::x), but is separate because it implies different semantics
-// Specifically, there are cases when it's useful to treat a wN register as an xN register *after* it has been assigned a value
-// Since all A64 instructions that write to wN implicitly zero the top half, this works when we need zero extension semantics
-// Crucially, this is *not* safe on an ABI boundary - an int parameter in wN register may have anything in its top half in certain cases
-// However, as long as our codegen doesn't use 32-bit truncation by using castReg x=>w, we can safely rely on this.
-constexpr RegisterA64 zextReg(RegisterA64 reg)
-{
-    LUAU_ASSERT(reg.kind == KindA64::w);
-
-    return RegisterA64{KindA64::x, reg.index};
-}
-
 constexpr RegisterA64 noreg{KindA64::none, 0};
 
 constexpr RegisterA64 w0{KindA64::w, 0};

CodeGen/src/AssemblyBuilderA64.cpp

@@ -105,7 +105,10 @@ void AssemblyBuilderA64::movk(RegisterA64 dst, uint16_t src, int shift)
 
 void AssemblyBuilderA64::add(RegisterA64 dst, RegisterA64 src1, RegisterA64 src2, int shift)
 {
-    placeSR3("add", dst, src1, src2, 0b00'01011, shift);
+    if (src1.kind == KindA64::x && src2.kind == KindA64::w)
+        placeER("add", dst, src1, src2, 0b00'01011, shift);
+    else
+        placeSR3("add", dst, src1, src2, 0b00'01011, shift);
 }
 
 void AssemblyBuilderA64::add(RegisterA64 dst, RegisterA64 src1, uint16_t src2)
@@ -115,7 +118,10 @@ void AssemblyBuilderA64::add(RegisterA64 dst, RegisterA64 src1, uint16_t src2)
 
 void AssemblyBuilderA64::sub(RegisterA64 dst, RegisterA64 src1, RegisterA64 src2, int shift)
 {
-    placeSR3("sub", dst, src1, src2, 0b10'01011, shift);
+    if (src1.kind == KindA64::x && src2.kind == KindA64::w)
+        placeER("sub", dst, src1, src2, 0b10'01011, shift);
+    else
+        placeSR3("sub", dst, src1, src2, 0b10'01011, shift);
 }
 
 void AssemblyBuilderA64::sub(RegisterA64 dst, RegisterA64 src1, uint16_t src2)
@@ -1075,6 +1081,22 @@ void AssemblyBuilderA64::placeBFM(const char* name, RegisterA64 dst, RegisterA64
     commit();
 }
 
+void AssemblyBuilderA64::placeER(const char* name, RegisterA64 dst, RegisterA64 src1, RegisterA64 src2, uint8_t op, int shift)
+{
+    if (logText)
+        log(name, dst, src1, src2, shift);
+
+    LUAU_ASSERT(dst.kind == KindA64::x && src1.kind == KindA64::x);
+    LUAU_ASSERT(src2.kind == KindA64::w);
+    LUAU_ASSERT(shift >= 0 && shift <= 4);
+
+    uint32_t sf = (dst.kind == KindA64::x) ? 0x80000000 : 0; // could be useful in the future for byte->word extends
+    int option = 0b010; // UXTW
+
+    place(dst.index | (src1.index << 5) | (shift << 10) | (option << 13) | (src2.index << 16) | (1 << 21) | (op << 24) | sf);
+    commit();
+}
+
 void AssemblyBuilderA64::place(uint32_t word)
 {
     LUAU_ASSERT(codePos < codeEnd);
@@ -1167,7 +1189,9 @@ void AssemblyBuilderA64::log(const char* opcode, RegisterA64 dst, RegisterA64 sr
     log(src1);
     text.append(",");
     log(src2);
-    if (shift > 0)
+    if (src1.kind == KindA64::x && src2.kind == KindA64::w)
+        logAppend(" UXTW #%d", shift);
+    else if (shift > 0)
         logAppend(" LSL #%d", shift);
     else if (shift < 0)
         logAppend(" LSR #%d", -shift);

CodeGen/src/CodeAllocator.cpp

@@ -71,10 +71,12 @@ static uint8_t* allocatePagesImpl(size_t size)
     LUAU_ASSERT(size == alignToPageSize(size));
 
 #ifdef __APPLE__
-    return (uint8_t*)mmap(nullptr, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON | MAP_JIT, -1, 0);
+    void* result = mmap(nullptr, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON | MAP_JIT, -1, 0);
 #else
-    return (uint8_t*)mmap(nullptr, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON, -1, 0);
+    void* result = mmap(nullptr, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON, -1, 0);
 #endif
+
+    return (result == MAP_FAILED) ? nullptr : static_cast<uint8_t*>(result);
 }
 
 static void freePagesImpl(uint8_t* mem, size_t size)

CodeGen/src/CodeGenLower.h

@@ -74,7 +74,11 @@ inline bool lowerImpl(AssemblyBuilder& build, IrLowering& lowering, IrFunction&
             return (a.kind == IrBlockKind::Fallback) < (b.kind == IrBlockKind::Fallback);
 
         // Try to order by instruction order
-        return a.sortkey < b.sortkey;
+        if (a.sortkey != b.sortkey)
+            return a.sortkey < b.sortkey;
+
+        // Chains of blocks are merged together by having the same sort key and consecutive chain key
+        return a.chainkey < b.chainkey;
     });
 
     // For each IR instruction that begins a bytecode instruction, which bytecode instruction is it?
@@ -100,6 +104,9 @@ inline bool lowerImpl(AssemblyBuilder& build, IrLowering& lowering, IrFunction&
     IrBlock dummy;
     dummy.start = ~0u;
 
+    // Make sure entry block is first
+    LUAU_ASSERT(sortedBlocks[0] == 0);
+
     for (size_t i = 0; i < sortedBlocks.size(); ++i)
     {
         uint32_t blockIndex = sortedBlocks[i];
@@ -137,6 +144,11 @@ inline bool lowerImpl(AssemblyBuilder& build, IrLowering& lowering, IrFunction&
 
         IrBlock& nextBlock = getNextBlock(function, sortedBlocks, dummy, i);
 
+        // Optimizations often propagate information between blocks
+        // To make sure the register and spill state is correct when blocks are lowered, we check that sorted block order matches the expected one
+        if (block.expectedNextBlock != ~0u)
+            LUAU_ASSERT(function.getBlockIndex(nextBlock) == block.expectedNextBlock);
+
         for (uint32_t index = block.start; index <= block.finish; index++)
         {
             LUAU_ASSERT(index < function.instructions.size());

CodeGen/src/EmitCommonX64.cpp

@@ -147,13 +147,14 @@ void callSetTable(IrRegAllocX64& regs, AssemblyBuilderX64& build, int rb, Operan
     emitUpdateBase(build);
 }
 
-void checkObjectBarrierConditions(AssemblyBuilderX64& build, RegisterX64 tmp, RegisterX64 object, int ra, int ratag, Label& skip)
+void checkObjectBarrierConditions(AssemblyBuilderX64& build, RegisterX64 tmp, RegisterX64 object, IrOp ra, int ratag, Label& skip)
 {
     // Barrier should've been optimized away if we know that it's not collectable, checking for correctness
     if (ratag == -1 || !isGCO(ratag))
     {
         // iscollectable(ra)
-        build.cmp(luauRegTag(ra), LUA_TSTRING);
+        OperandX64 tag = (ra.kind == IrOpKind::VmReg) ? luauRegTag(vmRegOp(ra)) : luauConstantTag(vmConstOp(ra));
+        build.cmp(tag, LUA_TSTRING);
         build.jcc(ConditionX64::Less, skip);
     }
 
@@ -162,12 +163,14 @@ void checkObjectBarrierConditions(AssemblyBuilderX64& build, RegisterX64 tmp, Re
     build.jcc(ConditionX64::Zero, skip);
 
     // iswhite(gcvalue(ra))
-    build.mov(tmp, luauRegValue(ra));
+    OperandX64 value = (ra.kind == IrOpKind::VmReg) ? luauRegValue(vmRegOp(ra)) : luauConstantValue(vmConstOp(ra));
+    build.mov(tmp, value);
     build.test(byte[tmp + offsetof(GCheader, marked)], bit2mask(WHITE0BIT, WHITE1BIT));
     build.jcc(ConditionX64::Zero, skip);
 }
 
-void callBarrierObject(IrRegAllocX64& regs, AssemblyBuilderX64& build, RegisterX64 object, IrOp objectOp, int ra, int ratag)
+
+void callBarrierObject(IrRegAllocX64& regs, AssemblyBuilderX64& build, RegisterX64 object, IrOp objectOp, IrOp ra, int ratag)
 {
     Label skip;
 

CodeGen/src/EmitCommonX64.h

@@ -202,8 +202,8 @@ void callArithHelper(IrRegAllocX64& regs, AssemblyBuilderX64& build, int ra, int
 void callLengthHelper(IrRegAllocX64& regs, AssemblyBuilderX64& build, int ra, int rb);
 void callGetTable(IrRegAllocX64& regs, AssemblyBuilderX64& build, int rb, OperandX64 c, int ra);
 void callSetTable(IrRegAllocX64& regs, AssemblyBuilderX64& build, int rb, OperandX64 c, int ra);
-void checkObjectBarrierConditions(AssemblyBuilderX64& build, RegisterX64 tmp, RegisterX64 object, int ra, int ratag, Label& skip);
-void callBarrierObject(IrRegAllocX64& regs, AssemblyBuilderX64& build, RegisterX64 object, IrOp objectOp, int ra, int ratag);
+void checkObjectBarrierConditions(AssemblyBuilderX64& build, RegisterX64 tmp, RegisterX64 object, IrOp ra, int ratag, Label& skip);
+void callBarrierObject(IrRegAllocX64& regs, AssemblyBuilderX64& build, RegisterX64 object, IrOp objectOp, IrOp ra, int ratag);
 void callBarrierTableFast(IrRegAllocX64& regs, AssemblyBuilderX64& build, RegisterX64 table, IrOp tableOp);
 void callStepGc(IrRegAllocX64& regs, AssemblyBuilderX64& build);
 

CodeGen/src/IrAnalysis.cpp

@@ -257,7 +257,7 @@ static void visitVmRegDefsUses(T& visitor, IrFunction& function, const IrBlock&
             break;
         case IrCmd::BARRIER_OBJ:
         case IrCmd::BARRIER_TABLE_FORWARD:
-            visitor.use(inst.b);
+            visitor.maybeUse(inst.b);
             break;
         case IrCmd::CLOSE_UPVALS:
             // Closing an upvalue should be counted as a register use (it copies the fresh register value)

CodeGen/src/IrBuilder.cpp

@@ -333,6 +333,9 @@ void IrBuilder::translateInst(LuauOpcode op, const Instruction* pc, int i)
     case LOP_DIV:
         translateInstBinary(*this, pc, i, TM_DIV);
         break;
+    case LOP_IDIV:
+        translateInstBinary(*this, pc, i, TM_IDIV);
+        break;
     case LOP_MOD:
         translateInstBinary(*this, pc, i, TM_MOD);
         break;
@@ -351,6 +354,9 @@ void IrBuilder::translateInst(LuauOpcode op, const Instruction* pc, int i)
     case LOP_DIVK:
         translateInstBinaryK(*this, pc, i, TM_DIV);
         break;
+    case LOP_IDIVK:
+        translateInstBinaryK(*this, pc, i, TM_IDIV);
+        break;
     case LOP_MODK:
         translateInstBinaryK(*this, pc, i, TM_MOD);
         break;

CodeGen/src/IrDump.cpp

@@ -125,6 +125,8 @@ const char* getCmdName(IrCmd cmd)
         return "MUL_NUM";
     case IrCmd::DIV_NUM:
         return "DIV_NUM";
+    case IrCmd::IDIV_NUM:
+        return "IDIV_NUM";
     case IrCmd::MOD_NUM:
         return "MOD_NUM";
     case IrCmd::MIN_NUM:
@@ -169,6 +171,8 @@ const char* getCmdName(IrCmd cmd)
         return "JUMP_SLOT_MATCH";
     case IrCmd::TABLE_LEN:
         return "TABLE_LEN";
+    case IrCmd::TABLE_SETNUM:
+        return "TABLE_SETNUM";
     case IrCmd::STRING_LEN:
         return "STRING_LEN";
     case IrCmd::NEW_TABLE:
@@ -229,6 +233,8 @@ const char* getCmdName(IrCmd cmd)
         return "CHECK_SLOT_MATCH";
     case IrCmd::CHECK_NODE_NO_NEXT:
         return "CHECK_NODE_NO_NEXT";
+    case IrCmd::CHECK_NODE_VALUE:
+        return "CHECK_NODE_VALUE";
     case IrCmd::INTERRUPT:
         return "INTERRUPT";
     case IrCmd::CHECK_GC:

CodeGen/src/IrLoweringA64.cpp

@@ -58,30 +58,6 @@ inline ConditionA64 getConditionFP(IrCondition cond)
     }
 }
 
-static void checkObjectBarrierConditions(AssemblyBuilderA64& build, RegisterA64 object, RegisterA64 temp, int ra, int ratag, Label& skip)
-{
-    RegisterA64 tempw = castReg(KindA64::w, temp);
-
-    // Barrier should've been optimized away if we know that it's not collectable, checking for correctness
-    if (ratag == -1 || !isGCO(ratag))
-    {
-        // iscollectable(ra)
-        build.ldr(tempw, mem(rBase, ra * sizeof(TValue) + offsetof(TValue, tt)));
-        build.cmp(tempw, LUA_TSTRING);
-        build.b(ConditionA64::Less, skip);
-    }
-
-    // isblack(obj2gco(o))
-    build.ldrb(tempw, mem(object, offsetof(GCheader, marked)));
-    build.tbz(tempw, BLACKBIT, skip);
-
-    // iswhite(gcvalue(ra))
-    build.ldr(temp, mem(rBase, ra * sizeof(TValue) + offsetof(TValue, value)));
-    build.ldrb(tempw, mem(temp, offsetof(GCheader, marked)));
-    build.tst(tempw, bit2mask(WHITE0BIT, WHITE1BIT));
-    build.b(ConditionA64::Equal, skip); // Equal = Zero after tst
-}
-
 static void emitAddOffset(AssemblyBuilderA64& build, RegisterA64 dst, RegisterA64 src, size_t offset)
 {
     LUAU_ASSERT(dst != src);
@@ -98,6 +74,47 @@ static void emitAddOffset(AssemblyBuilderA64& build, RegisterA64 dst, RegisterA6
     }
 }
 
+static void checkObjectBarrierConditions(AssemblyBuilderA64& build, RegisterA64 object, RegisterA64 temp, IrOp ra, int ratag, Label& skip)
+{
+    RegisterA64 tempw = castReg(KindA64::w, temp);
+    AddressA64 addr = temp;
+
+    // iscollectable(ra)
+    if (ratag == -1 || !isGCO(ratag))
+    {
+        if (ra.kind == IrOpKind::VmReg)
+        {
+            addr = mem(rBase, vmRegOp(ra) * sizeof(TValue) + offsetof(TValue, tt));
+        }
+        else if (ra.kind == IrOpKind::VmConst)
+        {
+            emitAddOffset(build, temp, rConstants, vmConstOp(ra) * sizeof(TValue) + offsetof(TValue, tt));
+        }
+
+        build.ldr(tempw, addr);
+        build.cmp(tempw, LUA_TSTRING);
+        build.b(ConditionA64::Less, skip);
+    }
+
+    // isblack(obj2gco(o))
+    build.ldrb(tempw, mem(object, offsetof(GCheader, marked)));
+    build.tbz(tempw, BLACKBIT, skip);
+
+    // iswhite(gcvalue(ra))
+    if (ra.kind == IrOpKind::VmReg)
+    {
+        addr = mem(rBase, vmRegOp(ra) * sizeof(TValue) + offsetof(TValue, value));
+    }
+    else if (ra.kind == IrOpKind::VmConst)
+    {
+        emitAddOffset(build, temp, rConstants, vmConstOp(ra) * sizeof(TValue) + offsetof(TValue, value));
+    }
+    build.ldr(temp, addr);
+    build.ldrb(tempw, mem(temp, offsetof(GCheader, marked)));
+    build.tst(tempw, bit2mask(WHITE0BIT, WHITE1BIT));
+    build.b(ConditionA64::Equal, skip); // Equal = Zero after tst
+}
+
 static void emitAbort(AssemblyBuilderA64& build, Label& abort)
 {
     Label skip;
@@ -242,7 +259,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
 
         if (inst.b.kind == IrOpKind::Inst)
         {
-            build.add(inst.regA64, inst.regA64, zextReg(regOp(inst.b)), kTValueSizeLog2);
+            build.add(inst.regA64, inst.regA64, regOp(inst.b), kTValueSizeLog2);
         }
         else if (inst.b.kind == IrOpKind::Constant)
         {
@@ -271,6 +288,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         RegisterA64 temp1 = regs.allocTemp(KindA64::x);
         RegisterA64 temp1w = castReg(KindA64::w, temp1);
         RegisterA64 temp2 = regs.allocTemp(KindA64::w);
+        RegisterA64 temp2x = castReg(KindA64::x, temp2);
 
         // note: since the stride of the load is the same as the destination register size, we can range check the array index, not the byte offset
         if (uintOp(inst.b) <= AddressA64::kMaxOffset)
@@ -288,7 +306,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
 
         // note: this may clobber inst.a, so it's important that we don't use it after this
         build.ldr(inst.regA64, mem(regOp(inst.a), offsetof(Table, node)));
-        build.add(inst.regA64, inst.regA64, zextReg(temp2), kLuaNodeSizeLog2);
+        build.add(inst.regA64, inst.regA64, temp2x, kLuaNodeSizeLog2); // "zero extend" temp2 to get a larger shift (top 32 bits are zero)
         break;
     }
     case IrCmd::GET_HASH_NODE_ADDR:
@@ -296,6 +314,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         inst.regA64 = regs.allocReuse(KindA64::x, index, {inst.a});
         RegisterA64 temp1 = regs.allocTemp(KindA64::w);
         RegisterA64 temp2 = regs.allocTemp(KindA64::w);
+        RegisterA64 temp2x = castReg(KindA64::x, temp2);
 
         // hash & ((1 << lsizenode) - 1) == hash & ~(-1 << lsizenode)
         build.mov(temp1, -1);
@@ -306,7 +325,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
 
         // note: this may clobber inst.a, so it's important that we don't use it after this
         build.ldr(inst.regA64, mem(regOp(inst.a), offsetof(Table, node)));
-        build.add(inst.regA64, inst.regA64, zextReg(temp2), kLuaNodeSizeLog2);
+        build.add(inst.regA64, inst.regA64, temp2x, kLuaNodeSizeLog2); // "zero extend" temp2 to get a larger shift (top 32 bits are zero)
         break;
     }
     case IrCmd::GET_CLOSURE_UPVAL_ADDR:
@@ -477,6 +496,15 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         build.fdiv(inst.regA64, temp1, temp2);
         break;
     }
+    case IrCmd::IDIV_NUM:
+    {
+        inst.regA64 = regs.allocReuse(KindA64::d, index, {inst.a, inst.b});
+        RegisterA64 temp1 = tempDouble(inst.a);
+        RegisterA64 temp2 = tempDouble(inst.b);
+        build.fdiv(inst.regA64, temp1, temp2);
+        build.frintm(inst.regA64, inst.regA64);
+        break;
+    }
     case IrCmd::MOD_NUM:
     {
         inst.regA64 = regs.allocReg(KindA64::d, index); // can't allocReuse because both A and B are used twice
@@ -604,9 +632,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
 
         emitUpdateBase(build);
 
-        // since w0 came from a call, we need to move it so that we don't violate zextReg safety contract
-        inst.regA64 = regs.allocReg(KindA64::w, index);
-        build.mov(inst.regA64, w0);
+        inst.regA64 = regs.takeReg(w0, index);
         break;
     }
     case IrCmd::JUMP:
@@ -750,8 +776,8 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         build.mov(x0, reg);
         build.ldr(x1, mem(rNativeContext, offsetof(NativeContext, luaH_getn)));
         build.blr(x1);
-        inst.regA64 = regs.allocReg(KindA64::d, index);
-        build.scvtf(inst.regA64, w0);
+
+        inst.regA64 = regs.takeReg(w0, index);
         break;
     }
     case IrCmd::STRING_LEN:
@@ -761,6 +787,33 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         build.ldr(inst.regA64, mem(regOp(inst.a), offsetof(TString, len)));
         break;
     }
+    case IrCmd::TABLE_SETNUM:
+    {
+        // note: we need to call regOp before spill so that we don't do redundant reloads
+        RegisterA64 table = regOp(inst.a);
+        RegisterA64 key = regOp(inst.b);
+        RegisterA64 temp = regs.allocTemp(KindA64::w);
+
+        regs.spill(build, index, {table, key});
+
+        if (w1 != key)
+        {
+            build.mov(x1, table);
+            build.mov(w2, key);
+        }
+        else
+        {
+            build.mov(temp, w1);
+            build.mov(x1, table);
+            build.mov(w2, temp);
+        }
+
+        build.mov(x0, rState);
+        build.ldr(x3, mem(rNativeContext, offsetof(NativeContext, luaH_setnum)));
+        build.blr(x3);
+        inst.regA64 = regs.takeReg(x0, index);
+        break;   
+    }
     case IrCmd::NEW_TABLE:
     {
         regs.spill(build, index);
@@ -854,8 +907,6 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         inst.regA64 = regs.allocReg(KindA64::w, index);
         RegisterA64 temp = tempDouble(inst.a);
         build.fcvtzs(castReg(KindA64::x, inst.regA64), temp);
-        // truncation needs to clear high bits to preserve zextReg safety contract
-        build.mov(inst.regA64, inst.regA64);
         break;
     }
     case IrCmd::ADJUST_STACK_TO_REG:
@@ -870,7 +921,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         else if (inst.b.kind == IrOpKind::Inst)
         {
             build.add(temp, rBase, uint16_t(vmRegOp(inst.a) * sizeof(TValue)));
-            build.add(temp, temp, zextReg(regOp(inst.b)), kTValueSizeLog2);
+            build.add(temp, temp, regOp(inst.b), kTValueSizeLog2);
             build.str(temp, mem(rState, offsetof(lua_State, top)));
         }
         else
@@ -919,9 +970,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         build.ldr(x6, mem(rNativeContext, offsetof(NativeContext, luauF_table) + uintOp(inst.a) * sizeof(luau_FastFunction)));
         build.blr(x6);
 
-        // since w0 came from a call, we need to move it so that we don't violate zextReg safety contract
-        inst.regA64 = regs.allocReg(KindA64::w, index);
-        build.mov(inst.regA64, w0);
+        inst.regA64 = regs.takeReg(w0, index);
         break;
     }
     case IrCmd::CHECK_FASTCALL_RES:
@@ -1063,7 +1112,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         if (inst.c.kind == IrOpKind::Undef || isGCO(tagOp(inst.c)))
         {
             Label skip;
-            checkObjectBarrierConditions(build, temp1, temp2, vmRegOp(inst.b), inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c), skip);
+            checkObjectBarrierConditions(build, temp1, temp2, inst.b, inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c), skip);
 
             size_t spills = regs.spill(build, index, {temp1});
 
@@ -1244,6 +1293,17 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         finalizeTargetLabel(inst.b, fresh);
         break;
     }
+    case IrCmd::CHECK_NODE_VALUE:
+    {
+        Label fresh; // used when guard aborts execution or jumps to a VM exit
+        RegisterA64 temp = regs.allocTemp(KindA64::w);
+
+        build.ldr(temp, mem(regOp(inst.a), offsetof(LuaNode, val.tt)));
+        LUAU_ASSERT(LUA_TNIL == 0);
+        build.cbz(temp, getTargetLabel(inst.b, fresh));
+        finalizeTargetLabel(inst.b, fresh);
+        break;
+    }
     case IrCmd::INTERRUPT:
     {
         regs.spill(build, index);
@@ -1288,7 +1348,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         RegisterA64 temp = regs.allocTemp(KindA64::x);
 
         Label skip;
-        checkObjectBarrierConditions(build, regOp(inst.a), temp, vmRegOp(inst.b), inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c), skip);
+        checkObjectBarrierConditions(build, regOp(inst.a), temp, inst.b, inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c), skip);
 
         RegisterA64 reg = regOp(inst.a); // note: we need to call regOp before spill so that we don't do redundant reloads
         size_t spills = regs.spill(build, index, {reg});
@@ -1332,13 +1392,14 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         RegisterA64 temp = regs.allocTemp(KindA64::x);
 
         Label skip;
-        checkObjectBarrierConditions(build, regOp(inst.a), temp, vmRegOp(inst.b), inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c), skip);
+        checkObjectBarrierConditions(build, regOp(inst.a), temp, inst.b, inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c), skip);
 
         RegisterA64 reg = regOp(inst.a); // note: we need to call regOp before spill so that we don't do redundant reloads
+        AddressA64 addr = tempAddr(inst.b, offsetof(TValue, value));
         size_t spills = regs.spill(build, index, {reg});
         build.mov(x1, reg);
         build.mov(x0, rState);
-        build.ldr(x2, mem(rBase, vmRegOp(inst.b) * sizeof(TValue) + offsetof(TValue, value)));
+        build.ldr(x2, addr);
         build.ldr(x3, mem(rNativeContext, offsetof(NativeContext, luaC_barriertable)));
         build.blr(x3);
 
@@ -1829,7 +1890,7 @@ void IrLoweringA64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         LUAU_ASSERT(sizeof(TString*) == 8);
 
         if (inst.a.kind == IrOpKind::Inst)
-            build.add(inst.regA64, rGlobalState, zextReg(regOp(inst.a)), 3);
+            build.add(inst.regA64, rGlobalState, regOp(inst.a), 3);
         else if (inst.a.kind == IrOpKind::Constant)
             build.add(inst.regA64, rGlobalState, uint16_t(tagOp(inst.a)) * 8);
         else

CodeGen/src/IrLoweringX64.cpp

@@ -407,6 +407,22 @@ void IrLoweringX64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
             build.vdivsd(inst.regX64, regOp(inst.a), memRegDoubleOp(inst.b));
         }
         break;
+    case IrCmd::IDIV_NUM:
+        inst.regX64 = regs.allocRegOrReuse(SizeX64::xmmword, index, {inst.a, inst.b});
+
+        if (inst.a.kind == IrOpKind::Constant)
+        {
+            ScopedRegX64 tmp{regs, SizeX64::xmmword};
+
+            build.vmovsd(tmp.reg, memRegDoubleOp(inst.a));
+            build.vdivsd(inst.regX64, tmp.reg, memRegDoubleOp(inst.b));
+        }
+        else
+        {
+            build.vdivsd(inst.regX64, regOp(inst.a), memRegDoubleOp(inst.b));
+        }
+        build.vroundsd(inst.regX64, inst.regX64, inst.regX64, RoundingModeX64::RoundToNegativeInfinity);
+        break;
     case IrCmd::MOD_NUM:
     {
         inst.regX64 = regs.allocRegOrReuse(SizeX64::xmmword, index, {inst.a, inst.b});
@@ -697,9 +713,17 @@ void IrLoweringX64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         IrCallWrapperX64 callWrap(regs, build, index);
         callWrap.addArgument(SizeX64::qword, regOp(inst.a), inst.a);
         callWrap.call(qword[rNativeContext + offsetof(NativeContext, luaH_getn)]);
-
-        inst.regX64 = regs.allocReg(SizeX64::xmmword, index);
-        build.vcvtsi2sd(inst.regX64, inst.regX64, eax);
+        inst.regX64 = regs.takeReg(eax, index);
+        break;
+    }
+    case IrCmd::TABLE_SETNUM:
+    {
+        IrCallWrapperX64 callWrap(regs, build, index);
+        callWrap.addArgument(SizeX64::qword, rState);
+        callWrap.addArgument(SizeX64::qword, regOp(inst.a), inst.a);
+        callWrap.addArgument(SizeX64::dword, regOp(inst.b), inst.b);
+        callWrap.call(qword[rNativeContext + offsetof(NativeContext, luaH_setnum)]);
+        inst.regX64 = regs.takeReg(rax, index);
         break;
     }
     case IrCmd::STRING_LEN:
@@ -997,7 +1021,7 @@ void IrLoweringX64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         tmp1.free();
 
         if (inst.c.kind == IrOpKind::Undef || isGCO(tagOp(inst.c)))
-            callBarrierObject(regs, build, tmp2.release(), {}, vmRegOp(inst.b), inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c));
+            callBarrierObject(regs, build, tmp2.release(), {}, inst.b, inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c));
         break;
     }
     case IrCmd::CHECK_TAG:
@@ -1106,6 +1130,12 @@ void IrLoweringX64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         jumpOrAbortOnUndef(ConditionX64::NotZero, inst.b, next);
         break;
     }
+    case IrCmd::CHECK_NODE_VALUE:
+    {
+        build.cmp(dword[regOp(inst.a) + offsetof(LuaNode, val) + offsetof(TValue, tt)], LUA_TNIL);
+        jumpOrAbortOnUndef(ConditionX64::Equal, inst.b, next);
+        break;
+    }
     case IrCmd::INTERRUPT:
     {
         unsigned pcpos = uintOp(inst.a);
@@ -1132,7 +1162,7 @@ void IrLoweringX64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         callStepGc(regs, build);
         break;
     case IrCmd::BARRIER_OBJ:
-        callBarrierObject(regs, build, regOp(inst.a), inst.a, vmRegOp(inst.b), inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c));
+        callBarrierObject(regs, build, regOp(inst.a), inst.a, inst.b, inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c));
         break;
     case IrCmd::BARRIER_TABLE_BACK:
         callBarrierTableFast(regs, build, regOp(inst.a), inst.a);
@@ -1142,7 +1172,8 @@ void IrLoweringX64::lowerInst(IrInst& inst, uint32_t index, const IrBlock& next)
         Label skip;
 
         ScopedRegX64 tmp{regs, SizeX64::qword};
-        checkObjectBarrierConditions(build, tmp.reg, regOp(inst.a), vmRegOp(inst.b), inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c), skip);
+
+        checkObjectBarrierConditions(build, tmp.reg, regOp(inst.a), inst.b, inst.c.kind == IrOpKind::Undef ? -1 : tagOp(inst.c), skip);
 
         {
             ScopedSpills spillGuard(regs);

CodeGen/src/IrRegAllocA64.cpp

@@ -70,9 +70,9 @@ static int getReloadOffset(IrCmd cmd)
     LUAU_UNREACHABLE();
 }
 
-static AddressA64 getReloadAddress(const IrFunction& function, const IrInst& inst)
+static AddressA64 getReloadAddress(const IrFunction& function, const IrInst& inst, bool limitToCurrentBlock)
 {
-    IrOp location = function.findRestoreOp(inst);
+    IrOp location = function.findRestoreOp(inst, limitToCurrentBlock);
 
     if (location.kind == IrOpKind::VmReg)
         return mem(rBase, vmRegOp(location) * sizeof(TValue) + getReloadOffset(inst.cmd));
@@ -99,7 +99,7 @@ static void restoreInst(AssemblyBuilderA64& build, uint32_t& freeSpillSlots, IrF
     else
     {
         LUAU_ASSERT(!inst.spilled && inst.needsReload);
-        AddressA64 addr = getReloadAddress(function, function.instructions[s.inst]);
+        AddressA64 addr = getReloadAddress(function, function.instructions[s.inst], /*limitToCurrentBlock*/ false);
         LUAU_ASSERT(addr.base != xzr);
         build.ldr(reg, addr);
     }
@@ -321,7 +321,7 @@ size_t IrRegAllocA64::spill(AssemblyBuilderA64& build, uint32_t index, std::init
             {
                 // instead of spilling the register to never reload it, we assume the register is not needed anymore
             }
-            else if (getReloadAddress(function, def).base != xzr)
+            else if (getReloadAddress(function, def, /*limitToCurrentBlock*/ true).base != xzr)
             {
                 // instead of spilling the register to stack, we can reload it from VM stack/constants
                 // we still need to record the spill for restore(start) to work

CodeGen/src/IrRegAllocX64.cpp

@@ -338,7 +338,9 @@ unsigned IrRegAllocX64::findSpillStackSlot(IrValueKind valueKind)
 
 IrOp IrRegAllocX64::getRestoreOp(const IrInst& inst) const
 {
-    if (IrOp location = function.findRestoreOp(inst); location.kind == IrOpKind::VmReg || location.kind == IrOpKind::VmConst)
+    // When restoring the value, we allow cross-block restore because we have commited to the target location at spill time
+    if (IrOp location = function.findRestoreOp(inst, /*limitToCurrentBlock*/ false);
+        location.kind == IrOpKind::VmReg || location.kind == IrOpKind::VmConst)
         return location;
 
     return IrOp();
@@ -346,11 +348,16 @@ IrOp IrRegAllocX64::getRestoreOp(const IrInst& inst) const
 
 bool IrRegAllocX64::hasRestoreOp(const IrInst& inst) const
 {
-    return getRestoreOp(inst).kind != IrOpKind::None;
+    // When checking if value has a restore operation to spill it, we only allow it in the same block
+    IrOp location = function.findRestoreOp(inst, /*limitToCurrentBlock*/ true);
+
+    return location.kind == IrOpKind::VmReg || location.kind == IrOpKind::VmConst;
 }
 
 OperandX64 IrRegAllocX64::getRestoreAddress(const IrInst& inst, IrOp restoreOp)
 {
+    LUAU_ASSERT(restoreOp.kind != IrOpKind::None);
+
     switch (getCmdValueKind(inst.cmd))
     {
     case IrValueKind::Unknown:

CodeGen/src/IrTranslateBuiltins.cpp

@@ -748,6 +748,28 @@ static BuiltinImplResult translateBuiltinVector(IrBuilder& build, int nparams, i
     return {BuiltinImplType::Full, 1};
 }
 
+static BuiltinImplResult translateBuiltinTableInsert(IrBuilder& build, int nparams, int ra, int arg, IrOp args, int nresults, int pcpos)
+{
+    if (nparams != 2 || nresults > 0)
+        return {BuiltinImplType::None, -1};
+
+    build.loadAndCheckTag(build.vmReg(arg), LUA_TTABLE, build.vmExit(pcpos));
+
+    IrOp table = build.inst(IrCmd::LOAD_POINTER, build.vmReg(arg));
+    build.inst(IrCmd::CHECK_READONLY, table, build.vmExit(pcpos));
+
+    IrOp pos = build.inst(IrCmd::ADD_INT, build.inst(IrCmd::TABLE_LEN, table), build.constInt(1));
+
+    IrOp setnum = build.inst(IrCmd::TABLE_SETNUM, table, pos);
+
+    IrOp va = build.inst(IrCmd::LOAD_TVALUE, args);
+    build.inst(IrCmd::STORE_TVALUE, setnum, va);
+
+    build.inst(IrCmd::BARRIER_TABLE_FORWARD, table, args, build.undef());
+
+    return {BuiltinImplType::Full, 0};
+}
+
 static BuiltinImplResult translateBuiltinStringLen(IrBuilder& build, int nparams, int ra, int arg, IrOp args, int nresults, int pcpos)
 {
     if (nparams < 1 || nresults > 1)
@@ -849,6 +871,8 @@ BuiltinImplResult translateBuiltin(IrBuilder& build, int bfid, int ra, int arg,
         return translateBuiltinTypeof(build, nparams, ra, arg, args, nresults);
     case LBF_VECTOR:
         return translateBuiltinVector(build, nparams, ra, arg, args, nresults, pcpos);
+    case LBF_TABLE_INSERT:
+        return translateBuiltinTableInsert(build, nparams, ra, arg, args, nresults, pcpos);
     case LBF_STRING_LEN:
         return translateBuiltinStringLen(build, nparams, ra, arg, args, nresults, pcpos);
     default:

CodeGen/src/IrTranslation.cpp

@@ -382,6 +382,9 @@ static void translateInstBinaryNumeric(IrBuilder& build, int ra, int rb, int rc,
         case TM_DIV:
             result = build.inst(IrCmd::DIV_NUM, vb, vc);
             break;
+        case TM_IDIV:
+            result = build.inst(IrCmd::IDIV_NUM, vb, vc);
+            break;
         case TM_MOD:
             result = build.inst(IrCmd::MOD_NUM, vb, vc);
             break;
@@ -472,8 +475,9 @@ void translateInstLength(IrBuilder& build, const Instruction* pc, int pcpos)
     build.inst(IrCmd::CHECK_NO_METATABLE, vb, fallback);
 
     IrOp va = build.inst(IrCmd::TABLE_LEN, vb);
+    IrOp vai = build.inst(IrCmd::INT_TO_NUM, va);
 
-    build.inst(IrCmd::STORE_DOUBLE, build.vmReg(ra), va);
+    build.inst(IrCmd::STORE_DOUBLE, build.vmReg(ra), vai);
     build.inst(IrCmd::STORE_TAG, build.vmReg(ra), build.constTag(LUA_TNUMBER));
 
     IrOp next = build.blockAtInst(pcpos + 1);
@@ -554,7 +558,7 @@ IrOp translateFastCallN(IrBuilder& build, const Instruction* pc, int pcpos, bool
 
     IrOp builtinArgs = args;
 
-    if (customArgs.kind == IrOpKind::VmConst)
+    if (customArgs.kind == IrOpKind::VmConst && bfid != LBF_TABLE_INSERT)
     {
         TValue protok = build.function.proto->k[customArgs.index];
 
@@ -976,7 +980,7 @@ void translateInstGetTableKS(IrBuilder& build, const Instruction* pc, int pcpos)
 
     IrOp vb = build.inst(IrCmd::LOAD_POINTER, build.vmReg(rb));
 
-    IrOp addrSlotEl = build.inst(IrCmd::GET_SLOT_NODE_ADDR, vb, build.constUint(pcpos));
+    IrOp addrSlotEl = build.inst(IrCmd::GET_SLOT_NODE_ADDR, vb, build.constUint(pcpos), build.vmConst(aux));
 
     build.inst(IrCmd::CHECK_SLOT_MATCH, addrSlotEl, build.vmConst(aux), fallback);
 
@@ -1003,7 +1007,7 @@ void translateInstSetTableKS(IrBuilder& build, const Instruction* pc, int pcpos)
 
     IrOp vb = build.inst(IrCmd::LOAD_POINTER, build.vmReg(rb));
 
-    IrOp addrSlotEl = build.inst(IrCmd::GET_SLOT_NODE_ADDR, vb, build.constUint(pcpos));
+    IrOp addrSlotEl = build.inst(IrCmd::GET_SLOT_NODE_ADDR, vb, build.constUint(pcpos), build.vmConst(aux));
 
     build.inst(IrCmd::CHECK_SLOT_MATCH, addrSlotEl, build.vmConst(aux), fallback);
     build.inst(IrCmd::CHECK_READONLY, vb, fallback);
@@ -1028,7 +1032,7 @@ void translateInstGetGlobal(IrBuilder& build, const Instruction* pc, int pcpos)
     IrOp fallback = build.block(IrBlockKind::Fallback);
 
     IrOp env = build.inst(IrCmd::LOAD_ENV);
-    IrOp addrSlotEl = build.inst(IrCmd::GET_SLOT_NODE_ADDR, env, build.constUint(pcpos));
+    IrOp addrSlotEl = build.inst(IrCmd::GET_SLOT_NODE_ADDR, env, build.constUint(pcpos), build.vmConst(aux));
 
     build.inst(IrCmd::CHECK_SLOT_MATCH, addrSlotEl, build.vmConst(aux), fallback);
 
@@ -1050,7 +1054,7 @@ void translateInstSetGlobal(IrBuilder& build, const Instruction* pc, int pcpos)
     IrOp fallback = build.block(IrBlockKind::Fallback);
 
     IrOp env = build.inst(IrCmd::LOAD_ENV);
-    IrOp addrSlotEl = build.inst(IrCmd::GET_SLOT_NODE_ADDR, env, build.constUint(pcpos));
+    IrOp addrSlotEl = build.inst(IrCmd::GET_SLOT_NODE_ADDR, env, build.constUint(pcpos), build.vmConst(aux));
 
     build.inst(IrCmd::CHECK_SLOT_MATCH, addrSlotEl, build.vmConst(aux), fallback);
     build.inst(IrCmd::CHECK_READONLY, env, fallback);
@@ -1141,7 +1145,7 @@ void translateInstNamecall(IrBuilder& build, const Instruction* pc, int pcpos)
     build.loadAndCheckTag(indexPtr, LUA_TTABLE, fallback);
     IrOp index = build.inst(IrCmd::LOAD_POINTER, indexPtr);
 
-    IrOp addrIndexNodeEl = build.inst(IrCmd::GET_SLOT_NODE_ADDR, index, build.constUint(pcpos));
+    IrOp addrIndexNodeEl = build.inst(IrCmd::GET_SLOT_NODE_ADDR, index, build.constUint(pcpos), build.vmConst(aux));
     build.inst(IrCmd::CHECK_SLOT_MATCH, addrIndexNodeEl, build.vmConst(aux), fallback);
 
     // TODO: original 'table' was clobbered by a call inside 'FASTGETTM'

CodeGen/src/IrUtils.cpp

@@ -54,6 +54,7 @@ IrValueKind getCmdValueKind(IrCmd cmd)
     case IrCmd::SUB_NUM:
     case IrCmd::MUL_NUM:
     case IrCmd::DIV_NUM:
+    case IrCmd::IDIV_NUM:
     case IrCmd::MOD_NUM:
     case IrCmd::MIN_NUM:
     case IrCmd::MAX_NUM:
@@ -79,7 +80,9 @@ IrValueKind getCmdValueKind(IrCmd cmd)
     case IrCmd::JUMP_SLOT_MATCH:
         return IrValueKind::None;
     case IrCmd::TABLE_LEN:
-        return IrValueKind::Double;
+        return IrValueKind::Int;
+    case IrCmd::TABLE_SETNUM:
+        return IrValueKind::Pointer;
     case IrCmd::STRING_LEN:
         return IrValueKind::Int;
     case IrCmd::NEW_TABLE:
@@ -119,6 +122,7 @@ IrValueKind getCmdValueKind(IrCmd cmd)
     case IrCmd::CHECK_ARRAY_SIZE:
     case IrCmd::CHECK_SLOT_MATCH:
     case IrCmd::CHECK_NODE_NO_NEXT:
+    case IrCmd::CHECK_NODE_VALUE:
     case IrCmd::INTERRUPT:
     case IrCmd::CHECK_GC:
     case IrCmd::BARRIER_OBJ:
@@ -464,6 +468,10 @@ void foldConstants(IrBuilder& build, IrFunction& function, IrBlock& block, uint3
         if (inst.a.kind == IrOpKind::Constant && inst.b.kind == IrOpKind::Constant)
             substitute(function, inst, build.constDouble(function.doubleOp(inst.a) / function.doubleOp(inst.b)));
         break;
+    case IrCmd::IDIV_NUM:
+        if (inst.a.kind == IrOpKind::Constant && inst.b.kind == IrOpKind::Constant)
+            substitute(function, inst, build.constDouble(luai_numidiv(function.doubleOp(inst.a), function.doubleOp(inst.b))));
+        break;
     case IrCmd::MOD_NUM:
         if (inst.a.kind == IrOpKind::Constant && inst.b.kind == IrOpKind::Constant)
             substitute(function, inst, build.constDouble(luai_nummod(function.doubleOp(inst.a), function.doubleOp(inst.b))));

CodeGen/src/IrValueLocationTracking.cpp

@@ -108,13 +108,14 @@ void IrValueLocationTracking::beforeInstLowering(IrInst& inst)
     case IrCmd::FINDUPVAL:
         break;
 
-        // These instrucitons read VmReg only after optimizeMemoryOperandsX64
+        // These instructions read VmReg only after optimizeMemoryOperandsX64
     case IrCmd::CHECK_TAG:
     case IrCmd::CHECK_TRUTHY:
     case IrCmd::ADD_NUM:
     case IrCmd::SUB_NUM:
     case IrCmd::MUL_NUM:
     case IrCmd::DIV_NUM:
+    case IrCmd::IDIV_NUM:
     case IrCmd::MOD_NUM:
     case IrCmd::MIN_NUM:
     case IrCmd::MAX_NUM:

CodeGen/src/NativeState.cpp

@@ -53,6 +53,7 @@ void initFunctions(NativeState& data)
     data.context.luaH_new = luaH_new;
     data.context.luaH_clone = luaH_clone;
     data.context.luaH_resizearray = luaH_resizearray;
+    data.context.luaH_setnum = luaH_setnum;
 
     data.context.luaC_barriertable = luaC_barriertable;
     data.context.luaC_barrierf = luaC_barrierf;

CodeGen/src/NativeState.h

@@ -44,6 +44,7 @@ struct NativeContext
     Table* (*luaH_new)(lua_State* L, int narray, int lnhash) = nullptr;
     Table* (*luaH_clone)(lua_State* L, Table* tt) = nullptr;
     void (*luaH_resizearray)(lua_State* L, Table* t, int nasize) = nullptr;
+    TValue* (*luaH_setnum)(lua_State* L, Table* t, int key);
 
     void (*luaC_barriertable)(lua_State* L, Table* t, GCObject* v) = nullptr;
     void (*luaC_barrierf)(lua_State* L, GCObject* o, GCObject* v) = nullptr;

CodeGen/src/OptimizeConstProp.cpp

@@ -13,7 +13,10 @@
 #include <vector>
 
 LUAU_FASTINTVARIABLE(LuauCodeGenMinLinearBlockPath, 3)
+LUAU_FASTINTVARIABLE(LuauCodeGenReuseSlotLimit, 64)
 LUAU_FASTFLAGVARIABLE(DebugLuauAbortingChecks, false)
+LUAU_FASTFLAGVARIABLE(LuauReuseHashSlots2, false)
+LUAU_FASTFLAGVARIABLE(LuauKeepVmapLinear, false)
 
 namespace Luau
 {
@@ -174,6 +177,10 @@ struct ConstPropState
     {
         for (int i = 0; i <= maxReg; ++i)
             invalidateHeap(regs[i]);
+
+        // If table memory has changed, we can't reuse previously computed and validated table slot lookups
+        getSlotNodeCache.clear();
+        checkSlotMatchCache.clear();
     }
 
     void invalidateHeap(RegisterInfo& reg)
@@ -190,6 +197,21 @@ struct ConstPropState
         inSafeEnv = false;
     }
 
+    void invalidateTableArraySize()
+    {
+        for (int i = 0; i <= maxReg; ++i)
+            invalidateTableArraySize(regs[i]);
+
+        // If table memory has changed, we can't reuse previously computed and validated table slot lookups
+        getSlotNodeCache.clear();
+        checkSlotMatchCache.clear();
+    }
+
+    void invalidateTableArraySize(RegisterInfo& reg)
+    {
+        reg.knownTableArraySize = -1;
+    }
+
     void createRegLink(uint32_t instIdx, IrOp regOp)
     {
         LUAU_ASSERT(!instLink.contains(instIdx));
@@ -367,6 +389,8 @@ struct ConstPropState
 
         instLink.clear();
         valueMap.clear();
+        getSlotNodeCache.clear();
+        checkSlotMatchCache.clear();
     }
 
     IrFunction& function;
@@ -384,6 +408,9 @@ struct ConstPropState
     DenseHashMap<uint32_t, RegisterLink> instLink{~0u};
 
     DenseHashMap<IrInst, uint32_t, IrInstHash, IrInstEq> valueMap;
+
+    std::vector<uint32_t> getSlotNodeCache;
+    std::vector<uint32_t> checkSlotMatchCache;
 };
 
 static void handleBuiltinEffects(ConstPropState& state, LuauBuiltinFunction bfid, uint32_t firstReturnReg, int nresults)
@@ -863,7 +890,25 @@ static void constPropInInst(ConstPropState& state, IrBuilder& build, IrFunction&
     case IrCmd::NOP:
     case IrCmd::LOAD_ENV:
     case IrCmd::GET_ARR_ADDR:
+        break;
     case IrCmd::GET_SLOT_NODE_ADDR:
+        if (!FFlag::LuauReuseHashSlots2)
+            break;
+
+        for (uint32_t prevIdx : state.getSlotNodeCache)
+        {
+            const IrInst& prev = function.instructions[prevIdx];
+
+            if (prev.a == inst.a && prev.c == inst.c)
+            {
+                substitute(function, inst, IrOp{IrOpKind::Inst, prevIdx});
+                return; // Break out from both the loop and the switch
+            }
+        }
+
+        if (int(state.getSlotNodeCache.size()) < FInt::LuauCodeGenReuseSlotLimit)
+            state.getSlotNodeCache.push_back(index);
+        break;
     case IrCmd::GET_HASH_NODE_ADDR:
     case IrCmd::GET_CLOSURE_UPVAL_ADDR:
         break;
@@ -873,6 +918,7 @@ static void constPropInInst(ConstPropState& state, IrBuilder& build, IrFunction&
     case IrCmd::SUB_NUM:
     case IrCmd::MUL_NUM:
     case IrCmd::DIV_NUM:
+    case IrCmd::IDIV_NUM:
     case IrCmd::MOD_NUM:
     case IrCmd::MIN_NUM:
     case IrCmd::MAX_NUM:
@@ -892,6 +938,10 @@ static void constPropInInst(ConstPropState& state, IrBuilder& build, IrFunction&
     case IrCmd::JUMP_EQ_POINTER:
     case IrCmd::JUMP_SLOT_MATCH:
     case IrCmd::TABLE_LEN:
+        break;
+    case IrCmd::TABLE_SETNUM:
+        state.invalidateTableArraySize();
+        break;
     case IrCmd::STRING_LEN:
     case IrCmd::NEW_TABLE:
     case IrCmd::DUP_TABLE:
@@ -938,7 +988,26 @@ static void constPropInInst(ConstPropState& state, IrBuilder& build, IrFunction&
         break;
     }
     case IrCmd::CHECK_SLOT_MATCH:
+        if (!FFlag::LuauReuseHashSlots2)
+            break;
+
+        for (uint32_t prevIdx : state.checkSlotMatchCache)
+        {
+            const IrInst& prev = function.instructions[prevIdx];
+
+            if (prev.a == inst.a && prev.b == inst.b)
+            {
+                // Only a check for 'nil' value is left
+                replace(function, block, index, {IrCmd::CHECK_NODE_VALUE, inst.a, inst.c});
+                return; // Break out from both the loop and the switch
+            }
+        }
+
+        if (int(state.checkSlotMatchCache.size()) < FInt::LuauCodeGenReuseSlotLimit)
+            state.checkSlotMatchCache.push_back(index);
+        break;
     case IrCmd::CHECK_NODE_NO_NEXT:
+    case IrCmd::CHECK_NODE_VALUE:
     case IrCmd::BARRIER_TABLE_BACK:
     case IrCmd::RETURN:
     case IrCmd::COVERAGE:
@@ -999,7 +1068,10 @@ static void constPropInInst(ConstPropState& state, IrBuilder& build, IrFunction&
         if (RegisterInfo* info = state.tryGetRegisterInfo(inst.b); info && info->knownTableArraySize >= 0)
             replace(function, inst.f, build.constUint(info->knownTableArraySize));
 
-        state.valueMap.clear(); // TODO: this can be relaxed when x64 emitInstSetList becomes aware of register allocator
+        // TODO: this can be relaxed when x64 emitInstSetList becomes aware of register allocator
+        state.valueMap.clear();
+        state.getSlotNodeCache.clear();
+        state.checkSlotMatchCache.clear();
         break;
     case IrCmd::CALL:
         state.invalidateRegistersFrom(vmRegOp(inst.a));
@@ -1012,7 +1084,11 @@ static void constPropInInst(ConstPropState& state, IrBuilder& build, IrFunction&
         break;
     case IrCmd::FORGLOOP:
         state.invalidateRegistersFrom(vmRegOp(inst.a) + 2); // Rn and Rn+1 are not modified
-        state.valueMap.clear();                             // TODO: this can be relaxed when x64 emitInstForGLoop becomes aware of register allocator
+
+        // TODO: this can be relaxed when x64 emitInstForGLoop becomes aware of register allocator
+        state.valueMap.clear();
+        state.getSlotNodeCache.clear();
+        state.checkSlotMatchCache.clear();
         break;
     case IrCmd::FORGLOOP_FALLBACK:
         state.invalidateRegistersFrom(vmRegOp(inst.a) + 2); // Rn and Rn+1 are not modified
@@ -1076,8 +1152,15 @@ static void constPropInBlock(IrBuilder& build, IrBlock& block, ConstPropState& s
         constPropInInst(state, build, function, block, inst, index);
     }
 
-    // Value numbering and load/store propagation is not performed between blocks
-    state.valueMap.clear();
+    if (!FFlag::LuauKeepVmapLinear)
+    {
+        // Value numbering and load/store propagation is not performed between blocks
+        state.valueMap.clear();
+
+        // Same for table slot data propagation
+        state.getSlotNodeCache.clear();
+        state.checkSlotMatchCache.clear();
+    }
 }
 
 static void constPropInBlockChain(IrBuilder& build, std::vector<uint8_t>& visited, IrBlock* block, ConstPropState& state)
@@ -1086,6 +1169,9 @@ static void constPropInBlockChain(IrBuilder& build, std::vector<uint8_t>& visite
 
     state.clear();
 
+    const uint32_t startSortkey = block->sortkey;
+    uint32_t chainPos = 0;
+
     while (block)
     {
         uint32_t blockIdx = function.getBlockIndex(*block);
@@ -1094,19 +1180,40 @@ static void constPropInBlockChain(IrBuilder& build, std::vector<uint8_t>& visite
 
         constPropInBlock(build, *block, state);
 
+        if (FFlag::LuauKeepVmapLinear)
+        {
+            // Value numbering and load/store propagation is not performed between blocks right now
+            // This is because cross-block value uses limit creation of linear block (restriction in collectDirectBlockJumpPath)
+            state.valueMap.clear();
+
+            // Same for table slot data propagation
+            state.getSlotNodeCache.clear();
+            state.checkSlotMatchCache.clear();
+        }
+
+        // Blocks in a chain are guaranteed to follow each other
+        // We force that by giving all blocks the same sorting key, but consecutive chain keys
+        block->sortkey = startSortkey;
+        block->chainkey = chainPos++;
+
         IrInst& termInst = function.instructions[block->finish];
 
         IrBlock* nextBlock = nullptr;
 
         // Unconditional jump into a block with a single user (current block) allows us to continue optimization
         // with the information we have gathered so far (unless we have already visited that block earlier)
-        if (termInst.cmd == IrCmd::JUMP && termInst.a.kind != IrOpKind::VmExit)
+        if (termInst.cmd == IrCmd::JUMP && termInst.a.kind == IrOpKind::Block)
         {
             IrBlock& target = function.blockOp(termInst.a);
             uint32_t targetIdx = function.getBlockIndex(target);
 
             if (target.useCount == 1 && !visited[targetIdx] && target.kind != IrBlockKind::Fallback)
+            {
+                // Make sure block ordering guarantee is checked at lowering time
+                block->expectedNextBlock = function.getBlockIndex(target);
+
                 nextBlock = &target;
+            }
         }
 
         block = nextBlock;
@@ -1134,7 +1241,7 @@ static std::vector<uint32_t> collectDirectBlockJumpPath(IrFunction& function, st
         IrBlock* nextBlock = nullptr;
 
         // A chain is made from internal blocks that were not a part of bytecode CFG
-        if (termInst.cmd == IrCmd::JUMP && termInst.a.kind != IrOpKind::VmExit)
+        if (termInst.cmd == IrCmd::JUMP && termInst.a.kind == IrOpKind::Block)
         {
             IrBlock& target = function.blockOp(termInst.a);
             uint32_t targetIdx = function.getBlockIndex(target);
@@ -1175,8 +1282,8 @@ static void tryCreateLinearBlock(IrBuilder& build, std::vector<uint8_t>& visited
     if (termInst.cmd != IrCmd::JUMP)
         return;
 
-    // And it can't be jump to a VM exit
-    if (termInst.a.kind == IrOpKind::VmExit)
+    // And it can't be jump to a VM exit or undef
+    if (termInst.a.kind != IrOpKind::Block)
         return;
 
     // And it has to jump to a block with more than one user
@@ -1196,14 +1303,14 @@ static void tryCreateLinearBlock(IrBuilder& build, std::vector<uint8_t>& visited
     // Initialize state with the knowledge of our current block
     state.clear();
 
-    // TODO: using values from the first block can cause 'live out' of the linear block predecessor to not have all required registers
     constPropInBlock(build, startingBlock, state);
 
     // Verify that target hasn't changed
     LUAU_ASSERT(function.instructions[startingBlock.finish].a.index == targetBlockIdx);
 
     // Note: using startingBlock after this line is unsafe as the reference may be reallocated by build.block() below
-    uint32_t startingInsn = startingBlock.start;
+    const uint32_t startingSortKey = startingBlock.sortkey;
+    const uint32_t startingChainKey = startingBlock.chainkey;
 
     // Create new linearized block into which we are going to redirect starting block jump
     IrOp newBlock = build.block(IrBlockKind::Linearized);
@@ -1213,7 +1320,11 @@ static void tryCreateLinearBlock(IrBuilder& build, std::vector<uint8_t>& visited
 
     // By default, blocks are ordered according to start instruction; we alter sort order to make sure linearized block is placed right after the
     // starting block
-    function.blocks[newBlock.index].sortkey = startingInsn + 1;
+    function.blocks[newBlock.index].sortkey = startingSortKey;
+    function.blocks[newBlock.index].chainkey = startingChainKey + 1;
+
+    // Make sure block ordering guarantee is checked at lowering time
+    function.blocks[blockIdx].expectedNextBlock = newBlock.index;
 
     replace(function, termInst.a, newBlock);
 
@@ -1252,6 +1363,12 @@ static void tryCreateLinearBlock(IrBuilder& build, std::vector<uint8_t>& visited
                 def.varargStart = pathDef.varargStart;
             }
         }
+
+        // Update predecessors
+        function.cfg.predecessorsOffsets.push_back(uint32_t(function.cfg.predecessors.size()));
+        function.cfg.predecessors.push_back(blockIdx);
+
+        // Updating successors will require visiting the instructions again and we don't have a current use for linearized block successor list
     }
 
     // Optimize our linear block

CodeGen/src/OptimizeFinalX64.cpp

@@ -58,6 +58,7 @@ static void optimizeMemoryOperandsX64(IrFunction& function, IrBlock& block)
         case IrCmd::SUB_NUM:
         case IrCmd::MUL_NUM:
         case IrCmd::DIV_NUM:
+        case IrCmd::IDIV_NUM:
         case IrCmd::MOD_NUM:
         case IrCmd::MIN_NUM:
         case IrCmd::MAX_NUM:

Common/include/Luau/Bytecode.h

@@ -44,7 +44,7 @@
 // Version 1: Baseline version for the open-source release. Supported until 0.521.
 // Version 2: Adds Proto::linedefined. Supported until 0.544.
 // Version 3: Adds FORGPREP/JUMPXEQK* and enhances AUX encoding for FORGLOOP. Removes FORGLOOP_NEXT/INEXT and JUMPIFEQK/JUMPIFNOTEQK. Currently supported.
-// Version 4: Adds Proto::flags and typeinfo. Currently supported.
+// Version 4: Adds Proto::flags, typeinfo, and floor division opcodes IDIV/IDIVK. Currently supported.
 
 // Bytecode opcode, part of the instruction header
 enum LuauOpcode
@@ -390,6 +390,18 @@ enum LuauOpcode
     LOP_JUMPXEQKN,
     LOP_JUMPXEQKS,
 
+    // IDIV: compute floor division between two source registers and put the result into target register
+    // A: target register
+    // B: source register 1
+    // C: source register 2
+    LOP_IDIV,
+
+    // IDIVK compute floor division between the source register and a constant and put the result into target register
+    // A: target register
+    // B: source register
+    // C: constant table index (0..255)
+    LOP_IDIVK,
+
     // Enum entry for number of opcodes, not a valid opcode by itself!
     LOP__COUNT
 };

Compiler/include/Luau/BytecodeBuilder.h

@@ -15,7 +15,6 @@ class BytecodeEncoder
 public:
     virtual ~BytecodeEncoder() {}
 
-    virtual uint8_t encodeOp(uint8_t op) = 0;
     virtual void encode(uint32_t* data, size_t count) = 0;
 };
 

Compiler/src/BytecodeBuilder.cpp

@@ -8,7 +8,8 @@
 #include <string.h>
 
 LUAU_FASTFLAGVARIABLE(BytecodeVersion4, false)
-LUAU_FASTFLAGVARIABLE(BytecodeEnc, false)
+
+LUAU_FASTFLAG(LuauFloorDivision)
 
 namespace Luau
 {
@@ -238,7 +239,7 @@ void BytecodeBuilder::endFunction(uint8_t maxstacksize, uint8_t numupvalues, uin
     // very approximate: 4 bytes per instruction for code, 1 byte for debug line, and 1-2 bytes for aux data like constants plus overhead
     func.data.reserve(32 + insns.size() * 7);
 
-    if (FFlag::BytecodeEnc && encoder)
+    if (encoder)
         encoder->encode(insns.data(), insns.size());
 
     writeFunction(func.data, currentFunction, flags);
@@ -625,29 +626,8 @@ void BytecodeBuilder::writeFunction(std::string& ss, uint32_t id, uint8_t flags)
     // instructions
     writeVarInt(ss, uint32_t(insns.size()));
 
-    if (encoder && !FFlag::BytecodeEnc)
-    {
-        for (size_t i = 0; i < insns.size();)
-        {
-            uint8_t op = LUAU_INSN_OP(insns[i]);
-            LUAU_ASSERT(op < LOP__COUNT);
-
-            int oplen = getOpLength(LuauOpcode(op));
-            uint8_t openc = encoder->encodeOp(op);
-
-            writeInt(ss, openc | (insns[i] & ~0xff));
-
-            for (int j = 1; j < oplen; ++j)
-                writeInt(ss, insns[i + j]);
-
-            i += oplen;
-        }
-    }
-    else
-    {
-        for (uint32_t insn : insns)
-            writeInt(ss, insn);
-    }
+    for (uint32_t insn : insns)
+        writeInt(ss, insn);
 
     // constants
     writeVarInt(ss, uint32_t(constants.size()));
@@ -1306,8 +1286,11 @@ void BytecodeBuilder::validateInstructions() const
         case LOP_SUB:
         case LOP_MUL:
         case LOP_DIV:
+        case LOP_IDIV:
         case LOP_MOD:
         case LOP_POW:
+            LUAU_ASSERT(FFlag::LuauFloorDivision || op != LOP_IDIV);
+
             VREG(LUAU_INSN_A(insn));
             VREG(LUAU_INSN_B(insn));
             VREG(LUAU_INSN_C(insn));
@@ -1317,8 +1300,11 @@ void BytecodeBuilder::validateInstructions() const
         case LOP_SUBK:
         case LOP_MULK:
         case LOP_DIVK:
+        case LOP_IDIVK:
         case LOP_MODK:
         case LOP_POWK:
+            LUAU_ASSERT(FFlag::LuauFloorDivision || op != LOP_IDIVK);
+
             VREG(LUAU_INSN_A(insn));
             VREG(LUAU_INSN_B(insn));
             VCONST(LUAU_INSN_C(insn), Number);
@@ -1885,6 +1871,12 @@ void BytecodeBuilder::dumpInstruction(const uint32_t* code, std::string& result,
         formatAppend(result, "DIV R%d R%d R%d\n", LUAU_INSN_A(insn), LUAU_INSN_B(insn), LUAU_INSN_C(insn));
         break;
 
+    case LOP_IDIV:
+        LUAU_ASSERT(FFlag::LuauFloorDivision);
+
+        formatAppend(result, "IDIV R%d R%d R%d\n", LUAU_INSN_A(insn), LUAU_INSN_B(insn), LUAU_INSN_C(insn));
+        break;
+
     case LOP_MOD:
         formatAppend(result, "MOD R%d R%d R%d\n", LUAU_INSN_A(insn), LUAU_INSN_B(insn), LUAU_INSN_C(insn));
         break;
@@ -1917,6 +1909,14 @@ void BytecodeBuilder::dumpInstruction(const uint32_t* code, std::string& result,
         result.append("]\n");
         break;
 
+    case LOP_IDIVK:
+        LUAU_ASSERT(FFlag::LuauFloorDivision);
+
+        formatAppend(result, "IDIVK R%d R%d K%d [", LUAU_INSN_A(insn), LUAU_INSN_B(insn), LUAU_INSN_C(insn));
+        dumpConstant(result, LUAU_INSN_C(insn));
+        result.append("]\n");
+        break;
+
     case LOP_MODK:
         formatAppend(result, "MODK R%d R%d K%d [", LUAU_INSN_A(insn), LUAU_INSN_B(insn), LUAU_INSN_C(insn));
         dumpConstant(result, LUAU_INSN_C(insn));

Compiler/src/Compiler.cpp

@@ -26,6 +26,8 @@ LUAU_FASTINTVARIABLE(LuauCompileInlineThreshold, 25)
 LUAU_FASTINTVARIABLE(LuauCompileInlineThresholdMaxBoost, 300)
 LUAU_FASTINTVARIABLE(LuauCompileInlineDepth, 5)
 
+LUAU_FASTFLAG(LuauFloorDivision)
+
 namespace Luau
 {
 
@@ -1019,6 +1021,11 @@ struct Compiler
         case AstExprBinary::Div:
             return k ? LOP_DIVK : LOP_DIV;
 
+        case AstExprBinary::FloorDiv:
+            LUAU_ASSERT(FFlag::LuauFloorDivision);
+
+            return k ? LOP_IDIVK : LOP_IDIV;
+
         case AstExprBinary::Mod:
             return k ? LOP_MODK : LOP_MOD;
 
@@ -1469,9 +1476,12 @@ struct Compiler
         case AstExprBinary::Sub:
         case AstExprBinary::Mul:
         case AstExprBinary::Div:
+        case AstExprBinary::FloorDiv:
         case AstExprBinary::Mod:
         case AstExprBinary::Pow:
         {
+            LUAU_ASSERT(FFlag::LuauFloorDivision || expr->op != AstExprBinary::FloorDiv);
+
             int32_t rc = getConstantNumber(expr->right);
 
             if (rc >= 0 && rc <= 255)
@@ -3192,9 +3202,12 @@ struct Compiler
         case AstExprBinary::Sub:
         case AstExprBinary::Mul:
         case AstExprBinary::Div:
+        case AstExprBinary::FloorDiv:
         case AstExprBinary::Mod:
         case AstExprBinary::Pow:
         {
+            LUAU_ASSERT(FFlag::LuauFloorDivision || stat->op != AstExprBinary::FloorDiv);
+
             if (var.kind != LValue::Kind_Local)
                 compileLValueUse(var, target, /* set= */ false);
 

Compiler/src/ConstantFolding.cpp

@@ -104,6 +104,14 @@ static void foldBinary(Constant& result, AstExprBinary::Op op, const Constant& l
         }
         break;
 
+    case AstExprBinary::FloorDiv:
+        if (la.type == Constant::Type_Number && ra.type == Constant::Type_Number)
+        {
+            result.type = Constant::Type_Number;
+            result.valueNumber = floor(la.valueNumber / ra.valueNumber);
+        }
+        break;
+
     case AstExprBinary::Mod:
         if (la.type == Constant::Type_Number && ra.type == Constant::Type_Number)
         {

Config/src/Config.cpp

@@ -4,6 +4,7 @@
 #include "Luau/Lexer.h"
 #include "Luau/StringUtils.h"
 
+LUAU_FASTFLAG(LuauFloorDivision)
 namespace Luau
 {
 
@@ -112,14 +113,23 @@ static void next(Lexer& lexer)
     lexer.next();
 
     // skip C-style comments as Lexer only understands Lua-style comments atm
-    while (lexer.current().type == '/')
+
+    if (FFlag::LuauFloorDivision)
     {
-        Lexeme peek = lexer.lookahead();
+        while (lexer.current().type == Luau::Lexeme::FloorDiv)
+            lexer.nextline();
+    }
+    else
+    {
+        while (lexer.current().type == '/')
+        {
+            Lexeme peek = lexer.lookahead();
 
-        if (peek.type != '/' || peek.location.begin != lexer.current().location.end)
-            break;
+            if (peek.type != '/' || peek.location.begin != lexer.current().location.end)
+                break;
 
-        lexer.nextline();
+            lexer.nextline();
+        }
     }
 }
 

LICENSE.txt

@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2019-2022 Roblox Corporation
+Copyright (c) 2019-2023 Roblox Corporation
 Copyright (c) 1994–2019 Lua.org, PUC-Rio.
 
 Permission is hereby granted, free of charge, to any person obtaining a copy of
@@ -19,4 +19,4 @@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE. 
+SOFTWARE.

Makefile

@@ -255,6 +255,7 @@ $(BUILD)/fuzz/protoprint.cpp.o: fuzz/luau.pb.cpp
 
 build/libprotobuf-mutator:
 	git clone https://github.com/google/libprotobuf-mutator build/libprotobuf-mutator
+	git -C build/libprotobuf-mutator checkout 212a7be1eb08e7f9c79732d2aab9b2097085d936
 	CXX= cmake -S build/libprotobuf-mutator -B build/libprotobuf-mutator $(DPROTOBUF)
 	make -C build/libprotobuf-mutator -j8
 

VM/include/lua.h

@@ -443,7 +443,7 @@ typedef struct lua_Callbacks lua_Callbacks;
 LUA_API lua_Callbacks* lua_callbacks(lua_State* L);
 
 /******************************************************************************
- * Copyright (c) 2019-2022 Roblox Corporation
+ * Copyright (c) 2019-2023 Roblox Corporation
  * Copyright (C) 1994-2008 Lua.org, PUC-Rio.  All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining

VM/src/lapi.cpp

@@ -38,7 +38,7 @@ const char* lua_ident = "$Lua: Lua 5.1.4 Copyright (C) 1994-2008 Lua.org, PUC-Ri
                         "$Authors: R. Ierusalimschy, L. H. de Figueiredo & W. Celes $\n"
                         "$URL: www.lua.org $\n";
 
-const char* luau_ident = "$Luau: Copyright (C) 2019-2022 Roblox Corporation $\n"
+const char* luau_ident = "$Luau: Copyright (C) 2019-2023 Roblox Corporation $\n"
                          "$URL: luau-lang.org $\n";
 
 #define api_checknelems(L, n) api_check(L, (n) <= (L->top - L->base))

VM/src/lnumutils.h

@@ -40,6 +40,13 @@ inline double luai_nummod(double a, double b)
 }
 LUAU_FASTMATH_END
 
+LUAU_FASTMATH_BEGIN
+inline double luai_numidiv(double a, double b)
+{
+    return floor(a / b);
+}
+LUAU_FASTMATH_END
+
 #define luai_num2int(i, d) ((i) = (int)(d))
 
 // On MSVC in 32-bit, double to unsigned cast compiles into a call to __dtoui3, so we invoke x87->int64 conversion path manually

VM/src/ltm.cpp

@@ -48,6 +48,7 @@ const char* const luaT_eventname[] = {
     "__sub",
     "__mul",
     "__div",
+    "__idiv",
     "__mod",
     "__pow",
     "__unm",

VM/src/ltm.h

@@ -27,6 +27,7 @@ typedef enum
     TM_SUB,
     TM_MUL,
     TM_DIV,
+    TM_IDIV,
     TM_MOD,
     TM_POW,
     TM_UNM,

VM/src/lvmexecute.cpp

@@ -103,7 +103,8 @@
         VM_DISPATCH_OP(LOP_LOADKX), VM_DISPATCH_OP(LOP_JUMPX), VM_DISPATCH_OP(LOP_FASTCALL), VM_DISPATCH_OP(LOP_COVERAGE), \
         VM_DISPATCH_OP(LOP_CAPTURE), VM_DISPATCH_OP(LOP_DEP_JUMPIFEQK), VM_DISPATCH_OP(LOP_DEP_JUMPIFNOTEQK), VM_DISPATCH_OP(LOP_FASTCALL1), \
         VM_DISPATCH_OP(LOP_FASTCALL2), VM_DISPATCH_OP(LOP_FASTCALL2K), VM_DISPATCH_OP(LOP_FORGPREP), VM_DISPATCH_OP(LOP_JUMPXEQKNIL), \
-        VM_DISPATCH_OP(LOP_JUMPXEQKB), VM_DISPATCH_OP(LOP_JUMPXEQKN), VM_DISPATCH_OP(LOP_JUMPXEQKS),
+        VM_DISPATCH_OP(LOP_JUMPXEQKB), VM_DISPATCH_OP(LOP_JUMPXEQKN), VM_DISPATCH_OP(LOP_JUMPXEQKS), VM_DISPATCH_OP(LOP_IDIV), \
+        VM_DISPATCH_OP(LOP_IDIVK),
 
 #if defined(__GNUC__) || defined(__clang__)
 #define VM_USE_CGOTO 1
@@ -1660,6 +1661,54 @@ reentry:
                 }
             }
 
+            VM_CASE(LOP_IDIV)
+            {
+                Instruction insn = *pc++;
+                StkId ra = VM_REG(LUAU_INSN_A(insn));
+                StkId rb = VM_REG(LUAU_INSN_B(insn));
+                StkId rc = VM_REG(LUAU_INSN_C(insn));
+
+                // fast-path
+                if (LUAU_LIKELY(ttisnumber(rb) && ttisnumber(rc)))
+                {
+                    setnvalue(ra, luai_numidiv(nvalue(rb), nvalue(rc)));
+                    VM_NEXT();
+                }
+                else if (ttisvector(rb) && ttisnumber(rc))
+                {
+                    const float* vb = vvalue(rb);
+                    float vc = cast_to(float, nvalue(rc));
+                    setvvalue(ra, float(luai_numidiv(vb[0], vc)), float(luai_numidiv(vb[1], vc)), float(luai_numidiv(vb[2], vc)),
+                        float(luai_numidiv(vb[3], vc)));
+                    VM_NEXT();
+                }
+                else
+                {
+                    // fast-path for userdata with C functions
+                    StkId rbc = ttisnumber(rb) ? rc : rb;
+                    const TValue* fn = 0;
+                    if (ttisuserdata(rbc) && (fn = luaT_gettmbyobj(L, rbc, TM_IDIV)) && ttisfunction(fn) && clvalue(fn)->isC)
+                    {
+                        // note: it's safe to push arguments past top for complicated reasons (see top of the file)
+                        LUAU_ASSERT(L->top + 3 < L->stack + L->stacksize);
+                        StkId top = L->top;
+                        setobj2s(L, top + 0, fn);
+                        setobj2s(L, top + 1, rb);
+                        setobj2s(L, top + 2, rc);
+                        L->top = top + 3;
+
+                        VM_PROTECT(luaV_callTM(L, 2, LUAU_INSN_A(insn)));
+                        VM_NEXT();
+                    }
+                    else
+                    {
+                        // slow-path, may invoke C/Lua via metamethods
+                        VM_PROTECT(luaV_doarith(L, ra, rb, rc, TM_IDIV));
+                        VM_NEXT();
+                    }
+                }
+            }
+
             VM_CASE(LOP_MOD)
             {
                 Instruction insn = *pc++;
@@ -1838,6 +1887,53 @@ reentry:
                 }
             }
 
+            VM_CASE(LOP_IDIVK)
+            {
+                Instruction insn = *pc++;
+                StkId ra = VM_REG(LUAU_INSN_A(insn));
+                StkId rb = VM_REG(LUAU_INSN_B(insn));
+                TValue* kv = VM_KV(LUAU_INSN_C(insn));
+
+                // fast-path
+                if (LUAU_LIKELY(ttisnumber(rb)))
+                {
+                    setnvalue(ra, luai_numidiv(nvalue(rb), nvalue(kv)));
+                    VM_NEXT();
+                }
+                else if (ttisvector(rb))
+                {
+                    const float* vb = vvalue(rb);
+                    float vc = cast_to(float, nvalue(kv));
+                    setvvalue(ra, float(luai_numidiv(vb[0], vc)), float(luai_numidiv(vb[1], vc)), float(luai_numidiv(vb[2], vc)),
+                        float(luai_numidiv(vb[3], vc)));
+                    VM_NEXT();
+                }
+                else
+                {
+                    // fast-path for userdata with C functions
+                    const TValue* fn = 0;
+                    if (ttisuserdata(rb) && (fn = luaT_gettmbyobj(L, rb, TM_IDIV)) && ttisfunction(fn) && clvalue(fn)->isC)
+                    {
+                        // note: it's safe to push arguments past top for complicated reasons (see top of the file)
+                        LUAU_ASSERT(L->top + 3 < L->stack + L->stacksize);
+                        StkId top = L->top;
+                        setobj2s(L, top + 0, fn);
+                        setobj2s(L, top + 1, rb);
+                        setobj2s(L, top + 2, kv);
+                        L->top = top + 3;
+
+                        VM_PROTECT(luaV_callTM(L, 2, LUAU_INSN_A(insn)));
+                        VM_NEXT();
+                    }
+                    else
+                    {
+                        // slow-path, may invoke C/Lua via metamethods
+                        VM_PROTECT(luaV_doarith(L, ra, rb, kv, TM_IDIV));
+                        VM_NEXT();
+                    }
+                }
+            }
+
             VM_CASE(LOP_MODK)
             {
                 Instruction insn = *pc++;

VM/src/lvmutils.cpp

@@ -394,6 +394,9 @@ void luaV_doarith(lua_State* L, StkId ra, const TValue* rb, const TValue* rc, TM
         case TM_DIV:
             setnvalue(ra, luai_numdiv(nb, nc));
             break;
+        case TM_IDIV:
+            setnvalue(ra, luai_numidiv(nb, nc));
+            break;
         case TM_MOD:
             setnvalue(ra, luai_nummod(nb, nc));
             break;
@@ -410,7 +413,12 @@ void luaV_doarith(lua_State* L, StkId ra, const TValue* rb, const TValue* rc, TM
     }
     else
     {
-        // vector operations that we support: v + v, v - v, v * v, s * v, v * s, v / v, s / v, v / s, -v
+        // vector operations that we support:
+        // v+v  v-v  -v    (add/sub/neg)
+        // v*v  s*v  v*s   (mul)
+        // v/v  s/v  v/s   (div)
+        // v//v s//v v//s  (floor div)
+
         const float* vb = luaV_tovector(rb);
         const float* vc = luaV_tovector(rc);
 
@@ -430,6 +438,10 @@ void luaV_doarith(lua_State* L, StkId ra, const TValue* rb, const TValue* rc, TM
             case TM_DIV:
                 setvvalue(ra, vb[0] / vc[0], vb[1] / vc[1], vb[2] / vc[2], vb[3] / vc[3]);
                 return;
+            case TM_IDIV:
+                setvvalue(ra, float(luai_numidiv(vb[0], vc[0])), float(luai_numidiv(vb[1], vc[1])), float(luai_numidiv(vb[2], vc[2])),
+                    float(luai_numidiv(vb[3], vc[3])));
+                return;
             case TM_UNM:
                 setvvalue(ra, -vb[0], -vb[1], -vb[2], -vb[3]);
                 return;
@@ -453,6 +465,10 @@ void luaV_doarith(lua_State* L, StkId ra, const TValue* rb, const TValue* rc, TM
                 case TM_DIV:
                     setvvalue(ra, vb[0] / nc, vb[1] / nc, vb[2] / nc, vb[3] / nc);
                     return;
+                case TM_IDIV:
+                    setvvalue(ra, float(luai_numidiv(vb[0], nc)), float(luai_numidiv(vb[1], nc)), float(luai_numidiv(vb[2], nc)),
+                        float(luai_numidiv(vb[3], nc)));
+                    return;
                 default:
                     break;
                 }
@@ -474,6 +490,10 @@ void luaV_doarith(lua_State* L, StkId ra, const TValue* rb, const TValue* rc, TM
                 case TM_DIV:
                     setvvalue(ra, nb / vc[0], nb / vc[1], nb / vc[2], nb / vc[3]);
                     return;
+                case TM_IDIV:
+                    setvvalue(ra, float(luai_numidiv(nb, vc[0])), float(luai_numidiv(nb, vc[1])), float(luai_numidiv(nb, vc[2])),
+                        float(luai_numidiv(nb, vc[3])));
+                    return;
                 default:
                     break;
                 }

bench/micro_tests/test_TableSort.lua

@@ -0,0 +1,22 @@
+local bench = script and require(script.Parent.bench_support) or require("bench_support")
+
+local arr_months = {"Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"}
+
+local arr_num = {}
+for i=1,100 do table.insert(arr_num, math.sin(i)) end
+
+local arr_numk = {}
+for i=1,10000 do table.insert(arr_numk, math.sin(i)) end
+
+function test(arr)
+    local t = table.create(#arr)
+
+    for i=1,1e6/#arr do
+        table.move(arr, 1, #arr, 1, t)
+        table.sort(t)
+    end
+end
+
+bench.runCode(function() test(arr_months) end, "table.sort: 12 strings")
+bench.runCode(function() test(arr_num) end, "table.sort: 100 numbers")
+bench.runCode(function() test(arr_numk) end, "table.sort: 10k numbers")

bench/tabulate.py

@@ -45,7 +45,7 @@ class TablePrinter(object):
     def _print_horizontal_separator(self):
         for i, align_width in enumerate(self._widths):
             if i > 0:
-                print('-+-', end='')
+                print('-|-', end='')
             print('-' * (align_width+1), end='')
         print()
         pass

fuzz/luau.proto

@@ -135,17 +135,18 @@ message ExprBinary {
         Sub = 1;
         Mul = 2;
         Div = 3;
-        Mod = 4;
-        Pow = 5;
-        Concat = 6;
-        CompareNe = 7;
-        CompareEq = 8;
-        CompareLt = 9;
-        CompareLe = 10;
-        CompareGt = 11;
-        CompareGe = 12;
-        And = 13;
-        Or = 14;
+        FloorDiv = 4;
+        Mod = 5;
+        Pow = 6;
+        Concat = 7;
+        CompareNe = 8;
+        CompareEq = 9;
+        CompareLt = 10;
+        CompareLe = 11;
+        CompareGt = 12;
+        CompareGe = 13;
+        And = 14;
+        Or = 15;
 	}
 
 	required Op op = 1;

fuzz/protoprint.cpp

@@ -495,6 +495,8 @@ struct ProtoToLuau
             source += " * ";
         else if (expr.op() == luau::ExprBinary::Div)
             source += " / ";
+        else if (expr.op() == luau::ExprBinary::FloorDiv)
+            source += " // ";
         else if (expr.op() == luau::ExprBinary::Mod)
             source += " % ";
         else if (expr.op() == luau::ExprBinary::Pow)

tests/AssemblyBuilderA64.test.cpp

@@ -107,6 +107,13 @@ TEST_CASE_FIXTURE(AssemblyBuilderA64Fixture, "Binary")
     SINGLE_COMPARE(cmp(w0, 42), 0x7100A81F);
 }
 
+TEST_CASE_FIXTURE(AssemblyBuilderA64Fixture, "BinaryExtended")
+{
+    // reg, reg
+    SINGLE_COMPARE(add(x0, x1, w2, 3), 0x8B224C20);
+    SINGLE_COMPARE(sub(x0, x1, w2, 3), 0xCB224C20);
+}
+
 TEST_CASE_FIXTURE(AssemblyBuilderA64Fixture, "BinaryImm")
 {
     // instructions
@@ -524,6 +531,8 @@ TEST_CASE("LogTest")
     build.ldr(x0, mem(x1, 1, AddressKindA64::pre));
     build.ldr(x0, mem(x1, 1, AddressKindA64::post));
 
+    build.add(x1, x2, w3, 3);
+
     build.setLabel(l);
     build.ret();
 
@@ -560,6 +569,7 @@ TEST_CASE("LogTest")
  ldr         x0,[x1,#1]
  ldr         x0,[x1,#1]!
  ldr         x0,[x1]!,#1
+ add         x1,x2,w3 UXTW #3
 .L1:
  ret
 )";

tests/Autocomplete.test.cpp

@@ -986,6 +986,33 @@ TEST_CASE_FIXTURE(ACFixture, "autocomplete_end_with_lambda")
     CHECK_EQ(ac.context, AutocompleteContext::Statement);
 }
 
+TEST_CASE_FIXTURE(ACFixture, "autocomplete_end_of_do_block")
+{
+    ScopedFastFlag sff{"LuauAutocompleteDoEnd", true};
+
+    check("do @1");
+
+    auto ac = autocomplete('1');
+
+    CHECK(ac.entryMap.count("end"));
+
+    check(R"(
+        function f()
+            do
+                @1
+        end
+        @2
+    )");
+
+    ac = autocomplete('1');
+
+    CHECK(ac.entryMap.count("end"));
+
+    ac = autocomplete('2');
+
+    CHECK(ac.entryMap.count("end"));
+}
+
 TEST_CASE_FIXTURE(ACFixture, "stop_at_first_stat_when_recommending_keywords")
 {
     check(R"(

tests/Conformance.test.cpp

@@ -24,6 +24,7 @@ extern bool codegen;
 extern int optimizationLevel;
 
 LUAU_FASTFLAG(LuauPCallDebuggerFix);
+LUAU_FASTFLAG(LuauFloorDivision);
 
 static lua_CompileOptions defaultOptions()
 {
@@ -280,6 +281,7 @@ TEST_CASE("Assert")
 
 TEST_CASE("Basic")
 {
+    ScopedFastFlag sffs{"LuauFloorDivision", true};
     runConformance("basic.lua");
 }
 
@@ -363,6 +365,7 @@ TEST_CASE("Errors")
 
 TEST_CASE("Events")
 {
+    ScopedFastFlag sffs{"LuauFloorDivision", true};
     runConformance("events.lua");
 }
 
@@ -444,6 +447,8 @@ TEST_CASE("Pack")
 
 TEST_CASE("Vector")
 {
+    ScopedFastFlag sffs{"LuauFloorDivision", true};
+
     lua_CompileOptions copts = defaultOptions();
     copts.vectorCtor = "vector";
 
@@ -1616,6 +1621,9 @@ static void pushInt64(lua_State* L, int64_t value)
 
 TEST_CASE("Userdata")
 {
+
+    ScopedFastFlag sffs{"LuauFloorDivision", true};
+
     runConformance("userdata.lua", [](lua_State* L) {
         // create metatable with all the metamethods
         lua_newtable(L);
@@ -1735,6 +1743,19 @@ TEST_CASE("Userdata")
             nullptr);
         lua_setfield(L, -2, "__div");
 
+        // __idiv
+        lua_pushcfunction(
+            L,
+            [](lua_State* L) {
+                // for testing we use different semantics here compared to __div: __idiv rounds to negative inf, __div truncates (rounds to zero)
+                // additionally, division loses precision here outside of 2^53 range
+                // we do not necessarily recommend this behavior in production code!
+                pushInt64(L, int64_t(floor(double(getInt64(L, 1)) / double(getInt64(L, 2)))));
+                return 1;
+            },
+            nullptr);
+        lua_setfield(L, -2, "__idiv");
+
         // __mod
         lua_pushcfunction(
             L,

tests/IrBuilder.test.cpp

@@ -5,6 +5,7 @@
 #include "Luau/IrUtils.h"
 #include "Luau/OptimizeConstProp.h"
 #include "Luau/OptimizeFinalX64.h"
+#include "ScopedFlags.h"
 
 #include "doctest.h"
 
@@ -1930,6 +1931,135 @@ bb_0:
 )");
 }
 
+TEST_CASE_FIXTURE(IrBuilderFixture, "DuplicateHashSlotChecks")
+{
+    ScopedFastFlag luauReuseHashSlots{"LuauReuseHashSlots2", true};
+
+    IrOp block = build.block(IrBlockKind::Internal);
+    IrOp fallback = build.block(IrBlockKind::Fallback);
+
+    build.beginBlock(block);
+
+    // This roughly corresponds to 'return t.a + t.a'
+    IrOp table1 = build.inst(IrCmd::LOAD_POINTER, build.vmReg(1));
+    IrOp slot1 = build.inst(IrCmd::GET_SLOT_NODE_ADDR, table1, build.constUint(3), build.vmConst(1));
+    build.inst(IrCmd::CHECK_SLOT_MATCH, slot1, build.vmConst(1), fallback);
+    IrOp value1 = build.inst(IrCmd::LOAD_TVALUE, slot1, build.constInt(0));
+    build.inst(IrCmd::STORE_TVALUE, build.vmReg(3), value1);
+
+    IrOp slot1b = build.inst(IrCmd::GET_SLOT_NODE_ADDR, table1, build.constUint(8), build.vmConst(1)); // This will be removed
+    build.inst(IrCmd::CHECK_SLOT_MATCH, slot1b, build.vmConst(1), fallback);                           // Key will be replaced with undef here
+    IrOp value1b = build.inst(IrCmd::LOAD_TVALUE, slot1b, build.constInt(0));
+    build.inst(IrCmd::STORE_TVALUE, build.vmReg(4), value1b);
+
+    IrOp a = build.inst(IrCmd::LOAD_DOUBLE, build.vmReg(3));
+    IrOp b = build.inst(IrCmd::LOAD_DOUBLE, build.vmReg(4));
+    IrOp sum = build.inst(IrCmd::ADD_NUM, a, b);
+    build.inst(IrCmd::STORE_DOUBLE, build.vmReg(2), sum);
+
+    build.inst(IrCmd::RETURN, build.vmReg(2), build.constUint(1));
+
+    build.beginBlock(fallback);
+    build.inst(IrCmd::RETURN, build.vmReg(0), build.constUint(1));
+
+    updateUseCounts(build.function);
+    constPropInBlockChains(build, true);
+
+    // In the future, we might even see duplicate identical TValue loads go away
+    // In the future, we might even see loads of different VM regs with the same value go away
+    CHECK("\n" + toString(build.function, /* includeUseInfo */ false) == R"(
+bb_0:
+   %0 = LOAD_POINTER R1
+   %1 = GET_SLOT_NODE_ADDR %0, 3u, K1
+   CHECK_SLOT_MATCH %1, K1, bb_fallback_1
+   %3 = LOAD_TVALUE %1, 0i
+   STORE_TVALUE R3, %3
+   CHECK_NODE_VALUE %1, bb_fallback_1
+   %7 = LOAD_TVALUE %1, 0i
+   STORE_TVALUE R4, %7
+   %9 = LOAD_DOUBLE R3
+   %10 = LOAD_DOUBLE R4
+   %11 = ADD_NUM %9, %10
+   STORE_DOUBLE R2, %11
+   RETURN R2, 1u
+
+bb_fallback_1:
+   RETURN R0, 1u
+
+)");
+}
+
+TEST_CASE_FIXTURE(IrBuilderFixture, "DuplicateHashSlotChecksAvoidNil")
+{
+    ScopedFastFlag luauReuseHashSlots{"LuauReuseHashSlots2", true};
+
+    IrOp block = build.block(IrBlockKind::Internal);
+    IrOp fallback = build.block(IrBlockKind::Fallback);
+
+    build.beginBlock(block);
+
+    IrOp table1 = build.inst(IrCmd::LOAD_POINTER, build.vmReg(1));
+    IrOp slot1 = build.inst(IrCmd::GET_SLOT_NODE_ADDR, table1, build.constUint(3), build.vmConst(1));
+    build.inst(IrCmd::CHECK_SLOT_MATCH, slot1, build.vmConst(1), fallback);
+    IrOp value1 = build.inst(IrCmd::LOAD_TVALUE, slot1, build.constInt(0));
+    build.inst(IrCmd::STORE_TVALUE, build.vmReg(3), value1);
+
+    IrOp table2 = build.inst(IrCmd::LOAD_POINTER, build.vmReg(2));
+    IrOp slot2 = build.inst(IrCmd::GET_SLOT_NODE_ADDR, table2, build.constUint(6), build.vmConst(1));
+    build.inst(IrCmd::CHECK_SLOT_MATCH, slot2, build.vmConst(1), fallback);
+    build.inst(IrCmd::CHECK_READONLY, table2, fallback);
+
+    build.inst(IrCmd::STORE_TAG, build.vmReg(4), build.constTag(tnil));
+    IrOp valueNil = build.inst(IrCmd::LOAD_TVALUE, build.vmReg(4));
+    build.inst(IrCmd::STORE_TVALUE, slot2, valueNil, build.constInt(0));
+
+    // In the future, we might get to track that value became 'nil' and that fallback will be taken
+    IrOp slot1b = build.inst(IrCmd::GET_SLOT_NODE_ADDR, table1, build.constUint(8), build.vmConst(1)); // This will be removed
+    build.inst(IrCmd::CHECK_SLOT_MATCH, slot1b, build.vmConst(1), fallback);                           // Key will be replaced with undef here
+    IrOp value1b = build.inst(IrCmd::LOAD_TVALUE, slot1b, build.constInt(0));
+    build.inst(IrCmd::STORE_TVALUE, build.vmReg(3), value1b);
+
+    IrOp slot2b = build.inst(IrCmd::GET_SLOT_NODE_ADDR, table2, build.constUint(11), build.vmConst(1)); // This will be removed
+    build.inst(IrCmd::CHECK_SLOT_MATCH, slot2b, build.vmConst(1), fallback);                            // Key will be replaced with undef here
+    build.inst(IrCmd::CHECK_READONLY, table2, fallback);
+
+    build.inst(IrCmd::STORE_SPLIT_TVALUE, slot2b, build.constTag(tnumber), build.constDouble(1), build.constInt(0));
+
+    build.inst(IrCmd::RETURN, build.vmReg(3), build.constUint(2));
+
+    build.beginBlock(fallback);
+    build.inst(IrCmd::RETURN, build.vmReg(1), build.constUint(2));
+
+    updateUseCounts(build.function);
+    constPropInBlockChains(build, true);
+
+    CHECK("\n" + toString(build.function, /* includeUseInfo */ false) == R"(
+bb_0:
+   %0 = LOAD_POINTER R1
+   %1 = GET_SLOT_NODE_ADDR %0, 3u, K1
+   CHECK_SLOT_MATCH %1, K1, bb_fallback_1
+   %3 = LOAD_TVALUE %1, 0i
+   STORE_TVALUE R3, %3
+   %5 = LOAD_POINTER R2
+   %6 = GET_SLOT_NODE_ADDR %5, 6u, K1
+   CHECK_SLOT_MATCH %6, K1, bb_fallback_1
+   CHECK_READONLY %5, bb_fallback_1
+   STORE_TAG R4, tnil
+   %10 = LOAD_TVALUE R4
+   STORE_TVALUE %6, %10, 0i
+   CHECK_NODE_VALUE %1, bb_fallback_1
+   %14 = LOAD_TVALUE %1, 0i
+   STORE_TVALUE R3, %14
+   CHECK_NODE_VALUE %6, bb_fallback_1
+   STORE_SPLIT_TVALUE %6, tnumber, 1, 0i
+   RETURN R3, 2u
+
+bb_fallback_1:
+   RETURN R1, 2u
+
+)");
+}
+
 TEST_SUITE_END();
 
 TEST_SUITE_BEGIN("Analysis");

tests/Parser.test.cpp

@@ -2970,4 +2970,40 @@ TEST_CASE_FIXTURE(Fixture, "unfinished_string_literal_types_get_reported_but_par
     CHECK_EQ(result.root->body.size, 2);
 }
 
+TEST_CASE_FIXTURE(Fixture, "do_block_with_no_end")
+{
+    ParseResult result = tryParse(R"(
+        do
+    )");
+
+    REQUIRE_EQ(1, result.errors.size());
+
+    AstStatBlock* stat0 = result.root->body.data[0]->as<AstStatBlock>();
+    REQUIRE(stat0);
+
+    CHECK(!stat0->hasEnd);
+}
+
+TEST_CASE_FIXTURE(Fixture, "parse_interpolated_string_with_lookahead_involved")
+{
+    ScopedFastFlag sff{"LuauLexerLookaheadRemembersBraceType", true};
+
+    ParseResult result = tryParse(R"(
+        local x = `{ {y} }`
+    )");
+
+    REQUIRE_MESSAGE(result.errors.empty(), result.errors[0].getMessage());
+}
+
+TEST_CASE_FIXTURE(Fixture, "parse_interpolated_string_with_lookahead_involved2")
+{
+    ScopedFastFlag sff{"LuauLexerLookaheadRemembersBraceType", true};
+
+    ParseResult result = tryParse(R"(
+        local x = `{ { y{} } }`
+    )");
+
+    REQUIRE_MESSAGE(result.errors.empty(), result.errors[0].getMessage());
+}
+
 TEST_SUITE_END();

tests/Subtyping.test.cpp

@@ -52,6 +52,35 @@ struct SubtypeFixture : Fixture
         return arena.addType(TableType{std::move(props), std::nullopt, {}, TableState::Sealed});
     }
 
+    // `&`
+    TypeId meet(TypeId a, TypeId b)
+    {
+        return arena.addType(IntersectionType{{a, b}});
+    }
+
+    // `|`
+    TypeId join(TypeId a, TypeId b)
+    {
+        return arena.addType(UnionType{{a, b}});
+    }
+
+    TypeId negate(TypeId ty)
+    {
+        return arena.addType(NegationType{ty});
+    }
+
+    TypeId cls(const std::string& name, std::optional<TypeId> parent = std::nullopt)
+    {
+        return arena.addType(ClassType{name, {}, parent.value_or(builtinTypes->classType), {}, {}, nullptr, ""});
+    }
+
+    TypeId cls(const std::string& name, ClassType::Props&& props)
+    {
+        TypeId ty = cls(name);
+        getMutable<ClassType>(ty)->props = std::move(props);
+        return ty;
+    }
+
     TypeId cyclicTable(std::function<void(TypeId, TableType*)>&& cb)
     {
         TypeId res = arena.addType(GenericType{});
@@ -61,6 +90,11 @@ struct SubtypeFixture : Fixture
         return res;
     }
 
+    TypeId meta(TableType::Props&& metaProps, TableType::Props&& tableProps = {})
+    {
+        return arena.addType(MetatableType{tbl(std::move(tableProps)), tbl(std::move(metaProps))});
+    }
+
     TypeId genericT = arena.addType(GenericType{"T"});
     TypeId genericU = arena.addType(GenericType{"U"});
 
@@ -77,8 +111,34 @@ struct SubtypeFixture : Fixture
     TypeId helloType2 = arena.addType(SingletonType{StringSingleton{"hello"}});
     TypeId worldType = arena.addType(SingletonType{StringSingleton{"world"}});
 
-    TypeId helloOrWorldType = arena.addType(UnionType{{helloType, worldType}});
-    TypeId trueOrFalseType = arena.addType(UnionType{{builtinTypes->trueType, builtinTypes->falseType}});
+    TypeId helloOrWorldType = join(helloType, worldType);
+    TypeId trueOrFalseType = join(builtinTypes->trueType, builtinTypes->falseType);
+
+    TypeId helloAndWorldType = meet(helloType, worldType);
+    TypeId booleanAndTrueType = meet(builtinTypes->booleanType, builtinTypes->trueType);
+
+    /**
+     * class
+     * \- Root
+     *    |- Child
+     *    |  |-GrandchildOne
+     *    |  \-GrandchildTwo
+     *    \- AnotherChild
+     *       |- AnotherGrandchildOne
+     *       \- AnotherGrandchildTwo
+    */
+    TypeId rootClass = cls("Root");
+    TypeId childClass = cls("Child", rootClass);
+    TypeId grandchildOneClass = cls("GrandchildOne", childClass);
+    TypeId grandchildTwoClass = cls("GrandchildTwo", childClass);
+    TypeId anotherChildClass = cls("AnotherChild", rootClass);
+    TypeId anotherGrandchildOneClass = cls("AnotherGrandchildOne", anotherChildClass);
+    TypeId anotherGrandchildTwoClass = cls("AnotherGrandchildTwo", anotherChildClass);
+
+    TypeId vec2Class = cls("Vec2", {
+        {"X", builtinTypes->numberType},
+        {"Y", builtinTypes->numberType},
+    });
 
     // "hello" | "hello"
     TypeId helloOrHelloType = arena.addType(UnionType{{helloType, helloType}});
@@ -86,12 +146,6 @@ struct SubtypeFixture : Fixture
     // () -> ()
     const TypeId nothingToNothingType = fn({}, {});
 
-    // ("hello") -> "world"
-    TypeId helloAndWorldType = arena.addType(IntersectionType{{helloType, worldType}});
-
-    // (boolean) -> true
-    TypeId booleanAndTrueType = arena.addType(IntersectionType{{builtinTypes->booleanType, builtinTypes->trueType}});
-
     // (number) -> string
     const TypeId numberToStringType = fn(
         {builtinTypes->numberType},
@@ -247,6 +301,11 @@ struct SubtypeFixture : Fixture
         builtinTypes->emptyTypePack,
         genericAs
     });
+
+    // { lower : string -> string }
+    TypeId tableWithLower = tbl(TableType::Props{{"lower", fn({builtinTypes->stringType}, {builtinTypes->stringType})}});
+    // { insaneThingNoScalarHas : () -> () }
+    TypeId tableWithoutScalarProp = tbl(TableType::Props{{"insaneThingNoScalarHas", fn({}, {})}});
 };
 
 #define CHECK_IS_SUBTYPE(left, right) \
@@ -620,6 +679,99 @@ TEST_CASE_FIXTURE(SubtypeFixture, "{x: <T>(T) -> ()} <: {x: <U>(U) -> ()}")
     );
 }
 
+TEST_CASE_FIXTURE(SubtypeFixture, "{ @metatable { x: number } } <: { @metatable {} }")
+{
+    CHECK_IS_SUBTYPE(
+        meta({{"x", builtinTypes->numberType}}),
+        meta({})
+    );
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "{ @metatable { x: number } } <!: { @metatable { x: boolean } }")
+{
+    CHECK_IS_NOT_SUBTYPE(
+        meta({{"x", builtinTypes->numberType}}),
+        meta({{"x", builtinTypes->booleanType}})
+    );
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "{ @metatable {} } <!: { @metatable { x: boolean } }")
+{
+    CHECK_IS_NOT_SUBTYPE(
+        meta({}),
+        meta({{"x", builtinTypes->booleanType}})
+    );
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "{ @metatable {} } <: {}")
+{
+    CHECK_IS_SUBTYPE(
+        meta({}),
+        tbl({})
+    );
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "{ @metatable { u: boolean }, x: number } <: { x: number }")
+{
+    CHECK_IS_SUBTYPE(
+        meta({{"u", builtinTypes->booleanType}}, {{"x", builtinTypes->numberType}}),
+        tbl({{"x", builtinTypes->numberType}})
+    );
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "{ @metatable { x: number } } <!: { x: number }")
+{
+    CHECK_IS_NOT_SUBTYPE(
+        meta({{"x", builtinTypes->numberType}}),
+        tbl({{"x", builtinTypes->numberType}})
+    );
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "Root <: class")
+{
+    CHECK_IS_SUBTYPE(rootClass, builtinTypes->classType);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "Child | AnotherChild <: class")
+{
+    CHECK_IS_SUBTYPE(join(childClass, anotherChildClass), builtinTypes->classType);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "Child | AnotherChild <: Child | AnotherChild")
+{
+    CHECK_IS_SUBTYPE(join(childClass, anotherChildClass), join(childClass, anotherChildClass));
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "Child | Root <: Root")
+{
+    CHECK_IS_SUBTYPE(join(childClass, rootClass), rootClass);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "Child & AnotherChild <: class")
+{
+    CHECK_IS_SUBTYPE(meet(childClass, anotherChildClass), builtinTypes->classType);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "Child & Root <: class")
+{
+    CHECK_IS_SUBTYPE(meet(childClass, rootClass), builtinTypes->classType);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "Child & ~Root <: class")
+{
+    CHECK_IS_SUBTYPE(meet(childClass, negate(rootClass)), builtinTypes->classType);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "Child & AnotherChild <: number")
+{
+    CHECK_IS_SUBTYPE(meet(childClass, anotherChildClass), builtinTypes->numberType);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "Child & ~GrandchildOne <!: number")
+{
+    CHECK_IS_NOT_SUBTYPE(meet(childClass, negate(grandchildOneClass)), builtinTypes->numberType);
+}
+
 TEST_CASE_FIXTURE(SubtypeFixture, "t1 where t1 = {trim: (t1) -> string} <: t2 where t2 = {trim: (t2) -> string}")
 {
     TypeId t1 = cyclicTable([&](TypeId ty, TableType* tt)
@@ -665,6 +817,84 @@ TEST_CASE_FIXTURE(SubtypeFixture, "t1 where t1 = {trim: (t1) -> t1} <!: t2 where
     CHECK_IS_NOT_SUBTYPE(t1, t2);
 }
 
+TEST_CASE_FIXTURE(SubtypeFixture, "Vec2 <: { X: number, Y: number }")
+{
+    TypeId xy = tbl({
+        {"X", builtinTypes->numberType},
+        {"Y", builtinTypes->numberType},
+    });
+
+    CHECK_IS_SUBTYPE(vec2Class, xy);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "Vec2 <: { X: number }")
+{
+    TypeId x = tbl({
+        {"X", builtinTypes->numberType},
+    });
+
+    CHECK_IS_SUBTYPE(vec2Class, x);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "{ X: number, Y: number } <!: Vec2")
+{
+    TypeId xy = tbl({
+        {"X", builtinTypes->numberType},
+        {"Y", builtinTypes->numberType},
+    });
+
+    CHECK_IS_NOT_SUBTYPE(xy, vec2Class);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "{ X: number } <!: Vec2")
+{
+    TypeId x = tbl({
+        {"X", builtinTypes->numberType},
+    });
+
+    CHECK_IS_NOT_SUBTYPE(x, vec2Class);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "table & { X: number, Y: number } <!: Vec2")
+{
+    TypeId x = tbl({
+        {"X", builtinTypes->numberType},
+        {"Y", builtinTypes->numberType},
+    });
+
+    CHECK_IS_NOT_SUBTYPE(meet(builtinTypes->tableType, x), vec2Class);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "Vec2 <!: table & { X: number, Y: number }")
+{
+    TypeId xy = tbl({
+        {"X", builtinTypes->numberType},
+        {"Y", builtinTypes->numberType},
+    });
+
+    CHECK_IS_NOT_SUBTYPE(vec2Class, meet(builtinTypes->tableType, xy));
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "\"hello\" <: { lower : (string) -> string }")
+{
+    CHECK_IS_SUBTYPE(helloType, tableWithLower);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "\"hello\" <!: { insaneThingNoScalarHas : () -> () }")
+{
+    CHECK_IS_NOT_SUBTYPE(helloType, tableWithoutScalarProp);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "string <: { lower : (string) -> string }")
+{
+    CHECK_IS_SUBTYPE(builtinTypes->stringType, tableWithLower);
+}
+
+TEST_CASE_FIXTURE(SubtypeFixture, "string <!: { insaneThingNoScalarHas : () -> () }")
+{
+    CHECK_IS_NOT_SUBTYPE(builtinTypes->stringType, tableWithoutScalarProp);
+}
+
 /*
  * <A>(A) -> A <: <X>(X) -> X
  *      A can be bound to X.

tests/ToString.test.cpp

@@ -833,14 +833,6 @@ TEST_CASE_FIXTURE(Fixture, "tostring_unsee_ttv_if_array")
 
 TEST_CASE_FIXTURE(Fixture, "tostring_error_mismatch")
 {
-    ScopedFastFlag sff[] = {
-        {"LuauIndentTypeMismatch", true},
-    };
-
-    ScopedFastInt sfi[] = {
-        {"LuauIndentTypeMismatchMaxTypeLength", 10},
-    };
-
     CheckResult result = check(R"(
 --!strict
    function f1() : {a : number, b : string, c : { d : number}}

tests/Transpiler.test.cpp

@@ -529,14 +529,17 @@ until c
     CHECK_EQ(code, transpile(code, {}, true).code);
 }
 
-TEST_CASE_FIXTURE(Fixture, "transpile_compound_assignmenr")
+TEST_CASE_FIXTURE(Fixture, "transpile_compound_assignment")
 {
+    ScopedFastFlag sffs{"LuauFloorDivision", true};
+
     std::string code = R"(
 local a = 1
 a += 2
 a -= 3
 a *= 4
 a /= 5
+a //= 5
 a %= 6
 a ^= 7
 a ..= ' - result'

tests/TypeInfer.aliases.test.cpp

@@ -189,9 +189,7 @@ TEST_CASE_FIXTURE(Fixture, "generic_aliases")
 {
     ScopedFastFlag sff[] = {
         {"DebugLuauDeferredConstraintResolution", true},
-        {"LuauIndentTypeMismatch", true},
     };
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
     CheckResult result = check(R"(
         type T<a> = { v: a }
         local x: T<number> = { v = 123 }
@@ -212,10 +210,7 @@ TEST_CASE_FIXTURE(Fixture, "dependent_generic_aliases")
 {
     ScopedFastFlag sff[] = {
         {"DebugLuauDeferredConstraintResolution", true},
-        {"LuauIndentTypeMismatch", true},
     };
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
 
     CheckResult result = check(R"(
         type T<a> = { v: a }

tests/TypeInfer.builtins.test.cpp

@@ -134,9 +134,7 @@ TEST_CASE_FIXTURE(BuiltinsFixture, "sort_with_bad_predicate")
 {
     ScopedFastFlag sff[] = {
         {"LuauAlwaysCommitInferencesOfFunctionCalls", true},
-        {"LuauIndentTypeMismatch", true},
     };
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
 
     CheckResult result = check(R"(
         --!strict

tests/TypeInfer.classes.test.cpp

@@ -369,9 +369,7 @@ TEST_CASE_FIXTURE(ClassFixture, "detailed_class_unification_error")
 {
     ScopedFastFlag sff[] = {
         {"LuauAlwaysCommitInferencesOfFunctionCalls", true},
-        {"LuauIndentTypeMismatch", true},
     };
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
     CheckResult result = check(R"(
 local function foo(v)
     return v.X :: number + string.len(v.Y)
@@ -457,8 +455,6 @@ TEST_CASE_FIXTURE(ClassFixture, "index_instance_property_nonstrict")
 
 TEST_CASE_FIXTURE(ClassFixture, "type_mismatch_invariance_required_for_error")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
     CheckResult result = check(R"(
 type A = { x: ChildClass }
 type B = { x: BaseClass }

tests/TypeInfer.functions.test.cpp

@@ -1095,9 +1095,6 @@ TEST_CASE_FIXTURE(Fixture, "return_type_by_overload")
 
 TEST_CASE_FIXTURE(BuiltinsFixture, "infer_anonymous_function_arguments")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     // Simple direct arg to arg propagation
     CheckResult result = check(R"(
 type Table = { x: number, y: number }
@@ -1342,8 +1339,6 @@ end
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_function_mismatch_arg_count")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
     CheckResult result = check(R"(
 type A = (number, number) -> string
 type B = (number) -> string
@@ -1364,9 +1359,6 @@ caused by:
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_function_mismatch_arg")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 type A = (number, number) -> string
 type B = (number, string) -> string
@@ -1388,9 +1380,6 @@ Type 'string' could not be converted into 'number')";
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_function_mismatch_ret_count")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 type A = (number, number) -> (number)
 type B = (number, number) -> (number, boolean)
@@ -1411,9 +1400,6 @@ caused by:
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_function_mismatch_ret")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 type A = (number, number) -> string
 type B = (number, number) -> number
@@ -1435,9 +1421,6 @@ Type 'string' could not be converted into 'number')";
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_function_mismatch_ret_mult")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 type A = (number, number) -> (number, string)
 type B = (number, number) -> (number, boolean)
@@ -1563,9 +1546,6 @@ TEST_CASE_FIXTURE(Fixture, "inferred_higher_order_functions_are_quantified_at_th
 
 TEST_CASE_FIXTURE(BuiltinsFixture, "function_decl_non_self_unsealed_overwrite")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 local t = { f = nil :: ((x: number) -> number)? }
 
@@ -1608,9 +1588,6 @@ TEST_CASE_FIXTURE(Fixture, "strict_mode_ok_with_missing_arguments")
 
 TEST_CASE_FIXTURE(Fixture, "function_statement_sealed_table_assignment_through_indexer")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 local t: {[string]: () -> number} = {}
 
@@ -1819,9 +1796,6 @@ foo(string.find("hello", "e"))
 
 TEST_CASE_FIXTURE(Fixture, "luau_subtyping_is_np_hard")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 --!strict
 
@@ -2017,10 +1991,8 @@ TEST_CASE_FIXTURE(Fixture, "function_exprs_are_generalized_at_signature_scope_no
 TEST_CASE_FIXTURE(BuiltinsFixture, "param_1_and_2_both_takes_the_same_generic_but_their_arguments_are_incompatible")
 {
     ScopedFastFlag sff[] = {
-        {"LuauIndentTypeMismatch", true},
         {"LuauAlwaysCommitInferencesOfFunctionCalls", true},
     };
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
 
     CheckResult result = check(R"(
         local function foo<a>(x: a, y: a?)

tests/TypeInfer.generics.test.cpp

@@ -713,9 +713,6 @@ end
 
 TEST_CASE_FIXTURE(Fixture, "generic_functions_should_be_memory_safe")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 --!strict
 -- At one point this produced a UAF

tests/TypeInfer.intersectionTypes.test.cpp

@@ -317,9 +317,6 @@ TEST_CASE_FIXTURE(Fixture, "table_intersection_write_sealed")
 
 TEST_CASE_FIXTURE(Fixture, "table_intersection_write_sealed_indirect")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
         type X = { x: (number) -> number }
         type Y = { y: (string) -> string }
@@ -350,9 +347,6 @@ caused by:
 
 TEST_CASE_FIXTURE(Fixture, "table_write_sealed_indirect")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     // After normalization, previous 'table_intersection_write_sealed_indirect' is identical to this one
     CheckResult result = check(R"(
     type XY = { x: (number) -> number, y: (string) -> string }
@@ -392,9 +386,6 @@ TEST_CASE_FIXTURE(BuiltinsFixture, "table_intersection_setmetatable")
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_intersection_part")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 type X = { x: number }
 type Y = { y: number }
@@ -482,9 +473,6 @@ TEST_CASE_FIXTURE(Fixture, "intersect_false_and_bool_and_false")
 
 TEST_CASE_FIXTURE(Fixture, "intersect_saturate_overloaded_functions")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
         local x : ((number?) -> number?) & ((string?) -> string?)
         local y : (nil) -> nil = x -- OK
@@ -501,8 +489,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "union_saturate_overloaded_functions")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
 
     CheckResult result = check(R"(
         local x : ((number) -> number) & ((string) -> string)
@@ -520,9 +506,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "intersection_of_tables")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
         local x : { p : number?, q : string? } & { p : number?, q : number?, r : number? }
         local y : { p : number?, q : nil, r : number? } = x -- OK
@@ -539,9 +522,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "intersection_of_tables_with_top_properties")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
         local x : { p : number?, q : any } & { p : unknown, q : string? }
         local y : { p : number?, q : string? } = x -- OK
@@ -594,9 +574,6 @@ TEST_CASE_FIXTURE(Fixture, "intersection_of_tables_with_never_properties")
 
 TEST_CASE_FIXTURE(Fixture, "overloaded_functions_returning_intersections")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
         local x : ((number?) -> ({ p : number } & { q : number })) & ((string?) -> ({ p : number } & { r : number }))
         local y : (nil) -> { p : number, q : number, r : number} = x -- OK
@@ -613,9 +590,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "overloaded_functions_mentioning_generic")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
       function f<a>()
         local x : ((number?) -> (a | number)) & ((string?) -> (a | string))
@@ -634,9 +608,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "overloaded_functions_mentioning_generics")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
       function f<a,b,c>()
         local x : ((a?) -> (a | b)) & ((c?) -> (b | c))
@@ -655,9 +626,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "overloaded_functions_mentioning_generic_packs")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
       function f<a...,b...>()
         local x : ((number?, a...) -> (number?, b...)) & ((string?, a...) -> (string?, b...))
@@ -676,9 +644,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "overloadeded_functions_with_unknown_result")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
       function f<a...,b...>()
         local x : ((number) -> number) & ((nil) -> unknown)
@@ -697,9 +662,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "overloadeded_functions_with_unknown_arguments")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
       function f<a...,b...>()
         local x : ((number) -> number?) & ((unknown) -> string?)
@@ -718,9 +680,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "overloadeded_functions_with_never_result")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
       function f<a...,b...>()
         local x : ((number) -> number) & ((nil) -> never)
@@ -739,9 +698,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "overloadeded_functions_with_never_arguments")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
       function f<a...,b...>()
         local x : ((number) -> number?) & ((never) -> string?)
@@ -760,9 +716,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "overloadeded_functions_with_overlapping_results_and_variadics")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
         local x : ((string?) -> (string | number)) & ((number?) -> ...number)
         local y : ((nil) -> (number, number?)) = x -- OK
@@ -809,9 +762,6 @@ TEST_CASE_FIXTURE(Fixture, "overloadeded_functions_with_weird_typepacks_2")
 
 TEST_CASE_FIXTURE(Fixture, "overloadeded_functions_with_weird_typepacks_3")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
       function f<a...>()
         local x : (() -> a...) & (() -> (number?,a...))
@@ -830,9 +780,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "overloadeded_functions_with_weird_typepacks_4")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
       function f<a...>()
         local x : ((a...) -> ()) & ((number,a...) -> number)

tests/TypeInfer.modules.test.cpp

@@ -389,9 +389,6 @@ type Table = typeof(tbl)
 
 TEST_CASE_FIXTURE(BuiltinsFixture, "module_type_conflict")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     fileResolver.source["game/A"] = R"(
 export type T = { x: number }
 return {}
@@ -420,9 +417,6 @@ Type 'number' could not be converted into 'string' in an invariant context)";
 
 TEST_CASE_FIXTURE(BuiltinsFixture, "module_type_conflict_instantiated")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     fileResolver.source["game/A"] = R"(
 export type Wrap<T> = { x: T }
 return {}

tests/TypeInfer.operators.test.cpp

@@ -12,6 +12,8 @@
 
 #include "doctest.h"
 
+#include "ScopedFlags.h"
+
 using namespace Luau;
 
 LUAU_FASTFLAG(DebugLuauDeferredConstraintResolution)
@@ -143,6 +145,24 @@ TEST_CASE_FIXTURE(Fixture, "some_primitive_binary_ops")
     CHECK_EQ("number", toString(requireType("c")));
 }
 
+TEST_CASE_FIXTURE(Fixture, "floor_division_binary_op")
+{
+    ScopedFastFlag sffs{"LuauFloorDivision", true};
+
+    CheckResult result = check(R"(
+        local a = 4 // 8
+        local b = -4 // 9 
+        local c = 9
+        c //= -6.5 
+    )");
+
+    LUAU_REQUIRE_NO_ERRORS(result);
+
+    CHECK_EQ("number", toString(requireType("a")));
+    CHECK_EQ("number", toString(requireType("b")));
+    CHECK_EQ("number", toString(requireType("c")));
+}
+
 TEST_CASE_FIXTURE(BuiltinsFixture, "typecheck_overloaded_multiply_that_is_an_intersection")
 {
     CheckResult result = check(R"(

tests/TypeInfer.provisional.test.cpp

@@ -787,9 +787,6 @@ TEST_CASE_FIXTURE(IsSubtypeFixture, "functions_with_mismatching_arity_but_any_is
 
 TEST_CASE_FIXTURE(Fixture, "assign_table_with_refined_property_with_a_similar_type_is_illegal")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
         local t: {x: number?} = {x = nil}
 
@@ -1050,4 +1047,42 @@ tbl:f3()
     }
 }
 
+// Ideally, unification with any will not cause a 2^n normalization of a function overload
+TEST_CASE_FIXTURE(BuiltinsFixture, "normalization_limit_in_unify_with_any")
+{
+    ScopedFastFlag sff[] = {
+        {"LuauTransitiveSubtyping", true},
+        {"DebugLuauDeferredConstraintResolution", true},
+    };
+
+    // With default limit, this test will take 10 seconds in NoOpt
+    ScopedFastInt luauNormalizeCacheLimit{"LuauNormalizeCacheLimit", 1000};
+
+    // Build a function type with a large overload set
+    const int parts = 100;
+    std::string source;
+
+    for (int i = 0; i < parts; i++)
+        formatAppend(source, "type T%d = { f%d: number }\n", i, i);
+
+    source += "type Instance = { new: (('s0', extra: Instance?) -> T0)";
+
+    for (int i = 1; i < parts; i++)
+        formatAppend(source, " & (('s%d', extra: Instance?) -> T%d)", i, i);
+
+    source += " }\n";
+
+    source += R"(
+local Instance: Instance = {} :: any
+
+local function foo(a: typeof(Instance.new)) return if a then 2 else 3 end
+
+foo(1 :: any)
+)";
+
+    CheckResult result = check(source);
+
+    LUAU_REQUIRE_ERRORS(result);
+}
+
 TEST_SUITE_END();

tests/TypeInfer.singletons.test.cpp

@@ -316,9 +316,6 @@ TEST_CASE_FIXTURE(Fixture, "table_properties_type_error_escapes")
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_tagged_union_mismatch_string")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 type Cat = { tag: 'cat', catfood: string }
 type Dog = { tag: 'dog', dogfood: string }
@@ -337,9 +334,6 @@ Table type 'a' not compatible with type 'Cat' because the former is missing fiel
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_tagged_union_mismatch_bool")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 type Good = { success: true, result: string }
 type Bad = { success: false, error: string }
@@ -360,9 +354,7 @@ TEST_CASE_FIXTURE(Fixture, "parametric_tagged_union_alias")
 {
     ScopedFastFlag sff[] = {
         {"DebugLuauDeferredConstraintResolution", true},
-        {"LuauIndentTypeMismatch", true},
     };
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
     CheckResult result = check(R"(
         type Ok<T> = {success: true, result: T}
         type Err<T> = {success: false, error: T}

tests/TypeInfer.tables.test.cpp

@@ -2083,8 +2083,6 @@ TEST_CASE_FIXTURE(BuiltinsFixture, "table_insert_should_cope_with_optional_prope
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_prop")
 {
-    ScopedFastFlag sff[] = {{"LuauIndentTypeMismatch", true}};
-    ScopedFastInt sfi[] = {{"LuauIndentTypeMismatchMaxTypeLength", 10}};
     CheckResult result = check(R"(
 type A = { x: number, y: number }
 type B = { x: number, y: string }
@@ -2103,8 +2101,6 @@ Type 'number' could not be converted into 'string' in an invariant context)";
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_prop_nested")
 {
-    ScopedFastFlag sff[] = {{"LuauIndentTypeMismatch", true}};
-    ScopedFastInt sfi[] = {{"LuauIndentTypeMismatchMaxTypeLength", 10}};
     CheckResult result = check(R"(
 type AS = { x: number, y: number }
 type BS = { x: number, y: string }
@@ -2129,9 +2125,6 @@ Type 'number' could not be converted into 'string' in an invariant context)";
 
 TEST_CASE_FIXTURE(BuiltinsFixture, "error_detailed_metatable_prop")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 local a1 = setmetatable({ x = 2, y = 3 }, { __call = function(s) end });
 local b1 = setmetatable({ x = 2, y = "hello" }, { __call = function(s) end });
@@ -2202,8 +2195,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_indexer_key")
 {
-    ScopedFastFlag sff[] = {{"LuauIndentTypeMismatch", true}};
-    ScopedFastInt sfi[] = {{"LuauIndentTypeMismatchMaxTypeLength", 10}};
     CheckResult result = check(R"(
         type A = { [number]: string }
         type B = { [string]: string }
@@ -2222,8 +2213,6 @@ Type 'number' could not be converted into 'string' in an invariant context)";
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_indexer_value")
 {
-    ScopedFastFlag sff[] = {{"LuauIndentTypeMismatch", true}};
-    ScopedFastInt sfi[] = {{"LuauIndentTypeMismatchMaxTypeLength", 10}};
     CheckResult result = check(R"(
         type A = { [number]: number }
         type B = { [number]: string }
@@ -2257,8 +2246,6 @@ a.p = { x = 9 }
 
 TEST_CASE_FIXTURE(Fixture, "explicitly_typed_table_error")
 {
-    ScopedFastFlag sff[] = {{"LuauIndentTypeMismatch", true}};
-    ScopedFastInt sfi[] = {{"LuauIndentTypeMismatchMaxTypeLength", 10}};
     CheckResult result = check(R"(
 --!strict
 type Super = { x : number }
@@ -3359,14 +3346,10 @@ TEST_CASE_FIXTURE(Fixture, "scalar_is_a_subtype_of_a_compatible_polymorphic_shap
 
 TEST_CASE_FIXTURE(Fixture, "scalar_is_not_a_subtype_of_a_compatible_polymorphic_shape_type")
 {
-    ScopedFastInt sfi[] = {{"LuauIndentTypeMismatchMaxTypeLength", 10}};
     ScopedFastFlag sff[] = {
         {"LuauAlwaysCommitInferencesOfFunctionCalls", true},
-        {"LuauIndentTypeMismatch", true},
     };
 
-
-
     CheckResult result = check(R"(
         local function f(s)
             return s:absolutely_no_scalar_has_this_method()
@@ -3422,8 +3405,6 @@ TEST_CASE_FIXTURE(Fixture, "a_free_shape_can_turn_into_a_scalar_if_it_is_compati
 
 TEST_CASE_FIXTURE(Fixture, "a_free_shape_cannot_turn_into_a_scalar_if_it_is_not_compatible")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
 
     CheckResult result = check(R"(
         local function f(s): string

tests/TypeInfer.test.cpp

@@ -992,9 +992,6 @@ end
 
 TEST_CASE_FIXTURE(Fixture, "cli_50041_committing_txnlog_in_apollo_client_error")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
         --!strict
         --!nolint

tests/TypeInfer.tryUnify.test.cpp

@@ -345,9 +345,7 @@ TEST_CASE_FIXTURE(TryUnifyFixture, "metatables_unify_against_shape_of_free_table
 {
     ScopedFastFlag sff[] = {
         {"LuauTransitiveSubtyping", true},
-        {"LuauIndentTypeMismatch", true},
     };
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
 
     TableType::Props freeProps{
         {"foo", {builtinTypes->numberType}},

tests/TypeInfer.typePacks.cpp

@@ -872,9 +872,6 @@ type R = { m: F<R> }
 
 TEST_CASE_FIXTURE(Fixture, "pack_tail_unification_check")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
-
     CheckResult result = check(R"(
 local a: () -> (number, ...string)
 local b: () -> (number, ...boolean)

tests/TypeInfer.unionTypes.test.cpp

@@ -459,8 +459,6 @@ local oh : boolean = t.y
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_union_part")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
     CheckResult result = check(R"(
 type X = { x: number }
 type Y = { y: number }
@@ -498,8 +496,6 @@ local a: XYZ = { w = 4 }
 
 TEST_CASE_FIXTURE(Fixture, "error_detailed_optional")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
     CheckResult result = check(R"(
 type X = { x: number }
 
@@ -532,8 +528,6 @@ TEST_CASE_FIXTURE(Fixture, "dont_allow_cyclic_unions_to_be_inferred")
 
 TEST_CASE_FIXTURE(BuiltinsFixture, "table_union_write_indirect")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
 
     CheckResult result = check(R"(
         type A = { x: number, y: (number) -> string } | { z: number, y: (number) -> string }
@@ -620,8 +614,6 @@ TEST_CASE_FIXTURE(Fixture, "union_of_functions_mentioning_generics")
 
 TEST_CASE_FIXTURE(Fixture, "union_of_functions_mentioning_generic_typepacks")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
     CheckResult result = check(R"(
       function f<a...>()
         local x : (number, a...) -> (number?, a...)
@@ -640,8 +632,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "union_of_functions_with_mismatching_arg_arities")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
     CheckResult result = check(R"(
         local x : (number) -> number?
         local y : ((number?) -> number) | ((number | string) -> nil) = x -- OK
@@ -658,8 +648,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "union_of_functions_with_mismatching_result_arities")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
 
     CheckResult result = check(R"(
         local x : () -> (number | string)
@@ -677,8 +665,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "union_of_functions_with_variadics")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
 
     CheckResult result = check(R"(
         local x : (...nil) -> (...number?)
@@ -696,8 +682,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "union_of_functions_with_mismatching_arg_variadics")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
 
     CheckResult result = check(R"(
         local x : (number) -> ()
@@ -715,8 +699,6 @@ could not be converted into
 
 TEST_CASE_FIXTURE(Fixture, "union_of_functions_with_mismatching_result_variadics")
 {
-    ScopedFastFlag sff{"LuauIndentTypeMismatch", true};
-    ScopedFastInt sfi{"LuauIndentTypeMismatchMaxTypeLength", 10};
 
     CheckResult result = check(R"(
         local x : () -> (number?, ...number)

tests/conformance/basic.lua

@@ -91,6 +91,15 @@ assert((function() local a = 1 a = a + 2 return a end)() == 3)
 assert((function() local a = 1 a = a - 2 return a end)() == -1)
 assert((function() local a = 1 a = a * 2 return a end)() == 2)
 assert((function() local a = 1 a = a / 2 return a end)() == 0.5)
+
+-- floor division should always round towards -Infinity
+assert((function() local a = 1 a = a // 2 return a end)() == 0)
+assert((function() local a = 3 a = a // 2 return a end)() == 1)
+assert((function() local a = 3.5 a = a // 2 return a end)() == 1)
+assert((function() local a = -1 a = a // 2 return a end)() == -1)
+assert((function() local a = -3 a = a // 2 return a end)() == -2)
+assert((function() local a = -3.5 a = a // 2 return a end)() == -2)
+
 assert((function() local a = 5 a = a % 2 return a end)() == 1)
 assert((function() local a = 3 a = a ^ 2 return a end)() == 9)
 assert((function() local a = 3 a = a ^ 3 return a end)() == 27)
@@ -494,6 +503,7 @@ local function vec3t(x, y, z)
         __sub = function(l, r) return vec3t(l.x - r.x, l.y - r.y, l.z - r.z) end,
         __mul = function(l, r) return type(r) == "number" and vec3t(l.x * r, l.y * r, l.z * r) or vec3t(l.x * r.x, l.y * r.y, l.z * r.z) end,
         __div = function(l, r) return type(r) == "number" and vec3t(l.x / r, l.y / r, l.z / r) or vec3t(l.x / r.x, l.y / r.y, l.z / r.z) end,
+        __idiv = function(l, r) return type(r) == "number" and vec3t(l.x // r, l.y // r, l.z // r) or vec3t(l.x // r.x, l.y // r.y, l.z // r.z) end,
         __unm = function(v) return vec3t(-v.x, -v.y, -v.z) end,
         __tostring = function(v) return string.format("%g, %g, %g", v.x, v.y, v.z) end
     })
@@ -504,10 +514,13 @@ assert((function() return tostring(vec3t(1,2,3) + vec3t(4,5,6)) end)() == "5, 7,
 assert((function() return tostring(vec3t(1,2,3) - vec3t(4,5,6)) end)() == "-3, -3, -3")
 assert((function() return tostring(vec3t(1,2,3) * vec3t(4,5,6)) end)() == "4, 10, 18")
 assert((function() return tostring(vec3t(1,2,3) / vec3t(2,4,8)) end)() == "0.5, 0.5, 0.375")
+assert((function() return tostring(vec3t(1,2,3) // vec3t(2,4,2)) end)() == "0, 0, 1")
+assert((function() return tostring(vec3t(1,2,3) // vec3t(-2,-4,-2)) end)() == "-1, -1, -2")
 
 -- reg vs constant
 assert((function() return tostring(vec3t(1,2,3) * 2) end)() == "2, 4, 6")
 assert((function() return tostring(vec3t(1,2,3) / 2) end)() == "0.5, 1, 1.5")
+assert((function() return tostring(vec3t(1,2,3) // 2) end)() == "0, 1, 1")
 
 -- unary
 assert((function() return tostring(-vec3t(1,2,3)) end)() == "-1, -2, -3")

tests/conformance/events.lua

@@ -107,6 +107,7 @@ t.__add = f("add")
 t.__sub = f("sub")
 t.__mul = f("mul")
 t.__div = f("div")
+t.__idiv = f("idiv")
 t.__mod = f("mod")
 t.__unm = f("unm")
 t.__pow = f("pow")
@@ -128,6 +129,8 @@ assert(a*a == a)
 assert(cap[0] == "mul" and cap[1] == a and cap[2] == a and cap[3]==nil)
 assert(a/0 == a)
 assert(cap[0] == "div" and cap[1] == a and cap[2] == 0 and cap[3]==nil)
+assert(a//0 == a)
+assert(cap[0] == "idiv" and cap[1] == a and cap[2] == 0 and cap[3]==nil)
 assert(a%2 == a)
 assert(cap[0] == "mod" and cap[1] == a and cap[2] == 2 and cap[3]==nil)
 assert(-a == a)

tests/conformance/math.lua

@@ -82,6 +82,7 @@ assert(not(1>1) and not(1>2) and (2>1))
 assert(not('a'>'a') and not('a'>'b') and ('b'>'a'))
 assert((1>=1) and not(1>=2) and (2>=1))
 assert(('a'>='a') and not('a'>='b') and ('b'>='a'))
+assert((unk and unk > 0) == nil) -- validate precedence between and and >
 
 -- testing mod operator
 assert(-4%3 == 2)

tests/conformance/native.lua

@@ -130,4 +130,45 @@ end
 
 assert(pcall(fuzzfail13) == true)
 
+local function arraySizeInv1()
+  local t = {1, 2, nil, nil, nil, nil, nil, nil, nil, true}
+
+  table.insert(t, 3)
+
+  return t[10]
+end
+
+assert(arraySizeInv1() == true)
+
+local function arraySizeInv2()
+  local t = {1, 2, nil, nil, nil, nil, nil, nil, nil, true}
+
+  local u = {a = t}
+  table.insert(u.a, 3) -- aliased modifiction of 't' register through other value
+
+  return t[10]
+end
+
+assert(arraySizeInv2() == true)
+
+local function nilInvalidatesSlot()
+  local function tabs()
+    local t = { x=1, y=2, z=3 }
+    setmetatable(t, { __index = function(t, k) return 42 end })
+    return t, t
+  end
+
+  local t1, t2 = tabs()
+
+  for i=1,2 do
+    local a = t1.x
+    t2.x = nil
+    local b = t1.x
+    t2.x = 1
+    assert(a == 1 and b == 42)
+  end
+end
+
+nilInvalidatesSlot()
+
 return('OK')

tests/conformance/userdata.lua

@@ -30,6 +30,12 @@ assert(int64(4) / 2 == int64(2))
 assert(int64(4) % 3 == int64(1))
 assert(int64(2) ^ 3 == int64(8))
 
+-- / and // round in different directions in our test implementation
+assert(int64(5) / int64(2) == int64(2))
+assert(int64(5) // int64(2) == int64(2))
+assert(int64(-5) / int64(2) == int64(-2))
+assert(int64(-5) // int64(2) == int64(-3))
+
 -- tostring
 assert(tostring(int64(2)) == "2")
 

tests/conformance/vector.lua

@@ -67,6 +67,18 @@ assert(vector(1, 2, 4) / '8' == vector(1/8, 1/4, 1/2));
 
 assert(-vector(1, 2, 4) == vector(-1, -2, -4));
 
+-- test floor division
+assert(vector(1, 3, 5) // 2 == vector(0, 1, 2))
+assert(vector(1, 3, 5) // val == vector(8, 24, 40))
+
+if vector_size == 4 then
+	assert(10 // vector(1, 2, 3, 4) == vector(10, 5, 3, 2))
+	assert(vector(10, 9, 8, 7) // vector(1, 2, 3, 4) == vector(10, 4, 2, 1))
+else
+	assert(10 // vector(1, 2, 3) == vector(10, 5, 3))
+	assert(vector(10, 9, 8) // vector(1, 2, 3) == vector(10, 4, 2))
+end
+
 -- test NaN comparison
 local nanv = vector(0/0, 0/0, 0/0)
 assert(nanv ~= nanv);