Enable matrices in HLSL

clang/include/clang/Driver/Options.td

@@ -4352,7 +4352,7 @@ defm ptrauth_init_fini_address_discrimination : OptInCC1FFlag<"ptrauth-init-fini
 def fenable_matrix : Flag<["-"], "fenable-matrix">, Group<f_Group>,
     Visibility<[ClangOption, CC1Option]>,
     HelpText<"Enable matrix data type and related builtin functions">,
-    MarshallingInfoFlag<LangOpts<"MatrixTypes">>;
+    MarshallingInfoFlag<LangOpts<"MatrixTypes">, hlsl.KeyPath>;
 
 defm raw_string_literals : BoolFOption<"raw-string-literals",
     LangOpts<"RawStringLiterals">, Default<std#".hasRawStringLiterals()">,

clang/include/clang/Sema/HLSLExternalSemaSource.h

@@ -28,6 +28,7 @@ class HLSLExternalSemaSource : public ExternalSemaSource {
   llvm::DenseMap<CXXRecordDecl *, CompletionFunction> Completions;
 
   void defineHLSLVectorAlias();
+  void defineHLSLMatrixAlias();
   void defineTrivialHLSLTypes();
   void defineHLSLTypesWithForwardDeclarations();
 

clang/lib/AST/TypePrinter.cpp

@@ -852,10 +852,18 @@ void TypePrinter::printExtVectorAfter(const ExtVectorType *T, raw_ostream &OS) {
 
 void TypePrinter::printConstantMatrixBefore(const ConstantMatrixType *T,
                                             raw_ostream &OS) {
+  if (Policy.UseHLSLTypes)
+    OS << "matrix<";
   printBefore(T->getElementType(), OS);
-  OS << " __attribute__((matrix_type(";
+  if (!Policy.UseHLSLTypes)
+    OS << " __attribute__((matrix_type(";
+  else
+    OS << ", ";
   OS << T->getNumRows() << ", " << T->getNumColumns();
-  OS << ")))";
+  if (!Policy.UseHLSLTypes)
+    OS << ")))";
+  else
+    OS << ">";
 }
 
 void TypePrinter::printConstantMatrixAfter(const ConstantMatrixType *T,
@@ -865,16 +873,25 @@ void TypePrinter::printConstantMatrixAfter(const ConstantMatrixType *T,
 
 void TypePrinter::printDependentSizedMatrixBefore(
     const DependentSizedMatrixType *T, raw_ostream &OS) {
+  if (Policy.UseHLSLTypes)
+    OS << "matrix<";
   printBefore(T->getElementType(), OS);
-  OS << " __attribute__((matrix_type(";
-  if (T->getRowExpr()) {
-    T->getRowExpr()->printPretty(OS, nullptr, Policy);
-  }
+  if (!Policy.UseHLSLTypes)
+    OS << " __attribute__((matrix_type(";
+  else
+    OS << ", ";
+
+  if (Expr *E = T->getRowExpr())
+    E->printPretty(OS, nullptr, Policy);
   OS << ", ";
-  if (T->getColumnExpr()) {
-    T->getColumnExpr()->printPretty(OS, nullptr, Policy);
-  }
-  OS << ")))";
+  if (Expr *E = T->getColumnExpr())
+    E->printPretty(OS, nullptr, Policy);
+
+  OS << ", ";
+  if (!Policy.UseHLSLTypes)
+    OS << ")))";
+  else
+    OS << ">";
 }
 
 void TypePrinter::printDependentSizedMatrixAfter(

clang/lib/Headers/hlsl/hlsl_basic_types.h

@@ -115,6 +115,238 @@ typedef vector<float64_t, 2> float64_t2;
 typedef vector<float64_t, 3> float64_t3;
 typedef vector<float64_t, 4> float64_t4;
 
+#ifdef __HLSL_ENABLE_16_BIT
+typedef matrix<int16_t, 1, 1> int16_t1x1;
+typedef matrix<int16_t, 1, 2> int16_t1x2;
+typedef matrix<int16_t, 1, 3> int16_t1x3;
+typedef matrix<int16_t, 1, 4> int16_t1x4;
+typedef matrix<int16_t, 2, 1> int16_t2x1;
+typedef matrix<int16_t, 2, 2> int16_t2x2;
+typedef matrix<int16_t, 2, 3> int16_t2x3;
+typedef matrix<int16_t, 2, 4> int16_t2x4;
+typedef matrix<int16_t, 3, 1> int16_t3x1;
+typedef matrix<int16_t, 3, 2> int16_t3x2;
+typedef matrix<int16_t, 3, 3> int16_t3x3;
+typedef matrix<int16_t, 3, 4> int16_t3x4;
+typedef matrix<int16_t, 4, 1> int16_t4x1;
+typedef matrix<int16_t, 4, 2> int16_t4x2;
+typedef matrix<int16_t, 4, 3> int16_t4x3;
+typedef matrix<int16_t, 4, 4> int16_t4x4;
+typedef matrix<uint16_t, 1, 1> uint16_t1x1;
+typedef matrix<uint16_t, 1, 2> uint16_t1x2;
+typedef matrix<uint16_t, 1, 3> uint16_t1x3;
+typedef matrix<uint16_t, 1, 4> uint16_t1x4;
+typedef matrix<uint16_t, 2, 1> uint16_t2x1;
+typedef matrix<uint16_t, 2, 2> uint16_t2x2;
+typedef matrix<uint16_t, 2, 3> uint16_t2x3;
+typedef matrix<uint16_t, 2, 4> uint16_t2x4;
+typedef matrix<uint16_t, 3, 1> uint16_t3x1;
+typedef matrix<uint16_t, 3, 2> uint16_t3x2;
+typedef matrix<uint16_t, 3, 3> uint16_t3x3;
+typedef matrix<uint16_t, 3, 4> uint16_t3x4;
+typedef matrix<uint16_t, 4, 1> uint16_t4x1;
+typedef matrix<uint16_t, 4, 2> uint16_t4x2;
+typedef matrix<uint16_t, 4, 3> uint16_t4x3;
+typedef matrix<uint16_t, 4, 4> uint16_t4x4;
+#endif
+typedef matrix<int, 1, 1> int1x1;
+typedef matrix<int, 1, 2> int1x2;
+typedef matrix<int, 1, 3> int1x3;
+typedef matrix<int, 1, 4> int1x4;
+typedef matrix<int, 2, 1> int2x1;
+typedef matrix<int, 2, 2> int2x2;
+typedef matrix<int, 2, 3> int2x3;
+typedef matrix<int, 2, 4> int2x4;
+typedef matrix<int, 3, 1> int3x1;
+typedef matrix<int, 3, 2> int3x2;
+typedef matrix<int, 3, 3> int3x3;
+typedef matrix<int, 3, 4> int3x4;
+typedef matrix<int, 4, 1> int4x1;
+typedef matrix<int, 4, 2> int4x2;
+typedef matrix<int, 4, 3> int4x3;
+typedef matrix<int, 4, 4> int4x4;
+typedef matrix<uint, 1, 1> uint1x1;
+typedef matrix<uint, 1, 2> uint1x2;
+typedef matrix<uint, 1, 3> uint1x3;
+typedef matrix<uint, 1, 4> uint1x4;
+typedef matrix<uint, 2, 1> uint2x1;
+typedef matrix<uint, 2, 2> uint2x2;
+typedef matrix<uint, 2, 3> uint2x3;
+typedef matrix<uint, 2, 4> uint2x4;
+typedef matrix<uint, 3, 1> uint3x1;
+typedef matrix<uint, 3, 2> uint3x2;
+typedef matrix<uint, 3, 3> uint3x3;
+typedef matrix<uint, 3, 4> uint3x4;
+typedef matrix<uint, 4, 1> uint4x1;
+typedef matrix<uint, 4, 2> uint4x2;
+typedef matrix<uint, 4, 3> uint4x3;
+typedef matrix<uint, 4, 4> uint4x4;
+typedef matrix<int32_t, 1, 1> int32_t1x1;
+typedef matrix<int32_t, 1, 2> int32_t1x2;
+typedef matrix<int32_t, 1, 3> int32_t1x3;
+typedef matrix<int32_t, 1, 4> int32_t1x4;
+typedef matrix<int32_t, 2, 1> int32_t2x1;
+typedef matrix<int32_t, 2, 2> int32_t2x2;
+typedef matrix<int32_t, 2, 3> int32_t2x3;
+typedef matrix<int32_t, 2, 4> int32_t2x4;
+typedef matrix<int32_t, 3, 1> int32_t3x1;
+typedef matrix<int32_t, 3, 2> int32_t3x2;
+typedef matrix<int32_t, 3, 3> int32_t3x3;
+typedef matrix<int32_t, 3, 4> int32_t3x4;
+typedef matrix<int32_t, 4, 1> int32_t4x1;
+typedef matrix<int32_t, 4, 2> int32_t4x2;
+typedef matrix<int32_t, 4, 3> int32_t4x3;
+typedef matrix<int32_t, 4, 4> int32_t4x4;
+typedef matrix<uint32_t, 1, 1> uint32_t1x1;
+typedef matrix<uint32_t, 1, 2> uint32_t1x2;
+typedef matrix<uint32_t, 1, 3> uint32_t1x3;
+typedef matrix<uint32_t, 1, 4> uint32_t1x4;
+typedef matrix<uint32_t, 2, 1> uint32_t2x1;
+typedef matrix<uint32_t, 2, 2> uint32_t2x2;
+typedef matrix<uint32_t, 2, 3> uint32_t2x3;
+typedef matrix<uint32_t, 2, 4> uint32_t2x4;
+typedef matrix<uint32_t, 3, 1> uint32_t3x1;
+typedef matrix<uint32_t, 3, 2> uint32_t3x2;
+typedef matrix<uint32_t, 3, 3> uint32_t3x3;
+typedef matrix<uint32_t, 3, 4> uint32_t3x4;
+typedef matrix<uint32_t, 4, 1> uint32_t4x1;
+typedef matrix<uint32_t, 4, 2> uint32_t4x2;
+typedef matrix<uint32_t, 4, 3> uint32_t4x3;
+typedef matrix<uint32_t, 4, 4> uint32_t4x4;
+typedef matrix<int64_t, 1, 1> int64_t1x1;
+typedef matrix<int64_t, 1, 2> int64_t1x2;
+typedef matrix<int64_t, 1, 3> int64_t1x3;
+typedef matrix<int64_t, 1, 4> int64_t1x4;
+typedef matrix<int64_t, 2, 1> int64_t2x1;
+typedef matrix<int64_t, 2, 2> int64_t2x2;
+typedef matrix<int64_t, 2, 3> int64_t2x3;
+typedef matrix<int64_t, 2, 4> int64_t2x4;
+typedef matrix<int64_t, 3, 1> int64_t3x1;
+typedef matrix<int64_t, 3, 2> int64_t3x2;
+typedef matrix<int64_t, 3, 3> int64_t3x3;
+typedef matrix<int64_t, 3, 4> int64_t3x4;
+typedef matrix<int64_t, 4, 1> int64_t4x1;
+typedef matrix<int64_t, 4, 2> int64_t4x2;
+typedef matrix<int64_t, 4, 3> int64_t4x3;
+typedef matrix<int64_t, 4, 4> int64_t4x4;
+typedef matrix<uint64_t, 1, 1> uint64_t1x1;
+typedef matrix<uint64_t, 1, 2> uint64_t1x2;
+typedef matrix<uint64_t, 1, 3> uint64_t1x3;
+typedef matrix<uint64_t, 1, 4> uint64_t1x4;
+typedef matrix<uint64_t, 2, 1> uint64_t2x1;
+typedef matrix<uint64_t, 2, 2> uint64_t2x2;
+typedef matrix<uint64_t, 2, 3> uint64_t2x3;
+typedef matrix<uint64_t, 2, 4> uint64_t2x4;
+typedef matrix<uint64_t, 3, 1> uint64_t3x1;
+typedef matrix<uint64_t, 3, 2> uint64_t3x2;
+typedef matrix<uint64_t, 3, 3> uint64_t3x3;
+typedef matrix<uint64_t, 3, 4> uint64_t3x4;
+typedef matrix<uint64_t, 4, 1> uint64_t4x1;
+typedef matrix<uint64_t, 4, 2> uint64_t4x2;
+typedef matrix<uint64_t, 4, 3> uint64_t4x3;
+typedef matrix<uint64_t, 4, 4> uint64_t4x4;
+
+typedef matrix<half, 1, 1> half1x1;
+typedef matrix<half, 1, 2> half1x2;
+typedef matrix<half, 1, 3> half1x3;
+typedef matrix<half, 1, 4> half1x4;
+typedef matrix<half, 2, 1> half2x1;
+typedef matrix<half, 2, 2> half2x2;
+typedef matrix<half, 2, 3> half2x3;
+typedef matrix<half, 2, 4> half2x4;
+typedef matrix<half, 3, 1> half3x1;
+typedef matrix<half, 3, 2> half3x2;
+typedef matrix<half, 3, 3> half3x3;
+typedef matrix<half, 3, 4> half3x4;
+typedef matrix<half, 4, 1> half4x1;
+typedef matrix<half, 4, 2> half4x2;
+typedef matrix<half, 4, 3> half4x3;
+typedef matrix<half, 4, 4> half4x4;
+typedef matrix<float, 1, 1> float1x1;
+typedef matrix<float, 1, 2> float1x2;
+typedef matrix<float, 1, 3> float1x3;
+typedef matrix<float, 1, 4> float1x4;
+typedef matrix<float, 2, 1> float2x1;
+typedef matrix<float, 2, 2> float2x2;
+typedef matrix<float, 2, 3> float2x3;
+typedef matrix<float, 2, 4> float2x4;
+typedef matrix<float, 3, 1> float3x1;
+typedef matrix<float, 3, 2> float3x2;
+typedef matrix<float, 3, 3> float3x3;
+typedef matrix<float, 3, 4> float3x4;
+typedef matrix<float, 4, 1> float4x1;
+typedef matrix<float, 4, 2> float4x2;
+typedef matrix<float, 4, 3> float4x3;
+typedef matrix<float, 4, 4> float4x4;
+typedef matrix<double, 1, 1> double1x1;
+typedef matrix<double, 1, 2> double1x2;
+typedef matrix<double, 1, 3> double1x3;
+typedef matrix<double, 1, 4> double1x4;
+typedef matrix<double, 2, 1> double2x1;
+typedef matrix<double, 2, 2> double2x2;
+typedef matrix<double, 2, 3> double2x3;
+typedef matrix<double, 2, 4> double2x4;
+typedef matrix<double, 3, 1> double3x1;
+typedef matrix<double, 3, 2> double3x2;
+typedef matrix<double, 3, 3> double3x3;
+typedef matrix<double, 3, 4> double3x4;
+typedef matrix<double, 4, 1> double4x1;
+typedef matrix<double, 4, 2> double4x2;
+typedef matrix<double, 4, 3> double4x3;
+typedef matrix<double, 4, 4> double4x4;
+
+#ifdef __HLSL_ENABLE_16_BIT
+typedef matrix<float16_t, 1, 1> float16_t1x1;
+typedef matrix<float16_t, 1, 2> float16_t1x2;
+typedef matrix<float16_t, 1, 3> float16_t1x3;
+typedef matrix<float16_t, 1, 4> float16_t1x4;
+typedef matrix<float16_t, 2, 1> float16_t2x1;
+typedef matrix<float16_t, 2, 2> float16_t2x2;
+typedef matrix<float16_t, 2, 3> float16_t2x3;
+typedef matrix<float16_t, 2, 4> float16_t2x4;
+typedef matrix<float16_t, 3, 1> float16_t3x1;
+typedef matrix<float16_t, 3, 2> float16_t3x2;
+typedef matrix<float16_t, 3, 3> float16_t3x3;
+typedef matrix<float16_t, 3, 4> float16_t3x4;
+typedef matrix<float16_t, 4, 1> float16_t4x1;
+typedef matrix<float16_t, 4, 2> float16_t4x2;
+typedef matrix<float16_t, 4, 3> float16_t4x3;
+typedef matrix<float16_t, 4, 4> float16_t4x4;
+#endif
+
+typedef matrix<float32_t, 1, 1> float32_t1x1;
+typedef matrix<float32_t, 1, 2> float32_t1x2;
+typedef matrix<float32_t, 1, 3> float32_t1x3;
+typedef matrix<float32_t, 1, 4> float32_t1x4;
+typedef matrix<float32_t, 2, 1> float32_t2x1;
+typedef matrix<float32_t, 2, 2> float32_t2x2;
+typedef matrix<float32_t, 2, 3> float32_t2x3;
+typedef matrix<float32_t, 2, 4> float32_t2x4;
+typedef matrix<float32_t, 3, 1> float32_t3x1;
+typedef matrix<float32_t, 3, 2> float32_t3x2;
+typedef matrix<float32_t, 3, 3> float32_t3x3;
+typedef matrix<float32_t, 3, 4> float32_t3x4;
+typedef matrix<float32_t, 4, 1> float32_t4x1;
+typedef matrix<float32_t, 4, 2> float32_t4x2;
+typedef matrix<float32_t, 4, 3> float32_t4x3;
+typedef matrix<float32_t, 4, 4> float32_t4x4;
+typedef matrix<float64_t, 1, 1> float64_t1x1;
+typedef matrix<float64_t, 1, 2> float64_t1x2;
+typedef matrix<float64_t, 1, 3> float64_t1x3;
+typedef matrix<float64_t, 1, 4> float64_t1x4;
+typedef matrix<float64_t, 2, 1> float64_t2x1;
+typedef matrix<float64_t, 2, 2> float64_t2x2;
+typedef matrix<float64_t, 2, 3> float64_t2x3;
+typedef matrix<float64_t, 2, 4> float64_t2x4;
+typedef matrix<float64_t, 3, 1> float64_t3x1;
+typedef matrix<float64_t, 3, 2> float64_t3x2;
+typedef matrix<float64_t, 3, 3> float64_t3x3;
+typedef matrix<float64_t, 3, 4> float64_t3x4;
+typedef matrix<float64_t, 4, 1> float64_t4x1;
+typedef matrix<float64_t, 4, 2> float64_t4x2;
+typedef matrix<float64_t, 4, 3> float64_t4x3;
+typedef matrix<float64_t, 4, 4> float64_t4x4;
+
 } // namespace hlsl
 
 #endif //_HLSL_HLSL_BASIC_TYPES_H_

clang/lib/Sema/HLSLExternalSemaSource.cpp

@@ -459,8 +459,81 @@ void HLSLExternalSemaSource::defineHLSLVectorAlias() {
   HLSLNamespace->addDecl(Template);
 }
 
+void HLSLExternalSemaSource::defineHLSLMatrixAlias() {
+  ASTContext &AST = SemaPtr->getASTContext();
+
+  llvm::SmallVector<NamedDecl *> TemplateParams;
+
+  auto *TypeParam = TemplateTypeParmDecl::Create(
+      AST, HLSLNamespace, SourceLocation(), SourceLocation(), 0, 0,
+      &AST.Idents.get("element", tok::TokenKind::identifier), false, false);
+  TypeParam->setDefaultArgument(
+      AST, SemaPtr->getTrivialTemplateArgumentLoc(
+               TemplateArgument(AST.FloatTy), QualType(), SourceLocation()));
+
+  TemplateParams.emplace_back(TypeParam);
+
+  // these should be 64 bit to be consistent with other clang matrices.
+  auto *RowsParam = NonTypeTemplateParmDecl::Create(
+      AST, HLSLNamespace, SourceLocation(), SourceLocation(), 0, 1,
+      &AST.Idents.get("rows_count", tok::TokenKind::identifier), AST.IntTy,
+      false, AST.getTrivialTypeSourceInfo(AST.IntTy));
+  llvm::APInt RVal(AST.getIntWidth(AST.IntTy), 4);
+  TemplateArgument RDefault(AST, llvm::APSInt(std::move(RVal)), AST.IntTy,
+                            /*IsDefaulted=*/true);
+  RowsParam->setDefaultArgument(
+      AST, SemaPtr->getTrivialTemplateArgumentLoc(RDefault, AST.IntTy,
+                                                  SourceLocation(), RowsParam));
+  TemplateParams.emplace_back(RowsParam);
+
+  auto *ColsParam = NonTypeTemplateParmDecl::Create(
+      AST, HLSLNamespace, SourceLocation(), SourceLocation(), 0, 2,
+      &AST.Idents.get("cols_count", tok::TokenKind::identifier), AST.IntTy,
+      false, AST.getTrivialTypeSourceInfo(AST.IntTy));
+  llvm::APInt CVal(AST.getIntWidth(AST.IntTy), 4);
+  TemplateArgument CDefault(AST, llvm::APSInt(std::move(CVal)), AST.IntTy,
+                            /*IsDefaulted=*/true);
+  ColsParam->setDefaultArgument(
+      AST, SemaPtr->getTrivialTemplateArgumentLoc(CDefault, AST.IntTy,
+                                                  SourceLocation(), ColsParam));
+  TemplateParams.emplace_back(RowsParam);
+
+  auto *ParamList =
+      TemplateParameterList::Create(AST, SourceLocation(), SourceLocation(),
+                                    TemplateParams, SourceLocation(), nullptr);
+
+  IdentifierInfo &II = AST.Idents.get("matrix", tok::TokenKind::identifier);
+
+  QualType AliasType = AST.getDependentSizedMatrixType(
+      AST.getTemplateTypeParmType(0, 0, false, TypeParam),
+      DeclRefExpr::Create(
+          AST, NestedNameSpecifierLoc(), SourceLocation(), RowsParam, false,
+          DeclarationNameInfo(RowsParam->getDeclName(), SourceLocation()),
+          AST.IntTy, VK_LValue),
+      DeclRefExpr::Create(
+          AST, NestedNameSpecifierLoc(), SourceLocation(), ColsParam, false,
+          DeclarationNameInfo(ColsParam->getDeclName(), SourceLocation()),
+          AST.IntTy, VK_LValue),
+      SourceLocation());
+
+  auto *Record = TypeAliasDecl::Create(AST, HLSLNamespace, SourceLocation(),
+                                       SourceLocation(), &II,
+                                       AST.getTrivialTypeSourceInfo(AliasType));
+  Record->setImplicit(true);
+
+  auto *Template =
+      TypeAliasTemplateDecl::Create(AST, HLSLNamespace, SourceLocation(),
+                                    Record->getIdentifier(), ParamList, Record);
+
+  Record->setDescribedAliasTemplate(Template);
+  Template->setImplicit(true);
+  Template->setLexicalDeclContext(Record->getDeclContext());
+  HLSLNamespace->addDecl(Template);
+}
+
 void HLSLExternalSemaSource::defineTrivialHLSLTypes() {
   defineHLSLVectorAlias();
+  defineHLSLMatrixAlias();
 }
 
 /// Set up common members and attributes for buffer types