Allow native vectors for LLVM operations (microsoft#7155)

Disables various forms of scalarization and vector elimination to permit
vectors to pass through to final DXIL when used in native LLVM
operations and loading/storing.

Introduces a few vector manipulation llvm instructions to DXIL allowing
for them to appear in output DXIL.

Skips passes for 6.9 that scalarize, convert to arrays, or otherwise
eliminate vectors.
This eliminates the element-by-element extraction, application, 
and reconstitution of the vectors to operators.
In many cases, this required plumbing the shader model information to
passes that didn't have it before and also the recreation of dxil version
information from metadata where necessary.

Many changes were needed for the MatrixBitcastLower pass related to
linking to avoid converting matrix vectors, but also to perform the
conversion if a shader was compiled for 6.9+, but then linked to a
earlier target.
This now adapts to the linker target to either preserve vectors for 6.9
or arrays for previous versions.
This requires running the DynamicIndexing VectorToArray pass during
linking since 6_x and 6_9+ will fail to run this in the initial compile,
but will still need to lower vectors to arrays. This required making the pass
particularly robust to different sources of version information as compiling,
linking, and running optimization in isolation each require retrieval from
a different source. The latter two sources are facilitated with a dxilutil function.

Ternary conditional/select operators were element extracted in codegen.
Removing this allows 6.9 to preserve the vectors, but also maintains
behavior for previous shader models because the operations get
scalarized later anyway. This was in the region of work to allow short
circuiting, but the effect of that is to introduce the select and skip the later
code that implements short circuiting for supported cases. Test confirm that
no short circuiting is introduced for native vectors.

Adds extensive tests for these operations using different types and
sizes and testing them appropriately. Booleans produce significantly
different code, so they get their own test.

Vec1s have some special treatment as they are not allowed in final dxil,
so they still need to be scalarized. This requires value specific conditionals
in transformation passes. Testing confirms that this is done. 

Fixes microsoft#7123

Author: Greg Roth

include/dxc/DXIL/DxilInstructions.h

@@ -645,6 +645,42 @@ struct LlvmInst_VAArg {
   bool isAllowed() const { return false; }
 };
 
+/// This instruction extracts from vector
+struct LlvmInst_ExtractElement {
+  llvm::Instruction *Instr;
+  // Construction and identification
+  LlvmInst_ExtractElement(llvm::Instruction *pInstr) : Instr(pInstr) {}
+  operator bool() const {
+    return Instr->getOpcode() == llvm::Instruction::ExtractElement;
+  }
+  // Validation support
+  bool isAllowed() const { return true; }
+};
+
+/// This instruction inserts into vector
+struct LlvmInst_InsertElement {
+  llvm::Instruction *Instr;
+  // Construction and identification
+  LlvmInst_InsertElement(llvm::Instruction *pInstr) : Instr(pInstr) {}
+  operator bool() const {
+    return Instr->getOpcode() == llvm::Instruction::InsertElement;
+  }
+  // Validation support
+  bool isAllowed() const { return true; }
+};
+
+/// This instruction Shuffle two vectors
+struct LlvmInst_ShuffleVector {
+  llvm::Instruction *Instr;
+  // Construction and identification
+  LlvmInst_ShuffleVector(llvm::Instruction *pInstr) : Instr(pInstr) {}
+  operator bool() const {
+    return Instr->getOpcode() == llvm::Instruction::ShuffleVector;
+  }
+  // Validation support
+  bool isAllowed() const { return true; }
+};
+
 /// This instruction extracts from aggregate
 struct LlvmInst_ExtractValue {
   llvm::Instruction *Instr;

include/dxc/DXIL/DxilMetadataHelper.h

@@ -427,6 +427,8 @@ class DxilMDHelper {
   // Dxil version.
   void EmitDxilVersion(unsigned Major, unsigned Minor);
   void LoadDxilVersion(unsigned &Major, unsigned &Minor);
+  static bool LoadDxilVersion(const llvm::Module *pModule, unsigned &Major,
+                              unsigned &Minor);
 
   // Validator version.
   void EmitValidatorVersion(unsigned Major, unsigned Minor);

include/dxc/DXIL/DxilUtil.h

@@ -223,6 +223,10 @@ bool DeleteDeadAllocas(llvm::Function &F);
 llvm::Value *GEPIdxToOffset(llvm::GetElementPtrInst *GEP,
                             llvm::IRBuilder<> &Builder, hlsl::OP *OP,
                             const llvm::DataLayout &DL);
+
+// Passes back Dxil version of the given module on true return.
+bool LoadDxilVersion(const llvm::Module *M, unsigned &Major, unsigned &Minor);
+
 } // namespace dxilutil
 
 } // namespace hlsl

lib/DXIL/DxilMetadataHelper.cpp

@@ -177,17 +177,28 @@ void DxilMDHelper::EmitDxilVersion(unsigned Major, unsigned Minor) {
   pDxilVersionMD->addOperand(MDNode::get(m_Ctx, MDVals));
 }
 
-void DxilMDHelper::LoadDxilVersion(unsigned &Major, unsigned &Minor) {
-  NamedMDNode *pDxilVersionMD = m_pModule->getNamedMetadata(kDxilVersionMDName);
-  IFTBOOL(pDxilVersionMD != nullptr, DXC_E_INCORRECT_DXIL_METADATA);
-  IFTBOOL(pDxilVersionMD->getNumOperands() == 1, DXC_E_INCORRECT_DXIL_METADATA);
+// Load dxil version from metadata contained in pModule.
+// Returns true and passes result through
+// the dxil major/minor version params if valid.
+// Returns false if metadata is missing or invalid.
+bool DxilMDHelper::LoadDxilVersion(const Module *pModule, unsigned &Major,
+                                   unsigned &Minor) {
+  NamedMDNode *pDxilVersionMD = pModule->getNamedMetadata(kDxilVersionMDName);
+  IFRBOOL(pDxilVersionMD != nullptr, false);
+  IFRBOOL(pDxilVersionMD->getNumOperands() == 1, false);
 
   MDNode *pVersionMD = pDxilVersionMD->getOperand(0);
-  IFTBOOL(pVersionMD->getNumOperands() == kDxilVersionNumFields,
-          DXC_E_INCORRECT_DXIL_METADATA);
+  IFRBOOL(pVersionMD->getNumOperands() == kDxilVersionNumFields, false);
 
   Major = ConstMDToUint32(pVersionMD->getOperand(kDxilVersionMajorIdx));
   Minor = ConstMDToUint32(pVersionMD->getOperand(kDxilVersionMinorIdx));
+
+  return true;
+}
+
+void DxilMDHelper::LoadDxilVersion(unsigned &Major, unsigned &Minor) {
+  IFTBOOL(LoadDxilVersion(m_pModule, Major, Minor),
+          DXC_E_INCORRECT_DXIL_METADATA);
 }
 
 //

lib/DXIL/DxilUtil.cpp

@@ -1415,5 +1415,18 @@ bool DeleteDeadAllocas(llvm::Function &F) {
   return Changed;
 }
 
+// Retrieve dxil version in the given module.
+// Where the module doesn't already have a Dxil module,
+// it identifies and returns the version info from the metatdata.
+// Returns false where none of that works, but that shouldn't happen much.
+bool LoadDxilVersion(const Module *M, unsigned &Major, unsigned &Minor) {
+  if (M->HasDxilModule()) {
+    M->GetDxilModule().GetShaderModel()->GetDxilVersion(Major, Minor);
+    return true;
+  }
+  // No module, try metadata.
+  return DxilMDHelper::LoadDxilVersion(M, Major, Minor);
+}
+
 } // namespace dxilutil
 } // namespace hlsl

lib/DxilValidation/DxilValidation.cpp

@@ -2193,6 +2193,9 @@ static bool ValidateType(Type *Ty, ValidationContext &ValCtx,
     return true;
 
   if (Ty->isVectorTy()) {
+    if (Ty->getVectorNumElements() > 1 &&
+        ValCtx.DxilMod.GetShaderModel()->IsSM69Plus())
+      return true;
     ValCtx.EmitTypeError(Ty, ValidationRule::TypesNoVector);
     return false;
   }
@@ -2669,6 +2672,23 @@ static bool IsLLVMInstructionAllowedForLib(Instruction &I,
   }
 }
 
+// Shader model specific checks for valid LLVM instructions.
+// Currently only checks for pre 6.9 usage of vector operations.
+// Returns false if shader model is pre 6.9 and I represents a vector
+// operation. Returns true otherwise.
+static bool IsLLVMInstructionAllowedForShaderModel(Instruction &I,
+                                                   ValidationContext &ValCtx) {
+  if (ValCtx.DxilMod.GetShaderModel()->IsSM69Plus())
+    return true;
+  unsigned OpCode = I.getOpcode();
+  if (OpCode == Instruction::InsertElement ||
+      OpCode == Instruction::ExtractElement ||
+      OpCode == Instruction::ShuffleVector)
+    return false;
+
+  return true;
+}
+
 static void ValidateFunctionBody(Function *F, ValidationContext &ValCtx) {
   bool SupportsMinPrecision =
       ValCtx.DxilMod.GetGlobalFlags() & DXIL::kEnableMinPrecision;
@@ -2691,7 +2711,8 @@ static void ValidateFunctionBody(Function *F, ValidationContext &ValCtx) {
       }
 
       // Instructions must be allowed.
-      if (!IsLLVMInstructionAllowed(I)) {
+      if (!IsLLVMInstructionAllowed(I) ||
+          !IsLLVMInstructionAllowedForShaderModel(I, ValCtx)) {
         if (!IsLLVMInstructionAllowedForLib(I, ValCtx)) {
           ValCtx.EmitInstrError(&I, ValidationRule::InstrAllowed);
           continue;

lib/HLSL/DxilLinker.cpp

@@ -1255,6 +1255,12 @@ void DxilLinkJob::RunPreparePass(Module &M) {
   // For static global handle.
   PM.add(createLowerStaticGlobalIntoAlloca());
 
+  // Change dynamic indexing vector to array where vectors aren't
+  // supported, but might be there from the initial compile.
+  if (!pSM->IsSM69Plus())
+    PM.add(
+        createDynamicIndexingVectorToArrayPass(false /* ReplaceAllVector */));
+
   // Remove MultiDimArray from function call arg.
   PM.add(createMultiDimArrayToOneDimArrayPass());
 

lib/HLSL/HLMatrixBitcastLowerPass.cpp

@@ -76,13 +76,17 @@ Type *TryLowerMatTy(Type *Ty) {
 }
 
 class MatrixBitcastLowerPass : public FunctionPass {
+  bool SupportsVectors = false;
 
 public:
   static char ID; // Pass identification, replacement for typeid
   explicit MatrixBitcastLowerPass() : FunctionPass(ID) {}
 
   StringRef getPassName() const override { return "Matrix Bitcast lower"; }
   bool runOnFunction(Function &F) override {
+    DxilModule &DM = F.getParent()->GetOrCreateDxilModule();
+    SupportsVectors = DM.GetShaderModel()->IsSM69Plus();
+
     bool bUpdated = false;
     std::unordered_set<BitCastInst *> matCastSet;
     for (auto blkIt = F.begin(); blkIt != F.end(); ++blkIt) {
@@ -100,7 +104,6 @@ class MatrixBitcastLowerPass : public FunctionPass {
       }
     }
 
-    DxilModule &DM = F.getParent()->GetOrCreateDxilModule();
     // Remove bitcast which has CallInst user.
     if (DM.GetShaderModel()->IsLib()) {
       for (auto it = matCastSet.begin(); it != matCastSet.end();) {
@@ -185,18 +188,19 @@ void MatrixBitcastLowerPass::lowerMatrix(Instruction *M, Value *A) {
     User *U = *(it++);
     if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(U)) {
       Type *EltTy = GEP->getType()->getPointerElementType();
-      if (HLMatrixType::isa(EltTy)) {
+      if (HLMatrixType MatTy = HLMatrixType::dyn_cast(EltTy)) {
         // Change gep matrixArray, 0, index
         // into
         //   gep oneDimArray, 0, index * matSize
         IRBuilder<> Builder(GEP);
         SmallVector<Value *, 2> idxList(GEP->idx_begin(), GEP->idx_end());
         DXASSERT(idxList.size() == 2,
                  "else not one dim matrix array index to matrix");
-
-        HLMatrixType MatTy = HLMatrixType::cast(EltTy);
-        Value *matSize = Builder.getInt32(MatTy.getNumElements());
-        idxList.back() = Builder.CreateMul(idxList.back(), matSize);
+        unsigned NumElts = MatTy.getNumElements();
+        if (!SupportsVectors || NumElts == 1) {
+          Value *MatSize = Builder.getInt32(NumElts);
+          idxList.back() = Builder.CreateMul(idxList.back(), MatSize);
+        }
         Value *NewGEP = Builder.CreateGEP(A, idxList);
         lowerMatrix(GEP, NewGEP);
         DXASSERT(GEP->user_empty(), "else lower matrix fail");
@@ -211,13 +215,23 @@ void MatrixBitcastLowerPass::lowerMatrix(Instruction *M, Value *A) {
     } else if (LoadInst *LI = dyn_cast<LoadInst>(U)) {
       if (VectorType *Ty = dyn_cast<VectorType>(LI->getType())) {
         IRBuilder<> Builder(LI);
-        Value *zeroIdx = Builder.getInt32(0);
-        unsigned vecSize = Ty->getNumElements();
-        Value *NewVec = UndefValue::get(LI->getType());
-        for (unsigned i = 0; i < vecSize; i++) {
-          Value *GEP = CreateEltGEP(A, i, zeroIdx, Builder);
-          Value *Elt = Builder.CreateLoad(GEP);
-          NewVec = Builder.CreateInsertElement(NewVec, Elt, i);
+        Value *NewVec = nullptr;
+        unsigned VecSize = Ty->getVectorNumElements();
+        if (SupportsVectors && VecSize > 1) {
+          // Create a replacement load using the vector pointer.
+          Instruction *NewLd = LI->clone();
+          unsigned VecIdx = NewLd->getNumOperands() - 1;
+          NewLd->setOperand(VecIdx, A);
+          Builder.Insert(NewLd);
+          NewVec = NewLd;
+        } else {
+          Value *zeroIdx = Builder.getInt32(0);
+          NewVec = UndefValue::get(LI->getType());
+          for (unsigned i = 0; i < VecSize; i++) {
+            Value *GEP = CreateEltGEP(A, i, zeroIdx, Builder);
+            Value *Elt = Builder.CreateLoad(GEP);
+            NewVec = Builder.CreateInsertElement(NewVec, Elt, i);
+          }
         }
         LI->replaceAllUsesWith(NewVec);
         LI->eraseFromParent();
@@ -228,12 +242,20 @@ void MatrixBitcastLowerPass::lowerMatrix(Instruction *M, Value *A) {
       Value *V = ST->getValueOperand();
       if (VectorType *Ty = dyn_cast<VectorType>(V->getType())) {
         IRBuilder<> Builder(LI);
-        Value *zeroIdx = Builder.getInt32(0);
-        unsigned vecSize = Ty->getNumElements();
-        for (unsigned i = 0; i < vecSize; i++) {
-          Value *GEP = CreateEltGEP(A, i, zeroIdx, Builder);
-          Value *Elt = Builder.CreateExtractElement(V, i);
-          Builder.CreateStore(Elt, GEP);
+        if (SupportsVectors && Ty->getVectorNumElements() > 1) {
+          // Create a replacement store using the vector pointer.
+          Instruction *NewSt = ST->clone();
+          unsigned VecIdx = NewSt->getNumOperands() - 1;
+          NewSt->setOperand(VecIdx, A);
+          Builder.Insert(NewSt);
+        } else {
+          Value *zeroIdx = Builder.getInt32(0);
+          unsigned vecSize = Ty->getNumElements();
+          for (unsigned i = 0; i < vecSize; i++) {
+            Value *GEP = CreateEltGEP(A, i, zeroIdx, Builder);
+            Value *Elt = Builder.CreateExtractElement(V, i);
+            Builder.CreateStore(Elt, GEP);
+          }
         }
         ST->eraseFromParent();
       } else {

lib/HLSL/HLModule.cpp

@@ -604,6 +604,9 @@ MDTuple *HLModule::EmitHLResources() {
 
 void HLModule::LoadHLResources(const llvm::MDOperand &MDO) {
   const llvm::MDTuple *pSRVs, *pUAVs, *pCBuffers, *pSamplers;
+  // No resources. Nothing to do.
+  if (MDO.get() == nullptr)
+    return;
   m_pMDHelper->GetDxilResources(MDO, pSRVs, pUAVs, pCBuffers, pSamplers);
 
   // Load SRV records.

lib/Transforms/Scalar/LowerTypePasses.cpp

@@ -8,6 +8,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "dxc/DXIL/DxilConstants.h"
+#include "dxc/DXIL/DxilModule.h"
 #include "dxc/DXIL/DxilOperations.h"
 #include "dxc/DXIL/DxilUtil.h"
 #include "dxc/HLSL/HLModule.h"
@@ -180,10 +181,12 @@ bool LowerTypePass::runOnModule(Module &M) {
 namespace {
 class DynamicIndexingVectorToArray : public LowerTypePass {
   bool ReplaceAllVectors;
+  bool SupportsVectors;
 
 public:
   explicit DynamicIndexingVectorToArray(bool ReplaceAll = false)
-      : LowerTypePass(ID), ReplaceAllVectors(ReplaceAll) {}
+      : LowerTypePass(ID), ReplaceAllVectors(ReplaceAll),
+        SupportsVectors(false) {}
   static char ID; // Pass identification, replacement for typeid
   void applyOptions(PassOptions O) override;
   void dumpConfig(raw_ostream &OS) override;
@@ -194,6 +197,7 @@ class DynamicIndexingVectorToArray : public LowerTypePass {
   Type *lowerType(Type *Ty) override;
   Constant *lowerInitVal(Constant *InitVal, Type *NewTy) override;
   StringRef getGlobalPrefix() override { return ".v"; }
+  void initialize(Module &M) override;
 
 private:
   bool HasVectorDynamicIndexing(Value *V);
@@ -207,6 +211,18 @@ class DynamicIndexingVectorToArray : public LowerTypePass {
   void ReplaceAddrSpaceCast(ConstantExpr *CE, Value *A, IRBuilder<> &Builder);
 };
 
+void DynamicIndexingVectorToArray::initialize(Module &M) {
+  // Set vector support according to available Dxil version.
+  // Use HLModule or metadata for version info.
+  // Otherwise retrieve from dxil module or metadata.
+  unsigned Major = 0, Minor = 0;
+  if (M.HasHLModule())
+    M.GetHLModule().GetShaderModel()->GetDxilVersion(Major, Minor);
+  else
+    dxilutil::LoadDxilVersion(&M, Major, Minor);
+  SupportsVectors = (Major == 1 && Minor >= 9);
+}
+
 void DynamicIndexingVectorToArray::applyOptions(PassOptions O) {
   GetPassOptionBool(O, "ReplaceAllVectors", &ReplaceAllVectors,
                     ReplaceAllVectors);
@@ -306,9 +322,21 @@ void DynamicIndexingVectorToArray::ReplaceStaticIndexingOnVector(Value *V) {
 }
 
 bool DynamicIndexingVectorToArray::needToLower(Value *V) {
+  bool MustReplaceVector = ReplaceAllVectors;
   Type *Ty = V->getType()->getPointerElementType();
-  if (dyn_cast<VectorType>(Ty)) {
-    if (isa<GlobalVariable>(V) || ReplaceAllVectors) {
+
+  if (ArrayType *AT = dyn_cast<ArrayType>(Ty)) {
+    // Array must be replaced even without dynamic indexing to remove vector
+    // type in dxil.
+    MustReplaceVector = true;
+    Ty = dxilutil::GetArrayEltTy(AT);
+  }
+
+  if (isa<VectorType>(Ty)) {
+    // Only needed for 2+ vectors where native vectors unsupported.
+    if (SupportsVectors && Ty->getVectorNumElements() > 1)
+      return false;
+    if (isa<GlobalVariable>(V) || MustReplaceVector) {
       return true;
     }
     // Don't lower local vector which only static indexing.
@@ -319,12 +347,6 @@ bool DynamicIndexingVectorToArray::needToLower(Value *V) {
       ReplaceStaticIndexingOnVector(V);
       return false;
     }
-  } else if (ArrayType *AT = dyn_cast<ArrayType>(Ty)) {
-    // Array must be replaced even without dynamic indexing to remove vector
-    // type in dxil.
-    // TODO: optimize static array index in later pass.
-    Type *EltTy = dxilutil::GetArrayEltTy(AT);
-    return isa<VectorType>(EltTy);
   }
   return false;
 }