Executive summary: getTypeSize -> getTypeStoreSize / getABITypeSize.

The meaning of getTypeSize was not clear - clarifying it is important
now that we have x86 long double and arbitrary precision integers.
The issue with long double is that it requires 80 bits, and this is
not a multiple of its alignment.  This gives a primitive type for
which getTypeSize differed from getABITypeSize.  For arbitrary precision
integers it is even worse: there is the minimum number of bits needed to
hold the type (eg: 36 for an i36), the maximum number of bits that will
be overwriten when storing the type (40 bits for i36) and the ABI size
(i.e. the storage size rounded up to a multiple of the alignment; 64 bits
for i36).

This patch removes getTypeSize (not really - it is still there but
deprecated to allow for a gradual transition).  Instead there is:

(1) getTypeSizeInBits - a number of bits that suffices to hold all
values of the type.  For a primitive type, this is the minimum number
of bits.  For an i36 this is 36 bits.  For x86 long double it is 80.
This corresponds to gcc's TYPE_PRECISION.

(2) getTypeStoreSizeInBits - the maximum number of bits that is
written when storing the type (or read when reading it).  For an
i36 this is 40 bits, for an x86 long double it is 80 bits.  This
is the size alias analysis is interested in (getTypeStoreSize
returns the number of bytes).  There doesn't seem to be anything
corresponding to this in gcc.

(3) getABITypeSizeInBits - this is getTypeStoreSizeInBits rounded
up to a multiple of the alignment.  For an i36 this is 64, for an
x86 long double this is 96 or 128 depending on the OS.  This is the
spacing between consecutive elements when you form an array out of
this type (getABITypeSize returns the number of bytes).  This is
TYPE_SIZE in gcc.

Since successive elements in a SequentialType (arrays, pointers
and vectors) need to be aligned, the spacing between them will be
given by getABITypeSize.  This means that the size of an array
is the length times the getABITypeSize.  It also means that GEP
computations need to use getABITypeSize when computing offsets.
Furthermore, if an alloca allocates several elements at once then
these too need to be aligned, so the size of the alloca has to be
the number of elements multiplied by getABITypeSize.  Logically
speaking this doesn't have to be the case when allocating just
one element, but it is simpler to also use getABITypeSize in this
case.  So alloca's and mallocs should use getABITypeSize.  Finally,
since gcc's only notion of size is that given by getABITypeSize, if
you want to output assembler etc the same as gcc then getABITypeSize
is the size you want.

Since a store will overwrite no more than getTypeStoreSize bytes,
and a read will read no more than that many bytes, this is the
notion of size appropriate for alias analysis calculations.

In this patch I have corrected all type size uses except some of
those in ScalarReplAggregates, lib/Codegen, lib/Target (the hard
cases).  I will get around to auditing these too at some point,
but I could do with some help.

Finally, I made one change which I think wise but others might
consider pointless and suboptimal: in an unpacked struct the
amount of space allocated for a field is now given by the ABI
size rather than getTypeStoreSize.  I did this because every
other place that reserves memory for a type (eg: alloca) now
uses getABITypeSize, and I didn't want to make an exception
for unpacked structs, i.e. I did it to make things more uniform.
This only effects structs containing long doubles and arbitrary
precision integers.  If someone wants to pack these types more
tightly they can always use a packed struct.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@43620 91177308-0d34-0410-b5e6-96231b3b80d8

8 files changed, 26 insertions, 29 deletions

diff --git a/lib/Transforms/IPO/ArgumentPromotion.cpp b/lib/Transforms/IPO/ArgumentPromotion.cpp
index 85b29f871f..7479c8ee67 100644
--- a/lib/Transforms/IPO/ArgumentPromotion.cpp
+++ b/lib/Transforms/IPO/ArgumentPromotion.cpp
@@ -277,7 +277,7 @@ bool ArgPromotion::isSafeToPromoteArgument(Argument *Arg) const {
 
     const PointerType *LoadTy =
       cast<PointerType>(Load->getOperand(0)->getType());
-    unsigned LoadSize = (unsigned)TD.getTypeSize(LoadTy->getElementType());
+    unsigned LoadSize = (unsigned)TD.getTypeStoreSize(LoadTy->getElementType());
 
     if (AA.canInstructionRangeModify(BB->front(), *Load, Arg, LoadSize))
       return false;  // Pointer is invalidated!
diff --git a/lib/Transforms/IPO/GlobalOpt.cpp b/lib/Transforms/IPO/GlobalOpt.cpp
index 5d8f969f33..779b4a1871 100644
--- a/lib/Transforms/IPO/GlobalOpt.cpp
+++ b/lib/Transforms/IPO/GlobalOpt.cpp
@@ -1227,7 +1227,7 @@ static bool OptimizeOnceStoredGlobal(GlobalVariable *GV, Value *StoredOnceVal,
         // (2048 bytes currently), as we don't want to introduce a 16M global or
         // something.
         if (NElements->getZExtValue()*
-                     TD.getTypeSize(MI->getAllocatedType()) < 2048) {
+                     TD.getABITypeSize(MI->getAllocatedType()) < 2048) {
           GVI = OptimizeGlobalAddressOfMalloc(GV, MI);
           return true;
         }
diff --git a/lib/Transforms/Scalar/CodeGenPrepare.cpp b/lib/Transforms/Scalar/CodeGenPrepare.cpp
index c2a0787236..0dad42f8e2 100644
--- a/lib/Transforms/Scalar/CodeGenPrepare.cpp
+++ b/lib/Transforms/Scalar/CodeGenPrepare.cpp
@@ -634,7 +634,7 @@ static bool FindMaximalLegalAddressingMode(Value *Addr, const Type *AccessTy,
           cast<ConstantInt>(AddrInst->getOperand(i))->getZExtValue();
         ConstantOffset += SL->getElementOffset(Idx);
       } else {
-        uint64_t TypeSize = TD->getTypeSize(GTI.getIndexedType());
+        uint64_t TypeSize = TD->getABITypeSize(GTI.getIndexedType());
         if (ConstantInt *CI = dyn_cast<ConstantInt>(AddrInst->getOperand(i))) {
           ConstantOffset += CI->getSExtValue()*TypeSize;
         } else if (TypeSize) {  // Scales of zero don't do anything.
diff --git a/lib/Transforms/Scalar/DeadStoreElimination.cpp b/lib/Transforms/Scalar/DeadStoreElimination.cpp
index 2e1d9ade0a..e5c557c349 100644
--- a/lib/Transforms/Scalar/DeadStoreElimination.cpp
+++ b/lib/Transforms/Scalar/DeadStoreElimination.cpp
@@ -137,8 +137,8 @@ bool DSE::runOnBasicBlock(BasicBlock &BB) {
              dep != MemoryDependenceAnalysis::NonLocal &&
              isa<StoreInst>(dep)) {
         if (dep != last ||
-             TD.getTypeSize(last->getOperand(0)->getType()) >
-             TD.getTypeSize(BBI->getOperand(0)->getType())) {
+             TD.getTypeStoreSize(last->getOperand(0)->getType()) >
+             TD.getTypeStoreSize(BBI->getOperand(0)->getType())) {
           dep = MD.getDependency(BBI, dep);
           continue;
         }
@@ -210,7 +210,7 @@ bool DSE::handleFreeWithNonTrivialDependency(FreeInst* F, Instruction* dep,
   
   Value* depPointer = dependency->getPointerOperand();
   const Type* depType = dependency->getOperand(0)->getType();
-  unsigned depPointerSize = TD.getTypeSize(depType);
+  unsigned depPointerSize = TD.getTypeStoreSize(depType);
   
   // Check for aliasing
   AliasAnalysis::AliasResult A = AA.alias(F->getPointerOperand(), ~0UL,
@@ -329,7 +329,7 @@ bool DSE::handleEndBlock(BasicBlock& BB,
         unsigned pointerSize = ~0UL;
         if (ConstantInt* C = dyn_cast<ConstantInt>((*I)->getArraySize()))
           pointerSize = C->getZExtValue() * \
-                        TD.getTypeSize((*I)->getAllocatedType());     
+                        TD.getABITypeSize((*I)->getAllocatedType());
         
         // See if the call site touches it
         AliasAnalysis::ModRefResult A = AA.getModRefInfo(CS, *I, pointerSize);
@@ -394,7 +394,7 @@ bool DSE::RemoveUndeadPointers(Value* killPointer,
     unsigned pointerSize = ~0UL;
     if (ConstantInt* C = dyn_cast<ConstantInt>((*I)->getArraySize()))
       pointerSize = C->getZExtValue() * \
-                    TD.getTypeSize((*I)->getAllocatedType());     
+                    TD.getABITypeSize((*I)->getAllocatedType());
       
     // See if this pointer could alias it
     AliasAnalysis::AliasResult A = AA.alias(*I, pointerSize,
diff --git a/lib/Transforms/Scalar/InstructionCombining.cpp b/lib/Transforms/Scalar/InstructionCombining.cpp
index 12523cd3fe..6ebf42a96d 100644
--- a/lib/Transforms/Scalar/InstructionCombining.cpp
+++ b/lib/Transforms/Scalar/InstructionCombining.cpp
@@ -4438,7 +4438,7 @@ static Value *EmitGEPOffset(User *GEP, Instruction &I, InstCombiner &IC) {
 
   for (unsigned i = 1, e = GEP->getNumOperands(); i != e; ++i, ++GTI) {
     Value *Op = GEP->getOperand(i);
-    uint64_t Size = TD.getTypeSize(GTI.getIndexedType()) & PtrSizeMask;
+    uint64_t Size = TD.getABITypeSize(GTI.getIndexedType()) & PtrSizeMask;
     if (ConstantInt *OpC = dyn_cast<ConstantInt>(Op)) {
       if (OpC->isZero()) continue;
       
@@ -4523,7 +4523,7 @@ Instruction *InstCombiner::FoldGEPICmp(User *GEPLHS, Value *RHS,
             return ReplaceInstUsesWith(I, UndefValue::get(I.getType()));
           if (C->isNullValue())
             EmitIt = false;
-          else if (TD->getTypeSize(GTI.getIndexedType()) == 0) {
+          else if (TD->getABITypeSize(GTI.getIndexedType()) == 0) {
             EmitIt = false;  // This is indexing into a zero sized array?
           } else if (isa<ConstantInt>(C))
             return ReplaceInstUsesWith(I, // No comparison is needed here.
@@ -6305,8 +6305,8 @@ Instruction *InstCombiner::PromoteCastOfAllocation(BitCastInst &CI,
   // same, we open the door to infinite loops of various kinds.
   if (!AI.hasOneUse() && CastElTyAlign == AllocElTyAlign) return 0;
 
-  uint64_t AllocElTySize = TD->getTypeSize(AllocElTy);
-  uint64_t CastElTySize = TD->getTypeSize(CastElTy);
+  uint64_t AllocElTySize = TD->getABITypeSize(AllocElTy);
+  uint64_t CastElTySize = TD->getABITypeSize(CastElTy);
   if (CastElTySize == 0 || AllocElTySize == 0) return 0;
 
   // See if we can satisfy the modulus by pulling a scale out of the array
@@ -6573,7 +6573,7 @@ Instruction *InstCombiner::commonPointerCastTransforms(CastInst &CI) {
           // is something like [0 x {int, int}]
           const Type *IntPtrTy = TD->getIntPtrType();
           int64_t FirstIdx = 0;
-          if (int64_t TySize = TD->getTypeSize(GEPIdxTy)) {
+          if (int64_t TySize = TD->getABITypeSize(GEPIdxTy)) {
             FirstIdx = Offset/TySize;
             Offset %= TySize;
           
@@ -6605,7 +6605,7 @@ Instruction *InstCombiner::commonPointerCastTransforms(CastInst &CI) {
               }
             } else if (isa<ArrayType>(GEPIdxTy) || isa<VectorType>(GEPIdxTy)) {
               const SequentialType *STy = cast<SequentialType>(GEPIdxTy);
-              if (uint64_t EltSize = TD->getTypeSize(STy->getElementType())) {
+              if (uint64_t EltSize = TD->getABITypeSize(STy->getElementType())){
                 NewIndices.push_back(ConstantInt::get(IntPtrTy,Offset/EltSize));
                 Offset %= EltSize;
               } else {
@@ -8644,7 +8644,7 @@ Instruction *InstCombiner::visitGetElementPtrInst(GetElementPtrInst &GEP) {
       // insert it.  This explicit cast can make subsequent optimizations more
       // obvious.
       Value *Op = GEP.getOperand(i);
-      if (TD->getTypeSize(Op->getType()) > TD->getPointerSize())
+      if (TD->getTypeSizeInBits(Op->getType()) > TD->getPointerSizeInBits())
         if (Constant *C = dyn_cast<Constant>(Op)) {
           GEP.setOperand(i, ConstantExpr::getTrunc(C, TD->getIntPtrType()));
           MadeChange = true;
@@ -8724,12 +8724,12 @@ Instruction *InstCombiner::visitGetElementPtrInst(GetElementPtrInst &GEP) {
           } else if (Constant *GO1C = dyn_cast<Constant>(GO1)) {
             GO1 = ConstantExpr::getIntegerCast(GO1C, SO1->getType(), true);
           } else {
-            unsigned PS = TD->getPointerSize();
-            if (TD->getTypeSize(SO1->getType()) == PS) {
+            unsigned PS = TD->getPointerSizeInBits();
+            if (TD->getTypeSizeInBits(SO1->getType()) == PS) {
               // Convert GO1 to SO1's type.
               GO1 = InsertCastToIntPtrTy(GO1, SO1->getType(), &GEP, this);
 
-            } else if (TD->getTypeSize(GO1->getType()) == PS) {
+            } else if (TD->getTypeSizeInBits(GO1->getType()) == PS) {
               // Convert SO1 to GO1's type.
               SO1 = InsertCastToIntPtrTy(SO1, GO1->getType(), &GEP, this);
             } else {
@@ -8818,8 +8818,8 @@ Instruction *InstCombiner::visitGetElementPtrInst(GetElementPtrInst &GEP) {
       const Type *SrcElTy = cast<PointerType>(X->getType())->getElementType();
       const Type *ResElTy=cast<PointerType>(PtrOp->getType())->getElementType();
       if (isa<ArrayType>(SrcElTy) &&
-          TD->getTypeSize(cast<ArrayType>(SrcElTy)->getElementType()) ==
-          TD->getTypeSize(ResElTy)) {
+          TD->getABITypeSize(cast<ArrayType>(SrcElTy)->getElementType()) ==
+          TD->getABITypeSize(ResElTy)) {
         Value *Idx[2];
         Idx[0] = Constant::getNullValue(Type::Int32Ty);
         Idx[1] = GEP.getOperand(1);
@@ -8837,7 +8837,7 @@ Instruction *InstCombiner::visitGetElementPtrInst(GetElementPtrInst &GEP) {
       if (isa<ArrayType>(SrcElTy) &&
           (ResElTy == Type::Int8Ty || ResElTy == Type::Int8Ty)) {
         uint64_t ArrayEltSize =
-            TD->getTypeSize(cast<ArrayType>(SrcElTy)->getElementType());
+            TD->getABITypeSize(cast<ArrayType>(SrcElTy)->getElementType());
         
         // Check to see if "tmp" is a scale by a multiple of ArrayEltSize.  We
         // allow either a mul, shift, or constant here.
@@ -8938,7 +8938,7 @@ Instruction *InstCombiner::visitAllocationInst(AllocationInst &AI) {
   // Note that we only do this for alloca's, because malloc should allocate and
   // return a unique pointer, even for a zero byte allocation.
   if (isa<AllocaInst>(AI) && AI.getAllocatedType()->isSized() &&
-      TD->getTypeSize(AI.getAllocatedType()) == 0)
+      TD->getABITypeSize(AI.getAllocatedType()) == 0)
     return ReplaceInstUsesWith(AI, Constant::getNullValue(AI.getType()));
 
   return 0;
diff --git a/lib/Transforms/Scalar/LICM.cpp b/lib/Transforms/Scalar/LICM.cpp
index 5311a7f04f..08c0a88851 100644
--- a/lib/Transforms/Scalar/LICM.cpp
+++ b/lib/Transforms/Scalar/LICM.cpp
@@ -366,7 +366,7 @@ bool LICM::canSinkOrHoistInst(Instruction &I) {
     // Don't hoist loads which have may-aliased stores in loop.
     unsigned Size = 0;
     if (LI->getType()->isSized())
-      Size = AA->getTargetData().getTypeSize(LI->getType());
+      Size = AA->getTargetData().getTypeStoreSize(LI->getType());
     return !pointerInvalidatedByLoop(LI->getOperand(0), Size);
   } else if (CallInst *CI = dyn_cast<CallInst>(&I)) {
     // Handle obvious cases efficiently.
diff --git a/lib/Transforms/Scalar/PredicateSimplifier.cpp b/lib/Transforms/Scalar/PredicateSimplifier.cpp
index 3723bcbb0a..e84f096fbb 100644
--- a/lib/Transforms/Scalar/PredicateSimplifier.cpp
+++ b/lib/Transforms/Scalar/PredicateSimplifier.cpp
@@ -1120,11 +1120,8 @@ namespace {
     uint32_t typeToWidth(const Type *Ty) const {
       if (TD)
         return TD->getTypeSizeInBits(Ty);
-
-      if (const IntegerType *ITy = dyn_cast<IntegerType>(Ty))
-        return ITy->getBitWidth();
-
-      return 0;
+      else
+        return Ty->getPrimitiveSizeInBits();
     }
 
     static bool isRelatedBy(const ConstantRange &CR1, const ConstantRange &CR2,
diff --git a/lib/Transforms/Utils/LowerAllocations.cpp b/lib/Transforms/Utils/LowerAllocations.cpp
index edc4c8a96f..b089cd6d8b 100644
--- a/lib/Transforms/Utils/LowerAllocations.cpp
+++ b/lib/Transforms/Utils/LowerAllocations.cpp
@@ -116,7 +116,7 @@ bool LowerAllocations::runOnBasicBlock(BasicBlock &BB) {
       // malloc(type) becomes sbyte *malloc(size)
       Value *MallocArg;
       if (LowerMallocArgToInteger)
-        MallocArg = ConstantInt::get(Type::Int64Ty, TD.getTypeSize(AllocTy));
+        MallocArg = ConstantInt::get(Type::Int64Ty, TD.getABITypeSize(AllocTy));
       else
         MallocArg = ConstantExpr::getSizeOf(AllocTy);
       MallocArg = ConstantExpr::getTruncOrBitCast(cast<Constant>(MallocArg),