implement an initial hack at a straight-line store -> memset optimization.

This fires dozens of times across spec and multisource, but I don't know
if it actually speeds stuff up.  Hopefully the testers will show something
nice :)



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48680 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Transforms/Scalar/GVN.cpp

@@ -33,6 +33,7 @@
 #include "llvm/Analysis/MemoryDependenceAnalysis.h"
 #include "llvm/Support/CFG.h"
 #include "llvm/Support/Compiler.h"
+#include "llvm/Support/GetElementPtrTypeIterator.h"
 #include "llvm/Target/TargetData.h"
 using namespace llvm;
 
@@ -1036,11 +1037,63 @@ static Value *isBytewiseValue(Value *V) {
   return 0;
 }
 
+static int64_t GetOffsetFromIndex(const GetElementPtrInst *GEP, unsigned Idx,
+                                  bool &VariableIdxFound, TargetData &TD) {
+  // Skip over the first indices.
+  gep_type_iterator GTI = gep_type_begin(GEP);
+  for (unsigned i = 1; i != Idx; ++i, ++GTI)
+    /*skip along*/;
+  
+  // Compute the offset implied by the rest of the indices.
+  int64_t Offset = 0;
+  for (unsigned i = Idx, e = GEP->getNumOperands(); i != e; ++i, ++GTI) {
+    ConstantInt *OpC = dyn_cast<ConstantInt>(GEP->getOperand(i));
+    if (OpC == 0)
+      return VariableIdxFound = true;
+    if (OpC->isZero()) continue;  // No offset.
+
+    // Handle struct indices, which add their field offset to the pointer.
+    if (const StructType *STy = dyn_cast<StructType>(*GTI)) {
+      Offset += TD.getStructLayout(STy)->getElementOffset(OpC->getZExtValue());
+      continue;
+    }
+    
+    // Otherwise, we have a sequential type like an array or vector.  Multiply
+    // the index by the ElementSize.
+    uint64_t Size = TD.getABITypeSize(GTI.getIndexedType());
+    Offset += Size*OpC->getSExtValue();
+  }
+
+  return Offset;
+}
+
 /// IsPointerAtOffset - Return true if Ptr1 is exactly provably equal to Ptr2
 /// plus the specified constant offset.  For example, Ptr1 might be &A[42], and
 /// Ptr2 might be &A[40] and Offset might be 8.
-static bool IsPointerAtOffset(Value *Ptr1, Value *Ptr2, uint64_t Offset) {
-  return true;
+static bool IsPointerAtOffset(Value *Ptr1, Value *Ptr2, uint64_t Offset,
+                              TargetData &TD) {
+  // Right now we handle the case when Ptr1/Ptr2 are both GEPs with an identical
+  // base.  After that base, they may have some number of common (and
+  // potentially variable) indices.  After that they handle some constant
+  // offset, which determines their offset from each other.  At this point, we
+  // handle no other case.
+  GetElementPtrInst *GEP1 = dyn_cast<GetElementPtrInst>(Ptr1);
+  GetElementPtrInst *GEP2 = dyn_cast<GetElementPtrInst>(Ptr2);
+  if (!GEP1 || !GEP2 || GEP1->getOperand(0) != GEP2->getOperand(0))
+    return false;
+  
+  // Skip any common indices and track the GEP types.
+  unsigned Idx = 1;
+  for (; Idx != GEP1->getNumOperands() && Idx != GEP2->getNumOperands(); ++Idx)
+    if (GEP1->getOperand(Idx) != GEP2->getOperand(Idx))
+      break;
+
+  bool VariableIdxFound = false;
+  int64_t Offset1 = GetOffsetFromIndex(GEP1, Idx, VariableIdxFound, TD);
+  int64_t Offset2 = GetOffsetFromIndex(GEP2, Idx, VariableIdxFound, TD);
+  if (VariableIdxFound) return false;
+  
+  return Offset1 == Offset2+(int64_t)Offset;
 }
 
 
@@ -1049,7 +1102,6 @@ static bool IsPointerAtOffset(Value *Ptr1, Value *Ptr2, uint64_t Offset) {
 /// neighboring locations of memory.  If it sees enough consequtive ones
 /// (currently 4) it attempts to merge them together into a memcpy/memset.
 bool GVN::processStore(StoreInst *SI, SmallVectorImpl<Instruction*> &toErase) {
-  return false;
   if (SI->isVolatile()) return false;
   
   // There are two cases that are interesting for this code to handle: memcpy
@@ -1078,7 +1130,7 @@ bool GVN::processStore(StoreInst *SI, SmallVectorImpl<Instruction*> &toErase) {
   SmallVector<StoreInst*, 16> Stores;
   Stores.push_back(SI);
   
-  BasicBlock::iterator BI = SI; ++BI;
+  BasicBlock::iterator BI = SI;
   for (++BI; !isa<TerminatorInst>(BI); ++BI) {
     if (isa<CallInst>(BI) || isa<InvokeInst>(BI)) { 
       // If the call is readnone, ignore it, otherwise bail out.  We don't even
@@ -1110,7 +1162,7 @@ bool GVN::processStore(StoreInst *SI, SmallVectorImpl<Instruction*> &toErase) {
     
     // If so, check to see if the store is before the current range or after it
     // in either case, extend the range, otherwise reject it.
-    if (IsPointerAtOffset(ThisPointer, StartPtr, BytesSoFar)) {
+    if (IsPointerAtOffset(ThisPointer, StartPtr, BytesSoFar, TD)) {
       // Okay, this extends the stored area on the end, just add to the bytes
       // so far and remember this store.
       BytesSoFar += AccessSize;
@@ -1118,7 +1170,7 @@ bool GVN::processStore(StoreInst *SI, SmallVectorImpl<Instruction*> &toErase) {
       continue;
     }
     
-    if (IsPointerAtOffset(StartPtr, ThisPointer, AccessSize)) {
+    if (IsPointerAtOffset(StartPtr, ThisPointer, AccessSize, TD)) {
       // Okay, the store is before the current range.  Reset our start pointer
       // and get new alignment info etc.
       BytesSoFar  += AccessSize;
@@ -1170,6 +1222,7 @@ bool GVN::processStore(StoreInst *SI, SmallVectorImpl<Instruction*> &toErase) {
   };
   new CallInst(F, Ops, Ops+4, "", InsertPt);
   
+  // Zap all the stores.
   toErase.append(Stores.begin(), Stores.end());
   
   ++NumMemSetInfer;

test/Transforms/GVN/form-memset.ll

@@ -0,0 +1,55 @@
+; RUN: llvm-as < %s | opt -gvn | llvm-dis | not grep store
+; RUN: llvm-as < %s | opt -gvn | llvm-dis | grep llvm.memset
+
+; All the stores in this example should be merged into a single memset.
+
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
+target triple = "i386-apple-darwin8"
+
+define void @foo(i8 signext  %c) nounwind  {
+entry:
+	%x = alloca [19 x i8]		; <[19 x i8]*> [#uses=20]
+	%tmp = getelementptr [19 x i8]* %x, i32 0, i32 0		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp, align 1
+	%tmp5 = getelementptr [19 x i8]* %x, i32 0, i32 1		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp5, align 1
+	%tmp9 = getelementptr [19 x i8]* %x, i32 0, i32 2		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp9, align 1
+	%tmp13 = getelementptr [19 x i8]* %x, i32 0, i32 3		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp13, align 1
+	%tmp17 = getelementptr [19 x i8]* %x, i32 0, i32 4		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp17, align 1
+	%tmp21 = getelementptr [19 x i8]* %x, i32 0, i32 5		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp21, align 1
+	%tmp25 = getelementptr [19 x i8]* %x, i32 0, i32 6		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp25, align 1
+	%tmp29 = getelementptr [19 x i8]* %x, i32 0, i32 7		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp29, align 1
+	%tmp33 = getelementptr [19 x i8]* %x, i32 0, i32 8		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp33, align 1
+	%tmp37 = getelementptr [19 x i8]* %x, i32 0, i32 9		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp37, align 1
+	%tmp41 = getelementptr [19 x i8]* %x, i32 0, i32 10		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp41, align 1
+	%tmp45 = getelementptr [19 x i8]* %x, i32 0, i32 11		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp45, align 1
+	%tmp49 = getelementptr [19 x i8]* %x, i32 0, i32 12		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp49, align 1
+	%tmp53 = getelementptr [19 x i8]* %x, i32 0, i32 13		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp53, align 1
+	%tmp57 = getelementptr [19 x i8]* %x, i32 0, i32 14		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp57, align 1
+	%tmp61 = getelementptr [19 x i8]* %x, i32 0, i32 15		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp61, align 1
+	%tmp65 = getelementptr [19 x i8]* %x, i32 0, i32 16		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp65, align 1
+	%tmp69 = getelementptr [19 x i8]* %x, i32 0, i32 17		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp69, align 1
+	%tmp73 = getelementptr [19 x i8]* %x, i32 0, i32 18		; <i8*> [#uses=1]
+	store i8 %c, i8* %tmp73, align 1
+	%tmp76 = call i32 (...)* @bar( [19 x i8]* %x ) nounwind 		; <i32> [#uses=0]
+	ret void
+}
+
+declare i32 @bar(...)
+