[llvm] [InstCombine] Canonicalize `gep T, (gep i8, base, C1), (Index + C2)` into `gep T, (gep i8, base, C1 + C2 * sizeof(T)), Index` (PR #76177)

[Nikita Popov via llvm-commits]

================
@@ -2317,6 +2317,43 @@ static Instruction *foldSelectGEP(GetElementPtrInst &GEP,
   return SelectInst::Create(Cond, NewTrueC, NewFalseC, "", nullptr, Sel);
 }
 
+// Canonicalization:
+// gep T, (gep i8, base, C1), (Index + C2) into
+// gep T, (gep i8, base, C1 + C2 * sizeof(T)), Index
+static Instruction *canonicalizeGEPOfConstGEPI8(GetElementPtrInst &GEP,
+                                                GEPOperator *Src,
+                                                InstCombinerImpl &IC) {
+  if (GEP.getNumIndices() != 1)
+    return nullptr;
+  auto &DL = IC.getDataLayout();
+  Value *Base;
+  const APInt *C1;
+  if (!match(Src, m_PtrAdd(m_Value(Base), m_APInt(C1))))
+    return nullptr;
+  Value *VarIndex;
+  const APInt *C2;
+  Type *PtrTy = Src->getType()->getScalarType();
+  unsigned IndexSizeInBits = DL.getIndexTypeSizeInBits(PtrTy);
+  if (!match(GEP.getOperand(1), m_AddLike(m_Value(VarIndex), m_APInt(C2))))
+    return nullptr;
+  if (C1->getBitWidth() != IndexSizeInBits ||
+      C2->getBitWidth() != IndexSizeInBits)
+    return nullptr;
+  Type *BaseType = GEP.getSourceElementType();
+  if (isa<ScalableVectorType>(BaseType))
+    return nullptr;
+  APInt TypeSize(IndexSizeInBits, DL.getTypeAllocSize(BaseType));
+  APInt NewOffset = TypeSize * *C2 + *C1;
+  if (NewOffset.isZero() ||
+      (Src->hasOneUse() && GEP.getOperand(1)->hasOneUse())) {
+    Value *GEPConst = IC.Builder.CreateGEP(IC.Builder.getInt8Ty(), Base,
----------------
nikic wrote:

```suggestion
    Value *GEPConst = IC.Builder.CreatePtrAdd(Base,
```

https://github.com/llvm/llvm-project/pull/76177