[llvm] 3e64db5 - [ConstantFolding] Support ptrtoaddr in cast folds (#162480)
via llvm-commits
llvm-commits at lists.llvm.org
Thu Oct 9 01:22:35 PDT 2025
Author: Nikita Popov
Date: 2025-10-09T08:22:31Z
New Revision: 3e64db583d1120f31a52d87a1d414c41e7e3d5cb
URL: https://github.com/llvm/llvm-project/commit/3e64db583d1120f31a52d87a1d414c41e7e3d5cb
DIFF: https://github.com/llvm/llvm-project/commit/3e64db583d1120f31a52d87a1d414c41e7e3d5cb.diff
LOG: [ConstantFolding] Support ptrtoaddr in cast folds (#162480)
We can support the same folds for as for ptrtoint. For the cast pair
fold we just need to use the address type instead. The gep based folds
were already operating on the address (aka index) width anyway.
Added:
Modified:
llvm/lib/Analysis/ConstantFolding.cpp
llvm/test/Transforms/InstCombine/ptrtoaddr.ll
Removed:
################################################################################
diff --git a/llvm/lib/Analysis/ConstantFolding.cpp b/llvm/lib/Analysis/ConstantFolding.cpp
index b744537aeb474..31546e628b226 100755
--- a/llvm/lib/Analysis/ConstantFolding.cpp
+++ b/llvm/lib/Analysis/ConstantFolding.cpp
@@ -1495,22 +1495,22 @@ Constant *llvm::ConstantFoldCastOperand(unsigned Opcode, Constant *C,
default:
llvm_unreachable("Missing case");
case Instruction::PtrToAddr:
- // TODO: Add some of the ptrtoint folds here as well.
- break;
case Instruction::PtrToInt:
if (auto *CE = dyn_cast<ConstantExpr>(C)) {
Constant *FoldedValue = nullptr;
- // If the input is a inttoptr, eliminate the pair. This requires knowing
+ // If the input is an inttoptr, eliminate the pair. This requires knowing
// the width of a pointer, so it can't be done in ConstantExpr::getCast.
if (CE->getOpcode() == Instruction::IntToPtr) {
- // zext/trunc the inttoptr to pointer size.
- FoldedValue = ConstantFoldIntegerCast(CE->getOperand(0),
- DL.getIntPtrType(CE->getType()),
+ // zext/trunc the inttoptr to pointer/address size.
+ Type *MidTy = Opcode == Instruction::PtrToInt
+ ? DL.getAddressType(CE->getType())
+ : DL.getIntPtrType(CE->getType());
+ FoldedValue = ConstantFoldIntegerCast(CE->getOperand(0), MidTy,
/*IsSigned=*/false, DL);
} else if (auto *GEP = dyn_cast<GEPOperator>(CE)) {
// If we have GEP, we can perform the following folds:
- // (ptrtoint (gep null, x)) -> x
- // (ptrtoint (gep (gep null, x), y) -> x + y, etc.
+ // (ptrtoint/ptrtoaddr (gep null, x)) -> x
+ // (ptrtoint/ptrtoaddr (gep (gep null, x), y) -> x + y, etc.
unsigned BitWidth = DL.getIndexTypeSizeInBits(GEP->getType());
APInt BaseOffset(BitWidth, 0);
auto *Base = cast<Constant>(GEP->stripAndAccumulateConstantOffsets(
@@ -1518,7 +1518,8 @@ Constant *llvm::ConstantFoldCastOperand(unsigned Opcode, Constant *C,
if (Base->isNullValue()) {
FoldedValue = ConstantInt::get(CE->getContext(), BaseOffset);
} else {
- // ptrtoint (gep i8, Ptr, (sub 0, V)) -> sub (ptrtoint Ptr), V
+ // ptrtoint/ptrtoaddr (gep i8, Ptr, (sub 0, V))
+ // -> sub (ptrtoint/ptrtoaddr Ptr), V
if (GEP->getNumIndices() == 1 &&
GEP->getSourceElementType()->isIntegerTy(8)) {
auto *Ptr = cast<Constant>(GEP->getPointerOperand());
@@ -1528,12 +1529,13 @@ Constant *llvm::ConstantFoldCastOperand(unsigned Opcode, Constant *C,
Sub->getOpcode() == Instruction::Sub &&
Sub->getOperand(0)->isNullValue())
FoldedValue = ConstantExpr::getSub(
- ConstantExpr::getPtrToInt(Ptr, IntIdxTy), Sub->getOperand(1));
+ ConstantExpr::getCast(Opcode, Ptr, IntIdxTy),
+ Sub->getOperand(1));
}
}
}
if (FoldedValue) {
- // Do a zext or trunc to get to the ptrtoint dest size.
+ // Do a zext or trunc to get to the ptrtoint/ptrtoaddr dest size.
return ConstantFoldIntegerCast(FoldedValue, DestTy, /*IsSigned=*/false,
DL);
}
diff --git a/llvm/test/Transforms/InstCombine/ptrtoaddr.ll b/llvm/test/Transforms/InstCombine/ptrtoaddr.ll
index 61b13312521d2..f76a1399f8f2a 100644
--- a/llvm/test/Transforms/InstCombine/ptrtoaddr.ll
+++ b/llvm/test/Transforms/InstCombine/ptrtoaddr.ll
@@ -1,6 +1,14 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
; RUN: opt < %s -passes=instcombine -S | FileCheck %s
-target datalayout = "p1:64:64:64:32"
+
+; The ptrtoaddr folds are also valid for pointers that have external state.
+target datalayout = "pe1:64:64:64:32"
+
+ at g = external global i8
+ at g2 = external global i8
+
+ at g.as1 = external addrspace(1) global i8
+ at g2.as1 = external addrspace(1) global i8
define i32 @ptrtoaddr_inttoptr_arg(i32 %a) {
; CHECK-LABEL: define i32 @ptrtoaddr_inttoptr_arg(
@@ -24,14 +32,14 @@ define i32 @ptrtoaddr_inttoptr() {
define i32 @ptrtoaddr_inttoptr_
diff _size1() {
; CHECK-LABEL: define i32 @ptrtoaddr_inttoptr_
diff _size1() {
-; CHECK-NEXT: ret i32 ptrtoaddr (ptr addrspace(1) inttoptr (i64 -1 to ptr addrspace(1)) to i32)
+; CHECK-NEXT: ret i32 -1
;
ret i32 ptrtoaddr (ptr addrspace(1) inttoptr (i64 -1 to ptr addrspace(1)) to i32)
}
define i32 @ptrtoaddr_inttoptr_
diff _size2() {
; CHECK-LABEL: define i32 @ptrtoaddr_inttoptr_
diff _size2() {
-; CHECK-NEXT: ret i32 ptrtoaddr (ptr addrspace(1) inttoptr (i16 -1 to ptr addrspace(1)) to i32)
+; CHECK-NEXT: ret i32 65535
;
ret i32 ptrtoaddr (ptr addrspace(1) inttoptr (i16 -1 to ptr addrspace(1)) to i32)
}
@@ -52,14 +60,42 @@ define i64 @ptr2addr2_inttoptr_noas2() {
define i64 @ptrtoaddr_inttoptr_noas_
diff _size1() {
; CHECK-LABEL: define i64 @ptrtoaddr_inttoptr_noas_
diff _size1() {
-; CHECK-NEXT: ret i64 ptrtoaddr (ptr inttoptr (i32 -1 to ptr) to i64)
+; CHECK-NEXT: ret i64 4294967295
;
ret i64 ptrtoaddr (ptr inttoptr (i32 -1 to ptr) to i64)
}
define i64 @ptrtoaddr_inttoptr_noas_
diff _size2() {
; CHECK-LABEL: define i64 @ptrtoaddr_inttoptr_noas_
diff _size2() {
-; CHECK-NEXT: ret i64 ptrtoaddr (ptr inttoptr (i128 -1 to ptr) to i64)
+; CHECK-NEXT: ret i64 -1
;
ret i64 ptrtoaddr (ptr inttoptr (i128 -1 to ptr) to i64)
}
+
+define i64 @ptrtoaddr_gep_null() {
+; CHECK-LABEL: define i64 @ptrtoaddr_gep_null() {
+; CHECK-NEXT: ret i64 42
+;
+ ret i64 ptrtoaddr (ptr getelementptr (i8, ptr null, i64 42) to i64)
+}
+
+define i32 @ptrtoaddr_gep_null_addrsize() {
+; CHECK-LABEL: define i32 @ptrtoaddr_gep_null_addrsize() {
+; CHECK-NEXT: ret i32 42
+;
+ ret i32 ptrtoaddr (ptr addrspace(1) getelementptr (i8, ptr addrspace(1) null, i32 42) to i32)
+}
+
+define i64 @ptrtoaddr_gep_sub() {
+; CHECK-LABEL: define i64 @ptrtoaddr_gep_sub() {
+; CHECK-NEXT: ret i64 sub (i64 ptrtoaddr (ptr @g to i64), i64 ptrtoaddr (ptr @g2 to i64))
+;
+ ret i64 ptrtoaddr (ptr getelementptr (i8, ptr @g, i64 sub (i64 0, i64 ptrtoaddr (ptr @g2 to i64))) to i64)
+}
+
+define i32 @ptrtoaddr_gep_sub_addrsize() {
+; CHECK-LABEL: define i32 @ptrtoaddr_gep_sub_addrsize() {
+; CHECK-NEXT: ret i32 sub (i32 ptrtoaddr (ptr addrspace(1) @g.as1 to i32), i32 ptrtoaddr (ptr addrspace(1) @g2.as1 to i32))
+;
+ ret i32 ptrtoaddr (ptr addrspace(1) getelementptr (i8, ptr addrspace(1) @g.as1, i32 sub (i32 0, i32 ptrtoaddr (ptr addrspace(1) @g2.as1 to i32))) to i32)
+}
More information about the llvm-commits
mailing list