Skip to content

[ConstantFolding] Avoid use of isNonIntegralPointerType() #159959

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 8 commits into from
Sep 23, 2025
Merged

[ConstantFolding] Avoid use of isNonIntegralPointerType() #159959

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
acd5695
[𝘀𝗽𝗿] changes to main this commit is based on
arichardson Sep 21, 2025
cccf6db
[𝘀𝗽𝗿] initial version
arichardson Sep 21, 2025
e866587
[𝘀𝗽𝗿] changes introduced through rebase
arichardson Sep 21, 2025
7005664
fix unused variable
arichardson Sep 21, 2025
690f308
fix typo in comment
arichardson Sep 21, 2025
d18eb5e
spelling
arichardson Sep 22, 2025
6dd1ea4
[𝘀𝗽𝗿] changes introduced through rebase
arichardson Sep 23, 2025
f705aa7
add vector test cases
arichardson Sep 23, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 7 additions & 7 deletions llvm/lib/Analysis/ConstantFolding.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -951,21 +951,21 @@ Constant *SymbolicallyEvaluateGEP(const GEPOperator *GEP,

// If the base value for this address is a literal integer value, fold the
// getelementptr to the resulting integer value casted to the pointer type.
APInt BasePtr(DL.getPointerTypeSizeInBits(Ptr->getType()), 0);
APInt BaseIntVal(DL.getPointerTypeSizeInBits(Ptr->getType()), 0);
if (auto *CE = dyn_cast<ConstantExpr>(Ptr)) {
if (CE->getOpcode() == Instruction::IntToPtr) {
if (auto *Base = dyn_cast<ConstantInt>(CE->getOperand(0)))
BasePtr = Base->getValue().zextOrTrunc(BasePtr.getBitWidth());
BaseIntVal = Base->getValue().zextOrTrunc(BaseIntVal.getBitWidth());
}
}

auto *PTy = cast<PointerType>(Ptr->getType());
if ((Ptr->isNullValue() || BasePtr != 0) &&
!DL.isNonIntegralPointerType(PTy)) {
if ((Ptr->isNullValue() || BaseIntVal != 0) &&
!DL.mustNotIntroduceIntToPtr(Ptr->getType())) {

// If the index size is smaller than the pointer size, add to the low
// bits only.
BasePtr.insertBits(BasePtr.trunc(BitWidth) + Offset, 0);
Constant *C = ConstantInt::get(Ptr->getContext(), BasePtr);
BaseIntVal.insertBits(BaseIntVal.trunc(BitWidth) + Offset, 0);
Constant *C = ConstantInt::get(Ptr->getContext(), BaseIntVal);
return ConstantExpr::getIntToPtr(C, ResTy);
}

Expand Down
16 changes: 0 additions & 16 deletions llvm/test/Transforms/InstSimplify/ConstProp/inttoptr-gep-index-width.ll
View file
Open in desktop

This file was deleted.

145 changes: 145 additions & 0 deletions llvm/test/Transforms/InstSimplify/ConstProp/inttoptr-gep-nonintegral.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,145 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt -S -passes=instsimplify < %s | FileCheck %s
;; Check that we do not create new inttoptr instructions for unstable pointers
;; or pointers with external state (even if the values are all constants).
;; NOTE: for all but the zero address space, the GEP should only modify the
;; low 8 bits of the pointer.
target datalayout = "p:16:16:16:16-p1:16:16:16:8-pu2:16:16:16:8-pe3:16:16:16:8"

define ptr @test_null_base_normal() {
; CHECK-LABEL: define ptr @test_null_base_normal() {
; CHECK-NEXT: ret ptr inttoptr (i16 -2 to ptr)
;
%gep = getelementptr i8, ptr null, i8 -2
ret ptr %gep
}
define ptr @test_inttoptr_base_normal() {
; CHECK-LABEL: define ptr @test_inttoptr_base_normal() {
; CHECK-NEXT: ret ptr null
;
%base = inttoptr i16 -1 to ptr
%gep = getelementptr i8, ptr %base, i8 1
ret ptr %gep
}

;; Transformation is fine for non-integral address space, but we can only change
;; the index bits: (i8 -2 == i16 254)
define ptr addrspace(1) @test_null_base_nonintegral() {
; CHECK-LABEL: define ptr addrspace(1) @test_null_base_nonintegral() {
; CHECK-NEXT: ret ptr addrspace(1) inttoptr (i16 254 to ptr addrspace(1))
;
%gep = getelementptr i8, ptr addrspace(1) null, i8 -2
ret ptr addrspace(1) %gep
}
define ptr addrspace(1) @test_inttoptr_base_nonintegral() {
; CHECK-LABEL: define ptr addrspace(1) @test_inttoptr_base_nonintegral() {
; CHECK-NEXT: ret ptr addrspace(1) inttoptr (i16 -256 to ptr addrspace(1))
;
%base = inttoptr i16 -1 to ptr addrspace(1)
%gep = getelementptr i8, ptr addrspace(1) %base, i8 1
ret ptr addrspace(1) %gep
}

;; For unstable pointers we should avoid any introduction of inttoptr
define ptr addrspace(2) @test_null_base_unstable() {
; CHECK-LABEL: define ptr addrspace(2) @test_null_base_unstable() {
; CHECK-NEXT: ret ptr addrspace(2) getelementptr (i8, ptr addrspace(2) null, i8 -2)
;
%gep = getelementptr i8, ptr addrspace(2) null, i8 -2
ret ptr addrspace(2) %gep
}
define ptr addrspace(2) @test_inttoptr_base_unstable() {
; CHECK-LABEL: define ptr addrspace(2) @test_inttoptr_base_unstable() {
; CHECK-NEXT: ret ptr addrspace(2) getelementptr (i8, ptr addrspace(2) inttoptr (i16 -1 to ptr addrspace(2)), i8 1)
;
%base = inttoptr i16 -1 to ptr addrspace(2)
%gep = getelementptr i8, ptr addrspace(2) %base, i8 1
ret ptr addrspace(2) %gep
}

;; The same is true for pointers with external state: no new inttoptr
define ptr addrspace(3) @test_null_base_external() {
; CHECK-LABEL: define ptr addrspace(3) @test_null_base_external() {
; CHECK-NEXT: ret ptr addrspace(3) getelementptr (i8, ptr addrspace(3) null, i8 -2)
;
%gep = getelementptr i8, ptr addrspace(3) null, i8 -2
ret ptr addrspace(3) %gep
}

define ptr addrspace(3) @test_inttoptr_base_external() {
; CHECK-LABEL: define ptr addrspace(3) @test_inttoptr_base_external() {
; CHECK-NEXT: ret ptr addrspace(3) getelementptr (i8, ptr addrspace(3) inttoptr (i16 -1 to ptr addrspace(3)), i8 1)
Copy link
Member Author

@arichardson arichardson Sep 21, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm not sure this really matters for CHERI, we could also emit the inttoptr (i16 -256 to ptr addrspace(1)) directly, but maybe it's cleaner to not attempt to perform any inttoptr based transforms?

;
%base = inttoptr i16 -1 to ptr addrspace(3)
%gep = getelementptr i8, ptr addrspace(3) %base, i8 1
ret ptr addrspace(3) %gep
}

define <2 x ptr> @test_vec_null_base_normal() {
; CHECK-LABEL: define <2 x ptr> @test_vec_null_base_normal() {
; CHECK-NEXT: ret <2 x ptr> getelementptr (i8, <2 x ptr> zeroinitializer, <2 x i16> <i16 -2, i16 -3>)
;
%gep = getelementptr i8, <2 x ptr> <ptr null, ptr null>, <2 x i8> <i8 -2, i8 -3>
ret <2 x ptr> %gep
}
define <2 x ptr> @test_vec_inttoptr_base_normal() {
; CHECK-LABEL: define <2 x ptr> @test_vec_inttoptr_base_normal() {
; CHECK-NEXT: ret <2 x ptr> getelementptr (i8, <2 x ptr> <ptr inttoptr (i16 -1 to ptr), ptr inttoptr (i16 -2 to ptr)>, <2 x i16> <i16 1, i16 2>)
;
%base = inttoptr <2 x i16> <i16 -1, i16 -2> to <2 x ptr>
%gep = getelementptr i8, <2 x ptr> %base, <2 x i8> <i8 1, i8 2>
ret <2 x ptr> %gep
}

;; Transformation is fine for non-integral address space, but we can only change
;; the index bits: (i8 -2 == i16 254)
define <2 x ptr addrspace(1)> @test_vec_null_base_nonintegral() {
; CHECK-LABEL: define <2 x ptr addrspace(1)> @test_vec_null_base_nonintegral() {
; CHECK-NEXT: ret <2 x ptr addrspace(1)> getelementptr (i8, <2 x ptr addrspace(1)> zeroinitializer, <2 x i8> <i8 -2, i8 -3>)
;
%gep = getelementptr i8, <2 x ptr addrspace(1)> <ptr addrspace(1) null, ptr addrspace(1) null>, <2 x i8> <i8 -2, i8 -3>
ret <2 x ptr addrspace(1)> %gep
}
define <2 x ptr addrspace(1)> @test_vec_inttoptr_base_nonintegral() {
; CHECK-LABEL: define <2 x ptr addrspace(1)> @test_vec_inttoptr_base_nonintegral() {
; CHECK-NEXT: ret <2 x ptr addrspace(1)> getelementptr (i8, <2 x ptr addrspace(1)> <ptr addrspace(1) inttoptr (i16 -1 to ptr addrspace(1)), ptr addrspace(1) inttoptr (i16 -2 to ptr addrspace(1))>, <2 x i8> <i8 1, i8 2>)
;
%base = inttoptr <2 x i16> <i16 -1, i16 -2> to <2 x ptr addrspace(1)>
%gep = getelementptr i8, <2 x ptr addrspace(1)> %base, <2 x i8> <i8 1, i8 2>
ret <2 x ptr addrspace(1)> %gep
}

;; For unstable pointers we should avoid any introduction of inttoptr
define <2 x ptr addrspace(2)> @test_vec_null_base_unstable() {
; CHECK-LABEL: define <2 x ptr addrspace(2)> @test_vec_null_base_unstable() {
; CHECK-NEXT: ret <2 x ptr addrspace(2)> getelementptr (i8, <2 x ptr addrspace(2)> zeroinitializer, <2 x i8> <i8 -2, i8 -3>)
;
%gep = getelementptr i8, <2 x ptr addrspace(2)> <ptr addrspace(2) null, ptr addrspace(2) null>, <2 x i8> <i8 -2, i8 -3>
ret <2 x ptr addrspace(2)> %gep
}
define <2 x ptr addrspace(2)> @test_vec_inttoptr_base_unstable() {
; CHECK-LABEL: define <2 x ptr addrspace(2)> @test_vec_inttoptr_base_unstable() {
; CHECK-NEXT: ret <2 x ptr addrspace(2)> getelementptr (i8, <2 x ptr addrspace(2)> <ptr addrspace(2) inttoptr (i16 -1 to ptr addrspace(2)), ptr addrspace(2) inttoptr (i16 -2 to ptr addrspace(2))>, <2 x i8> <i8 1, i8 2>)
;
%base = inttoptr <2 x i16> <i16 -1, i16 -2> to <2 x ptr addrspace(2)>
%gep = getelementptr i8, <2 x ptr addrspace(2)> %base, <2 x i8> <i8 1, i8 2>
ret <2 x ptr addrspace(2)> %gep
}

;; The same is true for pointers with external state: no new inttoptr
define <2 x ptr addrspace(3)> @test_vec_null_base_external() {
; CHECK-LABEL: define <2 x ptr addrspace(3)> @test_vec_null_base_external() {
; CHECK-NEXT: ret <2 x ptr addrspace(3)> getelementptr (i8, <2 x ptr addrspace(3)> zeroinitializer, <2 x i8> <i8 -2, i8 -3>)
;
%gep = getelementptr i8, <2 x ptr addrspace(3)> <ptr addrspace(3) null, ptr addrspace(3) null>, <2 x i8> <i8 -2, i8 -3>
ret <2 x ptr addrspace(3)> %gep
}

define <2 x ptr addrspace(3)> @test_vec_inttoptr_base_external() {
; CHECK-LABEL: define <2 x ptr addrspace(3)> @test_vec_inttoptr_base_external() {
; CHECK-NEXT: ret <2 x ptr addrspace(3)> getelementptr (i8, <2 x ptr addrspace(3)> <ptr addrspace(3) inttoptr (i16 -1 to ptr addrspace(3)), ptr addrspace(3) inttoptr (i16 -2 to ptr addrspace(3))>, <2 x i8> <i8 1, i8 2>)
;
%base = inttoptr <2 x i16> <i16 -1, i16 -2> to <2 x ptr addrspace(3)>
%gep = getelementptr i8, <2 x ptr addrspace(3)> %base, <2 x i8> <i8 1, i8 2>
ret <2 x ptr addrspace(3)> %gep
}
Loading