[PATCH] D47103: Implement strip.invariant.group
This revision was automatically updated to reflect the committed changes.
Prazek marked an inline comment as done.
Closed by commit rL336073: Implement strip.invariant.group (authored by Prazek,
committed by ).
Changed prior to commit:
https://reviews.llvm.org/D47103?vs=148313=153664#toc
Repository:
rL LLVM
https://reviews.llvm.org/D47103
Files:
llvm/trunk/docs/LangRef.rst
llvm/trunk/include/llvm/IR/IRBuilder.h
llvm/trunk/include/llvm/IR/Intrinsics.td
llvm/trunk/lib/Analysis/BasicAliasAnalysis.cpp
llvm/trunk/lib/Analysis/ConstantFolding.cpp
llvm/trunk/lib/Analysis/ValueTracking.cpp
llvm/trunk/lib/CodeGen/CodeGenPrepare.cpp
llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
llvm/trunk/lib/IR/Value.cpp
llvm/trunk/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
llvm/trunk/test/Analysis/ValueTracking/invariant.group.ll
llvm/trunk/test/CodeGen/Generic/intrinsics.ll
llvm/trunk/test/Other/invariant.group.ll
llvm/trunk/test/Other/launder.invariant.group.ll
llvm/trunk/test/Transforms/CodeGenPrepare/invariant.group.ll
llvm/trunk/test/Transforms/DeadStoreElimination/launder.invariant.group.ll
llvm/trunk/test/Transforms/FunctionAttrs/nocapture.ll
llvm/trunk/test/Transforms/GVN/invariant.group.ll
llvm/trunk/test/Transforms/GlobalOpt/invariant.group.barrier.ll
llvm/trunk/test/Transforms/GlobalOpt/invariant.group.ll
llvm/trunk/test/Transforms/InstCombine/invariant.group.ll
llvm/trunk/test/Transforms/NewGVN/invariant.group.ll
Index: llvm/trunk/lib/IR/Value.cpp
===
--- llvm/trunk/lib/IR/Value.cpp
+++ llvm/trunk/lib/IR/Value.cpp
@@ -521,7 +521,8 @@
// but it can't be marked with returned attribute, that's why it needs
// special case.
if (StripKind == PSK_ZeroIndicesAndAliasesAndInvariantGroups &&
-CS.getIntrinsicID() == Intrinsic::launder_invariant_group) {
+(CS.getIntrinsicID() == Intrinsic::launder_invariant_group ||
+ CS.getIntrinsicID() == Intrinsic::strip_invariant_group)) {
V = CS.getArgOperand(0);
continue;
}
Index: llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
===
--- llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
+++ llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
@@ -1437,6 +1437,7 @@
return true;
}
case Intrinsic::launder_invariant_group:
+ case Intrinsic::strip_invariant_group:
case Intrinsic::expect: {
unsigned ResultReg = getRegForValue(II->getArgOperand(0));
if (!ResultReg)
Index: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
===
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -5768,6 +5768,7 @@
case Intrinsic::annotation:
case Intrinsic::ptr_annotation:
case Intrinsic::launder_invariant_group:
+ case Intrinsic::strip_invariant_group:
// Drop the intrinsic, but forward the value
setValue(, getValue(I.getOperand(0)));
return nullptr;
Index: llvm/trunk/lib/CodeGen/CodeGenPrepare.cpp
===
--- llvm/trunk/lib/CodeGen/CodeGenPrepare.cpp
+++ llvm/trunk/lib/CodeGen/CodeGenPrepare.cpp
@@ -1702,6 +1702,7 @@
return true;
}
case Intrinsic::launder_invariant_group:
+case Intrinsic::strip_invariant_group:
II->replaceAllUsesWith(II->getArgOperand(0));
II->eraseFromParent();
return true;
Index: llvm/trunk/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
===
--- llvm/trunk/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
+++ llvm/trunk/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
@@ -457,6 +457,7 @@
case Intrinsic::invariant_start:
case Intrinsic::invariant_end:
case Intrinsic::launder_invariant_group:
+ case Intrinsic::strip_invariant_group:
case Intrinsic::objectsize:
return true;
default:
@@ -882,6 +883,7 @@
case Intrinsic::invariant_start:
case Intrinsic::invariant_end:
case Intrinsic::launder_invariant_group:
+case Intrinsic::strip_invariant_group:
Intr->eraseFromParent();
// FIXME: I think the invariant marker should still theoretically apply,
// but the intrinsics need to be changed to accept pointers with any
Index: llvm/trunk/lib/Analysis/ValueTracking.cpp
===
--- llvm/trunk/lib/Analysis/ValueTracking.cpp
+++ llvm/trunk/lib/Analysis/ValueTracking.cpp
@@ -3404,8 +3404,9 @@
}
bool llvm::isIntrinsicReturningPointerAliasingArgumentWithoutCapturing(
- ImmutableCallSite CS) {
- return CS.getIntrinsicID() == Intrinsic::launder_invariant_group;
+ImmutableCallSite CS) {
+ return CS.getIntrinsicID() ==
[PATCH] D47103: Implement strip.invariant.group
rsmith accepted this revision. rsmith added inline comments. This revision is now accepted and ready to land. Comment at: llvm/docs/LangRef.rst:12928 +established by ``invariant.group`` metadata no longer holds, to obtain a new pointer +value that does carries fresh invariant group information. It is an experimental intrinsic, which means that its semantics might change in the future. does carries -> carries Repository: rL LLVM https://reviews.llvm.org/D47103 ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D47103: Implement strip.invariant.group
Prazek added a comment. Friendly ping @hfinkel @rsmith Repository: rL LLVM https://reviews.llvm.org/D47103 ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D47103: Implement strip.invariant.group
Prazek added a comment. friendly ping Repository: rL LLVM https://reviews.llvm.org/D47103 ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D47103: Implement strip.invariant.group
rjmccall added inline comments. Comment at: clang/lib/CodeGen/CGExpr.cpp:3858 +} + } + Prazek wrote: > rjmccall wrote: > > Please add a comment explaining why this is necessary. (I'm actually not > > sure why it is, because surely the invariant groups we generate don't > > contain assumptions about memory from fields, right?) > Short answer: you can only make virtual calls on a dynamic pointer that > carries invariant.group and you can't do anything other because it could leak > the information about this pointer (which when used with comparison could > break devirtualization). Alright, sure. Repository: rL LLVM https://reviews.llvm.org/D47103 ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D47103: Implement strip.invariant.group
Prazek updated this revision to Diff 148313.
Prazek marked 2 inline comments as done.
Prazek added a comment.
Slitted commit into defining and using intrinsic
Repository:
rL LLVM
https://reviews.llvm.org/D47103
Files:
llvm/docs/LangRef.rst
llvm/include/llvm/IR/IRBuilder.h
llvm/include/llvm/IR/Intrinsics.td
llvm/lib/Analysis/BasicAliasAnalysis.cpp
llvm/lib/Analysis/ConstantFolding.cpp
llvm/lib/Analysis/ValueTracking.cpp
llvm/lib/CodeGen/CodeGenPrepare.cpp
llvm/lib/CodeGen/SelectionDAG/FastISel.cpp
llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
llvm/lib/IR/Value.cpp
llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
llvm/test/Analysis/ValueTracking/invariant.group.ll
llvm/test/CodeGen/Generic/intrinsics.ll
llvm/test/Other/invariant.group.ll
llvm/test/Other/launder.invariant.group.ll
llvm/test/Transforms/CodeGenPrepare/invariant.group.ll
llvm/test/Transforms/DeadStoreElimination/launder.invariant.group.ll
llvm/test/Transforms/FunctionAttrs/nocapture.ll
llvm/test/Transforms/GVN/invariant.group.ll
llvm/test/Transforms/GlobalOpt/invariant.group.barrier.ll
llvm/test/Transforms/GlobalOpt/invariant.group.ll
llvm/test/Transforms/InstCombine/invariant.group.ll
llvm/test/Transforms/NewGVN/invariant.group.ll
Index: llvm/test/Transforms/NewGVN/invariant.group.ll
===
--- llvm/test/Transforms/NewGVN/invariant.group.ll
+++ llvm/test/Transforms/NewGVN/invariant.group.ll
@@ -52,6 +52,19 @@
ret i8 %b
}
+; CHECK-LABEL: define i1 @proveEqualityForStrip(
+define i1 @proveEqualityForStrip(i8* %a) {
+; FIXME: The first call could be also removed by GVN. Right now
+; DCE removes it. The second call is CSE'd with the first one.
+; CHECK: %b1 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+ %b1 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+; CHECK-NOT: llvm.strip.invariant.group
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+ %r = icmp eq i8* %b1, %b2
+; CHECK: ret i1 true
+ ret i1 %r
+}
+
; CHECK-LABEL: define i8 @unoptimizable1() {
define i8 @unoptimizable1() {
entry:
Index: llvm/test/Transforms/InstCombine/invariant.group.ll
===
--- llvm/test/Transforms/InstCombine/invariant.group.ll
+++ llvm/test/Transforms/InstCombine/invariant.group.ll
@@ -1,5 +1,6 @@
; RUN: opt -instcombine -S < %s | FileCheck %s
+
; CHECK-LABEL: define i8* @simplifyNullLaunder()
define i8* @simplifyNullLaunder() {
; CHECK-NEXT: ret i8* null
@@ -29,6 +30,39 @@
ret i8 addrspace(42)* %b2
}
-
declare i8* @llvm.launder.invariant.group.p0i8(i8*)
declare i8 addrspace(42)* @llvm.launder.invariant.group.p42i8(i8 addrspace(42)*)
+
+
+; CHECK-LABEL: define i8* @simplifyNullStrip()
+define i8* @simplifyNullStrip() {
+; CHECK-NEXT: ret i8* null
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* null)
+ ret i8* %b2
+}
+
+; CHECK-LABEL: define i8 addrspace(42)* @dontsimplifyNullStripForDifferentAddrspace()
+define i8 addrspace(42)* @dontsimplifyNullStripForDifferentAddrspace() {
+; CHECK: %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* null)
+; CHECK: ret i8 addrspace(42)* %b2
+ %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* null)
+ ret i8 addrspace(42)* %b2
+}
+
+; CHECK-LABEL: define i8* @simplifyUndefStrip()
+define i8* @simplifyUndefStrip() {
+; CHECK-NEXT: ret i8* undef
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* undef)
+ ret i8* %b2
+}
+
+; CHECK-LABEL: define i8 addrspace(42)* @simplifyUndefStrip2()
+define i8 addrspace(42)* @simplifyUndefStrip2() {
+; CHECK-NEXT: ret i8 addrspace(42)* undef
+ %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* undef)
+ ret i8 addrspace(42)* %b2
+}
+
+declare i8* @llvm.strip.invariant.group.p0i8(i8*)
+declare i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)*)
+
Index: llvm/test/Transforms/GlobalOpt/invariant.group.ll
===
--- llvm/test/Transforms/GlobalOpt/invariant.group.ll
+++ llvm/test/Transforms/GlobalOpt/invariant.group.ll
@@ -27,46 +27,46 @@
define void @_optimizable() {
enter:
%valptr = alloca i32
-
+
%val = call i32 @TheAnswerToLifeTheUniverseAndEverything()
store i32 %val, i32* @tmp
store i32 %val, i32* %valptr
-
+
%0 = bitcast i32* %valptr to i8*
%barr = call i8* @llvm.launder.invariant.group(i8* %0)
%1 = bitcast i8* %barr to i32*
-
+
%val2 = load i32, i32* %1
store i32 %val2, i32* @tmp2
ret void
}
; We can't step through launder.invariant.group here, because that would change
; this load in @usage_of_globals()
-; val = load i32, i32* %ptrVal, !invariant.group !0
-; into
+; val = load i32, i32* %ptrVal, !invariant.group !0
+; into
; %val = load i32, i32* @tmp3, !invariant.group !0
-; and then we could assume that %val and %val2 to be the same,
[PATCH] D47103: Implement strip.invariant.group
Prazek added inline comments. Comment at: clang/lib/CodeGen/CGExpr.cpp:3858 +} + } + rjmccall wrote: > Please add a comment explaining why this is necessary. (I'm actually not > sure why it is, because surely the invariant groups we generate don't contain > assumptions about memory from fields, right?) Short answer: you can only make virtual calls on a dynamic pointer that carries invariant.group and you can't do anything other because it could leak the information about this pointer (which when used with comparison could break devirtualization). Repository: rL LLVM https://reviews.llvm.org/D47103 ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D47103: Implement strip.invariant.group
rjmccall added a comment. The changes to Clang generally seem reasonable, but I think you should split them into a separate commit from the commit that adds the intrinsic itself. Comment at: clang/lib/CodeGen/CGExpr.cpp:3858 +} + } + Please add a comment explaining why this is necessary. (I'm actually not sure why it is, because surely the invariant groups we generate don't contain assumptions about memory from fields, right?) Comment at: clang/lib/CodeGen/CGExprScalar.cpp:1626 + Src = Builder.CreateStripInvariantGroup(Src); +} + Again, comment. Comment at: clang/lib/CodeGen/CGExprScalar.cpp:1767 +if (CGF.CGM.getCodeGenOpts().StrictVTablePointers) + return Builder.CreateLaunderInvariantGroup(IntToPtr); + I think assigning to IntToPtr would be clearer here. You're changing the value returned, not introducing a reason to early-exit. And a comment, please, on all of these cases. Repository: rL LLVM https://reviews.llvm.org/D47103 ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D47103: Implement strip.invariant.group
amharc added inline comments.
Comment at: clang/include/clang/AST/DeclCXX.h:778
+ bool mayBeDynamicClass() const {
+return !isCompleteDefinition() || isDynamicClass();
xbolva00 wrote:
> maybeDynamicClass?
>
> https://github.com/llvm-mirror/llvm/search?utf8=%E2%9C%93=maybe=
I think `mayBe` is more appropriate, because it grammatically consists of 'may'
and 'be'. In this form it appears in multiple places in the codebase, too. A
few examples:
https://github.com/llvm-mirror/llvm/blob/0b24b74655b976aaba01b2f726659b8c745a16de/include/llvm/IR/GlobalValue.h#L131
https://github.com/llvm-mirror/clang/blob/647be32c6048e24f90f470c557e850597e5526c3/lib/AST/DeclCXX.cpp#L1800
https://github.com/llvm-mirror/llvm/blob/77438c3c98c0901fa9ef56f78bdefccedc16fcc4/lib/Analysis/ValueTracking.cpp#L3659
Repository:
rL LLVM
https://reviews.llvm.org/D47103
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D47103: Implement strip.invariant.group
Prazek updated this revision to Diff 147691.
Prazek added a comment.
rebase
Repository:
rL LLVM
https://reviews.llvm.org/D47103
Files:
clang/include/clang/AST/DeclCXX.h
clang/lib/CodeGen/CGExpr.cpp
clang/lib/CodeGen/CGExprScalar.cpp
clang/test/CodeGenCXX/strict-vtable-pointers.cpp
llvm/docs/LangRef.rst
llvm/include/llvm/IR/IRBuilder.h
llvm/include/llvm/IR/Intrinsics.td
llvm/lib/Analysis/BasicAliasAnalysis.cpp
llvm/lib/Analysis/ConstantFolding.cpp
llvm/lib/Analysis/ValueTracking.cpp
llvm/lib/CodeGen/CodeGenPrepare.cpp
llvm/lib/CodeGen/SelectionDAG/FastISel.cpp
llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
llvm/lib/IR/Value.cpp
llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
llvm/test/Analysis/ValueTracking/invariant.group.ll
llvm/test/CodeGen/Generic/intrinsics.ll
llvm/test/Other/invariant.group.ll
llvm/test/Other/launder.invariant.group.ll
llvm/test/Transforms/CodeGenPrepare/invariant.group.ll
llvm/test/Transforms/DeadStoreElimination/launder.invariant.group.ll
llvm/test/Transforms/FunctionAttrs/nocapture.ll
llvm/test/Transforms/GVN/invariant.group.ll
llvm/test/Transforms/GlobalOpt/invariant.group.barrier.ll
llvm/test/Transforms/GlobalOpt/invariant.group.ll
llvm/test/Transforms/InstCombine/invariant.group.ll
llvm/test/Transforms/NewGVN/invariant.group.ll
Index: llvm/test/Transforms/NewGVN/invariant.group.ll
===
--- llvm/test/Transforms/NewGVN/invariant.group.ll
+++ llvm/test/Transforms/NewGVN/invariant.group.ll
@@ -52,6 +52,19 @@
ret i8 %b
}
+; CHECK-LABEL: define i1 @proveEqualityForStrip(
+define i1 @proveEqualityForStrip(i8* %a) {
+; FIXME: The first call could be also removed by GVN. Right now
+; DCE removes it. The second call is CSE'd with the first one.
+; CHECK: %b1 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+ %b1 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+; CHECK-NOT: llvm.strip.invariant.group
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+ %r = icmp eq i8* %b1, %b2
+; CHECK: ret i1 true
+ ret i1 %r
+}
+
; CHECK-LABEL: define i8 @unoptimizable1() {
define i8 @unoptimizable1() {
entry:
Index: llvm/test/Transforms/InstCombine/invariant.group.ll
===
--- llvm/test/Transforms/InstCombine/invariant.group.ll
+++ llvm/test/Transforms/InstCombine/invariant.group.ll
@@ -1,5 +1,6 @@
; RUN: opt -instcombine -S < %s | FileCheck %s
+
; CHECK-LABEL: define i8* @simplifyNullLaunder()
define i8* @simplifyNullLaunder() {
; CHECK-NEXT: ret i8* null
@@ -29,6 +30,39 @@
ret i8 addrspace(42)* %b2
}
-
declare i8* @llvm.launder.invariant.group.p0i8(i8*)
declare i8 addrspace(42)* @llvm.launder.invariant.group.p42i8(i8 addrspace(42)*)
+
+
+; CHECK-LABEL: define i8* @simplifyNullStrip()
+define i8* @simplifyNullStrip() {
+; CHECK-NEXT: ret i8* null
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* null)
+ ret i8* %b2
+}
+
+; CHECK-LABEL: define i8 addrspace(42)* @dontsimplifyNullStripForDifferentAddrspace()
+define i8 addrspace(42)* @dontsimplifyNullStripForDifferentAddrspace() {
+; CHECK: %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* null)
+; CHECK: ret i8 addrspace(42)* %b2
+ %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* null)
+ ret i8 addrspace(42)* %b2
+}
+
+; CHECK-LABEL: define i8* @simplifyUndefStrip()
+define i8* @simplifyUndefStrip() {
+; CHECK-NEXT: ret i8* undef
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* undef)
+ ret i8* %b2
+}
+
+; CHECK-LABEL: define i8 addrspace(42)* @simplifyUndefStrip2()
+define i8 addrspace(42)* @simplifyUndefStrip2() {
+; CHECK-NEXT: ret i8 addrspace(42)* undef
+ %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* undef)
+ ret i8 addrspace(42)* %b2
+}
+
+declare i8* @llvm.strip.invariant.group.p0i8(i8*)
+declare i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)*)
+
Index: llvm/test/Transforms/GlobalOpt/invariant.group.ll
===
--- llvm/test/Transforms/GlobalOpt/invariant.group.ll
+++ llvm/test/Transforms/GlobalOpt/invariant.group.ll
@@ -27,46 +27,46 @@
define void @_optimizable() {
enter:
%valptr = alloca i32
-
+
%val = call i32 @TheAnswerToLifeTheUniverseAndEverything()
store i32 %val, i32* @tmp
store i32 %val, i32* %valptr
-
+
%0 = bitcast i32* %valptr to i8*
%barr = call i8* @llvm.launder.invariant.group(i8* %0)
%1 = bitcast i8* %barr to i32*
-
+
%val2 = load i32, i32* %1
store i32 %val2, i32* @tmp2
ret void
}
; We can't step through launder.invariant.group here, because that would change
; this load in @usage_of_globals()
-; val = load i32, i32* %ptrVal, !invariant.group !0
-; into
+; val = load i32, i32* %ptrVal, !invariant.group !0
+; into
; %val = load i32, i32* @tmp3, !invariant.group
[PATCH] D47103: Implement strip.invariant.group
Prazek updated this revision to Diff 147688.
Prazek added a comment.
After rebasing
Repository:
rL LLVM
https://reviews.llvm.org/D47103
Files:
clang/include/clang/AST/DeclCXX.h
clang/lib/CodeGen/CGExpr.cpp
clang/lib/CodeGen/CGExprScalar.cpp
clang/test/CodeGenCXX/strict-vtable-pointers.cpp
llvm/docs/LangRef.rst
llvm/include/llvm/IR/IRBuilder.h
llvm/include/llvm/IR/Intrinsics.td
llvm/lib/Analysis/BasicAliasAnalysis.cpp
llvm/lib/Analysis/ConstantFolding.cpp
llvm/lib/Analysis/ValueTracking.cpp
llvm/lib/CodeGen/CodeGenPrepare.cpp
llvm/lib/CodeGen/SelectionDAG/FastISel.cpp
llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
llvm/lib/IR/Value.cpp
llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
llvm/test/Analysis/ValueTracking/invariant.group.ll
llvm/test/CodeGen/Generic/intrinsics.ll
llvm/test/Other/invariant.group.ll
llvm/test/Other/launder.invariant.group.ll
llvm/test/Transforms/CodeGenPrepare/invariant.group.ll
llvm/test/Transforms/DeadStoreElimination/launder.invariant.group.ll
llvm/test/Transforms/FunctionAttrs/nocapture.ll
llvm/test/Transforms/GVN/invariant.group.ll
llvm/test/Transforms/GlobalOpt/invariant.group.barrier.ll
llvm/test/Transforms/GlobalOpt/invariant.group.ll
llvm/test/Transforms/InstCombine/invariant.group.ll
llvm/test/Transforms/NewGVN/invariant.group.ll
Index: llvm/test/Transforms/NewGVN/invariant.group.ll
===
--- llvm/test/Transforms/NewGVN/invariant.group.ll
+++ llvm/test/Transforms/NewGVN/invariant.group.ll
@@ -52,6 +52,19 @@
ret i8 %b
}
+; CHECK-LABEL: define i1 @proveEqualityForStrip(
+define i1 @proveEqualityForStrip(i8* %a) {
+; FIXME: The first call could be also removed by GVN. Right now
+; DCE removes it. The second call is CSE'd with the first one.
+; CHECK: %b1 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+ %b1 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+; CHECK-NOT: llvm.strip.invariant.group
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+ %r = icmp eq i8* %b1, %b2
+; CHECK: ret i1 true
+ ret i1 %r
+}
+
; CHECK-LABEL: define i8 @unoptimizable1() {
define i8 @unoptimizable1() {
entry:
Index: llvm/test/Transforms/InstCombine/invariant.group.ll
===
--- llvm/test/Transforms/InstCombine/invariant.group.ll
+++ llvm/test/Transforms/InstCombine/invariant.group.ll
@@ -1,5 +1,6 @@
; RUN: opt -instcombine -S < %s | FileCheck %s
+
; CHECK-LABEL: define i8* @simplifyNullLaunder()
define i8* @simplifyNullLaunder() {
; CHECK-NEXT: ret i8* null
@@ -29,6 +30,39 @@
ret i8 addrspace(42)* %b2
}
-
declare i8* @llvm.launder.invariant.group.p0i8(i8*)
declare i8 addrspace(42)* @llvm.launder.invariant.group.p42i8(i8 addrspace(42)*)
+
+
+; CHECK-LABEL: define i8* @simplifyNullStrip()
+define i8* @simplifyNullStrip() {
+; CHECK-NEXT: ret i8* null
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* null)
+ ret i8* %b2
+}
+
+; CHECK-LABEL: define i8 addrspace(42)* @dontsimplifyNullStripForDifferentAddrspace()
+define i8 addrspace(42)* @dontsimplifyNullStripForDifferentAddrspace() {
+; CHECK: %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* null)
+; CHECK: ret i8 addrspace(42)* %b2
+ %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* null)
+ ret i8 addrspace(42)* %b2
+}
+
+; CHECK-LABEL: define i8* @simplifyUndefStrip()
+define i8* @simplifyUndefStrip() {
+; CHECK-NEXT: ret i8* undef
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* undef)
+ ret i8* %b2
+}
+
+; CHECK-LABEL: define i8 addrspace(42)* @simplifyUndefStrip2()
+define i8 addrspace(42)* @simplifyUndefStrip2() {
+; CHECK-NEXT: ret i8 addrspace(42)* undef
+ %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* undef)
+ ret i8 addrspace(42)* %b2
+}
+
+declare i8* @llvm.strip.invariant.group.p0i8(i8*)
+declare i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)*)
+
Index: llvm/test/Transforms/GlobalOpt/invariant.group.ll
===
--- llvm/test/Transforms/GlobalOpt/invariant.group.ll
+++ llvm/test/Transforms/GlobalOpt/invariant.group.ll
@@ -27,46 +27,46 @@
define void @_optimizable() {
enter:
%valptr = alloca i32
-
+
%val = call i32 @TheAnswerToLifeTheUniverseAndEverything()
store i32 %val, i32* @tmp
store i32 %val, i32* %valptr
-
+
%0 = bitcast i32* %valptr to i8*
%barr = call i8* @llvm.launder.invariant.group(i8* %0)
%1 = bitcast i8* %barr to i32*
-
+
%val2 = load i32, i32* %1
store i32 %val2, i32* @tmp2
ret void
}
; We can't step through launder.invariant.group here, because that would change
; this load in @usage_of_globals()
-; val = load i32, i32* %ptrVal, !invariant.group !0
-; into
+; val = load i32, i32* %ptrVal, !invariant.group !0
+; into
; %val = load i32, i32* @tmp3,
[PATCH] D47103: Implement strip.invariant.group
Prazek updated this revision to Diff 147684.
Prazek added a comment.
Changed comments
Repository:
rL LLVM
https://reviews.llvm.org/D47103
Files:
clang/include/clang/AST/DeclCXX.h
clang/lib/CodeGen/CGExpr.cpp
clang/lib/CodeGen/CGExprScalar.cpp
clang/test/CodeGenCXX/strict-vtable-pointers.cpp
llvm/docs/LangRef.rst
llvm/include/llvm/IR/IRBuilder.h
llvm/include/llvm/IR/Intrinsics.td
llvm/lib/Analysis/BasicAliasAnalysis.cpp
llvm/lib/Analysis/CaptureTracking.cpp
llvm/lib/Analysis/ConstantFolding.cpp
llvm/lib/Analysis/ValueTracking.cpp
llvm/lib/CodeGen/CodeGenPrepare.cpp
llvm/lib/CodeGen/SelectionDAG/FastISel.cpp
llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
llvm/lib/IR/Value.cpp
llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
llvm/test/Analysis/ValueTracking/invariant.group.ll
llvm/test/CodeGen/Generic/intrinsics.ll
llvm/test/Other/invariant.group.ll
llvm/test/Other/launder.invariant.group.ll
llvm/test/Transforms/CodeGenPrepare/invariant.group.ll
llvm/test/Transforms/DeadStoreElimination/launder.invariant.group.ll
llvm/test/Transforms/FunctionAttrs/nocapture.ll
llvm/test/Transforms/GVN/invariant.group.ll
llvm/test/Transforms/GlobalOpt/invariant.group.barrier.ll
llvm/test/Transforms/GlobalOpt/invariant.group.ll
llvm/test/Transforms/InstCombine/invariant.group.ll
llvm/test/Transforms/NewGVN/invariant.group.ll
Index: llvm/test/Transforms/NewGVN/invariant.group.ll
===
--- llvm/test/Transforms/NewGVN/invariant.group.ll
+++ llvm/test/Transforms/NewGVN/invariant.group.ll
@@ -52,6 +52,19 @@
ret i8 %b
}
+; CHECK-LABEL: define i1 @proveEqualityForStrip(
+define i1 @proveEqualityForStrip(i8* %a) {
+; FIXME: The first call could be also removed by GVN. Right now
+; DCE removes it. The second call is CSE'd with the first one.
+; CHECK: %b1 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+ %b1 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+; CHECK-NOT: llvm.strip.invariant.group
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+ %r = icmp eq i8* %b1, %b2
+; CHECK: ret i1 true
+ ret i1 %r
+}
+
; CHECK-LABEL: define i8 @unoptimizable1() {
define i8 @unoptimizable1() {
entry:
Index: llvm/test/Transforms/InstCombine/invariant.group.ll
===
--- llvm/test/Transforms/InstCombine/invariant.group.ll
+++ llvm/test/Transforms/InstCombine/invariant.group.ll
@@ -1,5 +1,6 @@
; RUN: opt -instcombine -S < %s | FileCheck %s
+
; CHECK-LABEL: define i8* @simplifyNullLaunder()
define i8* @simplifyNullLaunder() {
; CHECK-NEXT: ret i8* null
@@ -29,6 +30,39 @@
ret i8 addrspace(42)* %b2
}
-
declare i8* @llvm.launder.invariant.group.p0i8(i8*)
declare i8 addrspace(42)* @llvm.launder.invariant.group.p42i8(i8 addrspace(42)*)
+
+
+; CHECK-LABEL: define i8* @simplifyNullStrip()
+define i8* @simplifyNullStrip() {
+; CHECK-NEXT: ret i8* null
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* null)
+ ret i8* %b2
+}
+
+; CHECK-LABEL: define i8 addrspace(42)* @dontsimplifyNullStripForDifferentAddrspace()
+define i8 addrspace(42)* @dontsimplifyNullStripForDifferentAddrspace() {
+; CHECK: %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* null)
+; CHECK: ret i8 addrspace(42)* %b2
+ %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* null)
+ ret i8 addrspace(42)* %b2
+}
+
+; CHECK-LABEL: define i8* @simplifyUndefStrip()
+define i8* @simplifyUndefStrip() {
+; CHECK-NEXT: ret i8* undef
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* undef)
+ ret i8* %b2
+}
+
+; CHECK-LABEL: define i8 addrspace(42)* @simplifyUndefStrip2()
+define i8 addrspace(42)* @simplifyUndefStrip2() {
+; CHECK-NEXT: ret i8 addrspace(42)* undef
+ %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* undef)
+ ret i8 addrspace(42)* %b2
+}
+
+declare i8* @llvm.strip.invariant.group.p0i8(i8*)
+declare i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)*)
+
Index: llvm/test/Transforms/GlobalOpt/invariant.group.ll
===
--- llvm/test/Transforms/GlobalOpt/invariant.group.ll
+++ llvm/test/Transforms/GlobalOpt/invariant.group.ll
@@ -27,46 +27,46 @@
define void @_optimizable() {
enter:
%valptr = alloca i32
-
+
%val = call i32 @TheAnswerToLifeTheUniverseAndEverything()
store i32 %val, i32* @tmp
store i32 %val, i32* %valptr
-
+
%0 = bitcast i32* %valptr to i8*
%barr = call i8* @llvm.launder.invariant.group(i8* %0)
%1 = bitcast i8* %barr to i32*
-
+
%val2 = load i32, i32* %1
store i32 %val2, i32* @tmp2
ret void
}
; We can't step through launder.invariant.group here, because that would change
; this load in @usage_of_globals()
-; val = load i32, i32* %ptrVal, !invariant.group !0
-; into
+; val = load i32, i32* %ptrVal, !invariant.group !0
+;
[PATCH] D47103: Implement strip.invariant.group
xbolva00 added inline comments.
Comment at: clang/include/clang/AST/DeclCXX.h:778
+ bool mayBeDynamicClass() const {
+return !isCompleteDefinition() || isDynamicClass();
maybeDynamicClass?
https://github.com/llvm-mirror/llvm/search?utf8=%E2%9C%93=maybe=
Repository:
rL LLVM
https://reviews.llvm.org/D47103
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[PATCH] D47103: Implement strip.invariant.group
Prazek updated this revision to Diff 147683.
Prazek added a comment.
introduced mayBeDynamicClass and added more tests
Repository:
rL LLVM
https://reviews.llvm.org/D47103
Files:
clang/include/clang/AST/DeclCXX.h
clang/lib/CodeGen/CGExpr.cpp
clang/lib/CodeGen/CGExprScalar.cpp
clang/test/CodeGenCXX/strict-vtable-pointers.cpp
llvm/docs/LangRef.rst
llvm/include/llvm/IR/IRBuilder.h
llvm/include/llvm/IR/Intrinsics.td
llvm/lib/Analysis/CaptureTracking.cpp
llvm/lib/Analysis/ConstantFolding.cpp
llvm/lib/Analysis/ValueTracking.cpp
llvm/lib/CodeGen/CodeGenPrepare.cpp
llvm/lib/CodeGen/SelectionDAG/FastISel.cpp
llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
llvm/lib/IR/Value.cpp
llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
llvm/test/Analysis/ValueTracking/invariant.group.ll
llvm/test/CodeGen/Generic/intrinsics.ll
llvm/test/Other/invariant.group.ll
llvm/test/Other/launder.invariant.group.ll
llvm/test/Transforms/CodeGenPrepare/invariant.group.ll
llvm/test/Transforms/DeadStoreElimination/launder.invariant.group.ll
llvm/test/Transforms/FunctionAttrs/nocapture.ll
llvm/test/Transforms/GVN/invariant.group.ll
llvm/test/Transforms/GlobalOpt/invariant.group.barrier.ll
llvm/test/Transforms/GlobalOpt/invariant.group.ll
llvm/test/Transforms/InstCombine/invariant.group.ll
llvm/test/Transforms/NewGVN/invariant.group.ll
Index: llvm/test/Transforms/NewGVN/invariant.group.ll
===
--- llvm/test/Transforms/NewGVN/invariant.group.ll
+++ llvm/test/Transforms/NewGVN/invariant.group.ll
@@ -52,6 +52,19 @@
ret i8 %b
}
+; CHECK-LABEL: define i1 @proveEqualityForStrip(
+define i1 @proveEqualityForStrip(i8* %a) {
+; FIXME: The first call could be also removed by GVN. Right now
+; DCE removes it. The second call is CSE'd with the first one.
+; CHECK: %b1 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+ %b1 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+; CHECK-NOT: llvm.strip.invariant.group
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* %a)
+ %r = icmp eq i8* %b1, %b2
+; CHECK: ret i1 true
+ ret i1 %r
+}
+
; CHECK-LABEL: define i8 @unoptimizable1() {
define i8 @unoptimizable1() {
entry:
Index: llvm/test/Transforms/InstCombine/invariant.group.ll
===
--- llvm/test/Transforms/InstCombine/invariant.group.ll
+++ llvm/test/Transforms/InstCombine/invariant.group.ll
@@ -1,5 +1,6 @@
; RUN: opt -instcombine -S < %s | FileCheck %s
+
; CHECK-LABEL: define i8* @simplifyNullLaunder()
define i8* @simplifyNullLaunder() {
; CHECK-NEXT: ret i8* null
@@ -29,6 +30,39 @@
ret i8 addrspace(42)* %b2
}
-
declare i8* @llvm.launder.invariant.group.p0i8(i8*)
declare i8 addrspace(42)* @llvm.launder.invariant.group.p42i8(i8 addrspace(42)*)
+
+
+; CHECK-LABEL: define i8* @simplifyNullStrip()
+define i8* @simplifyNullStrip() {
+; CHECK-NEXT: ret i8* null
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* null)
+ ret i8* %b2
+}
+
+; CHECK-LABEL: define i8 addrspace(42)* @dontsimplifyNullStripForDifferentAddrspace()
+define i8 addrspace(42)* @dontsimplifyNullStripForDifferentAddrspace() {
+; CHECK: %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* null)
+; CHECK: ret i8 addrspace(42)* %b2
+ %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* null)
+ ret i8 addrspace(42)* %b2
+}
+
+; CHECK-LABEL: define i8* @simplifyUndefStrip()
+define i8* @simplifyUndefStrip() {
+; CHECK-NEXT: ret i8* undef
+ %b2 = call i8* @llvm.strip.invariant.group.p0i8(i8* undef)
+ ret i8* %b2
+}
+
+; CHECK-LABEL: define i8 addrspace(42)* @simplifyUndefStrip2()
+define i8 addrspace(42)* @simplifyUndefStrip2() {
+; CHECK-NEXT: ret i8 addrspace(42)* undef
+ %b2 = call i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)* undef)
+ ret i8 addrspace(42)* %b2
+}
+
+declare i8* @llvm.strip.invariant.group.p0i8(i8*)
+declare i8 addrspace(42)* @llvm.strip.invariant.group.p42i8(i8 addrspace(42)*)
+
Index: llvm/test/Transforms/GlobalOpt/invariant.group.ll
===
--- llvm/test/Transforms/GlobalOpt/invariant.group.ll
+++ llvm/test/Transforms/GlobalOpt/invariant.group.ll
@@ -27,46 +27,46 @@
define void @_optimizable() {
enter:
%valptr = alloca i32
-
+
%val = call i32 @TheAnswerToLifeTheUniverseAndEverything()
store i32 %val, i32* @tmp
store i32 %val, i32* %valptr
-
+
%0 = bitcast i32* %valptr to i8*
%barr = call i8* @llvm.launder.invariant.group(i8* %0)
%1 = bitcast i8* %barr to i32*
-
+
%val2 = load i32, i32* %1
store i32 %val2, i32* @tmp2
ret void
}
; We can't step through launder.invariant.group here, because that would change
; this load in @usage_of_globals()
-; val = load i32, i32* %ptrVal, !invariant.group !0
-; into
+; val = load i32, i32* %ptrVal, !invariant.group !0
+; into
; %val
