[llvm] ValueTracking: Skip threadlocal.address in computeKnownBits (PR #88435)
Matthias Braun via llvm-commits
llvm-commits at lists.llvm.org
Fri Apr 12 14:55:01 PDT 2024
https://github.com/MatzeB updated https://github.com/llvm/llvm-project/pull/88435
>From b871d1c1900ca6112fbd537e723ada9a043fb049 Mon Sep 17 00:00:00 2001
From: Matthias Braun <matze at braunis.de>
Date: Thu, 11 Apr 2024 13:34:54 -0700
Subject: [PATCH 1/2] ValueTracking: Skip threadlocal.address in
computeKnownBits
---
llvm/lib/Analysis/ValueTracking.cpp | 4 +++
.../Analysis/ValueTracking/knownbits-tls.ll | 32 +++++++++++++++++++
2 files changed, 36 insertions(+)
create mode 100644 llvm/test/Analysis/ValueTracking/knownbits-tls.ll
diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 3a10de72a27562..c7aec8518044f3 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -1686,6 +1686,10 @@ static void computeKnownBitsFromOperator(const Operator *I,
Known.Zero.setBitsFrom(KnownZeroFirstBit);
break;
}
+ case Intrinsic::threadlocal_address: {
+ computeKnownBits(II->getArgOperand(0), Known, Depth + 1, Q);
+ break;
+ }
case Intrinsic::vscale: {
if (!II->getParent() || !II->getFunction())
break;
diff --git a/llvm/test/Analysis/ValueTracking/knownbits-tls.ll b/llvm/test/Analysis/ValueTracking/knownbits-tls.ll
new file mode 100644
index 00000000000000..203777c33f8486
--- /dev/null
+++ b/llvm/test/Analysis/ValueTracking/knownbits-tls.ll
@@ -0,0 +1,32 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -S -passes=instcombine < %s | FileCheck %s
+
+target datalayout = "p:64:64"
+
+ at tlsvar_align32 = thread_local global i32 3, align 32
+
+define i1 @lowbits_zero() {
+; CHECK-LABEL: define i1 @lowbits_zero() {
+; CHECK-NEXT: ret i1 true
+;
+ %p = call ptr @llvm.threadlocal.address(ptr @tlsvar_align32)
+ %p_int = ptrtoint ptr %p to i64
+ %lowbits = and i64 %p_int, 31
+ %lowbits_zero = icmp eq i64 %lowbits, 0
+ ret i1 %lowbits_zero
+}
+
+define i1 @lowbits_unknown() {
+; CHECK-LABEL: define i1 @lowbits_unknown() {
+; CHECK-NEXT: [[P:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @tlsvar_align32)
+; CHECK-NEXT: [[P_INT:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[LOWBITS:%.*]] = and i64 [[P_INT]], 32
+; CHECK-NEXT: [[LOWBITS_ZERO:%.*]] = icmp eq i64 [[LOWBITS]], 0
+; CHECK-NEXT: ret i1 [[LOWBITS_ZERO]]
+;
+ %p = call ptr @llvm.threadlocal.address(ptr @tlsvar_align32)
+ %p_int = ptrtoint ptr %p to i64
+ %lowbits = and i64 %p_int, 63
+ %lowbits_zero = icmp eq i64 %lowbits, 0
+ ret i1 %lowbits_zero
+}
>From 821b5edf1cc645ac45e7fda012bac9d93a0804c2 Mon Sep 17 00:00:00 2001
From: Matthias Braun <matze at braunis.de>
Date: Fri, 12 Apr 2024 14:18:52 -0700
Subject: [PATCH 2/2] Add aligment attribute in InstCombine instead
---
llvm/lib/Analysis/ValueTracking.cpp | 4 --
.../InstCombine/InstCombineCalls.cpp | 9 ++++
.../Analysis/ValueTracking/knownbits-tls.ll | 32 ---------------
.../InstCombine/threadlocal_address.ll | 41 +++++++++++++++++++
4 files changed, 50 insertions(+), 36 deletions(-)
delete mode 100644 llvm/test/Analysis/ValueTracking/knownbits-tls.ll
create mode 100644 llvm/test/Transforms/InstCombine/threadlocal_address.ll
diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index c7aec8518044f3..3a10de72a27562 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -1686,10 +1686,6 @@ static void computeKnownBitsFromOperator(const Operator *I,
Known.Zero.setBitsFrom(KnownZeroFirstBit);
break;
}
- case Intrinsic::threadlocal_address: {
- computeKnownBits(II->getArgOperand(0), Known, Depth + 1, Q);
- break;
- }
case Intrinsic::vscale: {
if (!II->getParent() || !II->getFunction())
break;
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index b1017f4d6bc179..b278c207787717 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -3407,6 +3407,15 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {
return I;
break;
}
+ case Intrinsic::threadlocal_address: {
+ Align MinAlign = getKnownAlignment(II->getArgOperand(0), DL, II, &AC, &DT);
+ MaybeAlign Align = II->getRetAlign();
+ if (MinAlign > Align.valueOrOne()) {
+ II->addRetAttr(Attribute::getWithAlignment(II->getContext(), MinAlign));
+ return II;
+ }
+ break;
+ }
default: {
// Handle target specific intrinsics
std::optional<Instruction *> V = targetInstCombineIntrinsic(*II);
diff --git a/llvm/test/Analysis/ValueTracking/knownbits-tls.ll b/llvm/test/Analysis/ValueTracking/knownbits-tls.ll
deleted file mode 100644
index 203777c33f8486..00000000000000
--- a/llvm/test/Analysis/ValueTracking/knownbits-tls.ll
+++ /dev/null
@@ -1,32 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
-; RUN: opt -S -passes=instcombine < %s | FileCheck %s
-
-target datalayout = "p:64:64"
-
- at tlsvar_align32 = thread_local global i32 3, align 32
-
-define i1 @lowbits_zero() {
-; CHECK-LABEL: define i1 @lowbits_zero() {
-; CHECK-NEXT: ret i1 true
-;
- %p = call ptr @llvm.threadlocal.address(ptr @tlsvar_align32)
- %p_int = ptrtoint ptr %p to i64
- %lowbits = and i64 %p_int, 31
- %lowbits_zero = icmp eq i64 %lowbits, 0
- ret i1 %lowbits_zero
-}
-
-define i1 @lowbits_unknown() {
-; CHECK-LABEL: define i1 @lowbits_unknown() {
-; CHECK-NEXT: [[P:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @tlsvar_align32)
-; CHECK-NEXT: [[P_INT:%.*]] = ptrtoint ptr [[P]] to i64
-; CHECK-NEXT: [[LOWBITS:%.*]] = and i64 [[P_INT]], 32
-; CHECK-NEXT: [[LOWBITS_ZERO:%.*]] = icmp eq i64 [[LOWBITS]], 0
-; CHECK-NEXT: ret i1 [[LOWBITS_ZERO]]
-;
- %p = call ptr @llvm.threadlocal.address(ptr @tlsvar_align32)
- %p_int = ptrtoint ptr %p to i64
- %lowbits = and i64 %p_int, 63
- %lowbits_zero = icmp eq i64 %lowbits, 0
- ret i1 %lowbits_zero
-}
diff --git a/llvm/test/Transforms/InstCombine/threadlocal_address.ll b/llvm/test/Transforms/InstCombine/threadlocal_address.ll
new file mode 100644
index 00000000000000..0c220d996839e9
--- /dev/null
+++ b/llvm/test/Transforms/InstCombine/threadlocal_address.ll
@@ -0,0 +1,41 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -o - -S %s -passes=instcombine | FileCheck %s
+
+ at tlsvar_a4 = thread_local global i32 4, align 4
+
+define void @func_increase_alignment() {
+; CHECK-LABEL: define void @func_increase_alignment() {
+; CHECK-NEXT: [[P:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr @tlsvar_a4)
+; CHECK-NEXT: store i32 42, ptr [[P]], align 2
+; CHECK-NEXT: ret void
+;
+ %p = call align 2 ptr @llvm.threadlocal.address(ptr @tlsvar_a4)
+ store i32 42, ptr %p, align 2
+ ret void
+}
+
+ at tlsvar_a32 = thread_local global i32 5, align 32
+
+define i1 @func_add_alignment() {
+; CHECK-LABEL: define i1 @func_add_alignment() {
+; CHECK-NEXT: ret i1 true
+;
+ %p = call ptr @llvm.threadlocal.address(ptr @tlsvar_a32)
+ %p_int = ptrtoint ptr %p to i32
+ %lowbits = and i32 %p_int, 31
+ %zero = icmp eq i32 %lowbits, 0
+ ret i1 %zero
+}
+
+ at tlsvar_a1 = thread_local global i8 6, align 1
+
+define i1 @func_dont_reduce_alignment() {
+; CHECK-LABEL: define i1 @func_dont_reduce_alignment() {
+; CHECK-NEXT: ret i1 true
+;
+ %p = call align 4 ptr @llvm.threadlocal.address(ptr @tlsvar_a1)
+ %p_int = ptrtoint ptr %p to i32
+ %lowbits = and i32 %p_int, 3
+ %zero = icmp eq i32 %lowbits, 0
+ ret i1 %zero
+}
More information about the llvm-commits
mailing list