summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorSanjay Patel <spatel@rotateright.com>2018-11-01 14:07:39 +0000
committerSanjay Patel <spatel@rotateright.com>2018-11-01 14:07:39 +0000
commit746ebb4ee8d1f2a2dad1fedbfd21d53673fe0fa4 (patch)
treede1d12854ecee06d995c91ad5d037d37637e26b6
parent056807b01ec5c930b945a91984cb551fc8f6a6d1 (diff)
downloadbcm5719-llvm-746ebb4ee8d1f2a2dad1fedbfd21d53673fe0fa4.tar.gz
bcm5719-llvm-746ebb4ee8d1f2a2dad1fedbfd21d53673fe0fa4.zip
[InstSimplify] fold icmp based on range of abs/nabs (2nd try)
This is retrying the fold from rL345717 (reverted at rL347780) ...with a fix for the miscompile demonstrated by PR39510: https://bugs.llvm.org/show_bug.cgi?id=39510 Original commit message: This is a fix for PR39475: https://bugs.llvm.org/show_bug.cgi?id=39475 We managed to get some of these patterns using computeKnownBits in https://reviews.llvm.org/D47041, but that can't be used for nabs(). Instead, put in some range-based logic, so we can fold both abs/nabs with icmp with a constant value. Alive proofs: https://rise4fun.com/Alive/21r Name: abs_nsw_is_positive %cmp = icmp slt i32 %x, 0 %negx = sub nsw i32 0, %x %abs = select i1 %cmp, i32 %negx, i32 %x %r = icmp sgt i32 %abs, -1 => %r = i1 true Name: abs_nsw_is_not_negative %cmp = icmp slt i32 %x, 0 %negx = sub nsw i32 0, %x %abs = select i1 %cmp, i32 %negx, i32 %x %r = icmp slt i32 %abs, 0 => %r = i1 false Name: nabs_is_negative_or_0 %cmp = icmp slt i32 %x, 0 %negx = sub i32 0, %x %nabs = select i1 %cmp, i32 %x, i32 %negx %r = icmp slt i32 %nabs, 1 => %r = i1 true Name: nabs_is_not_over_0 %cmp = icmp slt i32 %x, 0 %negx = sub i32 0, %x %nabs = select i1 %cmp, i32 %x, i32 %negx %r = icmp sgt i32 %nabs, 0 => %r = i1 false Differential Revision: https://reviews.llvm.org/D53844 llvm-svn: 345832
-rw-r--r--llvm/lib/Analysis/InstructionSimplify.cpp41
-rw-r--r--llvm/test/Transforms/InstSimplify/icmp-abs-nabs.ll90
2 files changed, 56 insertions, 75 deletions
diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index c4b076341fc..db929aa7059 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -2996,6 +2996,44 @@ static Value *simplifyICmpWithBinOp(CmpInst::Predicate Pred, Value *LHS,
return nullptr;
}
+static Value *simplifyICmpWithAbsNabs(CmpInst::Predicate Pred, Value *Op0,
+ Value *Op1) {
+ // We need a comparison with a constant.
+ const APInt *C;
+ if (!match(Op1, m_APInt(C)))
+ return nullptr;
+
+ // matchSelectPattern returns the negation part of an abs pattern in SP1.
+ // If the negate has an NSW flag, abs(INT_MIN) is undefined. Without that
+ // constraint, we can't make a contiguous range for the result of abs.
+ ICmpInst::Predicate AbsPred = ICmpInst::BAD_ICMP_PREDICATE;
+ Value *SP0, *SP1;
+ SelectPatternFlavor SPF = matchSelectPattern(Op0, SP0, SP1).Flavor;
+ if (SPF == SelectPatternFlavor::SPF_ABS &&
+ cast<Instruction>(SP1)->hasNoSignedWrap())
+ // The result of abs(X) is >= 0 (with nsw).
+ AbsPred = ICmpInst::ICMP_SGE;
+ if (SPF == SelectPatternFlavor::SPF_NABS)
+ // The result of -abs(X) is <= 0.
+ AbsPred = ICmpInst::ICMP_SLE;
+
+ if (AbsPred == ICmpInst::BAD_ICMP_PREDICATE)
+ return nullptr;
+
+ // If there is no intersection between abs/nabs and the range of this icmp,
+ // the icmp must be false. If the abs/nabs range is a subset of the icmp
+ // range, the icmp must be true.
+ APInt Zero = APInt::getNullValue(C->getBitWidth());
+ ConstantRange AbsRange = ConstantRange::makeExactICmpRegion(AbsPred, Zero);
+ ConstantRange CmpRange = ConstantRange::makeExactICmpRegion(Pred, *C);
+ if (AbsRange.intersectWith(CmpRange).isEmptySet())
+ return getFalse(GetCompareTy(Op0));
+ if (CmpRange.contains(AbsRange))
+ return getTrue(GetCompareTy(Op0));
+
+ return nullptr;
+}
+
/// Simplify integer comparisons where at least one operand of the compare
/// matches an integer min/max idiom.
static Value *simplifyICmpWithMinMax(CmpInst::Predicate Pred, Value *LHS,
@@ -3427,6 +3465,9 @@ static Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS,
if (Value *V = simplifyICmpWithMinMax(Pred, LHS, RHS, Q, MaxRecurse))
return V;
+ if (Value *V = simplifyICmpWithAbsNabs(Pred, LHS, RHS))
+ return V;
+
// Simplify comparisons of related pointers using a powerful, recursive
// GEP-walk when we have target data available..
if (LHS->getType()->isPointerTy())
diff --git a/llvm/test/Transforms/InstSimplify/icmp-abs-nabs.ll b/llvm/test/Transforms/InstSimplify/icmp-abs-nabs.ll
index c2e24de885f..41ffc33ef57 100644
--- a/llvm/test/Transforms/InstSimplify/icmp-abs-nabs.ll
+++ b/llvm/test/Transforms/InstSimplify/icmp-abs-nabs.ll
@@ -5,11 +5,7 @@
define i1 @abs_nsw_is_positive(i32 %x) {
; CHECK-LABEL: @abs_nsw_is_positive(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[X:%.*]], 0
-; CHECK-NEXT: [[NEGX:%.*]] = sub nsw i32 0, [[X]]
-; CHECK-NEXT: [[ABS:%.*]] = select i1 [[CMP]], i32 [[NEGX]], i32 [[X]]
-; CHECK-NEXT: [[R:%.*]] = icmp sgt i32 [[ABS]], -1
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 true
;
%cmp = icmp slt i32 %x, 0
%negx = sub nsw i32 0, %x
@@ -35,11 +31,7 @@ define i1 @abs_nsw_is_positive_sge(i32 %x) {
define i1 @abs_nsw_is_positive_reduced_range(i32 %x) {
; CHECK-LABEL: @abs_nsw_is_positive_reduced_range(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[X:%.*]], 0
-; CHECK-NEXT: [[NEGX:%.*]] = sub nsw i32 0, [[X]]
-; CHECK-NEXT: [[ABS:%.*]] = select i1 [[CMP]], i32 [[NEGX]], i32 [[X]]
-; CHECK-NEXT: [[R:%.*]] = icmp sgt i32 [[ABS]], -42
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 true
;
%cmp = icmp slt i32 %x, 0
%negx = sub nsw i32 0, %x
@@ -99,11 +91,7 @@ define i1 @abs_nsw_is_not_negative(i32 %x) {
define i1 @abs_nsw_is_not_negative_sle(i32 %x) {
; CHECK-LABEL: @abs_nsw_is_not_negative_sle(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[X:%.*]], 1
-; CHECK-NEXT: [[NEGX:%.*]] = sub nsw i32 0, [[X]]
-; CHECK-NEXT: [[ABS:%.*]] = select i1 [[CMP]], i32 [[NEGX]], i32 [[X]]
-; CHECK-NEXT: [[R:%.*]] = icmp sle i32 [[ABS]], -1
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 false
;
%cmp = icmp slt i32 %x, 1
%negx = sub nsw i32 0, %x
@@ -116,11 +104,7 @@ define i1 @abs_nsw_is_not_negative_sle(i32 %x) {
define i1 @abs_nsw_is_not_negative_reduced_range(i32 %x) {
; CHECK-LABEL: @abs_nsw_is_not_negative_reduced_range(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[X:%.*]], 0
-; CHECK-NEXT: [[NEGX:%.*]] = sub nsw i32 0, [[X]]
-; CHECK-NEXT: [[ABS:%.*]] = select i1 [[CMP]], i32 [[NEGX]], i32 [[X]]
-; CHECK-NEXT: [[R:%.*]] = icmp slt i32 [[ABS]], -24
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 false
;
%cmp = icmp slt i32 %x, 0
%negx = sub nsw i32 0, %x
@@ -167,11 +151,7 @@ define i1 @abs_nsw_is_not_negative_wrong_range(i32 %x) {
define i1 @nabs_is_negative_or_0(i32 %x) {
; CHECK-LABEL: @nabs_is_negative_or_0(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[X:%.*]], 0
-; CHECK-NEXT: [[NEGX:%.*]] = sub i32 0, [[X]]
-; CHECK-NEXT: [[NABS:%.*]] = select i1 [[CMP]], i32 [[X]], i32 [[NEGX]]
-; CHECK-NEXT: [[R:%.*]] = icmp slt i32 [[NABS]], 1
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 true
;
%cmp = icmp slt i32 %x, 0
%negx = sub i32 0, %x
@@ -184,11 +164,7 @@ define i1 @nabs_is_negative_or_0(i32 %x) {
define i1 @nabs_is_negative_or_0_sle(i32 %x) {
; CHECK-LABEL: @nabs_is_negative_or_0_sle(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[X:%.*]], 1
-; CHECK-NEXT: [[NEGX:%.*]] = sub i32 0, [[X]]
-; CHECK-NEXT: [[NABS:%.*]] = select i1 [[CMP]], i32 [[X]], i32 [[NEGX]]
-; CHECK-NEXT: [[R:%.*]] = icmp sle i32 [[NABS]], 0
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 true
;
%cmp = icmp slt i32 %x, 1
%negx = sub i32 0, %x
@@ -201,11 +177,7 @@ define i1 @nabs_is_negative_or_0_sle(i32 %x) {
define i1 @nabs_is_negative_or_0_reduced_range(i32 %x) {
; CHECK-LABEL: @nabs_is_negative_or_0_reduced_range(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[X:%.*]], 1
-; CHECK-NEXT: [[NEGX:%.*]] = sub i32 0, [[X]]
-; CHECK-NEXT: [[NABS:%.*]] = select i1 [[CMP]], i32 [[X]], i32 [[NEGX]]
-; CHECK-NEXT: [[R:%.*]] = icmp slt i32 [[NABS]], 421
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 true
;
%cmp = icmp slt i32 %x, 1
%negx = sub i32 0, %x
@@ -235,11 +207,7 @@ define i1 @nabs_is_negative_or_0_wrong_range(i32 %x) {
define i1 @nabs_is_not_over_0(i32 %x) {
; CHECK-LABEL: @nabs_is_not_over_0(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[X:%.*]], 0
-; CHECK-NEXT: [[NEGX:%.*]] = sub i32 0, [[X]]
-; CHECK-NEXT: [[NABS:%.*]] = select i1 [[CMP]], i32 [[X]], i32 [[NEGX]]
-; CHECK-NEXT: [[R:%.*]] = icmp sgt i32 [[NABS]], 0
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 false
;
%cmp = icmp slt i32 %x, 0
%negx = sub i32 0, %x
@@ -252,11 +220,7 @@ define i1 @nabs_is_not_over_0(i32 %x) {
define i1 @nabs_is_not_over_0_sle(i32 %x) {
; CHECK-LABEL: @nabs_is_not_over_0_sle(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[X:%.*]], 1
-; CHECK-NEXT: [[NEGX:%.*]] = sub i32 0, [[X]]
-; CHECK-NEXT: [[NABS:%.*]] = select i1 [[CMP]], i32 [[X]], i32 [[NEGX]]
-; CHECK-NEXT: [[R:%.*]] = icmp sge i32 [[NABS]], 1
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 false
;
%cmp = icmp slt i32 %x, 1
%negx = sub i32 0, %x
@@ -269,11 +233,7 @@ define i1 @nabs_is_not_over_0_sle(i32 %x) {
define i1 @nabs_is_not_over_0_reduced_range(i32 %x) {
; CHECK-LABEL: @nabs_is_not_over_0_reduced_range(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[X:%.*]], 1
-; CHECK-NEXT: [[NEGX:%.*]] = sub i32 0, [[X]]
-; CHECK-NEXT: [[NABS:%.*]] = select i1 [[CMP]], i32 [[X]], i32 [[NEGX]]
-; CHECK-NEXT: [[R:%.*]] = icmp sgt i32 [[NABS]], 4223
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 false
;
%cmp = icmp slt i32 %x, 1
%negx = sub i32 0, %x
@@ -318,11 +278,7 @@ define i1 @abs_nsw_is_positive_eq(i32 %x) {
define i1 @abs_nsw_is_positive_ult(i8 %x) {
; CHECK-LABEL: @abs_nsw_is_positive_ult(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i8 [[X:%.*]], 0
-; CHECK-NEXT: [[NEGX:%.*]] = sub nsw i8 0, [[X]]
-; CHECK-NEXT: [[ABS:%.*]] = select i1 [[CMP]], i8 [[NEGX]], i8 [[X]]
-; CHECK-NEXT: [[R:%.*]] = icmp ult i8 [[ABS]], -117
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 true
;
%cmp = icmp slt i8 %x, 0
%negx = sub nsw i8 0, %x
@@ -335,11 +291,7 @@ define i1 @abs_nsw_is_positive_ult(i8 %x) {
define i1 @abs_nsw_is_not_negative_ugt(i8 %x) {
; CHECK-LABEL: @abs_nsw_is_not_negative_ugt(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i8 [[X:%.*]], 0
-; CHECK-NEXT: [[NEGX:%.*]] = sub nsw i8 0, [[X]]
-; CHECK-NEXT: [[ABS:%.*]] = select i1 [[CMP]], i8 [[NEGX]], i8 [[X]]
-; CHECK-NEXT: [[R:%.*]] = icmp ugt i8 [[ABS]], 127
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 false
;
%cmp = icmp slt i8 %x, 0
%negx = sub nsw i8 0, %x
@@ -352,11 +304,7 @@ define i1 @abs_nsw_is_not_negative_ugt(i8 %x) {
define <2 x i1> @abs_nsw_is_not_negative_vec_splat(<2 x i32> %x) {
; CHECK-LABEL: @abs_nsw_is_not_negative_vec_splat(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt <2 x i32> [[X:%.*]], zeroinitializer
-; CHECK-NEXT: [[NEGX:%.*]] = sub nsw <2 x i32> zeroinitializer, [[X]]
-; CHECK-NEXT: [[ABS:%.*]] = select <2 x i1> [[CMP]], <2 x i32> [[NEGX]], <2 x i32> [[X]]
-; CHECK-NEXT: [[R:%.*]] = icmp slt <2 x i32> [[ABS]], <i32 -8, i32 -8>
-; CHECK-NEXT: ret <2 x i1> [[R]]
+; CHECK-NEXT: ret <2 x i1> zeroinitializer
;
%cmp = icmp slt <2 x i32> %x, zeroinitializer
%negx = sub nsw <2 x i32> zeroinitializer, %x
@@ -369,11 +317,7 @@ define <2 x i1> @abs_nsw_is_not_negative_vec_splat(<2 x i32> %x) {
define i1 @nabs_is_negative_or_0_ne(i8 %x) {
; CHECK-LABEL: @nabs_is_negative_or_0_ne(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt i8 [[X:%.*]], 0
-; CHECK-NEXT: [[NEGX:%.*]] = sub i8 0, [[X]]
-; CHECK-NEXT: [[NABS:%.*]] = select i1 [[CMP]], i8 [[X]], i8 [[NEGX]]
-; CHECK-NEXT: [[R:%.*]] = icmp ne i8 [[NABS]], 12
-; CHECK-NEXT: ret i1 [[R]]
+; CHECK-NEXT: ret i1 true
;
%cmp = icmp slt i8 %x, 0
%negx = sub i8 0, %x
@@ -386,11 +330,7 @@ define i1 @nabs_is_negative_or_0_ne(i8 %x) {
define <3 x i1> @nabs_is_not_over_0_sle_vec_splat(<3 x i33> %x) {
; CHECK-LABEL: @nabs_is_not_over_0_sle_vec_splat(
-; CHECK-NEXT: [[CMP:%.*]] = icmp slt <3 x i33> [[X:%.*]], <i33 1, i33 1, i33 1>
-; CHECK-NEXT: [[NEGX:%.*]] = sub <3 x i33> zeroinitializer, [[X]]
-; CHECK-NEXT: [[NABS:%.*]] = select <3 x i1> [[CMP]], <3 x i33> [[X]], <3 x i33> [[NEGX]]
-; CHECK-NEXT: [[R:%.*]] = icmp sge <3 x i33> [[NABS]], <i33 1, i33 1, i33 1>
-; CHECK-NEXT: ret <3 x i1> [[R]]
+; CHECK-NEXT: ret <3 x i1> zeroinitializer
;
%cmp = icmp slt <3 x i33> %x, <i33 1, i33 1, i33 1>
%negx = sub <3 x i33> zeroinitializer, %x
OpenPOWER on IntegriCloud