Recommit "[SCCP] Support NUW/NSW inference for all overflowing binary operators."

This reverts commit 43acb61a08fffada31fb2e20e45fcc8492ef76b9.

Recommit the patch after fixing the issue causing the revert in 4e607ec4987.
Extra tests have been added in 5c6cb61ad416a544.

Original commit message:

   Extend the NUW/NSW inference logic add in 72121a20cd and cdeaf5f28c3dc
    to all overflowing binary operators.

    Reviewed By: nikic

    Differential Revision: https://reviews.llvm.org/D142721

diff --git a/llvm/lib/Transforms/Utils/SCCPSolver.cpp b/llvm/lib/Transforms/Utils/SCCPSolver.cpp
index 80f11268ba139ebd6215034fe21b0cfefada80f8..3dd155ba3c9b2f6f14879a08963c24cd97375ab8 100644 (file)
--- a/llvm/lib/Transforms/Utils/SCCPSolver.cpp
+++ b/llvm/lib/Transforms/Utils/SCCPSolver.cpp
@@ -127,7 +127,7 @@ bool SCCPSolver::tryToReplaceWithConstant(Value *V) {
 static bool refineInstruction(SCCPSolver &Solver,
                               const SmallPtrSetImpl<Value *> &InsertedValues,
                               Instruction &Inst) {
-  if (Inst.getOpcode() != Instruction::Add)
+  if (!isa<OverflowingBinaryOperator>(Inst))
     return false;
 
   auto GetRange = [&Solver, &InsertedValues](Value *Op) {
@@ -145,7 +145,8 @@ static bool refineInstruction(SCCPSolver &Solver,
   bool Changed = false;
   if (!Inst.hasNoUnsignedWrap()) {
     auto NUWRange = ConstantRange::makeGuaranteedNoWrapRegion(
-        Instruction::Add, RangeB, OverflowingBinaryOperator::NoUnsignedWrap);
+        Instruction::BinaryOps(Inst.getOpcode()), RangeB,
+        OverflowingBinaryOperator::NoUnsignedWrap);
     if (NUWRange.contains(RangeA)) {
       Inst.setHasNoUnsignedWrap();
       Changed = true;
@@ -153,7 +154,7 @@ static bool refineInstruction(SCCPSolver &Solver,
   }
   if (!Inst.hasNoSignedWrap()) {
     auto NSWRange = ConstantRange::makeGuaranteedNoWrapRegion(
-        Instruction::Add, RangeB, OverflowingBinaryOperator::NoSignedWrap);
+        Instruction::BinaryOps(Inst.getOpcode()), RangeB, OverflowingBinaryOperator::NoSignedWrap);
     if (NSWRange.contains(RangeA)) {
       Inst.setHasNoSignedWrap();
       Changed = true;

diff --git a/llvm/test/Transforms/SCCP/ip-constant-ranges.ll b/llvm/test/Transforms/SCCP/ip-constant-ranges.ll
index ff5f15ef41dc08ae144a427e7233e0b279f56520..07068441ae4f4e81a1fa778f3d67d68b6fd44957 100644 (file)
--- a/llvm/test/Transforms/SCCP/ip-constant-ranges.ll
+++ b/llvm/test/Transforms/SCCP/ip-constant-ranges.ll
@@ -243,7 +243,7 @@ define internal i32 @recursive_f(i32 %i) {
 ; CHECK:       if.then:
 ; CHECK-NEXT:    br label [[RETURN:%.*]]
 ; CHECK:       if.else:
-; CHECK-NEXT:    [[SUB:%.*]] = sub nsw i32 [[I]], 1
+; CHECK-NEXT:    [[SUB:%.*]] = sub nuw nsw i32 [[I]], 1
 ; CHECK-NEXT:    [[CALL:%.*]] = call i32 @recursive_f(i32 [[SUB]])
 ; CHECK-NEXT:    [[ADD:%.*]] = add i32 [[I]], [[CALL]]
 ; CHECK-NEXT:    br label [[RETURN]]

diff --git a/llvm/test/Transforms/SCCP/ip-ranges-binaryops.ll b/llvm/test/Transforms/SCCP/ip-ranges-binaryops.ll
index d10a4dea73e0e728fcd66f63cfa1068a9f143f85..b9d5313dc215f92d89d76567968a06e53ee827ef 100644 (file)
--- a/llvm/test/Transforms/SCCP/ip-ranges-binaryops.ll
+++ b/llvm/test/Transforms/SCCP/ip-ranges-binaryops.ll
@@ -50,7 +50,7 @@ define i1 @caller.add() {
 ; x - y = [-190, -79)
 define internal i1 @f.sub(i32 %x, i32 %y) {
 ; CHECK-LABEL: @f.sub(
-; CHECK-NEXT:    [[A_1:%.*]] = sub i32 [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    [[A_1:%.*]] = sub nsw i32 [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[C_2:%.*]] = icmp sgt i32 [[A_1]], -81
 ; CHECK-NEXT:    [[C_4:%.*]] = icmp slt i32 [[A_1]], -189
 ; CHECK-NEXT:    [[C_5:%.*]] = icmp eq i32 [[A_1]], -150
@@ -94,7 +94,7 @@ define i1 @caller.sub() {
 ; x * y = [1000, 4001)
 define internal i1 @f.mul(i32 %x, i32 %y) {
 ; CHECK-LABEL: @f.mul(
-; CHECK-NEXT:    [[A_1:%.*]] = mul i32 [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    [[A_1:%.*]] = mul nuw nsw i32 [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[C_2:%.*]] = icmp sgt i32 [[A_1]], 3999
 ; CHECK-NEXT:    [[C_4:%.*]] = icmp slt i32 [[A_1]], 1001
 ; CHECK-NEXT:    [[C_5:%.*]] = icmp eq i32 [[A_1]], 1500

diff --git a/llvm/test/Transforms/SCCP/sub-nuw-nsw-flags.ll b/llvm/test/Transforms/SCCP/sub-nuw-nsw-flags.ll
index 3b3cd663bc5425a42a5346a465c58ba2b2468392..31c5575e14a4af26235e274b20547f32608aa5b8 100644 (file)
--- a/llvm/test/Transforms/SCCP/sub-nuw-nsw-flags.ll
+++ b/llvm/test/Transforms/SCCP/sub-nuw-nsw-flags.ll
@@ -5,7 +5,7 @@ define i8 @range_from_lshr(i8 %a) {
 ; CHECK-LABEL: @range_from_lshr(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[A_SHR:%.*]] = lshr i8 [[A:%.*]], 1
-; CHECK-NEXT:    [[SUB_1:%.*]] = sub i8 [[A_SHR]], 1
+; CHECK-NEXT:    [[SUB_1:%.*]] = sub nsw i8 [[A_SHR]], 1
 ; CHECK-NEXT:    [[SUB_2:%.*]] = sub i8 [[A_SHR]], -128
 ; CHECK-NEXT:    [[SUB_3:%.*]] = sub i8 [[A_SHR]], -127
 ; CHECK-NEXT:    [[SUB_4:%.*]] = sub i8 [[A_SHR]], -1
@@ -31,7 +31,7 @@ define i8 @sub_zero(i8 %a) {
 ; CHECK-LABEL: @sub_zero(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[SUB_1:%.*]] = sub i8 0, [[A:%.*]]
-; CHECK-NEXT:    [[SUB_2:%.*]] = sub i8 -1, [[A]]
+; CHECK-NEXT:    [[SUB_2:%.*]] = sub nuw nsw i8 -1, [[A]]
 ; CHECK-NEXT:    [[SUB_3:%.*]] = sub i8 1, [[A]]
 ; CHECK-NEXT:    [[SUB_4:%.*]] = sub i8 [[A]], -1
 ; CHECK-NEXT:    [[RES_1:%.*]] = xor i8 [[SUB_1]], [[SUB_2]]

diff --git a/llvm/test/Transforms/SCCP/ub-shift.ll b/llvm/test/Transforms/SCCP/ub-shift.ll
index e3a483ecce3bae85f033c775e69cbd7a6f443f62..ae2739d0445d0e940d64edb0364f4becc9bda039 100644 (file)
--- a/llvm/test/Transforms/SCCP/ub-shift.ll
+++ b/llvm/test/Transforms/SCCP/ub-shift.ll
@@ -5,7 +5,7 @@ define void @shift_undef_64(ptr %p) {
 ; CHECK-LABEL: @shift_undef_64(
 ; CHECK-NEXT:    store i64 0, ptr [[P:%.*]], align 4
 ; CHECK-NEXT:    store i64 -1, ptr [[P]], align 4
-; CHECK-NEXT:    [[R3:%.*]] = shl i64 -1, 4294967298
+; CHECK-NEXT:    [[R3:%.*]] = shl nuw nsw i64 -1, 4294967298
 ; CHECK-NEXT:    store i64 [[R3]], ptr [[P]], align 4
 ; CHECK-NEXT:    ret void
 ;
@@ -25,7 +25,7 @@ define void @shift_undef_65(ptr %p) {
 ; CHECK-LABEL: @shift_undef_65(
 ; CHECK-NEXT:    store i65 0, ptr [[P:%.*]], align 4
 ; CHECK-NEXT:    store i65 0, ptr [[P]], align 4
-; CHECK-NEXT:    [[R3:%.*]] = shl i65 1, -18446744073709551615
+; CHECK-NEXT:    [[R3:%.*]] = shl nuw nsw i65 1, -18446744073709551615
 ; CHECK-NEXT:    store i65 [[R3]], ptr [[P]], align 4
 ; CHECK-NEXT:    ret void
 ;
@@ -45,7 +45,7 @@ define void @shift_undef_256(ptr %p) {
 ; CHECK-LABEL: @shift_undef_256(
 ; CHECK-NEXT:    store i256 0, ptr [[P:%.*]], align 4
 ; CHECK-NEXT:    store i256 0, ptr [[P]], align 4
-; CHECK-NEXT:    [[R3:%.*]] = shl i256 1, 18446744073709551619
+; CHECK-NEXT:    [[R3:%.*]] = shl nuw nsw i256 1, 18446744073709551619
 ; CHECK-NEXT:    store i256 [[R3]], ptr [[P]], align 4
 ; CHECK-NEXT:    ret void
 ;
@@ -65,7 +65,7 @@ define void @shift_undef_511(ptr %p) {
 ; CHECK-LABEL: @shift_undef_511(
 ; CHECK-NEXT:    store i511 0, ptr [[P:%.*]], align 4
 ; CHECK-NEXT:    store i511 -1, ptr [[P]], align 4
-; CHECK-NEXT:    [[R3:%.*]] = shl i511 -3, 1208925819614629174706180
+; CHECK-NEXT:    [[R3:%.*]] = shl nuw nsw i511 -3, 1208925819614629174706180
 ; CHECK-NEXT:    store i511 [[R3]], ptr [[P]], align 4
 ; CHECK-NEXT:    ret void
 ;

diff --git a/llvm/test/Transforms/SCCP/widening.ll b/llvm/test/Transforms/SCCP/widening.ll
index 4485fb42fb4fbc2d5473e308b490b5c1f5bbf0b1..f482ed3a4e7f655d25596c632e89f59d43f38933 100644 (file)
--- a/llvm/test/Transforms/SCCP/widening.ll
+++ b/llvm/test/Transforms/SCCP/widening.ll
@@ -447,8 +447,8 @@ define void @foo(ptr %arg) {
 ; SCCP:       bb4:
 ; SCCP-NEXT:    [[TMP5:%.*]] = add i64 [[TMP2]], 3
 ; SCCP-NEXT:    [[TMP6:%.*]] = and i64 [[TMP5]], 3
-; SCCP-NEXT:    [[TMP7:%.*]] = sub i64 3, [[TMP6]]
-; SCCP-NEXT:    [[TMP8:%.*]] = shl i64 [[TMP7]], 1
+; SCCP-NEXT:    [[TMP7:%.*]] = sub nuw nsw i64 3, [[TMP6]]
+; SCCP-NEXT:    [[TMP8:%.*]] = shl nuw nsw i64 [[TMP7]], 1
 ; SCCP-NEXT:    [[TMP9:%.*]] = trunc i64 [[TMP8]] to i32
 ; SCCP-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
 ; SCCP-NEXT:    br label [[BB11:%.*]]
@@ -484,8 +484,8 @@ define void @foo(ptr %arg) {
 ; IPSCCP:       bb4:
 ; IPSCCP-NEXT:    [[TMP5:%.*]] = add i64 [[TMP2]], 3
 ; IPSCCP-NEXT:    [[TMP6:%.*]] = and i64 [[TMP5]], 3
-; IPSCCP-NEXT:    [[TMP7:%.*]] = sub i64 3, [[TMP6]]
-; IPSCCP-NEXT:    [[TMP8:%.*]] = shl i64 [[TMP7]], 1
+; IPSCCP-NEXT:    [[TMP7:%.*]] = sub nuw nsw i64 3, [[TMP6]]
+; IPSCCP-NEXT:    [[TMP8:%.*]] = shl nuw nsw i64 [[TMP7]], 1
 ; IPSCCP-NEXT:    [[TMP9:%.*]] = trunc i64 [[TMP8]] to i32
 ; IPSCCP-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
 ; IPSCCP-NEXT:    br label [[BB11:%.*]]