[llvm] 2e01cea - [IndVars] Recognize 'sub nuw' expressed as 'add' for widening

[Max Kazantsev via llvm-commits]

Author: Max Kazantsev
Date: 2020-11-12T10:51:29+07:00
New Revision: 2e01ceafaa1ceef2428182128c3654bab56a0a4f

URL: https://github.com/llvm/llvm-project/commit/2e01ceafaa1ceef2428182128c3654bab56a0a4f
DIFF: https://github.com/llvm/llvm-project/commit/2e01ceafaa1ceef2428182128c3654bab56a0a4f.diff

LOG: [IndVars] Recognize 'sub nuw' expressed as 'add' for widening

InstCombine canonicalizes 'sub nuw' instructions to 'add' without the
`nuw` flag. The typical case where we see it is decrementing induction
variables. For them, IndVars fails to prove that it's legal to widen them,
and inserts unprofitable `zext`'s.

This patch adds recognition of such pattern using SCEV.

Differential Revision: https://reviews.llvm.org/D89550
Reviewed By: fhahn, skatkov

Added: 
    

Modified: 
    llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
    llvm/test/Transforms/IndVarSimplify/widen-loop-comp.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
index 079ec7f0acd9..c7c081522f72 100644
--- a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
@@ -1539,8 +1539,26 @@ bool WidenIV::widenWithVariantUse(WidenIV::NarrowIVDefUse DU) {
   ExtendKind ExtKind = getExtendKind(NarrowDef);
   bool CanSignExtend = ExtKind == SignExtended && OBO->hasNoSignedWrap();
   bool CanZeroExtend = ExtKind == ZeroExtended && OBO->hasNoUnsignedWrap();
-  if (!CanSignExtend && !CanZeroExtend)
-    return false;
+  auto AnotherOpExtKind = ExtKind;
+  if (!CanSignExtend && !CanZeroExtend) {
+    // Because InstCombine turns 'sub nuw' to 'add' losing the no-wrap flag, we
+    // will most likely not see it. Let's try to prove it.
+    if (OpCode != Instruction::Add)
+      return false;
+    if (ExtKind != ZeroExtended)
+      return false;
+    const SCEV *LHS = SE->getSCEV(OBO->getOperand(0));
+    const SCEV *RHS = SE->getSCEV(OBO->getOperand(1));
+    if (!SE->isKnownNegative(RHS))
+      return false;
+    bool ProvedSubNUW = SE->isKnownPredicateAt(
+        ICmpInst::ICMP_UGE, LHS, SE->getNegativeSCEV(RHS), NarrowUse);
+    if (!ProvedSubNUW)
+      return false;
+    // In fact, our 'add' is 'sub nuw'. We will need to widen the 2nd operand as
+    // neg(zext(neg(op))), which is basically sext(op).
+    AnotherOpExtKind = SignExtended;
+  }
 
   // Verifying that Defining operand is an AddRec
   const SCEV *Op1 = SE->getSCEV(WideDef);
@@ -1564,11 +1582,11 @@ bool WidenIV::widenWithVariantUse(WidenIV::NarrowIVDefUse DU) {
   Value *LHS = (NarrowUse->getOperand(0) == NarrowDef)
                    ? WideDef
                    : createExtendInst(NarrowUse->getOperand(0), WideType,
-                                      ExtKind, NarrowUse);
+                                      AnotherOpExtKind, NarrowUse);
   Value *RHS = (NarrowUse->getOperand(1) == NarrowDef)
                    ? WideDef
                    : createExtendInst(NarrowUse->getOperand(1), WideType,
-                                      ExtKind, NarrowUse);
+                                      AnotherOpExtKind, NarrowUse);
 
   auto *NarrowBO = cast<BinaryOperator>(NarrowUse);
   auto *WideBO = BinaryOperator::Create(NarrowBO->getOpcode(), LHS, RHS,

diff  --git a/llvm/test/Transforms/IndVarSimplify/widen-loop-comp.ll b/llvm/test/Transforms/IndVarSimplify/widen-loop-comp.ll
index 07d31c49f6e9..48195a299fc8 100644
--- a/llvm/test/Transforms/IndVarSimplify/widen-loop-comp.ll
+++ b/llvm/test/Transforms/IndVarSimplify/widen-loop-comp.ll
@@ -657,13 +657,10 @@ define i32 @test13(i32 %start, i32* %p, i32* %q) {
 ; CHECK-NEXT:    [[COND:%.*]] = icmp eq i64 [[INDVARS_IV]], 0
 ; CHECK-NEXT:    br i1 [[COND]], label [[EXIT:%.*]], label [[BACKEDGE]]
 ; CHECK:       backedge:
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
-; CHECK-NEXT:    [[FOO:%.*]] = add i32 [[TMP1]], -1
-; CHECK-NEXT:    [[INDEX:%.*]] = zext i32 [[FOO]] to i64
-; CHECK-NEXT:    [[STORE_ADDR:%.*]] = getelementptr i32, i32* [[P:%.*]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP1:%.*]] = add nsw i64 [[INDVARS_IV]], -1
+; CHECK-NEXT:    [[STORE_ADDR:%.*]] = getelementptr i32, i32* [[P:%.*]], i64 [[TMP1]]
 ; CHECK-NEXT:    store i32 1, i32* [[STORE_ADDR]], align 4
-; CHECK-NEXT:    [[LOAD_ADDR:%.*]] = getelementptr i32, i32* [[Q:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[STOP:%.*]] = load i32, i32* [[Q]], align 4
+; CHECK-NEXT:    [[STOP:%.*]] = load i32, i32* [[Q:%.*]], align 4
 ; CHECK-NEXT:    [[LOOP_COND:%.*]] = icmp eq i32 [[STOP]], 0
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
 ; CHECK-NEXT:    br i1 [[LOOP_COND]], label [[LOOP]], label [[FAILURE:%.*]]