[PATCH] D90920: [IndVars] IV user should not prevent use widening

[Max Kazantsev via Phabricator via llvm-commits]

mkazantsev created this revision.
mkazantsev added reviewers: fhahn, lebedev.ri, skatkov, asbirlea.
Herald added subscribers: llvm-commits, hiraditya.
Herald added a project: LLVM.
mkazantsev requested review of this revision.

Sometimes the an instruction we are trying to widen is used by the IV
(which means the instruction is the IV increment). Currently this may
prevent its widening. We should ignore such user because it will be
dead once the transform is done anyways.


https://reviews.llvm.org/D90920

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


Index: llvm/test/Transforms/IndVarSimplify/widen-loop-comp.ll
===================================================================
--- llvm/test/Transforms/IndVarSimplify/widen-loop-comp.ll
+++ llvm/test/Transforms/IndVarSimplify/widen-loop-comp.ll
@@ -607,13 +607,10 @@
 ; 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:    [[IV_NEXT:%.*]] = add i32 [[TMP1]], -1
-; CHECK-NEXT:    [[INDEX:%.*]] = zext i32 [[IV_NEXT]] 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:%.*]]
Index: llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
===================================================================
--- llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
+++ llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
@@ -1567,7 +1567,11 @@
   if (!AddRecOp1 || AddRecOp1->getLoop() != L)
     return false;
 
+  // Check that all uses are either s/zext, or narrow def (in case of we are
+  // widening the IV increment).
   for (Use &U : NarrowUse->uses()) {
+    if (U.getUser() == NarrowDef)
+      continue;
     Instruction *User = nullptr;
     if (ExtKind == SignExtended)
       User = dyn_cast<SExtInst>(U.getUser());
@@ -1598,6 +1602,9 @@
   ExtendKindMap[NarrowUse] = ExtKind;
 
   for (Use &U : NarrowUse->uses()) {
+    // Ignore narrow def: it will be removed after the transform.
+    if (U.getUser() == NarrowDef)
+      continue;
     Instruction *User = nullptr;
     if (ExtKind == SignExtended)
       User = cast<SExtInst>(U.getUser());


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D90920.303389.patch
Type: text/x-patch
Size: 2300 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20201106/1f9168b3/attachment.bin>