[PATCH] D158422: llvm-reduce: Fix invalid reduction from breaking phi operands

[Matt Arsenault via Phabricator via llvm-commits]

arsenm created this revision.
arsenm added reviewers: aeubanks, regehr, dblaikie, fhahn, lebedev.ri, diegotf.
Herald added a subscriber: StephenFan.
Herald added a project: All.
arsenm requested review of this revision.
Herald added a subscriber: wdng.
Herald added a project: LLVM.

Fix "PHI node has multiple entries for the same basic block with
different incoming values!" when running operands-to-args.


https://reviews.llvm.org/D158422

Files:
  llvm/test/tools/llvm-reduce/operands-to-args-phi.ll
  llvm/tools/llvm-reduce/deltas/ReduceOperandsToArgs.cpp


Index: llvm/tools/llvm-reduce/deltas/ReduceOperandsToArgs.cpp
===================================================================
--- llvm/tools/llvm-reduce/deltas/ReduceOperandsToArgs.cpp
+++ llvm/tools/llvm-reduce/deltas/ReduceOperandsToArgs.cpp
@@ -160,7 +160,14 @@
   for (Use *Op : OpsToReplace) {
     Value *NewArg = OldValMap.lookup(Op->get());
     auto *NewUser = cast<Instruction>(VMap.lookup(Op->getUser()));
-    NewUser->setOperand(Op->getOperandNo(), NewArg);
+
+    if (PHINode *NewPhi = dyn_cast<PHINode>(NewUser)) {
+      PHINode *OldPhi = cast<PHINode>(Op->getUser());
+      BasicBlock *OldBB = OldPhi->getIncomingBlock(*Op);
+      NewPhi->setIncomingValueForBlock(cast<BasicBlock>(VMap.lookup(OldBB)),
+                                       NewArg);
+    } else
+      NewUser->setOperand(Op->getOperandNo(), NewArg);
   }
 
   // Replace all OldF uses with NewF.
Index: llvm/test/tools/llvm-reduce/operands-to-args-phi.ll
===================================================================
--- /dev/null
+++ llvm/test/tools/llvm-reduce/operands-to-args-phi.ll
@@ -0,0 +1,73 @@
+; RUN: llvm-reduce --abort-on-invalid-reduction %s -o %t --delta-passes=operands-to-args --test FileCheck --test-arg %s --test-arg --check-prefix=INTERESTING --test-arg --input-file
+; RUN: FileCheck -check-prefix=RESULT %s < %t
+
+; Make sure an invalid reduction isn't hit from only replacing one of
+; the values when a predecessor is listed multiple times in a phi.
+
+
+; RESULT: define void @fn(i1 %cmp11, i8 %p.2, i8 %k.2, i8 %p.4, i8 %k.0, i8 %k.4, i8 %p.0, i1 %tobool30, i8 %spec.select, i8 %spec.select1)
+
+; RESULT:     for.cond:
+; RESULT-NEXT:  %p.01 = phi i8 [ 0, %entry ], [ %p.2, %for.inc ]
+; RESULT-NEXT:  %k.02 = phi i8 [ 0, %entry ], [ %k.2, %for.inc ]
+
+; RESULT:     if.end26:
+; RESULT-NEXT:  %p.1 = phi i8 [ %p.4, %for.cond35 ], [ %k.0, %if.end ]
+; RESULT-NEXT:  %k.1 = phi i8 [ %k.4, %for.cond35 ], [ 0, %if.end ]
+
+; RESULT:     for.inc:
+; RESULT-NEXT:  %p.26 = phi i8 [ %spec.select, %if.end26 ], [ poison, %if.then13 ], [ %spec.select, %if.end26 ]
+; RESULT-NEXT:  %k.27 = phi i8 [ %spec.select1, %if.end26 ], [ 0, %if.then13 ], [ %spec.select1, %if.end26 ]
+
+
+; RESULT:      for.cond35:
+; RESULT-NEXT:   %p.48 = phi i8 [ 0, %if.then ], [ %k.0, %if.then13 ]
+; RESULT-NEXT:   %k.49 = phi i8 [ %k.0, %if.then ], [ 0, %if.then13 ]
+
+define void @fn(i1 %cmp11) {
+entry:
+  br label %for.cond
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %p.0 = phi i8 [ 0, %entry ], [ %p.2, %for.inc ]
+  %k.0 = phi i8 [ 0, %entry ], [ %k.2, %for.inc ]
+  br i1 %cmp11, label %if.then, label %if.end
+
+if.then:                                          ; preds = %for.cond
+  br label %for.cond35
+
+if.end:                                           ; preds = %for.cond
+  br i1 %cmp11, label %if.then13, label %if.end26
+
+if.then13:                                        ; preds = %if.end
+  br i1 %cmp11, label %for.inc, label %for.cond35
+
+if.end26:                                         ; preds = %for.cond35, %if.end
+  ; INTERESTING: %p.1 = phi i8
+  ; INTERESTING: %k.1 = phi i8
+
+  %p.1 = phi i8 [ %p.4, %for.cond35 ], [ %k.0, %if.end ]
+  %k.1 = phi i8 [ %k.4, %for.cond35 ], [ 0, %if.end ]
+  %tobool30 = icmp ne i8 %p.0, 0
+  %spec.select = select i1 false, i8 0, i8 %p.1
+  %spec.select1 = select i1 %tobool30, i8 %k.1, i8 0
+  br i1 false, label %for.inc, label %for.inc
+
+; INTERESTING: {{^}}for.inc:
+; INTERESTING: phi i8
+; INTERESTING-SAME: [ %spec.select{{[0-9]*}}, %if.end26 ]
+
+; INTERESTING: phi i8
+; INTERESTING-SAME: [ %spec.select{{[0-9]*}}, %if.end26 ]
+for.inc:                                          ; preds = %if.end26, %if.end26, %if.then13
+  %p.2 = phi i8 [ %spec.select, %if.end26 ], [ poison, %if.then13 ], [ %spec.select, %if.end26 ]
+  %k.2 = phi i8 [ %spec.select1, %if.end26 ], [ 0, %if.then13 ], [ %spec.select1, %if.end26 ]
+  %0 = load i32, ptr null, align 4
+  br label %for.cond
+
+for.cond35:                                       ; preds = %if.then13, %if.then
+  %p.4 = phi i8 [ 0, %if.then ], [ %k.0, %if.then13 ]
+  %k.4 = phi i8 [ %k.0, %if.then ], [ 0, %if.then13 ]
+  %tobool36 = icmp eq i32 0, 0
+  br label %if.end26
+}


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D158422.552006.patch
Type: text/x-patch
Size: 4243 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230821/4f1f86a6/attachment.bin>