From 2127b00cd5d4058bb315220ae64b021219f7ef5b Mon Sep 17 00:00:00 2001
From: Pete Cooper <peter_cooper@apple.com>
Date: Thu, 30 Apr 2015 23:57:47 +0000
Subject: [PATCH] [ARM] optimizeSelect should clear kill flags.

If we move an instruction from one block down to a MOVC and predicate it,
then the original instruction could be moved in to a loop.  In this case,
its invalid for any kill flags to remain on there.

Fails with -verfy-machineinstrs.

rdar://problem/20752113

llvm-svn: 236290
---
 llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp  |  7 ++++
 llvm/test/CodeGen/ARM/combine-movc-sub.ll | 63 +++++++++++++++++++++++++++++++
 2 files changed, 70 insertions(+)
 create mode 100644 llvm/test/CodeGen/ARM/combine-movc-sub.ll

diff --git a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
index e722ffb..9ce4247 100644
--- a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -1921,6 +1921,13 @@ ARMBaseInstrInfo::optimizeSelect(MachineInstr *MI,
   SeenMIs.insert(NewMI);
   SeenMIs.erase(DefMI);
 
+  // If MI is inside a loop, and DefMI is outside the loop, then kill flags on
+  // DefMI would be invalid when tranferred inside the loop.  Checking for a
+  // loop is expensive, but at least remove kill flags if they are in different
+  // BBs.
+  if (DefMI->getParent() != MI->getParent())
+    NewMI->clearKillInfo();
+
   // The caller will erase MI, but not DefMI.
   DefMI->eraseFromParent();
   return NewMI;
diff --git a/llvm/test/CodeGen/ARM/combine-movc-sub.ll b/llvm/test/CodeGen/ARM/combine-movc-sub.ll
new file mode 100644
index 0000000..6a5f7bd
--- /dev/null
+++ b/llvm/test/CodeGen/ARM/combine-movc-sub.ll
@@ -0,0 +1,63 @@
+; RUN: llc %s -o - -verify-machineinstrs | FileCheck %s
+
+target datalayout = "e-m:o-p:32:32-f64:32:64-v64:32:64-v128:32:128-a:0:32-n32-S32"
+target triple = "thumbv7s-apple-unknown"
+
+; The subtract instruction %3 will be optimized (combined and predicated) with the select
+; inside the loop.  In this case, the kill flag on the subtract should be removed or else
+; it will fail verification.
+
+%struct.PROOFSEARCH_HELP = type { %struct.LIST_HELP*, %struct.LIST_HELP*, %struct.LIST_HELP*, %struct.LIST_HELP*, %struct.SHARED_INDEX_NODE*, %struct.LIST_HELP*, %struct.SHARED_INDEX_NODE*, %struct.LIST_HELP*, %struct.SORTTHEORY_HELP*, %struct.SORTTHEORY_HELP*, %struct.SORTTHEORY_HELP*, %struct.SHARED_INDEX_NODE*, %struct.LIST_HELP*, i32*, i32*, %struct.LIST_HELP*, i32, i32, i32, i32, i32, i32, i32, i32 }
+%struct.SORTTHEORY_HELP = type { %struct.st*, [4000 x %struct.NODE_HELP*], %struct.LIST_HELP*, %struct.LIST_HELP*, i32 }
+%struct.st = type { %struct.subst*, %struct.LIST_HELP*, %struct.LIST_HELP*, i16, i16 }
+%struct.subst = type { %struct.subst*, i32, %struct.term* }
+%struct.term = type { i32, %union.anon, %struct.LIST_HELP*, i32, i32 }
+%union.anon = type { %struct.LIST_HELP* }
+%struct.NODE_HELP = type { %struct.LIST_HELP*, i32, i32, i32, %struct.LIST_HELP*, i32 }
+%struct.SHARED_INDEX_NODE = type { %struct.st*, [3001 x %struct.term*], [4000 x %struct.term*], i32 }
+%struct.LIST_HELP = type { %struct.LIST_HELP*, i8* }
+%struct.CLAUSE_HELP = type { i32, i32, i32, i32, i32*, i32, %struct.LIST_HELP*, %struct.LIST_HELP*, i32, i32, %struct.LITERAL_HELP**, i32, i32, i32, i32 }
+%struct.LITERAL_HELP = type { i32, i32, i32, %struct.CLAUSE_HELP*, %struct.term* }
+
+declare void @foo(%struct.PROOFSEARCH_HELP*, %struct.CLAUSE_HELP*)
+
+; CHECK-LABEL: @test
+; CHECK: it
+; CHECK-NEXT: sub
+
+define hidden fastcc %struct.LIST_HELP* @test(%struct.PROOFSEARCH_HELP* %Search, %struct.LIST_HELP* %ClauseList, i32 %Level, %struct.LIST_HELP** nocapture %New) {
+entry:
+  %cmp4.i.i = icmp ugt i32 %Level, 31
+  %0 = add i32 %Level, -32
+  %1 = lshr i32 %0, 5
+  %2 = shl nuw i32 %1, 5
+  %3 = sub i32 %0, %2
+  %4 = add nuw nsw i32 %1, 1
+  br label %for.body
+
+for.body:                                         ; preds = %for.inc, %entry
+  %Scan.038 = phi %struct.LIST_HELP* [ %ClauseList, %entry ], [ %9, %for.inc ]
+  %car.i33 = getelementptr inbounds %struct.LIST_HELP, %struct.LIST_HELP* %Scan.038, i32 0, i32 1
+  %5 = bitcast i8** %car.i33 to %struct.CLAUSE_HELP**
+  %6 = load %struct.CLAUSE_HELP*, %struct.CLAUSE_HELP** %5, align 4
+  %. = add i32 %4, 10
+  %.Level = select i1 %cmp4.i.i, i32 %3, i32 %Level
+  %splitfield.i = getelementptr inbounds %struct.CLAUSE_HELP, %struct.CLAUSE_HELP* %6, i32 0, i32 4
+  %7 = load i32*, i32** %splitfield.i, align 4
+  %arrayidx.i = getelementptr inbounds i32, i32* %7, i32 %.
+  %8 = load i32, i32* %arrayidx.i, align 4
+  %shl.i = shl i32 1, %.Level
+  %and.i = and i32 %8, %shl.i
+  %cmp4.i = icmp eq i32 %and.i, 0
+  br i1 %cmp4.i, label %for.inc, label %if.then
+
+if.then:                                          ; preds = %for.body
+  tail call void @foo(%struct.PROOFSEARCH_HELP* %Search, %struct.CLAUSE_HELP* %6)
+  store i8* null, i8** %car.i33, align 4
+  br label %for.inc
+
+for.inc:                                          ; preds = %if.then, %for.body
+  %cdr.i = getelementptr inbounds %struct.LIST_HELP, %struct.LIST_HELP* %Scan.038, i32 0, i32 0
+  %9 = load %struct.LIST_HELP*, %struct.LIST_HELP** %cdr.i, align 4
+  br label %for.body
+}
-- 
2.7.4