Index: lib/CodeGen/MachineLICM.cpp
===================================================================
--- lib/CodeGen/MachineLICM.cpp
+++ lib/CodeGen/MachineLICM.cpp
@@ -1336,6 +1336,8 @@
     // Otherwise, splice the instruction to the preheader.
     Preheader->splice(Preheader->getFirstTerminator(),MI->getParent(),MI);
 
+    MI->setDebugLoc(DebugLoc());
+
     // Update register pressure for BBs from header to this block.
     UpdateBackTraceRegPressure(MI);
 
Index: test/CodeGen/ARM/debug-info-branch-folding.ll
===================================================================
--- test/CodeGen/ARM/debug-info-branch-folding.ll
+++ test/CodeGen/ARM/debug-info-branch-folding.ll
@@ -3,11 +3,12 @@
 target triple = "thumbv7-apple-macosx10.6.7"
 
 ;CHECK: 	vadd.f32	q4, q8, q8
-;CHECK-NEXT: Ltmp1
-;CHECK-NEXT: LBB0_1
+;CHECK:         LBB0_1:
+;CHECK-NOT:     beq LBB0_1
 
-;CHECK:@DEBUG_VALUE: x <- %Q4{{$}}
-;CHECK-NEXT:@DEBUG_VALUE: y <- %Q4{{$}}
+;CHECK:         @DEBUG_VALUE: x <- %Q4{{$}}
+;CHECK-NEXT:    @DEBUG_VALUE: y <- %Q4{{$}}
+;CHECK:         beq LBB0_1
 
 
 @.str = external constant [13 x i8]
Index: test/DebugInfo/MIR/X86/mlicm-hoist.mir
===================================================================
--- test/DebugInfo/MIR/X86/mlicm-hoist.mir
+++ test/DebugInfo/MIR/X86/mlicm-hoist.mir
@@ -0,0 +1,203 @@
+--- |
+  ; RUN: llc -run-pass=machinelicm -o - %s | FileCheck %s
+  ; Line numbers should not be retained when loop invariant instructions are hoisted.
+  ; Doing so causes poor stepping bevavior.
+  ;
+  ; Created from:
+  ; int x;
+  ; int y;
+  ;
+  ; bool Process(const unsigned int *pData)
+  ; {
+  ;   int compare = 0;
+  ;   while (1)
+  ;   {
+  ;     int op = *pData++;
+  ;     switch (op)
+  ;     {
+  ;     case 1:
+  ;       x = *pData++;
+  ;       y = *pData++;
+  ;       break;
+  ;     }
+  ;   }
+  ;   return false;
+  ; }
+  ;
+  ; with clang -emit-llvm -S -O2 -g -o t.ll
+  ;      llc -relocation-model=pic -stop-after=machine-combine -O2 t.ll -o t.mir
+  ;
+  ; We check that when the machine loop invariant pass hoists instructions
+  ; out of a loop the debug information is not retained.
+  ;
+  ; CHECK-LABEL:  bb.0.entry:
+  ; CHECK:        MOV64rm %rip, 1, _, target-flags(x86-gotpcrel) @x, _ :: (load 8 from got)
+  ; CHECK:        MOV64rm %rip, 1, _, target-flags(x86-gotpcrel) @y, _ :: (load 8 from got)
+  ; CHECK-LABEL:  bb.1.while.cond:
+  ;
+  ; ModuleID = 'tx.ll'
+  source_filename = "t.cpp"
+  target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+  target triple = "x86_64-unknown-linux-gnu"
+
+  @x = local_unnamed_addr global i32 0, align 4, !dbg !0
+  @y = local_unnamed_addr global i32 0, align 4, !dbg !5
+
+  ; Function Attrs: noreturn nounwind uwtable
+  define zeroext i1 @_Z7ProcessPKj(i32* nocapture readonly %pData) local_unnamed_addr #0 !dbg !10 {
+  entry:
+    tail call void @llvm.dbg.value(metadata i32* %pData, i64 0, metadata !18, metadata !22), !dbg !23
+    tail call void @llvm.dbg.value(metadata i32 0, i64 0, metadata !19, metadata !22), !dbg !24
+    br label %while.cond, !dbg !25
+
+  while.cond:                                       ; preds = %while.cond, %sw.bb, %entry
+    %pData.addr.0 = phi i32* [ %pData, %entry ], [ %incdec.ptr2, %sw.bb ], [ %incdec.ptr, %while.cond ]
+    tail call void @llvm.dbg.value(metadata i32* %pData.addr.0, i64 0, metadata !18, metadata !22), !dbg !23
+    %incdec.ptr = getelementptr inbounds i32, i32* %pData.addr.0, i64 1, !dbg !26
+    tail call void @llvm.dbg.value(metadata i32* %incdec.ptr, i64 0, metadata !18, metadata !22), !dbg !23
+    %0 = load i32, i32* %pData.addr.0, align 4, !dbg !27, !tbaa !28
+    tail call void @llvm.dbg.value(metadata i32 %0, i64 0, metadata !20, metadata !22), !dbg !32
+    %cond = icmp eq i32 %0, 1, !dbg !33
+    br i1 %cond, label %sw.bb, label %while.cond, !dbg !33
+
+  sw.bb:                                            ; preds = %while.cond
+    %incdec.ptr1 = getelementptr inbounds i32, i32* %pData.addr.0, i64 2, !dbg !34
+    tail call void @llvm.dbg.value(metadata i32* %incdec.ptr1, i64 0, metadata !18, metadata !22), !dbg !23
+    %sunkaddr = ptrtoint i32* %pData.addr.0 to i64, !dbg !36
+    %sunkaddr1 = add i64 %sunkaddr, 4, !dbg !36
+    %sunkaddr2 = inttoptr i64 %sunkaddr1 to i32*, !dbg !36
+    %1 = load i32, i32* %sunkaddr2, align 4, !dbg !36, !tbaa !28
+    store i32 %1, i32* @x, align 4, !dbg !37, !tbaa !28
+    %incdec.ptr2 = getelementptr inbounds i32, i32* %pData.addr.0, i64 3, !dbg !38
+    tail call void @llvm.dbg.value(metadata i32* %incdec.ptr2, i64 0, metadata !18, metadata !22), !dbg !23
+    %2 = load i32, i32* %incdec.ptr1, align 4, !dbg !39, !tbaa !28
+    store i32 %2, i32* @y, align 4, !dbg !40, !tbaa !28
+    br label %while.cond, !dbg !41
+  }
+
+  ; Function Attrs: nounwind readnone
+  declare void @llvm.dbg.value(metadata, i64, metadata, metadata) #1
+
+  attributes #0 = { noreturn nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+  attributes #1 = { nounwind readnone }
+
+  !llvm.dbg.cu = !{!1}
+  !llvm.module.flags = !{!7, !8}
+  !llvm.ident = !{!9}
+
+  !0 = distinct !DIGlobalVariable(name: "x", scope: !1, file: !2, line: 1, type: !6, isLocal: false, isDefinition: true)
+  !1 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, file: !2, producer: "clang version 4.0.0 ", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !3, globals: !4)
+  !2 = !DIFile(filename: "t.cpp", directory: "/home/test")
+  !3 = !{}
+  !4 = !{!0, !5}
+  !5 = distinct !DIGlobalVariable(name: "y", scope: !1, file: !2, line: 2, type: !6, isLocal: false, isDefinition: true)
+  !6 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+  !7 = !{i32 2, !"Dwarf Version", i32 4}
+  !8 = !{i32 2, !"Debug Info Version", i32 3}
+  !9 = !{!"clang version 4.0.0 "}
+  !10 = distinct !DISubprogram(name: "Process", linkageName: "_Z7ProcessPKj", scope: !2, file: !2, line: 4, type: !11, isLocal: false, isDefinition: true, scopeLine: 5, flags: DIFlagPrototyped, isOptimized: true, unit: !1, variables: !17)
+  !11 = !DISubroutineType(types: !12)
+  !12 = !{!13, !14}
+  !13 = !DIBasicType(name: "bool", size: 8, encoding: DW_ATE_boolean)
+  !14 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !15, size: 64)
+  !15 = !DIDerivedType(tag: DW_TAG_const_type, baseType: !16)
+  !16 = !DIBasicType(name: "unsigned int", size: 32, encoding: DW_ATE_unsigned)
+  !17 = !{!18, !19, !20}
+  !18 = !DILocalVariable(name: "pData", arg: 1, scope: !10, file: !2, line: 4, type: !14)
+  !19 = !DILocalVariable(name: "compare", scope: !10, file: !2, line: 6, type: !6)
+  !20 = !DILocalVariable(name: "op", scope: !21, file: !2, line: 9, type: !6)
+  !21 = distinct !DILexicalBlock(scope: !10, file: !2, line: 8, column: 3)
+  !22 = !DIExpression()
+  !23 = !DILocation(line: 4, column: 34, scope: !10)
+  !24 = !DILocation(line: 6, column: 7, scope: !10)
+  !25 = !DILocation(line: 7, column: 3, scope: !10)
+  !26 = !DILocation(line: 9, column: 20, scope: !21)
+  !27 = !DILocation(line: 9, column: 14, scope: !21)
+  !28 = !{!29, !29, i64 0}
+  !29 = !{!"int", !30, i64 0}
+  !30 = !{!"omnipotent char", !31, i64 0}
+  !31 = !{!"Simple C++ TBAA"}
+  !32 = !DILocation(line: 9, column: 9, scope: !21)
+  !33 = !DILocation(line: 10, column: 5, scope: !21)
+  !34 = !DILocation(line: 13, column: 17, scope: !35)
+  !35 = distinct !DILexicalBlock(scope: !21, file: !2, line: 11, column: 5)
+  !36 = !DILocation(line: 13, column: 11, scope: !35)
+  !37 = !DILocation(line: 13, column: 9, scope: !35)
+  !38 = !DILocation(line: 14, column: 17, scope: !35)
+  !39 = !DILocation(line: 14, column: 11, scope: !35)
+  !40 = !DILocation(line: 14, column: 9, scope: !35)
+  !41 = !DILocation(line: 15, column: 7, scope: !35)
+
+...
+---
+name:            _Z7ProcessPKj
+alignment:       4
+exposesReturnsTwice: false
+legalized:       false
+regBankSelected: false
+selected:        false
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: gr64 }
+  - { id: 1, class: gr64 }
+  - { id: 2, class: gr64 }
+  - { id: 3, class: gr64 }
+  - { id: 4, class: gr32 }
+  - { id: 5, class: gr32 }
+  - { id: 6, class: gr32 }
+  - { id: 7, class: gr64 }
+  - { id: 8, class: gr32 }
+  - { id: 9, class: gr64 }
+liveins:
+  - { reg: '%rdi', virtual-reg: '%3' }
+frameInfo:
+  isFrameAddressTaken: false
+  isReturnAddressTaken: false
+  hasStackMap:     false
+  hasPatchPoint:   false
+  stackSize:       0
+  offsetAdjustment: 0
+  maxAlignment:    0
+  adjustsStack:    false
+  hasCalls:        false
+  maxCallFrameSize: 0
+  hasOpaqueSPAdjustment: false
+  hasVAStart:      false
+  hasMustTailInVarArgFunc: false
+body:             |
+  bb.0.entry:
+    successors: %bb.1.while.cond(0x80000000)
+    liveins: %rdi
+
+    DBG_VALUE debug-use %rdi, debug-use _, !18, !22, debug-location !23
+    %3 = COPY %rdi
+    DBG_VALUE debug-use %3, debug-use _, !18, !22, debug-location !23
+    DBG_VALUE 0, debug-use _, !19, !22, debug-location !24
+
+  bb.1.while.cond:
+    successors: %bb.2.sw.bb(0x40000000), %bb.1.while.cond(0x40000000)
+
+    %0 = PHI %3, %bb.0.entry, %1, %bb.1.while.cond, %2, %bb.2.sw.bb
+    DBG_VALUE debug-use %0, debug-use _, !18, !22, debug-location !23
+    %1 = ADD64ri8 %0, 4, implicit-def dead %eflags, debug-location !26
+    DBG_VALUE debug-use %1, debug-use _, !18, !22, debug-location !23
+    %4 = MOV32rm %0, 1, _, 0, _, debug-location !27 :: (load 4 from %ir.pData.addr.0, !tbaa !28)
+    DBG_VALUE debug-use %4, debug-use _, !20, !22, debug-location !32
+    %5 = SUB32ri8 %4, 1, implicit-def %eflags, debug-location !33
+    JNE_1 %bb.1.while.cond, implicit %eflags, debug-location !33
+    JMP_1 %bb.2.sw.bb, debug-location !33
+
+  bb.2.sw.bb:
+    successors: %bb.1.while.cond(0x80000000)
+
+    %6 = MOV32rm %0, 1, _, 4, _, debug-location !36 :: (load 4 from %ir.sunkaddr2, !tbaa !28)
+    %7 = MOV64rm %rip, 1, _, target-flags(x86-gotpcrel) @x, _, debug-location !37 :: (load 8 from got)
+    MOV32mr killed %7, 1, _, 0, _, killed %6, debug-location !37 :: (store 4 into @x, !tbaa !28)
+    %2 = ADD64ri8 %0, 12, implicit-def dead %eflags, debug-location !38
+    DBG_VALUE debug-use %2, debug-use _, !18, !22, debug-location !23
+    %8 = MOV32rm %0, 1, _, 8, _, debug-location !39 :: (load 4 from %ir.incdec.ptr1, !tbaa !28)
+    %9 = MOV64rm %rip, 1, _, target-flags(x86-gotpcrel) @y, _, debug-location !40 :: (load 8 from got)
+    MOV32mr killed %9, 1, _, 0, _, killed %8, debug-location !40 :: (store 4 into @y, !tbaa !28)
+    JMP_1 %bb.1.while.cond, debug-location !41
+
+...