diff --git a/lib/Target/ARM/ARMAsmPrinter.cpp b/lib/Target/ARM/ARMAsmPrinter.cpp
index e1a2cc57abf..478c2468381 100644
--- a/lib/Target/ARM/ARMAsmPrinter.cpp
+++ b/lib/Target/ARM/ARMAsmPrinter.cpp
@@ -945,6 +945,16 @@ void ARMAsmPrinter::EmitInstruction(const MachineInstr *MI) {
   unsigned Opc = MI->getOpcode();
   switch (Opc) {
   default: break;
+  case ARM::LDMIA_RET: {
+    // LDMIA_RET is just a normal LDMIA_UPD instruction that targets PC and as
+    // such has additional code-gen properties and scheduling information.
+    // To emit it, we just construct as normal and set the opcode to LDMIA_UPD.
+    MCInst TmpInst;
+    LowerARMMachineInstrToMCInst(MI, TmpInst, *this);
+    TmpInst.setOpcode(ARM::LDMIA_UPD);
+    OutStreamer.EmitInstruction(TmpInst);
+    return;
+  }
   case ARM::t2ADDrSPi:
   case ARM::t2ADDrSPi12:
   case ARM::t2SUBrSPi:
diff --git a/lib/Target/ARM/ARMInstrInfo.td b/lib/Target/ARM/ARMInstrInfo.td
index 2bfe11c955e..bb40e39903b 100644
--- a/lib/Target/ARM/ARMInstrInfo.td
+++ b/lib/Target/ARM/ARMInstrInfo.td
@@ -1928,16 +1928,10 @@ def : MnemonicAlias<"stm", "stmia">;
 // FIXME: Should pc be an implicit operand like PICADD, etc?
 let isReturn = 1, isTerminator = 1, isBarrier = 1, mayLoad = 1,
     hasExtraDefRegAllocReq = 1, isCodeGenOnly = 1 in
-// FIXME: Should be a pseudo-instruction.
-def LDMIA_RET : AXI4<(outs GPR:$wb), (ins GPR:$Rn, pred:$p,
-                                      reglist:$regs, variable_ops),
-                     IndexModeUpd, LdStMulFrm, IIC_iLoad_mBr,
-                     "ldmia${p}\t$Rn!, $regs",
-                     "$Rn = $wb", []> {
-  let Inst{24-23} = 0b01;       // Increment After
-  let Inst{21}    = 1;          // Writeback
-  let Inst{20}    = 1;          // Load
-}
+def LDMIA_RET : ARMPseudoInst<(outs GPR:$wb), (ins GPR:$Rn, pred:$p,
+                                               reglist:$regs, variable_ops),
+                     Size4Bytes, IIC_iLoad_mBr, []>,
+      RegConstraint<"$Rn = $wb">;
 
 //===----------------------------------------------------------------------===//
 //  Move Instructions.
diff --git a/test/CodeGen/ARM/2010-03-18-ldm-rtrn.ll b/test/CodeGen/ARM/2010-03-18-ldm-rtrn.ll
index d9e1a1486a3..fee86008ad7 100644
--- a/test/CodeGen/ARM/2010-03-18-ldm-rtrn.ll
+++ b/test/CodeGen/ARM/2010-03-18-ldm-rtrn.ll
@@ -6,7 +6,7 @@ define i32 @bar(i32 %a) nounwind {
 entry:
   %0 = tail call i32 @foo(i32 %a) nounwind ; <i32> [#uses=1]
   %1 = add nsw i32 %0, 3                          ; <i32> [#uses=1]
-; CHECK: ldmia	sp!, {r11, pc}
+; CHECK: pop {r11, pc}
 ; V4: pop
 ; V4-NEXT: mov pc, lr
   ret i32 %1
diff --git a/test/CodeGen/ARM/2010-10-25-ifcvt-ldm.ll b/test/CodeGen/ARM/2010-10-25-ifcvt-ldm.ll
index 163c9b030ec..32d350e9c8b 100644
--- a/test/CodeGen/ARM/2010-10-25-ifcvt-ldm.ll
+++ b/test/CodeGen/ARM/2010-10-25-ifcvt-ldm.ll
@@ -4,9 +4,9 @@
 ; was being treated as an instruction count.
 
 ; CHECK: push
-; CHECK: ldmia
-; CHECK: ldmia
-; CHECK: ldmia
+; CHECK: pop
+; CHECK: pop
+; CHECK: pop
 
 define i32 @test(i32 %x) {
 entry:
diff --git a/test/CodeGen/ARM/2010-11-29-PrologueBug.ll b/test/CodeGen/ARM/2010-11-29-PrologueBug.ll
index 8d7541feae9..e3c18cefd51 100644
--- a/test/CodeGen/ARM/2010-11-29-PrologueBug.ll
+++ b/test/CodeGen/ARM/2010-11-29-PrologueBug.ll
@@ -10,7 +10,7 @@ entry:
 ; ARM: bl _foo
 ; ARM: bl _foo
 ; ARM: bl _foo
-; ARM: ldmia sp!, {r7, pc}
+; ARM: pop {r7, pc}
 
 ; THUMB2: t:
 ; THUMB2: push
diff --git a/test/CodeGen/ARM/bx_fold.ll b/test/CodeGen/ARM/bx_fold.ll
index 09f1aae0a9f..5533038fb82 100644
--- a/test/CodeGen/ARM/bx_fold.ll
+++ b/test/CodeGen/ARM/bx_fold.ll
@@ -24,7 +24,7 @@ bb1:		; preds = %bb, %entry
 
 bb18:		; preds = %bb1
 ; CHECK-NOT: bx
-; CHECK: ldmia sp!
+; CHECK: pop
 	ret void
 }
 
diff --git a/test/CodeGen/ARM/code-placement.ll b/test/CodeGen/ARM/code-placement.ll
index 036598fc060..91ef6592522 100644
--- a/test/CodeGen/ARM/code-placement.ll
+++ b/test/CodeGen/ARM/code-placement.ll
@@ -72,7 +72,7 @@ bb2.preheader:                                    ; preds = %bb3, %bb.nph15
   br i1 %4, label %bb1, label %bb3
 
 ; CHECK: LBB1_[[RET]]: @ %bb5
-; CHECK: ldmia sp!
+; CHECK: pop
 bb5:                                              ; preds = %bb3, %entry
   %sum.1.lcssa = phi i32 [ 0, %entry ], [ %sum.0.lcssa, %bb3 ] ; <i32> [#uses=1]
   ret i32 %sum.1.lcssa
diff --git a/test/CodeGen/ARM/ifcvt10.ll b/test/CodeGen/ARM/ifcvt10.ll
index 75428ac2165..18f87bfc2e7 100644
--- a/test/CodeGen/ARM/ifcvt10.ll
+++ b/test/CodeGen/ARM/ifcvt10.ll
@@ -9,9 +9,9 @@ entry:
 ; CHECK: t:
 ; CHECK: vpop {d8}
 ; CHECK-NOT: vpopne
-; CHECK: ldmia sp!, {r7, pc}
+; CHECK: pop {r7, pc}
 ; CHECK: vpop {d8}
-; CHECK: ldmia sp!, {r7, pc}
+; CHECK: pop {r7, pc}
   br i1 undef, label %if.else, label %if.then
 
 if.then:                                          ; preds = %entry
diff --git a/test/CodeGen/ARM/ifcvt5.ll b/test/CodeGen/ARM/ifcvt5.ll
index bca2ae346a6..3615055f8b2 100644
--- a/test/CodeGen/ARM/ifcvt5.ll
+++ b/test/CodeGen/ARM/ifcvt5.ll
@@ -11,7 +11,7 @@ entry:
 
 define i32 @t1(i32 %a, i32 %b) {
 ; CHECK: t1:
-; CHECK: ldmialt sp!, {r7, pc}
+; CHECK: poplt {r7, pc}
 entry:
 	%tmp1 = icmp sgt i32 %a, 10		; <i1> [#uses=1]
 	br i1 %tmp1, label %cond_true, label %UnifiedReturnBlock
diff --git a/test/CodeGen/ARM/ifcvt6.ll b/test/CodeGen/ARM/ifcvt6.ll
index 5edf32fd1af..23276576855 100644
--- a/test/CodeGen/ARM/ifcvt6.ll
+++ b/test/CodeGen/ARM/ifcvt6.ll
@@ -3,7 +3,7 @@
 define void @foo(i32 %X, i32 %Y) {
 entry:
 ; CHECK: cmpne
-; CHECK: ldmiahi sp!
+; CHECK: pophi
 	%tmp1 = icmp ult i32 %X, 4		; <i1> [#uses=1]
 	%tmp4 = icmp eq i32 %Y, 0		; <i1> [#uses=1]
 	%tmp7 = or i1 %tmp4, %tmp1		; <i1> [#uses=1]
diff --git a/test/CodeGen/ARM/ifcvt7.ll b/test/CodeGen/ARM/ifcvt7.ll
index 62e13557cfd..476ed4d47c6 100644
--- a/test/CodeGen/ARM/ifcvt7.ll
+++ b/test/CodeGen/ARM/ifcvt7.ll
@@ -6,7 +6,7 @@
 define fastcc i32 @CountTree(%struct.quad_struct* %tree) {
 ; CHECK: cmpeq
 ; CHECK: moveq
-; CHECK: ldmiaeq sp!
+; CHECK: popeq
 entry:
 	br label %tailrecurse
 
diff --git a/test/CodeGen/ARM/ifcvt8.ll b/test/CodeGen/ARM/ifcvt8.ll
index 5fdfc4ea680..ca9a5c63cda 100644
--- a/test/CodeGen/ARM/ifcvt8.ll
+++ b/test/CodeGen/ARM/ifcvt8.ll
@@ -5,7 +5,7 @@
 declare void @abort()
 
 define fastcc void @t(%struct.SString* %word, i8 signext  %c) {
-; CHECK: ldmiane sp!
+; CHECK: popne
 entry:
 	%tmp1 = icmp eq %struct.SString* %word, null		; <i1> [#uses=1]
 	br i1 %tmp1, label %cond_true, label %cond_false
diff --git a/test/CodeGen/ARM/ldm.ll b/test/CodeGen/ARM/ldm.ll
index 2f1b85ebbb0..db78fd06ab2 100644
--- a/test/CodeGen/ARM/ldm.ll
+++ b/test/CodeGen/ARM/ldm.ll
@@ -5,9 +5,9 @@
 
 define i32 @t1() {
 ; CHECK: t1:
-; CHECK: ldmia
+; CHECK: pop
 ; V4T: t1:
-; V4T: ldmia
+; V4T: pop
         %tmp = load i32* getelementptr ([0 x i32]* @X, i32 0, i32 0)            ; <i32> [#uses=1]
         %tmp3 = load i32* getelementptr ([0 x i32]* @X, i32 0, i32 1)           ; <i32> [#uses=1]
         %tmp4 = tail call i32 @f1( i32 %tmp, i32 %tmp3 )                ; <i32> [#uses=1]
@@ -16,9 +16,9 @@ define i32 @t1() {
 
 define i32 @t2() {
 ; CHECK: t2:
-; CHECK: ldmia
+; CHECK: pop
 ; V4T: t2:
-; V4T: ldmia
+; V4T: pop
         %tmp = load i32* getelementptr ([0 x i32]* @X, i32 0, i32 2)            ; <i32> [#uses=1]
         %tmp3 = load i32* getelementptr ([0 x i32]* @X, i32 0, i32 3)           ; <i32> [#uses=1]
         %tmp5 = load i32* getelementptr ([0 x i32]* @X, i32 0, i32 4)           ; <i32> [#uses=1]
@@ -29,7 +29,7 @@ define i32 @t2() {
 define i32 @t3() {
 ; CHECK: t3:
 ; CHECK: ldmib
-; CHECK: ldmia sp!
+; CHECK: pop
 ; V4T: t3:
 ; V4T: ldmib
 ; V4T: pop