diff --git a/lib/Target/Hexagon/AsmParser/HexagonAsmParser.cpp b/lib/Target/Hexagon/AsmParser/HexagonAsmParser.cpp
index 6998383f08e..7c70a7e3d0a 100644
--- a/lib/Target/Hexagon/AsmParser/HexagonAsmParser.cpp
+++ b/lib/Target/Hexagon/AsmParser/HexagonAsmParser.cpp
@@ -2198,6 +2198,11 @@ int HexagonAsmParser::processInstruction(MCInst &Inst,
     if (!HexagonMCInstrInfo::mustExtend(*Inst.getOperand(0).getExpr()))
       Inst.setOpcode(Hexagon::S2_storerinewgp);
     break;
+  case Hexagon::A2_zxtb: {
+    Inst.setOpcode(Hexagon::A2_andir);
+    Inst.addOperand(MCOperand::createExpr(MCConstantExpr::create(255, Context)));
+    break;
+  }
   } // switch
 
   return Match_Success;
diff --git a/lib/Target/Hexagon/HexagonAsmPrinter.cpp b/lib/Target/Hexagon/HexagonAsmPrinter.cpp
index 54db5ad4374..01ba1ccd37f 100644
--- a/lib/Target/Hexagon/HexagonAsmPrinter.cpp
+++ b/lib/Target/Hexagon/HexagonAsmPrinter.cpp
@@ -282,6 +282,36 @@ void HexagonAsmPrinter::HexagonProcessInstruction(MCInst &Inst,
     break;
   }
 
+  case Hexagon::A2_tfrf: {
+    Inst.setOpcode(Hexagon::A2_paddif);
+    Inst.addOperand(MCOperand::createExpr(MCConstantExpr::create(0, OutContext)));
+    break;
+  }
+
+  case Hexagon::A2_tfrt: {
+    Inst.setOpcode(Hexagon::A2_paddit);
+    Inst.addOperand(MCOperand::createExpr(MCConstantExpr::create(0, OutContext)));
+    break;
+  }
+
+  case Hexagon::A2_tfrfnew: {
+    Inst.setOpcode(Hexagon::A2_paddifnew);
+    Inst.addOperand(MCOperand::createExpr(MCConstantExpr::create(0, OutContext)));
+    break;
+  }
+
+  case Hexagon::A2_tfrtnew: {
+    Inst.setOpcode(Hexagon::A2_padditnew);
+    Inst.addOperand(MCOperand::createExpr(MCConstantExpr::create(0, OutContext)));
+    break;
+  }
+
+  case Hexagon::A2_zxtb: {
+    Inst.setOpcode(Hexagon::A2_andir);
+    Inst.addOperand(MCOperand::createExpr(MCConstantExpr::create(255, OutContext)));
+    break;
+  }
+
   // "$dst = CONST64(#$src1)",
   case Hexagon::CONST64:
     if (!OutStreamer->hasRawTextSupport()) {
@@ -376,6 +406,9 @@ void HexagonAsmPrinter::HexagonProcessInstruction(MCInst &Inst,
     Rs.setReg(getHexagonRegisterPair(Rs.getReg(), RI));
     return;
   }
+  case Hexagon::PS_call_nr:
+    Inst.setOpcode(Hexagon::J2_call);
+    break;
   case Hexagon::S5_asrhub_rnd_sat_goodsyntax: {
     MCOperand &MO = MappedInst.getOperand(2);
     int64_t Imm;
diff --git a/lib/Target/Hexagon/HexagonInstrInfo.td b/lib/Target/Hexagon/HexagonInstrInfo.td
index c5719ad5b6d..16298cae083 100644
--- a/lib/Target/Hexagon/HexagonInstrInfo.td
+++ b/lib/Target/Hexagon/HexagonInstrInfo.td
@@ -401,13 +401,12 @@ def A2_tfril: T_tfr16<0>;
 def A2_tfrih: T_tfr16<1>;
 
 // Conditional transfer is an alias to conditional "Rd = add(Rs, #0)".
-let isPredicated = 1, hasNewValue = 1, opNewValue = 0 in
+let isPredicated = 1, hasNewValue = 1, isCodeGenOnly = 1, opNewValue = 0,
+    isPseudo = 1 in
 class T_tfr_pred<bit isPredNot, bit isPredNew>
   : ALU32Inst<(outs IntRegs:$dst),
               (ins PredRegs:$src1, IntRegs:$src2),
-              "if ("#!if(isPredNot, "!", "")#
-              "$src1"#!if(isPredNew, ".new", "")#
-              ") $dst = $src2"> {
+              ""> {
     bits<5> dst;
     bits<2> src1;
     bits<5> src2;
@@ -487,6 +486,11 @@ multiclass TFR64_base<string BaseName> {
   }
 }
 
+def A2_tfrfAlias : InstAlias<"if (!$Pu4) $Rd32=$Rs32", (A2_paddif IntRegs:$Rd32, PredRegs:$Pu4, IntRegs:$Rs32, 0)>;
+def A2_tfrfnewAlias : InstAlias<"if (!$Pu4.new) $Rd32=$Rs32", (A2_paddifnew IntRegs:$Rd32, PredRegs:$Pu4, IntRegs:$Rs32, 0)>;
+def A2_tfrtAlias : InstAlias<"if ($Pu4) $Rd32=$Rs32", (A2_paddit IntRegs:$Rd32, PredRegs:$Pu4, IntRegs:$Rs32, 0)>;
+def A2_tfrtnewAlias : InstAlias<"if ($Pu4.new) $Rd32=$Rs32", (A2_padditnew IntRegs:$Rd32, PredRegs:$Pu4, IntRegs:$Rs32, 0)>;
+
 let InputType = "imm", isExtendable = 1, isExtentSigned = 1, opExtentBits = 12,
     isMoveImm = 1, opExtendable = 2, BaseOpcode = "TFRI", CextOpcode = "TFR",
     hasSideEffects = 0, isPredicated = 1, hasNewValue = 1 in
@@ -699,19 +703,7 @@ defm zxth : ALU32_2op_base<"zxth", 0b110>, PredNewRel;
 
 let hasNewValue = 1, opNewValue = 0 in
 class T_ZXTB: ALU32Inst < (outs IntRegs:$Rd), (ins IntRegs:$Rs),
-  "$Rd = zxtb($Rs)", [] > { // Rd = and(Rs,255)
-    bits<5> Rd;
-    bits<5> Rs;
-    bits<10> s10 = 255;
-
-    let IClass = 0b0111;
-
-    let Inst{27-22} = 0b011000;
-    let Inst{4-0} = Rd;
-    let Inst{20-16} = Rs;
-    let Inst{21} = s10{9};
-    let Inst{13-5} = s10{8-0};
-}
+  "$Rd=zxtb($Rs)", [] >;
 
 //Rd=zxtb(Rs): assembler mapped to "Rd=and(Rs,#255)
 multiclass ZXTB_base <string mnemonic, bits<3> minOp> {
diff --git a/lib/Target/Hexagon/HexagonInstrInfoV3.td b/lib/Target/Hexagon/HexagonInstrInfoV3.td
index 225f9440507..5b7610a68af 100644
--- a/lib/Target/Hexagon/HexagonInstrInfoV3.td
+++ b/lib/Target/Hexagon/HexagonInstrInfoV3.td
@@ -61,7 +61,7 @@ multiclass T_Calls<bit CSR, string ExtStr> {
 defm J2_call: T_Calls<1, "">, PredRel;
 
 let isCodeGenOnly = 1, isCall = 1, hasSideEffects = 1,
-    Defs = VolatileV3.Regs in
+    Defs = VolatileV3.Regs, isPseudo = 1 in
 def PS_call_nr : T_Call<1, "">, PredRel;
 
 let isCodeGenOnly = 1, isCall = 1, hasSideEffects = 1,
diff --git a/test/CodeGen/Hexagon/swp-stages4.ll b/test/CodeGen/Hexagon/swp-stages4.ll
new file mode 100644
index 00000000000..cdd09845ba5
--- /dev/null
+++ b/test/CodeGen/Hexagon/swp-stages4.ll
@@ -0,0 +1,94 @@
+; RUN: llc -march=hexagon -mcpu=hexagonv5 -enable-pipeliner -pipeliner-max-stages=2 -disable-block-placement=0 -hexagon-bit=0 < %s | FileCheck %s
+
+; Test that we rename registers correctly for multiple stages when there is a
+; Phi and depends upon another Phi.
+
+; CHECK: = and
+; CHECK: = and
+; CHECK: = and
+; CHECK: [[REG0:(r[0-9]+)]] = and([[REG1:(r[0-9]+)]], #255)
+; CHECK-NOT: [[REG0]] = and([[REG1]], #255)
+; CHECK: loop0(.LBB0_[[LOOP:.]],
+; CHECK: .LBB0_[[LOOP]]:
+; CHECK: [[REG0]] += add
+; CHECK: = and
+; CHECK: = and
+; CHECK: [[REG0]] = and
+; CHECK: endloop
+
+; Function Attrs: nounwind
+define void @test(i8* noalias nocapture %src, i32 %srcWidth, i32 %srcHeight, i32 %srcStride, i8* noalias nocapture %dst, i32 %dstStride) #0 {
+entry:
+  %sub = add i32 %srcWidth, -1
+  %sub1 = add i32 %srcHeight, -1
+  %add.ptr = getelementptr inbounds i8, i8* %src, i32 %srcStride
+  %add.ptr.sum = mul i32 %srcStride, 2
+  %add.ptr2 = getelementptr inbounds i8, i8* %src, i32 %add.ptr.sum
+  br label %for.body.lr.ph
+
+for.body.lr.ph:
+  %0 = add i32 %srcHeight, -2
+  %1 = mul i32 %0, %dstStride
+  %2 = mul i32 %0, %srcStride
+  %3 = mul i32 %sub1, %srcStride
+  br label %for.cond
+
+for.cond:
+  %scevgep = getelementptr i8, i8* %dst, i32 %1
+  %scevgep220 = getelementptr i8, i8* %src, i32 %2
+  %scevgep221 = getelementptr i8, i8* %src, i32 %3
+  %arrayidx6 = getelementptr inbounds i8, i8* %src, i32 1
+  %add11 = add i32 %srcStride, 1
+  %arrayidx12 = getelementptr inbounds i8, i8* %src, i32 %add11
+  br label %for.body75.preheader
+
+for.body75.preheader:
+  %sri = load i8, i8* %arrayidx6, align 1
+  %sri224 = load i8, i8* %src, align 1
+  %sri227 = load i8, i8* %arrayidx12, align 1
+  %sri229 = load i8, i8* %add.ptr, align 1
+  br label %for.body75
+
+for.body75:
+  %j.0211 = phi i32 [ %add82, %for.body75 ], [ 1, %for.body75.preheader ]
+  %sr = phi i8 [ %4, %for.body75 ], [ %sri, %for.body75.preheader ]
+  %sr225 = phi i8 [ %sr, %for.body75 ], [ %sri224, %for.body75.preheader ]
+  %sr230 = phi i8 [ %5, %for.body75 ], [ %sri227, %for.body75.preheader ]
+  %sr231 = phi i8 [ %sr230, %for.body75 ], [ %sri229, %for.body75.preheader ]
+  %conv78 = zext i8 %sr225 to i32
+  %conv80 = zext i8 %sr to i32
+  %add81 = add nsw i32 %conv80, %conv78
+  %add82 = add i32 %j.0211, 1
+  %arrayidx83 = getelementptr inbounds i8, i8* %src, i32 %add82
+  %4 = load i8, i8* %arrayidx83, align 1, !tbaa !0
+  %conv84 = zext i8 %4 to i32
+  %add85 = add nsw i32 %add81, %conv84
+  %conv88 = zext i8 %sr231 to i32
+  %add89 = add nsw i32 %add85, %conv88
+  %conv91 = zext i8 %sr230 to i32
+  %add92 = add nsw i32 %add89, %conv91
+  %add.ptr.sum208 = add i32 %add82, %srcStride
+  %arrayidx94 = getelementptr inbounds i8, i8* %src, i32 %add.ptr.sum208
+  %5 = load i8, i8* %arrayidx94, align 1, !tbaa !0
+  %conv95 = zext i8 %5 to i32
+  %add96 = add nsw i32 %add92, %conv95
+  %mul97 = mul nsw i32 %add96, 7282
+  %add98 = add nsw i32 %mul97, 32768
+  %shr99209 = lshr i32 %add98, 16
+  %conv100 = trunc i32 %shr99209 to i8
+  %arrayidx101 = getelementptr inbounds i8, i8* %dst, i32 %j.0211
+  store i8 %conv100, i8* %arrayidx101, align 1, !tbaa !0
+  %exitcond = icmp eq i32 %add82, %sub
+  br i1 %exitcond, label %for.end104.loopexit, label %for.body75
+
+for.end104.loopexit:
+  br label %for.end104
+
+for.end104:
+  ret void
+}
+
+attributes #0 = { nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf"="true" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "unsafe-fp-math"="false" "use-soft-float"="false" }
+
+!0 = !{!"omnipotent char", !1}
+!1 = !{!"Simple C/C++ TBAA"}
diff --git a/test/CodeGen/Hexagon/swp-stages5.ll b/test/CodeGen/Hexagon/swp-stages5.ll
new file mode 100644
index 00000000000..f83aa32ae0a
--- /dev/null
+++ b/test/CodeGen/Hexagon/swp-stages5.ll
@@ -0,0 +1,78 @@
+; RUN: llc -march=hexagon -mcpu=hexagonv5 -enable-pipeliner -pipeliner-max-stages=2 -hexagon-bit=0 < %s | FileCheck %s
+
+; Very similar to swp-stages4.ll, but the pipelined schedule is a little
+; different.
+
+; CHECK: = memub(r{{[0-9]+}}++#1)
+; CHECK-DAG: [[REG0:(r[0-9]+)]] = memub(r{{[0-9]+}}++#1)
+; CHECK-DAG: loop0(.LBB0_[[LOOP:.]],
+; CHECK: .LBB0_[[LOOP]]:
+; CHECK: = and([[REG0]], #255)
+; CHECK: [[REG0]]{{[:0-9]*}} =
+; CHECK: endloop
+
+define void @fred(i8* noalias nocapture %src, i32 %srcWidth, i32 %srcHeight, i32 %srcStride, i8* noalias nocapture %dst, i32 %dstStride) #0 {
+entry:
+  %sub = add i32 %srcWidth, -1
+  %sub1 = add i32 %srcHeight, -1
+  %add.ptr = getelementptr inbounds i8, i8* %src, i32 %srcStride
+  %add.ptr.sum = mul i32 %srcStride, 2
+  %add.ptr2 = getelementptr inbounds i8, i8* %src, i32 %add.ptr.sum
+  %cmp212 = icmp ugt i32 %sub1, 1
+  br i1 %cmp212, label %for.body.lr.ph, label %for.end
+
+for.body.lr.ph:
+  br label %for.body74.preheader
+
+for.body74.preheader:
+  %0 = load i8, i8* %add.ptr, align 1, !tbaa !0
+  %arrayidx40 = getelementptr inbounds i8, i8* %add.ptr, i32 1
+  %1 = load i8, i8* %arrayidx40, align 1, !tbaa !0
+  %2 = load i8, i8* %add.ptr, align 1, !tbaa !0
+  %arrayidx46 = getelementptr inbounds i8, i8* %add.ptr, i32 1
+  %3 = load i8, i8* %arrayidx46, align 1, !tbaa !0
+  br label %for.body74
+
+for.body74:
+  %4 = phi i8 [ %9, %for.body74 ], [ %3, %for.body74.preheader ]
+  %5 = phi i8 [ %4, %for.body74 ], [ %2, %for.body74.preheader ]
+  %6 = phi i8 [ %8, %for.body74 ], [ %1, %for.body74.preheader ]
+  %7 = phi i8 [ %6, %for.body74 ], [ %0, %for.body74.preheader ]
+  %j.0211 = phi i32 [ %add81, %for.body74 ], [ 1, %for.body74.preheader ]
+  %conv77 = zext i8 %7 to i32
+  %conv79 = zext i8 %6 to i32
+  %add80 = add nsw i32 %conv79, %conv77
+  %add81 = add i32 %j.0211, 1
+  %arrayidx82 = getelementptr inbounds i8, i8* %src, i32 %add81
+  %8 = load i8, i8* %arrayidx82, align 1, !tbaa !0
+  %conv83 = zext i8 %8 to i32
+  %add84 = add nsw i32 %add80, %conv83
+  %conv87 = zext i8 %5 to i32
+  %add88 = add nsw i32 %add84, %conv87
+  %conv90 = zext i8 %4 to i32
+  %add91 = add nsw i32 %add88, %conv90
+  %arrayidx93 = getelementptr inbounds i8, i8* %add.ptr, i32 %add81
+  %9 = load i8, i8* %arrayidx93, align 1, !tbaa !0
+  %conv94 = zext i8 %9 to i32
+  %add95 = add nsw i32 %add91, %conv94
+  %mul96 = mul nsw i32 %add95, 7282
+  %add97 = add nsw i32 %mul96, 32768
+  %shr98208 = lshr i32 %add97, 16
+  %conv99 = trunc i32 %shr98208 to i8
+  %add.ptr5.sum209 = add i32 %j.0211, %dstStride
+  %arrayidx100 = getelementptr inbounds i8, i8* %dst, i32 %add.ptr5.sum209
+  store i8 %conv99, i8* %arrayidx100, align 1, !tbaa !0
+  %exitcond = icmp eq i32 %add81, %sub
+  br i1 %exitcond, label %for.end103.loopexit, label %for.body74
+
+for.end103.loopexit:
+  br label %for.end
+
+for.end:
+  ret void
+}
+
+attributes #0 = { nounwind }
+
+!0 = !{!"omnipotent char", !1}
+!1 = !{!"Simple C/C++ TBAA"}
diff --git a/test/MC/Hexagon/asmMap.s b/test/MC/Hexagon/asmMap.s
index 81bb8f31f02..f9dc0afc47c 100644
--- a/test/MC/Hexagon/asmMap.s
+++ b/test/MC/Hexagon/asmMap.s
@@ -540,10 +540,10 @@ r5=zxtb(r20)
 	p0=cmp.eq(r0,##179976360)
 }
 
-#CHECK: 74f9c00f { if (!p3) r15{{ *}}={{ *}}r25
+#CHECK: 74f9c00f { if (!p3) r15{{ *}} ={{ *}}add(r25, #0)
 if (!p3) r15=r25
 
-#CHECK: 7425c005 { if (p1) r5{{ *}}={{ *}}r5
+#CHECK: 7425c005 { if (p1) r5{{ *}}={{ *}}add(r5, #0)
 if (p1) r5=r5
 
 #CHECK: e9badae2 { r2{{ *}}={{ *}}vrcmpys(r27:26, r27:26):<<1:rnd:sat:raw:lo
@@ -561,7 +561,7 @@ if (!p2) r3:2=r7:6
 #CHECK: fd0dcc7e { if (p3) r31:30{{ *}}={{ *}}{{r13:12|combine\(r13, *r12\)}}
 if (p3) r31:30=r13:12
 
-#CHECK: 748ae015 if (!p0.new) r21{{ *}}={{ *}}r10
+#CHECK: 748ae015 if (!p0.new) r21{{ *}}={{ *}}add(r10, #0)
 {
   p0=cmp.eq(r23,##805633208)
   if (!p0.new) r21=r10
@@ -605,4 +605,4 @@ r19:18=vrcmpys(r11:10,r16):<<1:sat
 r21:20=memb_fifo(r2)
 
 #CHECK: 9056c01c { r29:28{{ *}}={{ *}}memh_fifo(r22{{ *}}+{{ *}}#0)
-r29:28=memh_fifo(r22)
\ No newline at end of file
+r29:28=memh_fifo(r22)