diff --git a/lib/Target/CellSPU/SPUFrameInfo.h b/lib/Target/CellSPU/SPUFrameInfo.h
index 99136c2..e8ca333 100644
--- a/lib/Target/CellSPU/SPUFrameInfo.h
+++ b/lib/Target/CellSPU/SPUFrameInfo.h
@@ -57,6 +57,10 @@
     static int fullSpillSize() {
       return (SPURegisterInfo::getNumArgRegs() * stackSlotSize());
     }
+    //! Convert frame index to stack offset
+    static int FItoStackOffset(int frame_index) {
+      return frame_index * stackSlotSize();
+    }
     //! Number of instructions required to overcome hint-for-branch latency
     /*!
       HBR (hint-for-branch) instructions can be inserted when, for example,
diff --git a/lib/Target/CellSPU/SPUISelDAGToDAG.cpp b/lib/Target/CellSPU/SPUISelDAGToDAG.cpp
index b81f277..49a5e86 100644
--- a/lib/Target/CellSPU/SPUISelDAGToDAG.cpp
+++ b/lib/Target/CellSPU/SPUISelDAGToDAG.cpp
@@ -17,6 +17,7 @@
 #include "SPUISelLowering.h"
 #include "SPUHazardRecognizers.h"
 #include "SPUFrameInfo.h"
+#include "SPURegisterNames.h"
 #include "llvm/CodeGen/MachineConstantPool.h"
 #include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/CodeGen/MachineFunction.h"
@@ -411,7 +412,10 @@
 bool 
 SPUDAGToDAGISel::SelectDForm2Addr(SDOperand Op, SDOperand N, SDOperand &Disp,
                                   SDOperand &Base) {
-  return DFormAddressPredicate(Op, N, Disp, Base, -(1 << 7), (1 << 7) - 1);
+  const int minDForm2Offset = -(1 << 7);
+  const int maxDForm2Offset = (1 << 7) - 1;
+  return DFormAddressPredicate(Op, N, Disp, Base, minDForm2Offset,
+                               maxDForm2Offset);
 }
 
 /*!
@@ -443,12 +447,13 @@
 
   if (Opc == ISD::FrameIndex) {
     // Stack frame index must be less than 512 (divided by 16):
-    FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(N);
+    FrameIndexSDNode *FIN = dyn_cast<FrameIndexSDNode>(N);
+    int FI = int(FIN->getIndex());
     DEBUG(cerr << "SelectDFormAddr: ISD::FrameIndex = "
-          << FI->getIndex() << "\n");
-    if (FI->getIndex() < maxOffset) {
+               << FI << "\n");
+    if (SPUFrameInfo::FItoStackOffset(FI) < maxOffset) {
       Base = CurDAG->getTargetConstant(0, PtrTy);
-      Index = CurDAG->getTargetFrameIndex(FI->getIndex(), PtrTy);
+      Index = CurDAG->getTargetFrameIndex(FI, PtrTy);
       return true;
     }
   } else if (Opc == ISD::ADD) {
@@ -467,13 +472,14 @@
       int32_t offset = int32_t(CN->getSignExtended());
 
       if (Op0.getOpcode() == ISD::FrameIndex) {
-        FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Op0);
+        FrameIndexSDNode *FIN = dyn_cast<FrameIndexSDNode>(Op0);
+        int FI = int(FIN->getIndex());
         DEBUG(cerr << "SelectDFormAddr: ISD::ADD offset = " << offset
-              << " frame index = " << FI->getIndex() << "\n");
+                   << " frame index = " << FI << "\n");
 
-        if (FI->getIndex() < maxOffset) {
+        if (SPUFrameInfo::FItoStackOffset(FI) < maxOffset) {
           Base = CurDAG->getTargetConstant(offset, PtrTy);
-          Index = CurDAG->getTargetFrameIndex(FI->getIndex(), PtrTy);
+          Index = CurDAG->getTargetFrameIndex(FI, PtrTy);
           return true;
         }
       } else if (offset > minOffset && offset < maxOffset) {
@@ -487,13 +493,14 @@
       int32_t offset = int32_t(CN->getSignExtended());
 
       if (Op1.getOpcode() == ISD::FrameIndex) {
-        FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Op1);
+        FrameIndexSDNode *FIN = dyn_cast<FrameIndexSDNode>(Op1);
+        int FI = int(FIN->getIndex());
         DEBUG(cerr << "SelectDFormAddr: ISD::ADD offset = " << offset
-              << " frame index = " << FI->getIndex() << "\n");
+                   << " frame index = " << FI << "\n");
 
-        if (FI->getIndex() < maxOffset) {
+        if (SPUFrameInfo::FItoStackOffset(FI) < maxOffset) {
           Base = CurDAG->getTargetConstant(offset, PtrTy);
-          Index = CurDAG->getTargetFrameIndex(FI->getIndex(), PtrTy);
+          Index = CurDAG->getTargetFrameIndex(FI, PtrTy);
           return true;
         }
       } else if (offset > minOffset && offset < maxOffset) {
@@ -583,17 +590,31 @@
   if (Opc >= ISD::BUILTIN_OP_END && Opc < SPUISD::FIRST_NUMBER) {
     return NULL;   // Already selected.
   } else if (Opc == ISD::FrameIndex) {
-    // Selects to AIr32 FI, 0 which in turn will become AIr32 SP, imm.
-    int FI = cast<FrameIndexSDNode>(N)->getIndex();
+    // Selects to (add $sp, FI * stackSlotSize)
+    int FI =
+      SPUFrameInfo::FItoStackOffset(cast<FrameIndexSDNode>(N)->getIndex());
     MVT::ValueType PtrVT = SPUtli.getPointerTy();
-    SDOperand Zero = CurDAG->getTargetConstant(0, PtrVT);
-    SDOperand TFI = CurDAG->getTargetFrameIndex(FI, PtrVT);
 
-    DEBUG(cerr << "SPUDAGToDAGISel: Replacing FrameIndex with AI32 <FI>, 0\n");
-    NewOpc = SPU::AIr32;
-    Ops[0] = TFI;
-    Ops[1] = Zero;
-    n_ops = 2;
+    // Adjust stack slot to actual offset in frame:
+    if (isS10Constant(FI)) {
+      DEBUG(cerr << "SPUDAGToDAGISel: Replacing FrameIndex with AIr32 $sp, "
+                 << FI
+                 << "\n");
+      NewOpc = SPU::AIr32;
+      Ops[0] = CurDAG->getRegister(SPU::R1, PtrVT);
+      Ops[1] = CurDAG->getTargetConstant(FI, PtrVT);
+      n_ops = 2;
+    } else {
+      DEBUG(cerr << "SPUDAGToDAGISel: Replacing FrameIndex with Ar32 $sp, "
+                 << FI
+                 << "\n");
+      NewOpc = SPU::Ar32;
+      Ops[0] = CurDAG->getRegister(SPU::R1, PtrVT);
+      Ops[1] = CurDAG->getConstant(FI, PtrVT);
+      n_ops = 2;
+
+      AddToISelQueue(Ops[1]);
+    }
   } else if (Opc == ISD::ZERO_EXTEND) {
     // (zero_extend:i16 (and:i8 <arg>, <const>))
     const SDOperand &Op1 = N->getOperand(0);
diff --git a/lib/Target/CellSPU/SPUISelLowering.cpp b/lib/Target/CellSPU/SPUISelLowering.cpp
index ca5638c..ad797dd 100644
--- a/lib/Target/CellSPU/SPUISelLowering.cpp
+++ b/lib/Target/CellSPU/SPUISelLowering.cpp
@@ -14,6 +14,7 @@
 #include "SPURegisterNames.h"
 #include "SPUISelLowering.h"
 #include "SPUTargetMachine.h"
+#include "SPUFrameInfo.h"
 #include "llvm/ADT/VectorExtras.h"
 #include "llvm/Analysis/ScalarEvolutionExpressions.h"
 #include "llvm/CodeGen/CallingConvLower.h"
@@ -514,6 +515,12 @@
       alignOffs = 0;
       prefSlotOffs = -vtm->prefslot_byte;
     }
+  } else if (basePtr.getOpcode() == ISD::FrameIndex) {
+    FrameIndexSDNode *FIN = cast<FrameIndexSDNode>(basePtr);
+    alignOffs = int(FIN->getIndex() * SPUFrameInfo::stackSlotSize());
+    prefSlotOffs = (int) (alignOffs & 0xf);
+    prefSlotOffs -= vtm->prefslot_byte;
+    basePtr = DAG.getRegister(SPU::R1, VT);
   } else {
     alignOffs = 0;
     prefSlotOffs = -vtm->prefslot_byte;
@@ -1912,7 +1919,7 @@
     SDOperand FSMBOp =
       DAG.getCopyToReg(Chain, FSMBIreg,
                        DAG.getNode(SPUISD::FSMBI, MVT::v8i16,
-                                   DAG.getConstant(0xcccc, MVT::i32)));
+                                   DAG.getConstant(0xcccc, MVT::i16)));
 
     SDOperand HHProd =
       DAG.getCopyToReg(FSMBOp, HiProdReg,
@@ -1956,7 +1963,7 @@
                   DAG.getNode(SPUISD::MPY, MVT::v8i16, rALH, rBLH), c8);
 
     SDOperand FSMBmask = DAG.getNode(SPUISD::FSMBI, MVT::v8i16,
-                                     DAG.getConstant(0x2222, MVT::i32));
+                                     DAG.getConstant(0x2222, MVT::i16));
 
     SDOperand LoProdParts =
       DAG.getNode(ISD::BIT_CONVERT, MVT::v4i32,
@@ -2271,6 +2278,7 @@
 
     assert(Op0VT == MVT::i32
            && "CellSPU: Zero/sign extending something other than i32");
+    DEBUG(cerr << "CellSPU: LowerI64Math custom lowering zero/sign/any extend\n");
 
     unsigned NewOpc = (Opc == ISD::SIGN_EXTEND
                       ? SPUISD::ROTBYTES_RIGHT_S
@@ -2764,7 +2772,7 @@
   }
   }
   // Otherwise, return unchanged.
-#if 0
+#if 1
   if (Result.Val) {
     DEBUG(cerr << "\nReplace.SPU: ");
     DEBUG(N->dump(&DAG));
@@ -2833,7 +2841,9 @@
                                                   APInt &KnownOne,
                                                   const SelectionDAG &DAG,
                                                   unsigned Depth ) const {
+#if 0
   const uint64_t uint64_sizebits = sizeof(uint64_t) * 8;
+#endif
 
   switch (Op.getOpcode()) {
   default:
@@ -2849,18 +2859,22 @@
 
   case SPUISD::PROMOTE_SCALAR: {
     SDOperand Op0 = Op.getOperand(0);
-    uint64_t InMask = MVT::getIntVTBitMask(Op0.getValueType());
-    KnownZero |= APInt(uint64_sizebits, ~InMask, false);
-    KnownOne |= APInt(uint64_sizebits, InMask, false);
+    MVT::ValueType Op0VT = Op0.getValueType();
+    unsigned Op0VTBits = MVT::getSizeInBits(Op0VT);
+    uint64_t InMask = MVT::getIntVTBitMask(Op0VT);
+    KnownZero |= APInt(Op0VTBits, ~InMask, false);
+    KnownOne |= APInt(Op0VTBits, InMask, false);
     break;
   }
     
   case SPUISD::LDRESULT:
   case SPUISD::EXTRACT_ELT0:
   case SPUISD::EXTRACT_ELT0_CHAINED: {
-    uint64_t InMask = MVT::getIntVTBitMask(Op.getValueType());
-    KnownZero |= APInt(uint64_sizebits, ~InMask, false);
-    KnownOne |= APInt(uint64_sizebits, InMask, false);
+    MVT::ValueType OpVT = Op.getValueType();
+    unsigned OpVTBits = MVT::getSizeInBits(OpVT);
+    uint64_t InMask = MVT::getIntVTBitMask(OpVT);
+    KnownZero |= APInt(OpVTBits, ~InMask, false);
+    KnownOne |= APInt(OpVTBits, InMask, false);
     break;
   }
 
@@ -2873,21 +2887,20 @@
   case MPYU:
   case MPYH:
   case MPYHH:
-  case SHLQUAD_L_BITS:
-  case SHLQUAD_L_BYTES:
-  case VEC_SHL:
-  case VEC_SRL:
-  case VEC_SRA:
-  case VEC_ROTL:
-  case VEC_ROTR:
-  case ROTQUAD_RZ_BYTES:
-  case ROTQUAD_RZ_BITS:
-  case ROTBYTES_RIGHT_S:
-  case ROTBYTES_LEFT:
-  case ROTBYTES_LEFT_CHAINED:
+  case SPUISD::SHLQUAD_L_BITS:
+  case SPUISD::SHLQUAD_L_BYTES:
+  case SPUISD::VEC_SHL:
+  case SPUISD::VEC_SRL:
+  case SPUISD::VEC_SRA:
+  case SPUISD::VEC_ROTL:
+  case SPUISD::VEC_ROTR:
+  case SPUISD::ROTQUAD_RZ_BYTES:
+  case SPUISD::ROTQUAD_RZ_BITS:
+  case SPUISD::ROTBYTES_RIGHT_S:
+  case SPUISD::ROTBYTES_LEFT:
+  case SPUISD::ROTBYTES_LEFT_CHAINED:
   case FSMBI:
   case SELB:
-  case SFPConstant:
   case FPInterp:
   case FPRecipEst:
   case SEXT32TO64:
@@ -2895,6 +2908,16 @@
   }
 }
 
+// LowerAsmOperandForConstraint
+void
+SPUTargetLowering::LowerAsmOperandForConstraint(SDOperand Op,
+                                                char ConstraintLetter,
+                                                std::vector<SDOperand> &Ops,
+                                                SelectionDAG &DAG) const {
+  // Default, for the time being, to the base class handler
+  TargetLowering::LowerAsmOperandForConstraint(Op, ConstraintLetter, Ops, DAG);
+}
+
 /// isLegalAddressImmediate - Return true if the integer value can be used
 /// as the offset of the target addressing mode.
 bool SPUTargetLowering::isLegalAddressImmediate(int64_t V, const Type *Ty) const {
diff --git a/lib/Target/CellSPU/SPUISelLowering.h b/lib/Target/CellSPU/SPUISelLowering.h
index 4e9f474..3c73aa5 100644
--- a/lib/Target/CellSPU/SPUISelLowering.h
+++ b/lib/Target/CellSPU/SPUISelLowering.h
@@ -125,6 +125,10 @@
       getRegForInlineAsmConstraint(const std::string &Constraint,
                                    MVT::ValueType VT) const;
 
+    void LowerAsmOperandForConstraint(SDOperand Op, char ConstraintLetter,
+                                      std::vector<SDOperand> &Ops,
+                                      SelectionDAG &DAG) const;
+
     /// isLegalAddressImmediate - Return true if the integer value can be used
     /// as the offset of the target addressing mode.
     virtual bool isLegalAddressImmediate(int64_t V, const Type *Ty) const;
diff --git a/lib/Target/CellSPU/SPUInstrInfo.td b/lib/Target/CellSPU/SPUInstrInfo.td
index 03efdc8..8e1933c 100644
--- a/lib/Target/CellSPU/SPUInstrInfo.td
+++ b/lib/Target/CellSPU/SPUInstrInfo.td
@@ -22,10 +22,10 @@
 //===----------------------------------------------------------------------===//
 
 let hasCtrlDep = 1, Defs = [R1], Uses = [R1] in {
-  def ADJCALLSTACKDOWN : Pseudo<(outs), (ins u16imm:$amt),
+  def ADJCALLSTACKDOWN : Pseudo<(outs), (ins u16imm_i32:$amt),
                                 "${:comment} ADJCALLSTACKDOWN",
                                 [(callseq_start imm:$amt)]>;
-  def ADJCALLSTACKUP   : Pseudo<(outs), (ins u16imm:$amt),
+  def ADJCALLSTACKUP   : Pseudo<(outs), (ins u16imm_i32:$amt),
                                 "${:comment} ADJCALLSTACKUP",
                                 [(callseq_end imm:$amt)]>;
 }
@@ -290,18 +290,34 @@
     "cwd\t$rT, $src", ShuffleOp,
     [(set (v4i32 VECREG:$rT), (SPUvecinsmask dform2_addr:$src))]>;
 
+def CWDf32 : RI7Form<0b01101111100, (outs VECREG:$rT), (ins memri7:$src),
+    "cwd\t$rT, $src", ShuffleOp,
+    [(set (v4f32 VECREG:$rT), (SPUvecinsmask dform2_addr:$src))]>;
+
 def CWX : RRForm<0b01101011100, (outs VECREG:$rT), (ins memrr:$src),
     "cwx\t$rT, $src", ShuffleOp,
     [(set (v4i32 VECREG:$rT), (SPUvecinsmask xform_addr:$src))]>;
 
+def CWXf32 : RRForm<0b01101011100, (outs VECREG:$rT), (ins memrr:$src),
+    "cwx\t$rT, $src", ShuffleOp,
+    [(set (v4f32 VECREG:$rT), (SPUvecinsmask xform_addr:$src))]>;
+
 def CDD : RI7Form<0b11101111100, (outs VECREG:$rT), (ins memri7:$src),
     "cdd\t$rT, $src", ShuffleOp,
     [(set (v2i64 VECREG:$rT), (SPUvecinsmask dform2_addr:$src))]>;
 
+def CDDf64 : RI7Form<0b11101111100, (outs VECREG:$rT), (ins memri7:$src),
+    "cdd\t$rT, $src", ShuffleOp,
+    [(set (v2f64 VECREG:$rT), (SPUvecinsmask dform2_addr:$src))]>;
+
 def CDX : RRForm<0b11101011100, (outs VECREG:$rT), (ins memrr:$src),
     "cdx\t$rT, $src", ShuffleOp,
     [(set (v2i64 VECREG:$rT), (SPUvecinsmask xform_addr:$src))]>;
 
+def CDXf64 : RRForm<0b11101011100, (outs VECREG:$rT), (ins memrr:$src),
+    "cdx\t$rT, $src", ShuffleOp,
+    [(set (v2f64 VECREG:$rT), (SPUvecinsmask xform_addr:$src))]>;
+
 //===----------------------------------------------------------------------===//
 // Constant formation:
 //===----------------------------------------------------------------------===//
@@ -367,10 +383,10 @@
 multiclass ImmLoadHalfwordUpper
 {
   def v2i64: ILHUVecInst<v2i64, u16imm_i64, immILHUvec_i64>;
-  def v4i32: ILHUVecInst<v4i32, u16imm, immILHUvec>;
+  def v4i32: ILHUVecInst<v4i32, u16imm_i32, immILHUvec>;
 
   def r64: ILHURegInst<R64C, u16imm_i64, hi16>;
-  def r32: ILHURegInst<R32C, u16imm, hi16>;
+  def r32: ILHURegInst<R32C, u16imm_i32, hi16>;
 
   // Loads the high portion of an address
   def hi: ILHURegInst<R32C, symbolHi, hi16>;
@@ -436,7 +452,7 @@
 multiclass ImmOrHalfwordLower
 {
   def v2i64: IOHLVecInst<v2i64, u16imm_i64>;
-  def v4i32: IOHLVecInst<v4i32, u16imm>;
+  def v4i32: IOHLVecInst<v4i32, u16imm_i32>;
 
   def r32: IOHLRegInst<R32C, i32imm>;
   def f32: IOHLRegInst<R32FP, f32imm>;
@@ -453,7 +469,7 @@
   RI16Form<0b101001100, (outs VECREG:$rT), (ins u16imm:$val),
           "fsmbi\t$rT, $val",
           SelectOp,
-          [(set (vectype VECREG:$rT), (SPUfsmbi (i32 immU16:$val)))]>;
+          [(set (vectype VECREG:$rT), (SPUfsmbi (i16 immU16:$val)))]>;
 
 multiclass FormSelectMaskBytesImm
 {
@@ -3873,6 +3889,13 @@
                        (SPUlo tconstpool:$in, 0)),
           (IOHLlo (ILHUhi tconstpool:$in), tconstpool:$in)>;
 
+def : Pat<(SPUindirect R32C:$sp, i32ImmSExt10:$imm),
+          (AIr32 R32C:$sp, i32ImmSExt10:$imm)>;
+
+def : Pat<(SPUindirect R32C:$sp, imm:$imm),
+          (Ar32 R32C:$sp,
+                (IOHLr32 (ILHUr32 (HI16 imm:$imm)), (LO16 imm:$imm)))>;
+
 def : Pat<(add (SPUhi tglobaladdr:$in, 0), (SPUlo tglobaladdr:$in, 0)),
           (IOHLlo (ILHUhi tglobaladdr:$in), tglobaladdr:$in)>;
 
diff --git a/lib/Target/CellSPU/SPUNodes.td b/lib/Target/CellSPU/SPUNodes.td
index e390345..00d0f94 100644
--- a/lib/Target/CellSPU/SPUNodes.td
+++ b/lib/Target/CellSPU/SPUNodes.td
@@ -59,7 +59,8 @@
 // FSMBI type constraints: There are several variations for the various
 // vector types (this avoids having to bit_convert all over the place.)
 def SPUfsmbi_type: SDTypeProfile<1, 1, [
-  /* SDTCisVT<1, i32> */ SDTCisInt<1>]>;
+  SDTCisInt<1>
+]>;
 
 // SELB type constraints:
 def SPUselb_type: SDTypeProfile<1, 3, [
diff --git a/lib/Target/CellSPU/SPUOperands.td b/lib/Target/CellSPU/SPUOperands.td
index 1e8eff5..aae79b5 100644
--- a/lib/Target/CellSPU/SPUOperands.td
+++ b/lib/Target/CellSPU/SPUOperands.td
@@ -509,7 +509,11 @@
   let PrintMethod = "printU16ImmOperand";
 }
 
-def u16imm : Operand<i32> {
+def u16imm_i32 : Operand<i32> {
+  let PrintMethod = "printU16ImmOperand";
+}
+
+def u16imm : Operand<i16> {
   let PrintMethod = "printU16ImmOperand";
 }