Thumb2 TBB and TBH instructions are only allowed at the end of IT blocks, not in...
[oota-llvm.git] / lib / VMCore / AutoUpgrade.cpp
index 64b0b518284e03e6e2a3fb5caa61641230bc690c..04221d461d7e9d22aee6eefb045f0d351758cf21 100644 (file)
 #include "llvm/AutoUpgrade.h"
 #include "llvm/Constants.h"
 #include "llvm/Function.h"
+#include "llvm/Instruction.h"
 #include "llvm/LLVMContext.h"
 #include "llvm/Module.h"
 #include "llvm/IntrinsicInst.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/Support/CallSite.h"
+#include "llvm/Support/CFG.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/IRBuilder.h"
 #include <cstring>
@@ -28,520 +32,97 @@ using namespace llvm;
 static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
   assert(F && "Illegal to upgrade a non-existent Function.");
 
-  // Get the Function's name.
-  const std::string& Name = F->getName();
-
-  // Convenience
-  const FunctionType *FTy = F->getFunctionType();
-
   // Quickly eliminate it, if it's not a candidate.
-  if (Name.length() <= 8 || Name[0] != 'l' || Name[1] != 'l' || 
-      Name[2] != 'v' || Name[3] != 'm' || Name[4] != '.')
+  StringRef Name = F->getName();
+  if (Name.size() <= 8 || !Name.startswith("llvm."))
     return false;
+  Name = Name.substr(5); // Strip off "llvm."
 
+  FunctionType *FTy = F->getFunctionType();
   Module *M = F->getParent();
-  switch (Name[5]) {
+  
+  switch (Name[0]) {
   default: break;
-  case 'a':
-    // This upgrades the llvm.atomic.lcs, llvm.atomic.las, llvm.atomic.lss,
-    // and atomics with default address spaces to their new names to their new
-    // function name (e.g. llvm.atomic.add.i32 => llvm.atomic.add.i32.p0i32)
-    if (Name.compare(5,7,"atomic.",7) == 0) {
-      if (Name.compare(12,3,"lcs",3) == 0) {
-        std::string::size_type delim = Name.find('.',12);
-        F->setName("llvm.atomic.cmp.swap" + Name.substr(delim) +
-                   ".p0" + Name.substr(delim+1));
-        NewFn = F;
-        return true;
-      }
-      else if (Name.compare(12,3,"las",3) == 0) {
-        std::string::size_type delim = Name.find('.',12);
-        F->setName("llvm.atomic.load.add"+Name.substr(delim)
-                   + ".p0" + Name.substr(delim+1));
-        NewFn = F;
-        return true;
-      }
-      else if (Name.compare(12,3,"lss",3) == 0) {
-        std::string::size_type delim = Name.find('.',12);
-        F->setName("llvm.atomic.load.sub"+Name.substr(delim)
-                   + ".p0" + Name.substr(delim+1));
-        NewFn = F;
-        return true;
-      }
-      else if (Name.rfind(".p") == std::string::npos) {
-        // We don't have an address space qualifier so this has be upgraded
-        // to the new name.  Copy the type name at the end of the intrinsic
-        // and add to it
-        std::string::size_type delim = Name.find_last_of('.');
-        assert(delim != std::string::npos && "can not find type");
-        F->setName(Name + ".p0" + Name.substr(delim+1));
-        NewFn = F;
-        return true;
-      }
-    } else if (Name.compare(5, 9, "arm.neon.", 9) == 0) {
-      if (((Name.compare(14, 5, "vmovl", 5) == 0 ||
-            Name.compare(14, 5, "vaddl", 5) == 0 ||
-            Name.compare(14, 5, "vsubl", 5) == 0 ||
-            Name.compare(14, 5, "vaddw", 5) == 0 ||
-            Name.compare(14, 5, "vsubw", 5) == 0 ||
-            Name.compare(14, 5, "vmull", 5) == 0 ||
-            Name.compare(14, 5, "vmlal", 5) == 0 ||
-            Name.compare(14, 5, "vmlsl", 5) == 0 ||
-            Name.compare(14, 5, "vabdl", 5) == 0 ||
-            Name.compare(14, 5, "vabal", 5) == 0) &&
-           (Name.compare(19, 2, "s.", 2) == 0 ||
-            Name.compare(19, 2, "u.", 2) == 0)) ||
-
-          (Name.compare(14, 4, "vaba", 4) == 0 &&
-           (Name.compare(18, 2, "s.", 2) == 0 ||
-            Name.compare(18, 2, "u.", 2) == 0)) ||
-
-          (Name.compare(14, 6, "vmovn.", 6) == 0)) {
-
-        // Calls to these are transformed into IR without intrinsics.
-        NewFn = 0;
-        return true;
-      }
-      // Old versions of NEON ld/st intrinsics are missing alignment arguments.
-      bool isVLd = (Name.compare(14, 3, "vld", 3) == 0);
-      bool isVSt = (Name.compare(14, 3, "vst", 3) == 0);
-      if (isVLd || isVSt) {
-        unsigned NumVecs = Name.at(17) - '0';
-        if (NumVecs == 0 || NumVecs > 4)
-          return false;
-        bool isLaneOp = (Name.compare(18, 5, "lane.", 5) == 0);
-        if (!isLaneOp && Name.at(18) != '.')
-          return false;
-        unsigned ExpectedArgs = 2; // for the address and alignment
-        if (isVSt || isLaneOp)
-          ExpectedArgs += NumVecs;
-        if (isLaneOp)
-          ExpectedArgs += 1; // for the lane number
-        unsigned NumP = FTy->getNumParams();
-        if (NumP != ExpectedArgs - 1)
-          return false;
-
-        // Change the name of the old (bad) intrinsic, because 
-        // its type is incorrect, but we cannot overload that name.
-        F->setName("");
-
-        // One argument is missing: add the alignment argument.
-        std::vector<const Type*> NewParams;
-        for (unsigned p = 0; p < NumP; ++p)
-          NewParams.push_back(FTy->getParamType(p));
-        NewParams.push_back(Type::getInt32Ty(F->getContext()));
-        FunctionType *NewFTy = FunctionType::get(FTy->getReturnType(),
-                                                 NewParams, false);
-        NewFn = cast<Function>(M->getOrInsertFunction(Name, NewFTy));
-        return true;
-      }
-    }
-    break;
-  case 'b':
-    //  This upgrades the name of the llvm.bswap intrinsic function to only use 
-    //  a single type name for overloading. We only care about the old format
-    //  'llvm.bswap.i*.i*', so check for 'bswap.' and then for there being 
-    //  a '.' after 'bswap.'
-    if (Name.compare(5,6,"bswap.",6) == 0) {
-      std::string::size_type delim = Name.find('.',11);
-      
-      if (delim != std::string::npos) {
-        //  Construct the new name as 'llvm.bswap' + '.i*'
-        F->setName(Name.substr(0,10)+Name.substr(delim));
-        NewFn = F;
-        return true;
-      }
+  case 'i':
+    //  This upgrades the old llvm.init.trampoline to the new
+    //  llvm.init.trampoline and llvm.adjust.trampoline pair.
+    if (Name == "init.trampoline") {
+      // The new llvm.init.trampoline returns nothing.
+      if (FTy->getReturnType()->isVoidTy())
+        break;
+
+      assert(FTy->getNumParams() == 3 && "old init.trampoline takes 3 args!");
+
+      // Change the name of the old intrinsic so that we can play with its type.
+      std::string NameTmp = F->getName();
+      F->setName("");
+      NewFn = cast<Function>(M->getOrInsertFunction(
+                               NameTmp,
+                               Type::getVoidTy(M->getContext()),
+                               FTy->getParamType(0), FTy->getParamType(1),
+                               FTy->getParamType(2), (Type *)0));
+      return true;
     }
-    break;
+  case 'p':
+    //  This upgrades the llvm.prefetch intrinsic to accept one more parameter,
+    //  which is a instruction / data cache identifier. The old version only
+    //  implicitly accepted the data version.
+    if (Name == "prefetch") {
+      // Don't do anything if it has the correct number of arguments already
+      if (FTy->getNumParams() == 4)
+        break;
 
-  case 'c':
-    //  We only want to fix the 'llvm.ct*' intrinsics which do not have the 
-    //  correct return type, so we check for the name, and then check if the 
-    //  return type does not match the parameter type.
-    if ( (Name.compare(5,5,"ctpop",5) == 0 ||
-          Name.compare(5,4,"ctlz",4) == 0 ||
-          Name.compare(5,4,"cttz",4) == 0) &&
-        FTy->getReturnType() != FTy->getParamType(0)) {
-      //  We first need to change the name of the old (bad) intrinsic, because 
-      //  its type is incorrect, but we cannot overload that name. We 
-      //  arbitrarily unique it here allowing us to construct a correctly named 
+      assert(FTy->getNumParams() == 3 && "old prefetch takes 3 args!");
+      //  We first need to change the name of the old (bad) intrinsic, because
+      //  its type is incorrect, but we cannot overload that name. We
+      //  arbitrarily unique it here allowing us to construct a correctly named
       //  and typed function below.
+      std::string NameTmp = F->getName();
       F->setName("");
-
-      //  Now construct the new intrinsic with the correct name and type. We 
-      //  leave the old function around in order to query its type, whatever it 
-      //  may be, and correctly convert up to the new type.
-      NewFn = cast<Function>(M->getOrInsertFunction(Name, 
+      NewFn = cast<Function>(M->getOrInsertFunction(NameTmp,
+                                                    FTy->getReturnType(),
                                                     FTy->getParamType(0),
-                                                    FTy->getParamType(0),
-                                                    (Type *)0));
+                                                    FTy->getParamType(1),
+                                                    FTy->getParamType(2),
+                                                    FTy->getParamType(2),
+                                                    (Type*)0));
       return true;
     }
-    break;
 
-  case 'e':
-    //  The old llvm.eh.selector.i32 is equivalent to the new llvm.eh.selector.
-    if (Name.compare("llvm.eh.selector.i32") == 0) {
-      F->setName("llvm.eh.selector");
-      NewFn = F;
-      return true;
-    }
-    //  The old llvm.eh.typeid.for.i32 is equivalent to llvm.eh.typeid.for.
-    if (Name.compare("llvm.eh.typeid.for.i32") == 0) {
-      F->setName("llvm.eh.typeid.for");
-      NewFn = F;
-      return true;
-    }
-    //  Convert the old llvm.eh.selector.i64 to a call to llvm.eh.selector.
-    if (Name.compare("llvm.eh.selector.i64") == 0) {
-      NewFn = Intrinsic::getDeclaration(M, Intrinsic::eh_selector);
-      return true;
-    }
-    //  Convert the old llvm.eh.typeid.for.i64 to a call to llvm.eh.typeid.for.
-    if (Name.compare("llvm.eh.typeid.for.i64") == 0) {
-      NewFn = Intrinsic::getDeclaration(M, Intrinsic::eh_typeid_for);
-      return true;
-    }
     break;
-
-  case 'm': {
-    // This upgrades the llvm.memcpy, llvm.memmove, and llvm.memset to the
-    // new format that allows overloading the pointer for different address
-    // space (e.g., llvm.memcpy.i16 => llvm.memcpy.p0i8.p0i8.i16)
-    const char* NewFnName = NULL;
-    if (Name.compare(5,8,"memcpy.i",8) == 0) {
-      if (Name[13] == '8')
-        NewFnName = "llvm.memcpy.p0i8.p0i8.i8";
-      else if (Name.compare(13,2,"16") == 0)
-        NewFnName = "llvm.memcpy.p0i8.p0i8.i16";
-      else if (Name.compare(13,2,"32") == 0)
-        NewFnName = "llvm.memcpy.p0i8.p0i8.i32";
-      else if (Name.compare(13,2,"64") == 0)
-        NewFnName = "llvm.memcpy.p0i8.p0i8.i64";
-    } else if (Name.compare(5,9,"memmove.i",9) == 0) {
-      if (Name[14] == '8')
-        NewFnName = "llvm.memmove.p0i8.p0i8.i8";
-      else if (Name.compare(14,2,"16") == 0)
-        NewFnName = "llvm.memmove.p0i8.p0i8.i16";
-      else if (Name.compare(14,2,"32") == 0)
-        NewFnName = "llvm.memmove.p0i8.p0i8.i32";
-      else if (Name.compare(14,2,"64") == 0)
-        NewFnName = "llvm.memmove.p0i8.p0i8.i64";
-    }
-    else if (Name.compare(5,8,"memset.i",8) == 0) {
-      if (Name[13] == '8')
-        NewFnName = "llvm.memset.p0i8.i8";
-      else if (Name.compare(13,2,"16") == 0)
-        NewFnName = "llvm.memset.p0i8.i16";
-      else if (Name.compare(13,2,"32") == 0)
-        NewFnName = "llvm.memset.p0i8.i32";
-      else if (Name.compare(13,2,"64") == 0)
-        NewFnName = "llvm.memset.p0i8.i64";
-    }
+  case 'x': {
+    const char *NewFnName = NULL;
+    // This fixes the poorly named crc32 intrinsics.
+    if (Name == "x86.sse42.crc32.8")
+      NewFnName = "llvm.x86.sse42.crc32.32.8";
+    else if (Name == "x86.sse42.crc32.16")
+      NewFnName = "llvm.x86.sse42.crc32.32.16";
+    else if (Name == "x86.sse42.crc32.32")
+      NewFnName = "llvm.x86.sse42.crc32.32.32";
+    else if (Name == "x86.sse42.crc64.8")
+      NewFnName = "llvm.x86.sse42.crc32.64.8";
+    else if (Name == "x86.sse42.crc64.64")
+      NewFnName = "llvm.x86.sse42.crc32.64.64";
+    
     if (NewFnName) {
-      NewFn = cast<Function>(M->getOrInsertFunction(NewFnName, 
-                                            FTy->getReturnType(),
-                                            FTy->getParamType(0),
-                                            FTy->getParamType(1),
-                                            FTy->getParamType(2),
-                                            FTy->getParamType(3),
-                                            Type::getInt1Ty(F->getContext()),
-                                            (Type *)0));
+      F->setName(NewFnName);
+      NewFn = F;
       return true;
     }
-    break;
-  }
-  case 'p':
-    //  This upgrades the llvm.part.select overloaded intrinsic names to only 
-    //  use one type specifier in the name. We only care about the old format
-    //  'llvm.part.select.i*.i*', and solve as above with bswap.
-    if (Name.compare(5,12,"part.select.",12) == 0) {
-      std::string::size_type delim = Name.find('.',17);
-      
-      if (delim != std::string::npos) {
-        //  Construct a new name as 'llvm.part.select' + '.i*'
-        F->setName(Name.substr(0,16)+Name.substr(delim));
-        NewFn = F;
-        return true;
-      }
-      break;
-    }
-
-    //  This upgrades the llvm.part.set intrinsics similarly as above, however 
-    //  we care about 'llvm.part.set.i*.i*.i*', but only the first two types 
-    //  must match. There is an additional type specifier after these two 
-    //  matching types that we must retain when upgrading.  Thus, we require 
-    //  finding 2 periods, not just one, after the intrinsic name.
-    if (Name.compare(5,9,"part.set.",9) == 0) {
-      std::string::size_type delim = Name.find('.',14);
-
-      if (delim != std::string::npos &&
-          Name.find('.',delim+1) != std::string::npos) {
-        //  Construct a new name as 'llvm.part.select' + '.i*.i*'
-        F->setName(Name.substr(0,13)+Name.substr(delim));
-        NewFn = F;
-        return true;
-      }
-      break;
-    }
-
-    break;
-  case 'x': 
-    // This fixes all MMX shift intrinsic instructions to take a
-    // x86_mmx instead of a v1i64, v2i32, v4i16, or v8i8.
-    if (Name.compare(5, 8, "x86.mmx.", 8) == 0) {
-      const Type *X86_MMXTy = VectorType::getX86_MMXTy(FTy->getContext());
-
-      if (Name.compare(13, 4, "padd", 4) == 0   ||
-          Name.compare(13, 4, "psub", 4) == 0   ||
-          Name.compare(13, 4, "pmul", 4) == 0   ||
-          Name.compare(13, 5, "pmadd", 5) == 0  ||
-          Name.compare(13, 4, "pand", 4) == 0   ||
-          Name.compare(13, 3, "por", 3) == 0    ||
-          Name.compare(13, 4, "pxor", 4) == 0   ||
-          Name.compare(13, 4, "pavg", 4) == 0   ||
-          Name.compare(13, 4, "pmax", 4) == 0   ||
-          Name.compare(13, 4, "pmin", 4) == 0   ||
-          Name.compare(13, 4, "psad", 4) == 0   ||
-          Name.compare(13, 4, "psll", 4) == 0   ||
-          Name.compare(13, 4, "psrl", 4) == 0   ||
-          Name.compare(13, 4, "psra", 4) == 0   ||
-          Name.compare(13, 4, "pack", 4) == 0   ||
-          Name.compare(13, 6, "punpck", 6) == 0 ||
-          Name.compare(13, 4, "pcmp", 4) == 0) {
-        assert(FTy->getNumParams() == 2 && "MMX intrinsic takes 2 args!");
-        const Type *SecondParamTy = X86_MMXTy;
-
-        if (Name.compare(13, 5, "pslli", 5) == 0 ||
-            Name.compare(13, 5, "psrli", 5) == 0 ||
-            Name.compare(13, 5, "psrai", 5) == 0)
-          SecondParamTy = FTy->getParamType(1);
-
-        // Don't do anything if it has the correct types.
-        if (FTy->getReturnType() == X86_MMXTy &&
-            FTy->getParamType(0) == X86_MMXTy &&
-            FTy->getParamType(1) == SecondParamTy)
-          break;
-
-        // We first need to change the name of the old (bad) intrinsic, because
-        // its type is incorrect, but we cannot overload that name. We
-        // arbitrarily unique it here allowing us to construct a correctly named
-        // and typed function below.
-        F->setName("");
-
-        // Now construct the new intrinsic with the correct name and type. We
-        // leave the old function around in order to query its type, whatever it
-        // may be, and correctly convert up to the new type.
-        NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                      X86_MMXTy, X86_MMXTy,
-                                                      SecondParamTy, (Type*)0));
-        return true;
-      }
-
-      if (Name.compare(13, 8, "maskmovq", 8) == 0) {
-        // Don't do anything if it has the correct types.
-        if (FTy->getParamType(0) == X86_MMXTy &&
-            FTy->getParamType(1) == X86_MMXTy)
-          break;
-
-        F->setName("");
-        NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                      FTy->getReturnType(),
-                                                      X86_MMXTy,
-                                                      X86_MMXTy,
-                                                      FTy->getParamType(2),
-                                                      (Type*)0));
-        return true;
-      }
-
-      if (Name.compare(13, 8, "pmovmskb", 8) == 0) {
-        if (FTy->getParamType(0) == X86_MMXTy)
-          break;
-
-        F->setName("");
-        NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                      FTy->getReturnType(),
-                                                      X86_MMXTy,
-                                                      (Type*)0));
-        return true;
-      }
-
-      if (Name.compare(13, 5, "movnt", 5) == 0) {
-        if (FTy->getParamType(1) == X86_MMXTy)
-          break;
-
-        F->setName("");
-        NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                      FTy->getReturnType(),
-                                                      FTy->getParamType(0),
-                                                      X86_MMXTy,
-                                                      (Type*)0));
-        return true;
-      }
-
-      if (Name.compare(13, 7, "palignr", 7) == 0) {
-        if (FTy->getReturnType() == X86_MMXTy &&
-            FTy->getParamType(0) == X86_MMXTy &&
-            FTy->getParamType(1) == X86_MMXTy)
-          break;
-
-        F->setName("");
-        NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                      X86_MMXTy,
-                                                      X86_MMXTy,
-                                                      X86_MMXTy,
-                                                      FTy->getParamType(2),
-                                                      (Type*)0));
-        return true;
-      }
-
-      if (Name.compare(13, 5, "pextr", 5) == 0) {
-        if (FTy->getParamType(0) == X86_MMXTy)
-          break;
-
-        F->setName("");
-        NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                      FTy->getReturnType(),
-                                                      X86_MMXTy,
-                                                      FTy->getParamType(1),
-                                                      (Type*)0));
-        return true;
-      }
-
-      if (Name.compare(13, 5, "pinsr", 5) == 0) {
-        if (FTy->getReturnType() == X86_MMXTy &&
-            FTy->getParamType(0) == X86_MMXTy)
-          break;
-
-        F->setName("");
-        NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                      X86_MMXTy,
-                                                      X86_MMXTy,
-                                                      FTy->getParamType(1),
-                                                      FTy->getParamType(2),
-                                                      (Type*)0));
-        return true;
-      }
-
-      if (Name.compare(13, 12, "cvtsi32.si64", 12) == 0) {
-        if (FTy->getReturnType() == X86_MMXTy)
-          break;
-
-        F->setName("");
-        NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                      X86_MMXTy,
-                                                      FTy->getParamType(0),
-                                                      (Type*)0));
-        return true;
-      }
 
-      if (Name.compare(13, 12, "cvtsi64.si32", 12) == 0) {
-        if (FTy->getParamType(0) == X86_MMXTy)
-          break;
-
-        F->setName("");
-        NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                      FTy->getReturnType(),
-                                                      X86_MMXTy,
-                                                      (Type*)0));
-        return true;
-      }
-
-      if (Name.compare(13, 8, "vec.init", 8) == 0) {
-        if (FTy->getReturnType() == X86_MMXTy)
-          break;
-
-        F->setName("");
-
-        if (Name.compare(21, 2, ".b", 2) == 0)
-          NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                        X86_MMXTy,
-                                                        FTy->getParamType(0),
-                                                        FTy->getParamType(1),
-                                                        FTy->getParamType(2),
-                                                        FTy->getParamType(3),
-                                                        FTy->getParamType(4),
-                                                        FTy->getParamType(5),
-                                                        FTy->getParamType(6),
-                                                        FTy->getParamType(7),
-                                                        (Type*)0));
-        else if (Name.compare(21, 2, ".w", 2) == 0)
-          NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                        X86_MMXTy,
-                                                        FTy->getParamType(0),
-                                                        FTy->getParamType(1),
-                                                        FTy->getParamType(2),
-                                                        FTy->getParamType(3),
-                                                        (Type*)0));
-        else if (Name.compare(21, 2, ".d", 2) == 0)
-          NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                        X86_MMXTy,
-                                                        FTy->getParamType(0),
-                                                        FTy->getParamType(1),
-                                                        (Type*)0));
-        return true;
-      }
-
-
-      if (Name.compare(13, 9, "vec.ext.d", 9) == 0) {
-        if (FTy->getReturnType() == X86_MMXTy &&
-            FTy->getParamType(0) == X86_MMXTy)
-          break;
-
-        F->setName("");
-        NewFn = cast<Function>(M->getOrInsertFunction(Name, 
-                                                      X86_MMXTy,
-                                                      X86_MMXTy,
-                                                      FTy->getParamType(1),
-                                                      (Type*)0));
-        return true;
-      }
-
-      if (Name.compare(13, 9, "emms", 4) == 0 ||
-          Name.compare(13, 9, "femms", 5) == 0) {
-        NewFn = 0;
-        break;
-      }
-
-      // We really shouldn't get here ever.
-      assert(0 && "Invalid MMX intrinsic!");
-      break;
-    } else if (Name.compare(5,17,"x86.sse2.loadh.pd",17) == 0 ||
-               Name.compare(5,17,"x86.sse2.loadl.pd",17) == 0 ||
-               Name.compare(5,16,"x86.sse2.movl.dq",16) == 0 ||
-               Name.compare(5,15,"x86.sse2.movs.d",15) == 0 ||
-               Name.compare(5,16,"x86.sse2.shuf.pd",16) == 0 ||
-               Name.compare(5,18,"x86.sse2.unpckh.pd",18) == 0 ||
-               Name.compare(5,18,"x86.sse2.unpckl.pd",18) == 0 ||
-               Name.compare(5,20,"x86.sse2.punpckh.qdq",20) == 0 ||
-               Name.compare(5,20,"x86.sse2.punpckl.qdq",20) == 0) {
-      // Calls to these intrinsics are transformed into ShuffleVector's.
-      NewFn = 0;
-      return true;
-    } else if (Name.compare(5, 16, "x86.sse41.pmulld", 16) == 0) {
-      // Calls to these intrinsics are transformed into vector multiplies.
-      NewFn = 0;
+    // Calls to these instructions are transformed into unaligned loads.
+    if (Name == "x86.sse.loadu.ps" || Name == "x86.sse2.loadu.dq" ||
+        Name == "x86.sse2.loadu.pd")
       return true;
-    } else if (Name.compare(5, 18, "x86.ssse3.palign.r", 18) == 0 ||
-               Name.compare(5, 22, "x86.ssse3.palign.r.128", 22) == 0) {
-      // Calls to these intrinsics are transformed into vector shuffles, shifts,
-      // or 0.
-      NewFn = 0;
-      return true;           
-    } else if (Name.compare(5, 17, "x86.ssse3.pshuf.w", 17) == 0) {
-      // This is an SSE/MMX instruction.
-      const Type *X86_MMXTy = VectorType::getX86_MMXTy(FTy->getContext());
-      NewFn =
-        cast<Function>(M->getOrInsertFunction("llvm.x86.sse.pshuf.w",
-                                              X86_MMXTy,
-                                              X86_MMXTy,
-                                              Type::getInt8Ty(F->getContext()),
-                                              (Type*)0));
+      
+    // Calls to these instructions are transformed into nontemporal stores.
+    if (Name == "x86.sse.movnt.ps"  || Name == "x86.sse2.movnt.dq" ||
+        Name == "x86.sse2.movnt.pd" || Name == "x86.sse2.movnt.i")
       return true;
-    }
 
     break;
   }
+  }
 
   //  This may not belong here. This function is effectively being overloaded 
   //  to both detect an intrinsic which needs upgrading, and to provide the 
@@ -563,105 +144,10 @@ bool llvm::UpgradeIntrinsicFunction(Function *F, Function *&NewFn) {
 }
 
 bool llvm::UpgradeGlobalVariable(GlobalVariable *GV) {
-  StringRef Name(GV->getName());
-
-  // We are only upgrading one symbol here.
-  if (Name == ".llvm.eh.catch.all.value") {
-    GV->setName("llvm.eh.catch.all.value");
-    return true;
-  }
-
+  // Nothing to do yet.
   return false;
 }
 
-/// ExtendNEONArgs - For NEON "long" and "wide" operations, where the results
-/// have vector elements twice as big as one or both source operands, do the
-/// sign- or zero-extension that used to be handled by intrinsics.  The
-/// extended values are returned via V0 and V1.
-static void ExtendNEONArgs(CallInst *CI, Value *Arg0, Value *Arg1,
-                           Value *&V0, Value *&V1) {
-  Function *F = CI->getCalledFunction();
-  const std::string& Name = F->getName();
-  bool isLong = (Name.at(18) == 'l');
-  bool isSigned = (Name.at(19) == 's');
-
-  if (isSigned) {
-    if (isLong)
-      V0 = new SExtInst(Arg0, CI->getType(), "", CI);
-    else
-      V0 = Arg0;
-    V1 = new SExtInst(Arg1, CI->getType(), "", CI);
-  } else {
-    if (isLong)
-      V0 = new ZExtInst(Arg0, CI->getType(), "", CI);
-    else
-      V0 = Arg0;
-    V1 = new ZExtInst(Arg1, CI->getType(), "", CI);
-  }
-}
-
-/// CallVABD - As part of expanding a call to one of the old NEON vabdl, vaba,
-/// or vabal intrinsics, construct a call to a vabd intrinsic.  Examine the
-/// name of the old intrinsic to determine whether to use a signed or unsigned
-/// vabd intrinsic.  Get the type from the old call instruction, adjusted for
-/// half-size vector elements if the old intrinsic was vabdl or vabal.
-static Instruction *CallVABD(CallInst *CI, Value *Arg0, Value *Arg1) {
-  Function *F = CI->getCalledFunction();
-  const std::string& Name = F->getName();
-  bool isLong = (Name.at(18) == 'l');
-  bool isSigned = (Name.at(isLong ? 19 : 18) == 's');
-
-  Intrinsic::ID intID;
-  if (isSigned)
-    intID = Intrinsic::arm_neon_vabds;
-  else
-    intID = Intrinsic::arm_neon_vabdu;
-
-  const Type *Ty = CI->getType();
-  if (isLong)
-    Ty = VectorType::getTruncatedElementVectorType(cast<const VectorType>(Ty));
-
-  Function *VABD = Intrinsic::getDeclaration(F->getParent(), intID, &Ty, 1);
-  Value *Operands[2];
-  Operands[0] = Arg0;
-  Operands[1] = Arg1;
-  return CallInst::Create(VABD, Operands, Operands+2, 
-                          "upgraded."+CI->getName(), CI);
-}
-
-/// ConstructNewCallInst - Construct a new CallInst with the signature of NewFn.
-static void ConstructNewCallInst(Function *NewFn, CallInst *OldCI,
-                                 Value **Operands, unsigned NumOps,
-                                 bool AssignName = true) {
-  // Construct a new CallInst.
-  CallInst *NewCI =
-    CallInst::Create(NewFn, Operands, Operands + NumOps,
-                     AssignName ? "upgraded." + OldCI->getName() : "", OldCI);
-
-  NewCI->setTailCall(OldCI->isTailCall());
-  NewCI->setCallingConv(OldCI->getCallingConv());
-
-  // Handle any uses of the old CallInst. If the type has changed, add a cast.
-  if (!OldCI->use_empty()) {
-    if (OldCI->getType() != NewCI->getType()) {
-      Function *OldFn = OldCI->getCalledFunction();
-      CastInst *RetCast =
-        CastInst::Create(CastInst::getCastOpcode(NewCI, true,
-                                                 OldFn->getReturnType(), true),
-                         NewCI, OldFn->getReturnType(), NewCI->getName(),OldCI);
-
-      // Replace all uses of the old call with the new cast which has the
-      // correct type.
-      OldCI->replaceAllUsesWith(RetCast);
-    } else {
-      OldCI->replaceAllUsesWith(NewCI);
-    }
-  }
-
-  // Clean up the old call now that it has been completely upgraded.
-  OldCI->eraseFromParent();
-}
-
 // UpgradeIntrinsicCall - Upgrade a call to an old intrinsic to be a call the 
 // upgraded intrinsic. All argument and return casting must be provided in 
 // order to seamlessly integrate with existing context.
@@ -673,281 +159,53 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
   assert(F && "CallInst has no function associated with it.");
 
   if (!NewFn) {
-    // Get the Function's name.
-    const std::string& Name = F->getName();
-
-    // Upgrade ARM NEON intrinsics.
-    if (Name.compare(5, 9, "arm.neon.", 9) == 0) {
-      Instruction *NewI;
-      Value *V0, *V1;
-      if (Name.compare(14, 7, "vmovls.", 7) == 0) {
-        NewI = new SExtInst(CI->getArgOperand(0), CI->getType(),
-                            "upgraded." + CI->getName(), CI);
-      } else if (Name.compare(14, 7, "vmovlu.", 7) == 0) {
-        NewI = new ZExtInst(CI->getArgOperand(0), CI->getType(),
-                            "upgraded." + CI->getName(), CI);
-      } else if (Name.compare(14, 4, "vadd", 4) == 0) {
-        ExtendNEONArgs(CI, CI->getArgOperand(0), CI->getArgOperand(1), V0, V1);
-        NewI = BinaryOperator::CreateAdd(V0, V1, "upgraded."+CI->getName(), CI);
-      } else if (Name.compare(14, 4, "vsub", 4) == 0) {
-        ExtendNEONArgs(CI, CI->getArgOperand(0), CI->getArgOperand(1), V0, V1);
-        NewI = BinaryOperator::CreateSub(V0, V1,"upgraded."+CI->getName(),CI);
-      } else if (Name.compare(14, 4, "vmul", 4) == 0) {
-        ExtendNEONArgs(CI, CI->getArgOperand(0), CI->getArgOperand(1), V0, V1);
-        NewI = BinaryOperator::CreateMul(V0, V1,"upgraded."+CI->getName(),CI);
-      } else if (Name.compare(14, 4, "vmla", 4) == 0) {
-        ExtendNEONArgs(CI, CI->getArgOperand(1), CI->getArgOperand(2), V0, V1);
-        Instruction *MulI = BinaryOperator::CreateMul(V0, V1, "", CI);
-        NewI = BinaryOperator::CreateAdd(CI->getArgOperand(0), MulI,
-                                         "upgraded."+CI->getName(), CI);
-      } else if (Name.compare(14, 4, "vmls", 4) == 0) {
-        ExtendNEONArgs(CI, CI->getArgOperand(1), CI->getArgOperand(2), V0, V1);
-        Instruction *MulI = BinaryOperator::CreateMul(V0, V1, "", CI);
-        NewI = BinaryOperator::CreateSub(CI->getArgOperand(0), MulI,
-                                         "upgraded."+CI->getName(), CI);
-      } else if (Name.compare(14, 4, "vabd", 4) == 0) {
-        NewI = CallVABD(CI, CI->getArgOperand(0), CI->getArgOperand(1));
-        NewI = new ZExtInst(NewI, CI->getType(), "upgraded."+CI->getName(), CI);
-      } else if (Name.compare(14, 4, "vaba", 4) == 0) {
-        NewI = CallVABD(CI, CI->getArgOperand(1), CI->getArgOperand(2));
-        if (Name.at(18) == 'l')
-          NewI = new ZExtInst(NewI, CI->getType(), "", CI);
-        NewI = BinaryOperator::CreateAdd(CI->getArgOperand(0), NewI,
-                                         "upgraded."+CI->getName(), CI);
-      } else if (Name.compare(14, 6, "vmovn.", 6) == 0) {
-        NewI = new TruncInst(CI->getArgOperand(0), CI->getType(),
-                             "upgraded." + CI->getName(), CI);
-      } else {
-        llvm_unreachable("Unknown arm.neon function for CallInst upgrade.");
-      }
-      // Replace any uses of the old CallInst.
-      if (!CI->use_empty())
-        CI->replaceAllUsesWith(NewI);
-      CI->eraseFromParent();
-      return;
-    }
-
-    bool isLoadH = false, isLoadL = false, isMovL = false;
-    bool isMovSD = false, isShufPD = false;
-    bool isUnpckhPD = false, isUnpcklPD = false;
-    bool isPunpckhQPD = false, isPunpcklQPD = false;
-    if (F->getName() == "llvm.x86.sse2.loadh.pd")
-      isLoadH = true;
-    else if (F->getName() == "llvm.x86.sse2.loadl.pd")
-      isLoadL = true;
-    else if (F->getName() == "llvm.x86.sse2.movl.dq")
-      isMovL = true;
-    else if (F->getName() == "llvm.x86.sse2.movs.d")
-      isMovSD = true;
-    else if (F->getName() == "llvm.x86.sse2.shuf.pd")
-      isShufPD = true;
-    else if (F->getName() == "llvm.x86.sse2.unpckh.pd")
-      isUnpckhPD = true;
-    else if (F->getName() == "llvm.x86.sse2.unpckl.pd")
-      isUnpcklPD = true;
-    else if (F->getName() ==  "llvm.x86.sse2.punpckh.qdq")
-      isPunpckhQPD = true;
-    else if (F->getName() ==  "llvm.x86.sse2.punpckl.qdq")
-      isPunpcklQPD = true;
-
-    if (isLoadH || isLoadL || isMovL || isMovSD || isShufPD ||
-        isUnpckhPD || isUnpcklPD || isPunpckhQPD || isPunpcklQPD) {
-      std::vector<Constant*> Idxs;
-      Value *Op0 = CI->getArgOperand(0);
-      ShuffleVectorInst *SI = NULL;
-      if (isLoadH || isLoadL) {
-        Value *Op1 = UndefValue::get(Op0->getType());
-        Value *Addr = new BitCastInst(CI->getArgOperand(1), 
-                                  Type::getDoublePtrTy(C),
-                                      "upgraded.", CI);
-        Value *Load = new LoadInst(Addr, "upgraded.", false, 8, CI);
-        Value *Idx = ConstantInt::get(Type::getInt32Ty(C), 0);
-        Op1 = InsertElementInst::Create(Op1, Load, Idx, "upgraded.", CI);
-
-        if (isLoadH) {
-          Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 0));
-          Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 2));
-        } else {
-          Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 2));
-          Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 1));
-        }
-        Value *Mask = ConstantVector::get(Idxs);
-        SI = new ShuffleVectorInst(Op0, Op1, Mask, "upgraded.", CI);
-      } else if (isMovL) {
-        Constant *Zero = ConstantInt::get(Type::getInt32Ty(C), 0);
-        Idxs.push_back(Zero);
-        Idxs.push_back(Zero);
-        Idxs.push_back(Zero);
-        Idxs.push_back(Zero);
-        Value *ZeroV = ConstantVector::get(Idxs);
-
-        Idxs.clear(); 
-        Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 4));
-        Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 5));
-        Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 2));
-        Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 3));
-        Value *Mask = ConstantVector::get(Idxs);
-        SI = new ShuffleVectorInst(ZeroV, Op0, Mask, "upgraded.", CI);
-      } else if (isMovSD ||
-                 isUnpckhPD || isUnpcklPD || isPunpckhQPD || isPunpcklQPD) {
-        Value *Op1 = CI->getArgOperand(1);
-        if (isMovSD) {
-          Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 2));
-          Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 1));
-        } else if (isUnpckhPD || isPunpckhQPD) {
-          Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 1));
-          Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 3));
-        } else {
-          Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 0));
-          Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), 2));
-        }
-        Value *Mask = ConstantVector::get(Idxs);
-        SI = new ShuffleVectorInst(Op0, Op1, Mask, "upgraded.", CI);
-      } else if (isShufPD) {
-        Value *Op1 = CI->getArgOperand(1);
-        unsigned MaskVal =
-                        cast<ConstantInt>(CI->getArgOperand(2))->getZExtValue();
-        Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C), MaskVal & 1));
-        Idxs.push_back(ConstantInt::get(Type::getInt32Ty(C),
-                                               ((MaskVal >> 1) & 1)+2));
-        Value *Mask = ConstantVector::get(Idxs);
-        SI = new ShuffleVectorInst(Op0, Op1, Mask, "upgraded.", CI);
-      }
-
-      assert(SI && "Unexpected!");
-
-      // Handle any uses of the old CallInst.
-      if (!CI->use_empty())
-        //  Replace all uses of the old call with the new cast which has the 
-        //  correct type.
-        CI->replaceAllUsesWith(SI);
-      
-      //  Clean up the old call now that it has been completely upgraded.
-      CI->eraseFromParent();
-    } else if (F->getName() == "llvm.x86.sse41.pmulld") {
-      // Upgrade this set of intrinsics into vector multiplies.
-      Instruction *Mul = BinaryOperator::CreateMul(CI->getArgOperand(0),
-                                                   CI->getArgOperand(1),
-                                                   CI->getName(),
-                                                   CI);
-      // Fix up all the uses with our new multiply.
-      if (!CI->use_empty())
-        CI->replaceAllUsesWith(Mul);
-        
-      // Remove upgraded multiply.
-      CI->eraseFromParent();
-    } else if (F->getName() == "llvm.x86.ssse3.palign.r") {
-      Value *Op1 = CI->getArgOperand(0);
-      Value *Op2 = CI->getArgOperand(1);
-      Value *Op3 = CI->getArgOperand(2);
-      unsigned shiftVal = cast<ConstantInt>(Op3)->getZExtValue();
-      Value *Rep;
+    if (F->getName() == "llvm.x86.sse.loadu.ps" ||
+        F->getName() == "llvm.x86.sse2.loadu.dq" ||
+        F->getName() == "llvm.x86.sse2.loadu.pd") {
+      // Convert to a native, unaligned load.
+      Type *VecTy = CI->getType();
+      Type *IntTy = IntegerType::get(C, 128);
       IRBuilder<> Builder(C);
       Builder.SetInsertPoint(CI->getParent(), CI);
 
-      // If palignr is shifting the pair of input vectors less than 9 bytes,
-      // emit a shuffle instruction.
-      if (shiftVal <= 8) {
-        const Type *IntTy = Type::getInt32Ty(C);
-        const Type *EltTy = Type::getInt8Ty(C);
-        const Type *VecTy = VectorType::get(EltTy, 8);
-        
-        Op2 = Builder.CreateBitCast(Op2, VecTy);
-        Op1 = Builder.CreateBitCast(Op1, VecTy);
-
-        llvm::SmallVector<llvm::Constant*, 8> Indices;
-        for (unsigned i = 0; i != 8; ++i)
-          Indices.push_back(ConstantInt::get(IntTy, shiftVal + i));
-
-        Value *SV = ConstantVector::get(Indices.begin(), Indices.size());
-        Rep = Builder.CreateShuffleVector(Op2, Op1, SV, "palignr");
-        Rep = Builder.CreateBitCast(Rep, F->getReturnType());
-      }
-
-      // If palignr is shifting the pair of input vectors more than 8 but less
-      // than 16 bytes, emit a logical right shift of the destination.
-      else if (shiftVal < 16) {
-        // MMX has these as 1 x i64 vectors for some odd optimization reasons.
-        const Type *EltTy = Type::getInt64Ty(C);
-        const Type *VecTy = VectorType::get(EltTy, 1);
+      Value *BC = Builder.CreateBitCast(CI->getArgOperand(0),
+                                        PointerType::getUnqual(IntTy),
+                                        "cast");
+      LoadInst *LI = Builder.CreateLoad(BC, CI->getName());
+      LI->setAlignment(1);      // Unaligned load.
+      BC = Builder.CreateBitCast(LI, VecTy, "new.cast");
 
-        Op1 = Builder.CreateBitCast(Op1, VecTy, "cast");
-        Op2 = ConstantInt::get(VecTy, (shiftVal-8) * 8);
-
-        // create i32 constant
-        Function *I =
-          Intrinsic::getDeclaration(F->getParent(), Intrinsic::x86_mmx_psrl_q);
-        Rep = Builder.CreateCall2(I, Op1, Op2, "palignr");
-      }
-
-      // If palignr is shifting the pair of vectors more than 32 bytes, emit zero.
-      else {
-        Rep = Constant::getNullValue(F->getReturnType());
-      }
-      
-      // Replace any uses with our new instruction.
+      // Fix up all the uses with our new load.
       if (!CI->use_empty())
-        CI->replaceAllUsesWith(Rep);
-        
-      // Remove upgraded instruction.
+        CI->replaceAllUsesWith(BC);
+
+      // Remove intrinsic.
       CI->eraseFromParent();
-      
-    } else if (F->getName() == "llvm.x86.ssse3.palign.r.128") {
-      Value *Op1 = CI->getArgOperand(0);
-      Value *Op2 = CI->getArgOperand(1);
-      Value *Op3 = CI->getArgOperand(2);
-      unsigned shiftVal = cast<ConstantInt>(Op3)->getZExtValue();
-      Value *Rep;
+    } else if (F->getName() == "llvm.x86.sse.movnt.ps" ||
+               F->getName() == "llvm.x86.sse2.movnt.dq" ||
+               F->getName() == "llvm.x86.sse2.movnt.pd" ||
+               F->getName() == "llvm.x86.sse2.movnt.i") {
       IRBuilder<> Builder(C);
       Builder.SetInsertPoint(CI->getParent(), CI);
 
-      // If palignr is shifting the pair of input vectors less than 17 bytes,
-      // emit a shuffle instruction.
-      if (shiftVal <= 16) {
-        const Type *IntTy = Type::getInt32Ty(C);
-        const Type *EltTy = Type::getInt8Ty(C);
-        const Type *VecTy = VectorType::get(EltTy, 16);
-        
-        Op2 = Builder.CreateBitCast(Op2, VecTy);
-        Op1 = Builder.CreateBitCast(Op1, VecTy);
-
-        llvm::SmallVector<llvm::Constant*, 16> Indices;
-        for (unsigned i = 0; i != 16; ++i)
-          Indices.push_back(ConstantInt::get(IntTy, shiftVal + i));
-
-        Value *SV = ConstantVector::get(Indices.begin(), Indices.size());
-        Rep = Builder.CreateShuffleVector(Op2, Op1, SV, "palignr");
-        Rep = Builder.CreateBitCast(Rep, F->getReturnType());
-      }
-
-      // If palignr is shifting the pair of input vectors more than 16 but less
-      // than 32 bytes, emit a logical right shift of the destination.
-      else if (shiftVal < 32) {
-        const Type *EltTy = Type::getInt64Ty(C);
-        const Type *VecTy = VectorType::get(EltTy, 2);
-        const Type *IntTy = Type::getInt32Ty(C);
+      Module *M = F->getParent();
+      SmallVector<Value *, 1> Elts;
+      Elts.push_back(ConstantInt::get(Type::getInt32Ty(C), 1));
+      MDNode *Node = MDNode::get(C, Elts);
 
-        Op1 = Builder.CreateBitCast(Op1, VecTy, "cast");
-        Op2 = ConstantInt::get(IntTy, (shiftVal-16) * 8);
+      Value *Arg0 = CI->getArgOperand(0);
+      Value *Arg1 = CI->getArgOperand(1);
 
-        // create i32 constant
-        Function *I =
-          Intrinsic::getDeclaration(F->getParent(), Intrinsic::x86_sse2_psrl_dq);
-        Rep = Builder.CreateCall2(I, Op1, Op2, "palignr");
-      }
+      // Convert the type of the pointer to a pointer to the stored type.
+      Value *BC = Builder.CreateBitCast(Arg0,
+                                        PointerType::getUnqual(Arg1->getType()),
+                                        "cast");
+      StoreInst *SI = Builder.CreateStore(Arg1, BC);
+      SI->setMetadata(M->getMDKindID("nontemporal"), Node);
+      SI->setAlignment(16);
 
-      // If palignr is shifting the pair of vectors more than 32 bytes, emit zero.
-      else {
-        Rep = Constant::getNullValue(F->getReturnType());
-      }
-      
-      // Replace any uses with our new instruction.
-      if (!CI->use_empty())
-        CI->replaceAllUsesWith(Rep);
-        
-      // Remove upgraded instruction.
+      // Remove intrinsic.
       CI->eraseFromParent();
-      
     } else {
       llvm_unreachable("Unknown function for CallInst upgrade.");
     }
@@ -955,375 +213,52 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
   }
 
   switch (NewFn->getIntrinsicID()) {
-  default: llvm_unreachable("Unknown function for CallInst upgrade.");
-  case Intrinsic::arm_neon_vld1:
-  case Intrinsic::arm_neon_vld2:
-  case Intrinsic::arm_neon_vld3:
-  case Intrinsic::arm_neon_vld4:
-  case Intrinsic::arm_neon_vst1:
-  case Intrinsic::arm_neon_vst2:
-  case Intrinsic::arm_neon_vst3:
-  case Intrinsic::arm_neon_vst4:
-  case Intrinsic::arm_neon_vld2lane:
-  case Intrinsic::arm_neon_vld3lane:
-  case Intrinsic::arm_neon_vld4lane:
-  case Intrinsic::arm_neon_vst2lane:
-  case Intrinsic::arm_neon_vst3lane:
-  case Intrinsic::arm_neon_vst4lane: {
-    // Add a default alignment argument of 1.
-    SmallVector<Value*, 8> Operands(CS.arg_begin(), CS.arg_end());
-    Operands.push_back(ConstantInt::get(Type::getInt32Ty(C), 1));
-    CallInst *NewCI = CallInst::Create(NewFn, Operands.begin(), Operands.end(),
+  case Intrinsic::prefetch: {
+    IRBuilder<> Builder(C);
+    Builder.SetInsertPoint(CI->getParent(), CI);
+    llvm::Type *I32Ty = llvm::Type::getInt32Ty(CI->getContext());
+
+    // Add the extra "data cache" argument
+    Value *Operands[4] = { CI->getArgOperand(0), CI->getArgOperand(1),
+                           CI->getArgOperand(2),
+                           llvm::ConstantInt::get(I32Ty, 1) };
+    CallInst *NewCI = CallInst::Create(NewFn, Operands,
                                        CI->getName(), CI);
     NewCI->setTailCall(CI->isTailCall());
     NewCI->setCallingConv(CI->getCallingConv());
-
     //  Handle any uses of the old CallInst.
     if (!CI->use_empty())
-      //  Replace all uses of the old call with the new cast which has the 
+      //  Replace all uses of the old call with the new cast which has the
       //  correct type.
       CI->replaceAllUsesWith(NewCI);
-    
+
     //  Clean up the old call now that it has been completely upgraded.
     CI->eraseFromParent();
     break;
-  }        
-
-  case Intrinsic::x86_mmx_padd_b:
-  case Intrinsic::x86_mmx_padd_w:
-  case Intrinsic::x86_mmx_padd_d:
-  case Intrinsic::x86_mmx_padd_q:
-  case Intrinsic::x86_mmx_padds_b:
-  case Intrinsic::x86_mmx_padds_w:
-  case Intrinsic::x86_mmx_paddus_b:
-  case Intrinsic::x86_mmx_paddus_w:
-  case Intrinsic::x86_mmx_psub_b:
-  case Intrinsic::x86_mmx_psub_w:
-  case Intrinsic::x86_mmx_psub_d:
-  case Intrinsic::x86_mmx_psub_q:
-  case Intrinsic::x86_mmx_psubs_b:
-  case Intrinsic::x86_mmx_psubs_w:
-  case Intrinsic::x86_mmx_psubus_b:
-  case Intrinsic::x86_mmx_psubus_w:
-  case Intrinsic::x86_mmx_pmulh_w:
-  case Intrinsic::x86_mmx_pmull_w:
-  case Intrinsic::x86_mmx_pmulhu_w:
-  case Intrinsic::x86_mmx_pmulu_dq:
-  case Intrinsic::x86_mmx_pmadd_wd:
-  case Intrinsic::x86_mmx_pand:
-  case Intrinsic::x86_mmx_pandn:
-  case Intrinsic::x86_mmx_por:
-  case Intrinsic::x86_mmx_pxor:
-  case Intrinsic::x86_mmx_pavg_b:
-  case Intrinsic::x86_mmx_pavg_w:
-  case Intrinsic::x86_mmx_pmaxu_b:
-  case Intrinsic::x86_mmx_pmaxs_w:
-  case Intrinsic::x86_mmx_pminu_b:
-  case Intrinsic::x86_mmx_pmins_w:
-  case Intrinsic::x86_mmx_psad_bw:
-  case Intrinsic::x86_mmx_psll_w:
-  case Intrinsic::x86_mmx_psll_d:
-  case Intrinsic::x86_mmx_psll_q:
-  case Intrinsic::x86_mmx_pslli_w:
-  case Intrinsic::x86_mmx_pslli_d:
-  case Intrinsic::x86_mmx_pslli_q:
-  case Intrinsic::x86_mmx_psrl_w:
-  case Intrinsic::x86_mmx_psrl_d:
-  case Intrinsic::x86_mmx_psrl_q:
-  case Intrinsic::x86_mmx_psrli_w:
-  case Intrinsic::x86_mmx_psrli_d:
-  case Intrinsic::x86_mmx_psrli_q:
-  case Intrinsic::x86_mmx_psra_w:
-  case Intrinsic::x86_mmx_psra_d:
-  case Intrinsic::x86_mmx_psrai_w:
-  case Intrinsic::x86_mmx_psrai_d:
-  case Intrinsic::x86_mmx_packsswb:
-  case Intrinsic::x86_mmx_packssdw:
-  case Intrinsic::x86_mmx_packuswb:
-  case Intrinsic::x86_mmx_punpckhbw:
-  case Intrinsic::x86_mmx_punpckhwd:
-  case Intrinsic::x86_mmx_punpckhdq:
-  case Intrinsic::x86_mmx_punpcklbw:
-  case Intrinsic::x86_mmx_punpcklwd:
-  case Intrinsic::x86_mmx_punpckldq:
-  case Intrinsic::x86_mmx_pcmpeq_b:
-  case Intrinsic::x86_mmx_pcmpeq_w:
-  case Intrinsic::x86_mmx_pcmpeq_d:
-  case Intrinsic::x86_mmx_pcmpgt_b:
-  case Intrinsic::x86_mmx_pcmpgt_w:
-  case Intrinsic::x86_mmx_pcmpgt_d: {
-    Value *Operands[2];
-    
-    // Cast the operand to the X86 MMX type.
-    Operands[0] = new BitCastInst(CI->getArgOperand(0), 
-                                  NewFn->getFunctionType()->getParamType(0),
-                                  "upgraded.", CI);
-
-    switch (NewFn->getIntrinsicID()) {
-    default:
-      // Cast to the X86 MMX type.
-      Operands[1] = new BitCastInst(CI->getArgOperand(1), 
-                                    NewFn->getFunctionType()->getParamType(1),
-                                    "upgraded.", CI);
-      break;
-    case Intrinsic::x86_mmx_pslli_w:
-    case Intrinsic::x86_mmx_pslli_d:
-    case Intrinsic::x86_mmx_pslli_q:
-    case Intrinsic::x86_mmx_psrli_w:
-    case Intrinsic::x86_mmx_psrli_d:
-    case Intrinsic::x86_mmx_psrli_q:
-    case Intrinsic::x86_mmx_psrai_w:
-    case Intrinsic::x86_mmx_psrai_d:
-      // These take an i32 as their second parameter.
-      Operands[1] = CI->getArgOperand(1);
-      break;
-    }
-
-    ConstructNewCallInst(NewFn, CI, Operands, 2);
-    break;
   }
-  case Intrinsic::x86_mmx_maskmovq: {
-    Value *Operands[3];
-
-    // Cast the operands to the X86 MMX type.
-    Operands[0] = new BitCastInst(CI->getArgOperand(0), 
-                                  NewFn->getFunctionType()->getParamType(0),
-                                  "upgraded.", CI);
-    Operands[1] = new BitCastInst(CI->getArgOperand(1), 
-                                  NewFn->getFunctionType()->getParamType(1),
-                                  "upgraded.", CI);
-    Operands[2] = CI->getArgOperand(2);
-
-    ConstructNewCallInst(NewFn, CI, Operands, 3, false);
-    break;
-  }
-  case Intrinsic::x86_mmx_pmovmskb: {
-    Value *Operands[1];
-
-    // Cast the operand to the X86 MMX type.
-    Operands[0] = new BitCastInst(CI->getArgOperand(0), 
-                                  NewFn->getFunctionType()->getParamType(0),
-                                  "upgraded.", CI);
-
-    ConstructNewCallInst(NewFn, CI, Operands, 1);
-    break;
-  }
-  case Intrinsic::x86_mmx_movnt_dq: {
-    Value *Operands[2];
+  case Intrinsic::init_trampoline: {
 
-    Operands[0] = CI->getArgOperand(0);
+    //  Transform
+    //    %tramp = call i8* llvm.init.trampoline (i8* x, i8* y, i8* z)
+    //  to
+    //    call void llvm.init.trampoline (i8* %x, i8* %y, i8* %z)
+    //    %tramp = call i8* llvm.adjust.trampoline (i8* %x)
 
-    // Cast the operand to the X86 MMX type.
-    Operands[1] = new BitCastInst(CI->getArgOperand(1),
-                                  NewFn->getFunctionType()->getParamType(1),
-                                  "upgraded.", CI);
+    Function *AdjustTrampolineFn =
+      cast<Function>(Intrinsic::getDeclaration(F->getParent(),
+                                               Intrinsic::adjust_trampoline));
 
-    ConstructNewCallInst(NewFn, CI, Operands, 2, false);
-    break;
-  }
-  case Intrinsic::x86_mmx_palignr_b: {
-    Value *Operands[3];
-
-    // Cast the operands to the X86 MMX type.
-    Operands[0] = new BitCastInst(CI->getArgOperand(0),
-                                  NewFn->getFunctionType()->getParamType(0),
-                                  "upgraded.", CI);
-    Operands[1] = new BitCastInst(CI->getArgOperand(1),
-                                  NewFn->getFunctionType()->getParamType(1),
-                                  "upgraded.", CI);
-    Operands[2] = CI->getArgOperand(2);
-
-    ConstructNewCallInst(NewFn, CI, Operands, 3);
-    break;
-  }
-  case Intrinsic::x86_mmx_pextr_w: {
-    Value *Operands[2];
-
-    // Cast the operands to the X86 MMX type.
-    Operands[0] = new BitCastInst(CI->getArgOperand(0),
-                                  NewFn->getFunctionType()->getParamType(0),
-                                  "upgraded.", CI);
-    Operands[1] = CI->getArgOperand(1);
-
-    ConstructNewCallInst(NewFn, CI, Operands, 2);
-    break;
-  }
-  case Intrinsic::x86_mmx_pinsr_w: {
-    Value *Operands[3];
-
-    // Cast the operands to the X86 MMX type.
-    Operands[0] = new BitCastInst(CI->getArgOperand(0),
-                                  NewFn->getFunctionType()->getParamType(0),
-                                  "upgraded.", CI);
-    Operands[1] = CI->getArgOperand(1);
-    Operands[2] = CI->getArgOperand(2);
-
-    ConstructNewCallInst(NewFn, CI, Operands, 3);
-    break;
-  }
-  case Intrinsic::x86_sse_pshuf_w: {
     IRBuilder<> Builder(C);
-    Builder.SetInsertPoint(CI->getParent(), CI);
-
-    // Cast the operand to the X86 MMX type.
-    Value *Operands[2];
-    Operands[0] =
-      Builder.CreateBitCast(CI->getArgOperand(0), 
-                            NewFn->getFunctionType()->getParamType(0),
-                            "upgraded.");
-    Operands[1] =
-      Builder.CreateTrunc(CI->getArgOperand(1),
-                          Type::getInt8Ty(C),
-                          "upgraded.");
-
-    ConstructNewCallInst(NewFn, CI, Operands, 2);
-    break;
-  }
-
-#if 0
-  case Intrinsic::x86_mmx_cvtsi32_si64: {
-    // The return type needs to be changed.
-    Value *Operands[1];
-    Operands[0] = CI->getArgOperand(0);
-    ConstructNewCallInst(NewFn, CI, Operands, 1);
-    break;
-  }
-  case Intrinsic::x86_mmx_cvtsi64_si32: {
-    Value *Operands[1];
+    Builder.SetInsertPoint(CI);
 
-    // Cast the operand to the X86 MMX type.
-    Operands[0] = new BitCastInst(CI->getArgOperand(0),
-                                  NewFn->getFunctionType()->getParamType(0),
-                                  "upgraded.", CI);
+    Builder.CreateCall3(NewFn, CI->getArgOperand(0), CI->getArgOperand(1),
+                        CI->getArgOperand(2));
 
-    ConstructNewCallInst(NewFn, CI, Operands, 1);
-    break;
-  }
-  case Intrinsic::x86_mmx_vec_init_b:
-  case Intrinsic::x86_mmx_vec_init_w:
-  case Intrinsic::x86_mmx_vec_init_d: {
-    // The return type needs to be changed.
-    Value *Operands[8];
-    unsigned NumOps = 0;
-
-    switch (NewFn->getIntrinsicID()) {
-    default: break;
-    case Intrinsic::x86_mmx_vec_init_b: NumOps = 8; break;
-    case Intrinsic::x86_mmx_vec_init_w: NumOps = 4; break;
-    case Intrinsic::x86_mmx_vec_init_d: NumOps = 2; break;
-    }
-
-    switch (NewFn->getIntrinsicID()) {
-    default: break;
-    case Intrinsic::x86_mmx_vec_init_b:
-      Operands[7] = CI->getArgOperand(7);
-      Operands[6] = CI->getArgOperand(6);
-      Operands[5] = CI->getArgOperand(5);
-      Operands[4] = CI->getArgOperand(4);
-      // FALLTHRU
-    case Intrinsic::x86_mmx_vec_init_w:
-      Operands[3] = CI->getArgOperand(3);
-      Operands[2] = CI->getArgOperand(2);
-      // FALLTHRU
-    case Intrinsic::x86_mmx_vec_init_d:
-      Operands[1] = CI->getArgOperand(1);
-      Operands[0] = CI->getArgOperand(0);
-      break;
-    }
-
-    ConstructNewCallInst(NewFn, CI, Operands, NumOps);
-    break;
-  }
-  case Intrinsic::x86_mmx_vec_ext_d: {
-    Value *Operands[2];
-
-    // Cast the operand to the X86 MMX type.
-    Operands[0] = new BitCastInst(CI->getArgOperand(0),
-                                  NewFn->getFunctionType()->getParamType(0),
-                                  "upgraded.", CI);
-    Operands[1] = CI->getArgOperand(1);
-
-    ConstructNewCallInst(NewFn, CI, Operands, 2);
-    break;
-  }
-#endif
-
-  case Intrinsic::ctlz:
-  case Intrinsic::ctpop:
-  case Intrinsic::cttz: {
-    //  Build a small vector of the original arguments.
-    SmallVector<Value*, 8> Operands(CS.arg_begin(), CS.arg_end());
-
-    //  Construct a new CallInst
-    CallInst *NewCI = CallInst::Create(NewFn, Operands.begin(), Operands.end(),
-                                       "upgraded."+CI->getName(), CI);
-    NewCI->setTailCall(CI->isTailCall());
-    NewCI->setCallingConv(CI->getCallingConv());
-
-    //  Handle any uses of the old CallInst.
-    if (!CI->use_empty()) {
-      //  Check for sign extend parameter attributes on the return values.
-      bool SrcSExt = NewFn->getAttributes().paramHasAttr(0, Attribute::SExt);
-      bool DestSExt = F->getAttributes().paramHasAttr(0, Attribute::SExt);
-      
-      //  Construct an appropriate cast from the new return type to the old.
-      CastInst *RetCast = CastInst::Create(
-                            CastInst::getCastOpcode(NewCI, SrcSExt,
-                                                    F->getReturnType(),
-                                                    DestSExt),
-                            NewCI, F->getReturnType(),
-                            NewCI->getName(), CI);
-      NewCI->moveBefore(RetCast);
-
-      //  Replace all uses of the old call with the new cast which has the 
-      //  correct type.
-      CI->replaceAllUsesWith(RetCast);
-    }
-
-    //  Clean up the old call now that it has been completely upgraded.
-    CI->eraseFromParent();
-  }
-  break;
-  case Intrinsic::eh_selector:
-  case Intrinsic::eh_typeid_for: {
-    // Only the return type changed.
-    SmallVector<Value*, 8> Operands(CS.arg_begin(), CS.arg_end());
-    CallInst *NewCI = CallInst::Create(NewFn, Operands.begin(), Operands.end(),
-                                       "upgraded." + CI->getName(), CI);
-    NewCI->setTailCall(CI->isTailCall());
-    NewCI->setCallingConv(CI->getCallingConv());
-
-    //  Handle any uses of the old CallInst.
-    if (!CI->use_empty()) {
-      //  Construct an appropriate cast from the new return type to the old.
-      CastInst *RetCast =
-        CastInst::Create(CastInst::getCastOpcode(NewCI, true,
-                                                 F->getReturnType(), true),
-                         NewCI, F->getReturnType(), NewCI->getName(), CI);
-      CI->replaceAllUsesWith(RetCast);
-    }
-    CI->eraseFromParent();
-  }
-  break;
-  case Intrinsic::memcpy:
-  case Intrinsic::memmove:
-  case Intrinsic::memset: {
-    // Add isVolatile
-    const llvm::Type *I1Ty = llvm::Type::getInt1Ty(CI->getContext());
-    Value *Operands[5] = { CI->getArgOperand(0), CI->getArgOperand(1),
-                           CI->getArgOperand(2), CI->getArgOperand(3),
-                           llvm::ConstantInt::get(I1Ty, 0) };
-    CallInst *NewCI = CallInst::Create(NewFn, Operands, Operands+5,
-                                       CI->getName(), CI);
-    NewCI->setTailCall(CI->isTailCall());
-    NewCI->setCallingConv(CI->getCallingConv());
-    //  Handle any uses of the old CallInst.
+    CallInst *AdjustCall = Builder.CreateCall(AdjustTrampolineFn,
+                                              CI->getArgOperand(0),
+                                              CI->getName());
     if (!CI->use_empty())
-      //  Replace all uses of the old call with the new cast which has the 
-      //  correct type.
-      CI->replaceAllUsesWith(NewCI);
-    
-    //  Clean up the old call now that it has been completely upgraded.
+      CI->replaceAllUsesWith(AdjustCall);
     CI->eraseFromParent();
     break;
   }
@@ -1337,13 +272,13 @@ void llvm::UpgradeCallsToIntrinsic(Function* F) {
   assert(F && "Illegal attempt to upgrade a non-existent intrinsic.");
 
   // Upgrade the function and check if it is a totaly new function.
-  FunctionNewFn;
+  Function *NewFn;
   if (UpgradeIntrinsicFunction(F, NewFn)) {
     if (NewFn != F) {
       // Replace all uses to the old function with the new one if necessary.
       for (Value::use_iterator UI = F->use_begin(), UE = F->use_end();
            UI != UE; ) {
-        if (CallInstCI = dyn_cast<CallInst>(*UI++))
+        if (CallInst *CI = dyn_cast<CallInst>(*UI++))
           UpgradeIntrinsicCall(CI, NewFn);
       }
       // Remove old function, no longer used, from the module.
@@ -1356,37 +291,27 @@ void llvm::UpgradeCallsToIntrinsic(Function* F) {
 /// If an llvm.dbg.declare intrinsic is invalid, then this function simply
 /// strips that use.
 void llvm::CheckDebugInfoIntrinsics(Module *M) {
-
-
   if (Function *FuncStart = M->getFunction("llvm.dbg.func.start")) {
-    while (!FuncStart->use_empty()) {
-      CallInst *CI = cast<CallInst>(FuncStart->use_back());
-      CI->eraseFromParent();
-    }
+    while (!FuncStart->use_empty())
+      cast<CallInst>(FuncStart->use_back())->eraseFromParent();
     FuncStart->eraseFromParent();
   }
   
   if (Function *StopPoint = M->getFunction("llvm.dbg.stoppoint")) {
-    while (!StopPoint->use_empty()) {
-      CallInst *CI = cast<CallInst>(StopPoint->use_back());
-      CI->eraseFromParent();
-    }
+    while (!StopPoint->use_empty())
+      cast<CallInst>(StopPoint->use_back())->eraseFromParent();
     StopPoint->eraseFromParent();
   }
 
   if (Function *RegionStart = M->getFunction("llvm.dbg.region.start")) {
-    while (!RegionStart->use_empty()) {
-      CallInst *CI = cast<CallInst>(RegionStart->use_back());
-      CI->eraseFromParent();
-    }
+    while (!RegionStart->use_empty())
+      cast<CallInst>(RegionStart->use_back())->eraseFromParent();
     RegionStart->eraseFromParent();
   }
 
   if (Function *RegionEnd = M->getFunction("llvm.dbg.region.end")) {
-    while (!RegionEnd->use_empty()) {
-      CallInst *CI = cast<CallInst>(RegionEnd->use_back());
-      CI->eraseFromParent();
-    }
+    while (!RegionEnd->use_empty())
+      cast<CallInst>(RegionEnd->use_back())->eraseFromParent();
     RegionEnd->eraseFromParent();
   }
   
@@ -1404,3 +329,249 @@ void llvm::CheckDebugInfoIntrinsics(Module *M) {
     }
   }
 }
+
+/// FindExnAndSelIntrinsics - Find the eh_exception and eh_selector intrinsic
+/// calls reachable from the unwind basic block.
+static void FindExnAndSelIntrinsics(BasicBlock *BB, CallInst *&Exn,
+                                    CallInst *&Sel,
+                                    SmallPtrSet<BasicBlock*, 8> &Visited) {
+  if (!Visited.insert(BB)) return;
+
+  for (BasicBlock::iterator
+         I = BB->begin(), E = BB->end(); I != E; ++I) {
+    if (CallInst *CI = dyn_cast<CallInst>(I)) {
+      switch (CI->getCalledFunction()->getIntrinsicID()) {
+      default: break;
+      case Intrinsic::eh_exception:
+        assert(!Exn && "Found more than one eh.exception call!");
+        Exn = CI;
+        break;
+      case Intrinsic::eh_selector:
+        assert(!Sel && "Found more than one eh.selector call!");
+        Sel = CI;
+        break;
+      }
+
+      if (Exn && Sel) return;
+    }
+  }
+
+  if (Exn && Sel) return;
+
+  for (succ_iterator I = succ_begin(BB), E = succ_end(BB); I != E; ++I) {
+    FindExnAndSelIntrinsics(*I, Exn, Sel, Visited);
+    if (Exn && Sel) return;
+  }
+}
+
+/// TransferClausesToLandingPadInst - Transfer the exception handling clauses
+/// from the eh_selector call to the new landingpad instruction.
+static void TransferClausesToLandingPadInst(LandingPadInst *LPI,
+                                            CallInst *EHSel) {
+  LLVMContext &Context = LPI->getContext();
+  unsigned N = EHSel->getNumArgOperands();
+
+  for (unsigned i = N - 1; i > 1; --i) {
+    if (const ConstantInt *CI = dyn_cast<ConstantInt>(EHSel->getArgOperand(i))){
+      unsigned FilterLength = CI->getZExtValue();
+      unsigned FirstCatch = i + FilterLength + !FilterLength;
+      assert(FirstCatch <= N && "Invalid filter length");
+
+      if (FirstCatch < N)
+        for (unsigned j = FirstCatch; j < N; ++j) {
+          Value *Val = EHSel->getArgOperand(j);
+          if (!Val->hasName() || Val->getName() != "llvm.eh.catch.all.value") {
+            LPI->addClause(EHSel->getArgOperand(j));
+          } else {
+            GlobalVariable *GV = cast<GlobalVariable>(Val);
+            LPI->addClause(GV->getInitializer());
+          }
+        }
+
+      if (!FilterLength) {
+        // Cleanup.
+        LPI->setCleanup(true);
+      } else {
+        // Filter.
+        SmallVector<Constant *, 4> TyInfo;
+        TyInfo.reserve(FilterLength - 1);
+        for (unsigned j = i + 1; j < FirstCatch; ++j)
+          TyInfo.push_back(cast<Constant>(EHSel->getArgOperand(j)));
+        ArrayType *AType =
+          ArrayType::get(!TyInfo.empty() ? TyInfo[0]->getType() :
+                         PointerType::getUnqual(Type::getInt8Ty(Context)),
+                         TyInfo.size());
+        LPI->addClause(ConstantArray::get(AType, TyInfo));
+      }
+
+      N = i;
+    }
+  }
+
+  if (N > 2)
+    for (unsigned j = 2; j < N; ++j) {
+      Value *Val = EHSel->getArgOperand(j);
+      if (!Val->hasName() || Val->getName() != "llvm.eh.catch.all.value") {
+        LPI->addClause(EHSel->getArgOperand(j));
+      } else {
+        GlobalVariable *GV = cast<GlobalVariable>(Val);
+        LPI->addClause(GV->getInitializer());
+      }
+    }
+}
+
+/// This function upgrades the old pre-3.0 exception handling system to the new
+/// one. N.B. This will be removed in 3.1.
+void llvm::UpgradeExceptionHandling(Module *M) {
+  Function *EHException = M->getFunction("llvm.eh.exception");
+  Function *EHSelector = M->getFunction("llvm.eh.selector");
+  if (!EHException || !EHSelector)
+    return;
+
+  LLVMContext &Context = M->getContext();
+  Type *ExnTy = PointerType::getUnqual(Type::getInt8Ty(Context));
+  Type *SelTy = Type::getInt32Ty(Context);
+  Type *LPadSlotTy = StructType::get(ExnTy, SelTy, NULL);
+
+  // This map links the invoke instruction with the eh.exception and eh.selector
+  // calls associated with it.
+  DenseMap<InvokeInst*, std::pair<Value*, Value*> > InvokeToIntrinsicsMap;
+  for (Module::iterator
+         I = M->begin(), E = M->end(); I != E; ++I) {
+    Function &F = *I;
+
+    for (Function::iterator
+           II = F.begin(), IE = F.end(); II != IE; ++II) {
+      BasicBlock *BB = &*II;
+      InvokeInst *Inst = dyn_cast<InvokeInst>(BB->getTerminator());
+      if (!Inst) continue;
+      BasicBlock *UnwindDest = Inst->getUnwindDest();
+      if (UnwindDest->isLandingPad()) continue; // Already converted.
+
+      SmallPtrSet<BasicBlock*, 8> Visited;
+      CallInst *Exn = 0;
+      CallInst *Sel = 0;
+      FindExnAndSelIntrinsics(UnwindDest, Exn, Sel, Visited);
+      assert(Exn && Sel && "Cannot find eh.exception and eh.selector calls!");
+      InvokeToIntrinsicsMap[Inst] = std::make_pair(Exn, Sel);
+    }
+  }
+
+  // This map stores the slots where the exception object and selector value are
+  // stored within a function.
+  DenseMap<Function*, std::pair<Value*, Value*> > FnToLPadSlotMap;
+  SmallPtrSet<Instruction*, 32> DeadInsts;
+  for (DenseMap<InvokeInst*, std::pair<Value*, Value*> >::iterator
+         I = InvokeToIntrinsicsMap.begin(), E = InvokeToIntrinsicsMap.end();
+       I != E; ++I) {
+    InvokeInst *Invoke = I->first;
+    BasicBlock *UnwindDest = Invoke->getUnwindDest();
+    Function *F = UnwindDest->getParent();
+    std::pair<Value*, Value*> EHIntrinsics = I->second;
+    CallInst *Exn = cast<CallInst>(EHIntrinsics.first);
+    CallInst *Sel = cast<CallInst>(EHIntrinsics.second);
+
+    // Store the exception object and selector value in the entry block.
+    Value *ExnSlot = 0;
+    Value *SelSlot = 0;
+    if (!FnToLPadSlotMap[F].first) {
+      BasicBlock *Entry = &F->front();
+      ExnSlot = new AllocaInst(ExnTy, "exn", Entry->getTerminator());
+      SelSlot = new AllocaInst(SelTy, "sel", Entry->getTerminator());
+      FnToLPadSlotMap[F] = std::make_pair(ExnSlot, SelSlot);
+    } else {
+      ExnSlot = FnToLPadSlotMap[F].first;
+      SelSlot = FnToLPadSlotMap[F].second;
+    }
+
+    if (!UnwindDest->getSinglePredecessor()) {
+      // The unwind destination doesn't have a single predecessor. Create an
+      // unwind destination which has only one predecessor.
+      BasicBlock *NewBB = BasicBlock::Create(Context, "new.lpad",
+                                             UnwindDest->getParent());
+      BranchInst::Create(UnwindDest, NewBB);
+      Invoke->setUnwindDest(NewBB);
+
+      // Fix up any PHIs in the original unwind destination block.
+      for (BasicBlock::iterator
+             II = UnwindDest->begin(); isa<PHINode>(II); ++II) {
+        PHINode *PN = cast<PHINode>(II);
+        int Idx = PN->getBasicBlockIndex(Invoke->getParent());
+        if (Idx == -1) continue;
+        PN->setIncomingBlock(Idx, NewBB);
+      }
+
+      UnwindDest = NewBB;
+    }
+
+    IRBuilder<> Builder(Context);
+    Builder.SetInsertPoint(UnwindDest, UnwindDest->getFirstInsertionPt());
+
+    Value *PersFn = Sel->getArgOperand(1);
+    LandingPadInst *LPI = Builder.CreateLandingPad(LPadSlotTy, PersFn, 0);
+    Value *LPExn = Builder.CreateExtractValue(LPI, 0);
+    Value *LPSel = Builder.CreateExtractValue(LPI, 1);
+    Builder.CreateStore(LPExn, ExnSlot);
+    Builder.CreateStore(LPSel, SelSlot);
+
+    TransferClausesToLandingPadInst(LPI, Sel);
+
+    DeadInsts.insert(Exn);
+    DeadInsts.insert(Sel);
+  }
+
+  // Replace the old intrinsic calls with the values from the landingpad
+  // instruction(s). These values were stored in allocas for us to use here.
+  for (DenseMap<InvokeInst*, std::pair<Value*, Value*> >::iterator
+         I = InvokeToIntrinsicsMap.begin(), E = InvokeToIntrinsicsMap.end();
+       I != E; ++I) {
+    std::pair<Value*, Value*> EHIntrinsics = I->second;
+    CallInst *Exn = cast<CallInst>(EHIntrinsics.first);
+    CallInst *Sel = cast<CallInst>(EHIntrinsics.second);
+    BasicBlock *Parent = Exn->getParent();
+
+    std::pair<Value*,Value*> ExnSelSlots = FnToLPadSlotMap[Parent->getParent()];
+
+    IRBuilder<> Builder(Context);
+    Builder.SetInsertPoint(Parent, Exn);
+    LoadInst *LPExn = Builder.CreateLoad(ExnSelSlots.first, "exn.load");
+    LoadInst *LPSel = Builder.CreateLoad(ExnSelSlots.second, "sel.load");
+
+    Exn->replaceAllUsesWith(LPExn);
+    Sel->replaceAllUsesWith(LPSel);
+  }
+
+  // Remove the dead instructions.
+  for (SmallPtrSet<Instruction*, 32>::iterator
+         I = DeadInsts.begin(), E = DeadInsts.end(); I != E; ++I) {
+    Instruction *Inst = *I;
+    Inst->eraseFromParent();
+  }
+
+  // Replace calls to "llvm.eh.resume" with the 'resume' instruction. Load the
+  // exception and selector values from the stored place.
+  Function *EHResume = M->getFunction("llvm.eh.resume");
+  if (!EHResume) return;
+
+  while (!EHResume->use_empty()) {
+    CallInst *Resume = cast<CallInst>(EHResume->use_back());
+    BasicBlock *BB = Resume->getParent();
+
+    IRBuilder<> Builder(Context);
+    Builder.SetInsertPoint(BB, Resume);
+
+    Value *LPadVal =
+      Builder.CreateInsertValue(UndefValue::get(LPadSlotTy),
+                                Resume->getArgOperand(0), 0, "lpad.val");
+    LPadVal = Builder.CreateInsertValue(LPadVal, Resume->getArgOperand(1),
+                                        1, "lpad.val");
+    Builder.CreateResume(LPadVal);
+
+    // Remove all instructions after the 'resume.'
+    BasicBlock::iterator I = Resume;
+    while (I != BB->end()) {
+      Instruction *Inst = &*I++;
+      Inst->eraseFromParent();
+    }
+  }
+}