Update compiler-rt to 3.9.0 release, and update the build glue for

libcompiler_rt and libclang_rt.

18 files changed, 1557 insertions, 578 deletions

diff --git a/contrib/compiler-rt/lib/profile/GCDAProfiling.c b/contrib/compiler-rt/lib/profile/GCDAProfiling.c
index 2338761ae1ab..2756084f5fd3 100644
--- a/contrib/compiler-rt/lib/profile/GCDAProfiling.c
+++ b/contrib/compiler-rt/lib/profile/GCDAProfiling.c
@@ -20,6 +20,8 @@
 |*
 \*===----------------------------------------------------------------------===*/
 
+#include "InstrProfilingInternal.h"
+#include "InstrProfilingPort.h"
 #include "InstrProfilingUtil.h"
 
 #include <errno.h>
@@ -35,7 +37,11 @@
 #include <sys/file.h>
 #endif
 
-#define I386_FREEBSD (defined(__FreeBSD__) && defined(__i386__))
+#if defined(__FreeBSD__) && defined(__i386__)
+#define I386_FREEBSD 1
+#else
+#define I386_FREEBSD 0
+#endif
 
 #if !defined(_MSC_VER) && !I386_FREEBSD
 #include <stdint.h>
@@ -166,44 +172,16 @@ static uint64_t read_64bit_value() {
 
 static char *mangle_filename(const char *orig_filename) {
   char *new_filename;
-  size_t filename_len, prefix_len;
+  size_t prefix_len;
   int prefix_strip;
-  int level = 0;
-  const char *fname, *ptr;
-  const char *prefix = getenv("GCOV_PREFIX");
-  const char *prefix_strip_str = getenv("GCOV_PREFIX_STRIP");
+  const char *prefix = lprofGetPathPrefix(&prefix_strip, &prefix_len);
 
-  if (prefix == NULL || prefix[0] == '\0')
+  if (prefix == NULL)
     return strdup(orig_filename);
 
-  if (prefix_strip_str) {
-    prefix_strip = atoi(prefix_strip_str);
-
-    /* Negative GCOV_PREFIX_STRIP values are ignored */
-    if (prefix_strip < 0)
-      prefix_strip = 0;
-  } else {
-    prefix_strip = 0;
-  }
-
-  fname = orig_filename;
-  for (level = 0, ptr = fname + 1; level < prefix_strip; ++ptr) {
-    if (*ptr == '\0')
-      break;
-    if (*ptr != '/')
-      continue;
-    fname = ptr;
-    ++level;
-  }
-
-  filename_len = strlen(fname);
-  prefix_len = strlen(prefix);
-  new_filename = malloc(prefix_len + 1 + filename_len + 1);
-  memcpy(new_filename, prefix, prefix_len);
-
-  if (prefix[prefix_len - 1] != '/')
-    new_filename[prefix_len++] = '/';
-  memcpy(new_filename + prefix_len, fname, filename_len + 1);
+  new_filename = malloc(prefix_len + 1 + strlen(orig_filename) + 1);
+  lprofApplyPathPrefix(new_filename, orig_filename, prefix, prefix_len,
+                       prefix_strip);
 
   return new_filename;
 }
@@ -482,8 +460,8 @@ void llvm_gcda_end_file() {
       unmap_file();
     }
 
-    fclose(output_file);
     flock(fd, LOCK_UN);
+    fclose(output_file);
     output_file = NULL;
     write_buffer = NULL;
   }
@@ -507,7 +485,7 @@ void llvm_register_writeout_function(writeout_fn fn) {
   }
 }
 
-void llvm_writeout_files() {
+void llvm_writeout_files(void) {
   struct writeout_fn_node *curr = writeout_fn_head;
 
   while (curr) {
@@ -516,7 +494,7 @@ void llvm_writeout_files() {
   }
 }
 
-void llvm_delete_writeout_function_list() {
+void llvm_delete_writeout_function_list(void) {
   while (writeout_fn_head) {
     struct writeout_fn_node *node = writeout_fn_head;
     writeout_fn_head = writeout_fn_head->next;
@@ -548,7 +526,7 @@ void __gcov_flush() {
   }
 }
 
-void llvm_delete_flush_function_list() {
+void llvm_delete_flush_function_list(void) {
   while (flush_fn_head) {
     struct flush_fn_node *node = flush_fn_head;
     flush_fn_head = flush_fn_head->next;
diff --git a/contrib/compiler-rt/lib/profile/InstrProfData.inc b/contrib/compiler-rt/lib/profile/InstrProfData.inc
index 33c7d94aea2a..93d14ac4f6f9 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfData.inc
+++ b/contrib/compiler-rt/lib/profile/InstrProfData.inc
@@ -57,6 +57,12 @@
  *
 \*===----------------------------------------------------------------------===*/
 
+/* Functions marked with INSTR_PROF_VISIBILITY must have hidden visibility in
+ * the compiler runtime. */
+#ifndef INSTR_PROF_VISIBILITY
+#define INSTR_PROF_VISIBILITY
+#endif
+
 /* INSTR_PROF_DATA start. */
 /* Definition of member fields of the per-function control structure. */
 #ifndef INSTR_PROF_DATA
@@ -64,29 +70,57 @@
 #else
 #define INSTR_PROF_DATA_DEFINED
 #endif
-
-INSTR_PROF_DATA(const uint32_t, llvm::Type::getInt32Ty(Ctx), NameSize, \
-                ConstantInt::get(llvm::Type::getInt32Ty(Ctx), \
-                NamePtr->getType()->getPointerElementType()->getArrayNumElements()))
-INSTR_PROF_DATA(const uint32_t, llvm::Type::getInt32Ty(Ctx), NumCounters, \
-                ConstantInt::get(llvm::Type::getInt32Ty(Ctx), NumCounters))
+INSTR_PROF_DATA(const uint64_t, llvm::Type::getInt64Ty(Ctx), NameRef, \
+                ConstantInt::get(llvm::Type::getInt64Ty(Ctx), \
+		IndexedInstrProf::ComputeHash(getPGOFuncNameVarInitializer(Inc->getName()))))
 INSTR_PROF_DATA(const uint64_t, llvm::Type::getInt64Ty(Ctx), FuncHash, \
                 ConstantInt::get(llvm::Type::getInt64Ty(Ctx), \
                 Inc->getHash()->getZExtValue()))
-INSTR_PROF_DATA(const IntPtrT, llvm::Type::getInt8PtrTy(Ctx), NamePtr, \
-                ConstantExpr::getBitCast(NamePtr, llvm::Type::getInt8PtrTy(Ctx)))
 INSTR_PROF_DATA(const IntPtrT, llvm::Type::getInt64PtrTy(Ctx), CounterPtr, \
                 ConstantExpr::getBitCast(CounterPtr, \
                 llvm::Type::getInt64PtrTy(Ctx)))
+/* This is used to map function pointers for the indirect call targets to
+ * function name hashes during the conversion from raw to merged profile
+ * data.
+ */
 INSTR_PROF_DATA(const IntPtrT, llvm::Type::getInt8PtrTy(Ctx), FunctionPointer, \
                 FunctionAddr)
 INSTR_PROF_DATA(IntPtrT, llvm::Type::getInt8PtrTy(Ctx), Values, \
-                ConstantPointerNull::get(Int8PtrTy))
+                ValuesPtrExpr)
+INSTR_PROF_DATA(const uint32_t, llvm::Type::getInt32Ty(Ctx), NumCounters, \
+                ConstantInt::get(llvm::Type::getInt32Ty(Ctx), NumCounters))
 INSTR_PROF_DATA(const uint16_t, Int16ArrayTy, NumValueSites[IPVK_Last+1], \
                 ConstantArray::get(Int16ArrayTy, Int16ArrayVals))
 #undef INSTR_PROF_DATA
 /* INSTR_PROF_DATA end. */
 
+
+/* This is an internal data structure used by value profiler. It
+ * is defined here to allow serialization code sharing by LLVM
+ * to be used in unit test.
+ *
+ * typedef struct ValueProfNode {
+ *   // InstrProfValueData VData;
+ *   uint64_t Value;
+ *   uint64_t Count;
+ *   struct ValueProfNode *Next;
+ * } ValueProfNode;
+ */
+/* INSTR_PROF_VALUE_NODE start. */
+#ifndef INSTR_PROF_VALUE_NODE
+#define INSTR_PROF_VALUE_NODE(Type, LLVMType, Name, Initializer)
+#else
+#define INSTR_PROF_DATA_DEFINED
+#endif
+INSTR_PROF_VALUE_NODE(uint64_t, llvm::Type::getInt64Ty(Ctx), Value, \
+                      ConstantInt::get(llvm::Type::GetInt64Ty(Ctx), 0))
+INSTR_PROF_VALUE_NODE(uint64_t, llvm::Type::getInt64Ty(Ctx), Count, \
+                      ConstantInt::get(llvm::Type::GetInt64Ty(Ctx), 0))
+INSTR_PROF_VALUE_NODE(PtrToNodeT, llvm::Type::getInt8PtrTy(Ctx), Next, \
+                      ConstantInt::get(llvm::Type::GetInt8PtrTy(Ctx), 0))
+#undef INSTR_PROF_VALUE_NODE
+/* INSTR_PROF_VALUE_NODE end. */
+
 /* INSTR_PROF_RAW_HEADER  start */
 /* Definition of member fields of the raw profile header data structure. */
 #ifndef INSTR_PROF_RAW_HEADER
@@ -102,8 +136,6 @@ INSTR_PROF_RAW_HEADER(uint64_t, NamesSize,  NamesSize)
 INSTR_PROF_RAW_HEADER(uint64_t, CountersDelta, (uintptr_t)CountersBegin)
 INSTR_PROF_RAW_HEADER(uint64_t, NamesDelta, (uintptr_t)NamesBegin)
 INSTR_PROF_RAW_HEADER(uint64_t, ValueKindLast, IPVK_Last)
-INSTR_PROF_RAW_HEADER(uint64_t, ValueDataSize, ValueDataSize)
-INSTR_PROF_RAW_HEADER(uint64_t, ValueDataDelta, (uintptr_t)ValueDataBegin)
 #undef INSTR_PROF_RAW_HEADER
 /* INSTR_PROF_RAW_HEADER  end */
 
@@ -132,6 +164,15 @@ VALUE_PROF_FUNC_PARAM(uint32_t, CounterIndex, Type::getInt32Ty(Ctx))
 #else
 #define INSTR_PROF_DATA_DEFINED
 #endif
+/* For indirect function call value profiling, the addresses of the target
+ * functions are profiled by the instrumented code. The target addresses are
+ * written in the raw profile data and converted to target function name's MD5
+ * hash by the profile reader during deserialization.  Typically, this happens
+ * when the the raw profile data is read during profile merging.
+ *
+ * For this remapping the ProfData is used.  ProfData contains both the function
+ * name hash and the function address.
+ */
 VALUE_PROF_KIND(IPVK_IndirectCallTarget, 0)
 /* These two kinds must be the last to be
  * declared. This is to make sure the string
@@ -153,12 +194,18 @@ VALUE_PROF_KIND(IPVK_Last, IPVK_IndirectCallTarget)
 #else
 #define INSTR_PROF_DATA_DEFINED
 #endif
+#ifdef COVMAP_V1
 COVMAP_FUNC_RECORD(const IntPtrT, llvm::Type::getInt8PtrTy(Ctx), \
                    NamePtr, llvm::ConstantExpr::getBitCast(NamePtr, \
                    llvm::Type::getInt8PtrTy(Ctx)))
 COVMAP_FUNC_RECORD(const uint32_t, llvm::Type::getInt32Ty(Ctx), NameSize, \
-                   llvm::ConstantInt::get(llvm::Type::getInt32Ty(Ctx),\
+                   llvm::ConstantInt::get(llvm::Type::getInt32Ty(Ctx), \
                    NameValue.size()))
+#else
+COVMAP_FUNC_RECORD(const int64_t, llvm::Type::getInt64Ty(Ctx), NameRef, \
+                   llvm::ConstantInt::get(llvm::Type::getInt64Ty(Ctx), \
+	           llvm::IndexedInstrProf::ComputeHash(NameValue)))
+#endif
 COVMAP_FUNC_RECORD(const uint32_t, llvm::Type::getInt32Ty(Ctx), DataSize, \
                    llvm::ConstantInt::get(llvm::Type::getInt32Ty(Ctx),\
                    CoverageMapping.size()))
@@ -182,7 +229,7 @@ COVMAP_HEADER(uint32_t, Int32Ty, FilenamesSize, \
 COVMAP_HEADER(uint32_t, Int32Ty, CoverageSize, \
               llvm::ConstantInt::get(Int32Ty, CoverageMappingSize))
 COVMAP_HEADER(uint32_t, Int32Ty, Version, \
-              llvm::ConstantInt::get(Int32Ty, CoverageMappingCurrentVersion))
+              llvm::ConstantInt::get(Int32Ty, CovMapVersion::CurrentVersion))
 #undef COVMAP_HEADER
 /* COVMAP_HEADER end.  */
 
@@ -281,16 +328,15 @@ typedef struct ValueProfData {
   static std::unique_ptr<ValueProfData>
   serializeFrom(const InstrProfRecord &Record);
   /*!
-   * Check the integrity of the record. Return the error code when
-   * an error is detected, otherwise return instrprof_error::success.
+   * Check the integrity of the record.
    */
-  instrprof_error checkIntegrity();
+  Error checkIntegrity();
   /*!
    * Return a pointer to \c ValueProfileData instance ready to be read.
    * All data in the instance are properly byte swapped. The input
    * data is assumed to be in little endian order.
    */
-  static ErrorOr<std::unique_ptr<ValueProfData>>
+  static Expected<std::unique_ptr<ValueProfData>>
   getValueProfData(const unsigned char *SrcBuffer,
                    const unsigned char *const SrcBufferEnd,
                    support::endianness SrcDataEndianness);
@@ -343,46 +389,18 @@ typedef struct ValueProfRecordClosure {
    */
   uint64_t (*RemapValueData)(uint32_t, uint64_t Value);
   void (*GetValueForSite)(const void *R, InstrProfValueData *Dst, uint32_t K,
-                          uint32_t S, uint64_t (*Mapper)(uint32_t, uint64_t));
+                          uint32_t S);
   ValueProfData *(*AllocValueProfData)(size_t TotalSizeInBytes);
 } ValueProfRecordClosure;
 
-/*
- * A wrapper struct that represents value profile runtime data.
- * Like InstrProfRecord class which is used by profiling host tools,
- * ValueProfRuntimeRecord also implements the abstract intefaces defined in
- * ValueProfRecordClosure so that the runtime data can be serialized using
- * shared C implementation. In this structure, NumValueSites and Nodes
- * members are the primary fields while other fields hold the derived
- * information for fast implementation of closure interfaces.
- */
-typedef struct ValueProfRuntimeRecord {
-  /* Number of sites for each value profile kind.  */
-  const uint16_t *NumValueSites;
-  /* An array of linked-list headers. The size of of the array is the
-   * total number of value profile sites : sum(NumValueSites[*])). Each
-   * linked-list stores the values profiled for a value profile site. */
-  ValueProfNode **Nodes;
-
-  /* Total number of value profile kinds which have at least one
-   *  value profile sites. */
-  uint32_t NumValueKinds;
-  /* An array recording the number of values tracked at each site.
-   * The size of the array is TotalNumValueSites. */
-  uint8_t *SiteCountArray[IPVK_Last + 1];
-  ValueProfNode **NodesKind[IPVK_Last + 1];
-} ValueProfRuntimeRecord;
-
-/* Forward declarations of C interfaces.  */
-int initializeValueProfRuntimeRecord(ValueProfRuntimeRecord *RuntimeRecord,
-                                     const uint16_t *NumValueSites,
-                                     ValueProfNode **Nodes);
-void finalizeValueProfRuntimeRecord(ValueProfRuntimeRecord *RuntimeRecord);
-uint32_t getValueProfDataSizeRT(const ValueProfRuntimeRecord *Record);
-ValueProfData *
-serializeValueProfDataFromRT(const ValueProfRuntimeRecord *Record,
-                             ValueProfData *Dst);
-uint32_t getNumValueKindsRT(const void *R);
+INSTR_PROF_VISIBILITY ValueProfRecord *
+getFirstValueProfRecord(ValueProfData *VPD);
+INSTR_PROF_VISIBILITY ValueProfRecord *
+getValueProfRecordNext(ValueProfRecord *VPR);
+INSTR_PROF_VISIBILITY InstrProfValueData *
+getValueProfRecordValueData(ValueProfRecord *VPR);
+INSTR_PROF_VISIBILITY uint32_t
+getValueProfRecordHeaderSize(uint32_t NumValueSites);
 
 #undef INSTR_PROF_VALUE_PROF_DATA
 #endif  /* INSTR_PROF_VALUE_PROF_DATA */
@@ -392,8 +410,10 @@ uint32_t getNumValueKindsRT(const void *R);
 #define INSTR_PROF_DATA_DEFINED
 #ifdef __cplusplus
 #define INSTR_PROF_INLINE inline
+#define INSTR_PROF_NULLPTR nullptr
 #else
 #define INSTR_PROF_INLINE
+#define INSTR_PROF_NULLPTR NULL
 #endif
 
 #ifndef offsetof
@@ -404,7 +424,7 @@ uint32_t getNumValueKindsRT(const void *R);
  * \brief Return the \c ValueProfRecord header size including the
  * padding bytes.
  */
-INSTR_PROF_INLINE
+INSTR_PROF_VISIBILITY INSTR_PROF_INLINE
 uint32_t getValueProfRecordHeaderSize(uint32_t NumValueSites) {
   uint32_t Size = offsetof(ValueProfRecord, SiteCountArray) +
                   sizeof(uint8_t) * NumValueSites;
@@ -417,7 +437,7 @@ uint32_t getValueProfRecordHeaderSize(uint32_t NumValueSites) {
  * \brief Return the total size of the value profile record including the
  * header and the value data.
  */
-INSTR_PROF_INLINE
+INSTR_PROF_VISIBILITY INSTR_PROF_INLINE
 uint32_t getValueProfRecordSize(uint32_t NumValueSites,
                                 uint32_t NumValueData) {
   return getValueProfRecordHeaderSize(NumValueSites) +
@@ -427,7 +447,7 @@ uint32_t getValueProfRecordSize(uint32_t NumValueSites,
 /*!
  * \brief Return the pointer to the start of value data array.
  */
-INSTR_PROF_INLINE
+INSTR_PROF_VISIBILITY INSTR_PROF_INLINE
 InstrProfValueData *getValueProfRecordValueData(ValueProfRecord *This) {
   return (InstrProfValueData *)((char *)This + getValueProfRecordHeaderSize(
                                                    This->NumValueSites));
@@ -436,7 +456,7 @@ InstrProfValueData *getValueProfRecordValueData(ValueProfRecord *This) {
 /*!
  * \brief Return the total number of value data for \c This record.
  */
-INSTR_PROF_INLINE
+INSTR_PROF_VISIBILITY INSTR_PROF_INLINE
 uint32_t getValueProfRecordNumValueData(ValueProfRecord *This) {
   uint32_t NumValueData = 0;
   uint32_t I;
@@ -448,7 +468,7 @@ uint32_t getValueProfRecordNumValueData(ValueProfRecord *This) {
 /*!
  * \brief Use this method to advance to the next \c This \c ValueProfRecord.
  */
-INSTR_PROF_INLINE
+INSTR_PROF_VISIBILITY INSTR_PROF_INLINE
 ValueProfRecord *getValueProfRecordNext(ValueProfRecord *This) {
   uint32_t NumValueData = getValueProfRecordNumValueData(This);
   return (ValueProfRecord *)((char *)This +
@@ -459,7 +479,7 @@ ValueProfRecord *getValueProfRecordNext(ValueProfRecord *This) {
 /*!
  * \brief Return the first \c ValueProfRecord instance.
  */
-INSTR_PROF_INLINE
+INSTR_PROF_VISIBILITY INSTR_PROF_INLINE
 ValueProfRecord *getFirstValueProfRecord(ValueProfData *This) {
   return (ValueProfRecord *)((char *)This + sizeof(ValueProfData));
 }
@@ -470,13 +490,11 @@ ValueProfRecord *getFirstValueProfRecord(ValueProfData *This) {
  * Return the total size in bytes of the on-disk value profile data
  * given the data stored in Record.
  */
-uint32_t getValueProfDataSize(ValueProfRecordClosure *Closure) {
+INSTR_PROF_VISIBILITY uint32_t
+getValueProfDataSize(ValueProfRecordClosure *Closure) {
   uint32_t Kind;
   uint32_t TotalSize = sizeof(ValueProfData);
   const void *Record = Closure->Record;
-  uint32_t NumValueKinds = Closure->GetNumValueKinds(Record);
-  if (NumValueKinds == 0)
-    return TotalSize;
 
   for (Kind = IPVK_First; Kind <= IPVK_Last; Kind++) {
     uint32_t NumValueSites = Closure->GetNumValueSites(Record, Kind);
@@ -492,9 +510,10 @@ uint32_t getValueProfDataSize(ValueProfRecordClosure *Closure) {
  * Extract value profile data of a function for the profile kind \c ValueKind
  * from the \c Closure and serialize the data into \c This record instance.
  */
-void serializeValueProfRecordFrom(ValueProfRecord *This,
-                                  ValueProfRecordClosure *Closure,
-                                  uint32_t ValueKind, uint32_t NumValueSites) {
+INSTR_PROF_VISIBILITY void
+serializeValueProfRecordFrom(ValueProfRecord *This,
+                             ValueProfRecordClosure *Closure,
+                             uint32_t ValueKind, uint32_t NumValueSites) {
   uint32_t S;
   const void *Record = Closure->Record;
   This->Kind = ValueKind;
@@ -504,8 +523,7 @@ void serializeValueProfRecordFrom(ValueProfRecord *This,
   for (S = 0; S < NumValueSites; S++) {
     uint32_t ND = Closure->GetNumValueDataForSite(Record, ValueKind, S);
     This->SiteCountArray[S] = ND;
-    Closure->GetValueForSite(Record, DstVD, ValueKind, S,
-                             Closure->RemapValueData);
+    Closure->GetValueForSite(Record, DstVD, ValueKind, S);
     DstVD += ND;
   }
 }
@@ -513,12 +531,16 @@ void serializeValueProfRecordFrom(ValueProfRecord *This,
 /*!
  * Extract value profile data of a function  from the \c Closure
  * and serialize the data into \c DstData if it is not NULL or heap
- * memory allocated by the \c Closure's allocator method.
+ * memory allocated by the \c Closure's allocator method. If \c
+ * DstData is not null, the caller is expected to set the TotalSize
+ * in DstData.
  */
-ValueProfData *serializeValueProfDataFrom(ValueProfRecordClosure *Closure,
-                                          ValueProfData *DstData) {
+INSTR_PROF_VISIBILITY ValueProfData *
+serializeValueProfDataFrom(ValueProfRecordClosure *Closure,
+                           ValueProfData *DstData) {
   uint32_t Kind;
-  uint32_t TotalSize = getValueProfDataSize(Closure);
+  uint32_t TotalSize =
+      DstData ? DstData->TotalSize : getValueProfDataSize(Closure);
 
   ValueProfData *VPD =
       DstData ? DstData : Closure->AllocValueProfData(TotalSize);
@@ -536,144 +558,15 @@ ValueProfData *serializeValueProfDataFrom(ValueProfRecordClosure *Closure,
   return VPD;
 }
 
-/*
- * The value profiler runtime library stores the value profile data
- * for a given function in \c NumValueSites and \c Nodes structures.
- * \c ValueProfRuntimeRecord class is used to encapsulate the runtime
- * profile data and provides fast interfaces to retrieve the profile
- * information. This interface is used to initialize the runtime record
- * and pre-compute the information needed for efficient implementation
- * of callbacks required by ValueProfRecordClosure class.
- */
-int initializeValueProfRuntimeRecord(ValueProfRuntimeRecord *RuntimeRecord,
-                                     const uint16_t *NumValueSites,
-                                     ValueProfNode **Nodes) {
-  unsigned I, J, S = 0, NumValueKinds = 0;
-  RuntimeRecord->NumValueSites = NumValueSites;
-  RuntimeRecord->Nodes = Nodes;
-  for (I = 0; I <= IPVK_Last; I++) {
-    uint16_t N = NumValueSites[I];
-    if (!N) {
-      RuntimeRecord->SiteCountArray[I] = 0;
-      continue;
-    }
-    NumValueKinds++;
-    RuntimeRecord->SiteCountArray[I] = (uint8_t *)calloc(N, 1);
-    if (!RuntimeRecord->SiteCountArray[I])
-      return 1;
-    RuntimeRecord->NodesKind[I] = Nodes ? &Nodes[S] : NULL;
-    for (J = 0; J < N; J++) {
-      /* Compute value count for each site. */
-      uint32_t C = 0;
-      ValueProfNode *Site = Nodes ? RuntimeRecord->NodesKind[I][J] : NULL;
-      while (Site) {
-        C++;
-        Site = Site->Next;
-      }
-      if (C > UCHAR_MAX)
-        C = UCHAR_MAX;
-      RuntimeRecord->SiteCountArray[I][J] = C;
-    }
-    S += N;
-  }
-  RuntimeRecord->NumValueKinds = NumValueKinds;
-  return 0;
-}
-
-void finalizeValueProfRuntimeRecord(ValueProfRuntimeRecord *RuntimeRecord) {
-  unsigned I;
-  for (I = 0; I <= IPVK_Last; I++) {
-    if (RuntimeRecord->SiteCountArray[I])
-      free(RuntimeRecord->SiteCountArray[I]);
-  }
-}
-
-/* ValueProfRecordClosure Interface implementation for
- * ValueProfDataRuntimeRecord.  */
-uint32_t getNumValueKindsRT(const void *R) {
-  return ((const ValueProfRuntimeRecord *)R)->NumValueKinds;
-}
-
-uint32_t getNumValueSitesRT(const void *R, uint32_t VK) {
-  return ((const ValueProfRuntimeRecord *)R)->NumValueSites[VK];
-}
-
-uint32_t getNumValueDataForSiteRT(const void *R, uint32_t VK, uint32_t S) {
-  const ValueProfRuntimeRecord *Record = (const ValueProfRuntimeRecord *)R;
-  return Record->SiteCountArray[VK][S];
-}
-
-uint32_t getNumValueDataRT(const void *R, uint32_t VK) {
-  unsigned I, S = 0;
-  const ValueProfRuntimeRecord *Record = (const ValueProfRuntimeRecord *)R;
-  if (Record->SiteCountArray[VK] == 0)
-    return 0;
-  for (I = 0; I < Record->NumValueSites[VK]; I++)
-    S += Record->SiteCountArray[VK][I];
-  return S;
-}
-
-void getValueForSiteRT(const void *R, InstrProfValueData *Dst, uint32_t VK,
-                       uint32_t S, uint64_t (*Mapper)(uint32_t, uint64_t)) {
-  unsigned I, N = 0;
-  const ValueProfRuntimeRecord *Record = (const ValueProfRuntimeRecord *)R;
-  N = getNumValueDataForSiteRT(R, VK, S);
-  if (N == 0)
-    return;
-  ValueProfNode *VNode = Record->NodesKind[VK][S];
-  for (I = 0; I < N; I++) {
-    Dst[I] = VNode->VData;
-    VNode = VNode->Next;
-  }
-}
-
-ValueProfData *allocValueProfDataRT(size_t TotalSizeInBytes) {
-  return (ValueProfData *)calloc(TotalSizeInBytes, 1);
-}
-
-static ValueProfRecordClosure RTRecordClosure = {0,
-                                                 getNumValueKindsRT,
-                                                 getNumValueSitesRT,
-                                                 getNumValueDataRT,
-                                                 getNumValueDataForSiteRT,
-                                                 0,
-                                                 getValueForSiteRT,
-                                                 allocValueProfDataRT};
-
-/*
- * Return the size of ValueProfData structure to store data
- * recorded in the runtime record.
- */
-uint32_t getValueProfDataSizeRT(const ValueProfRuntimeRecord *Record) {
-  RTRecordClosure.Record = Record;
-  return getValueProfDataSize(&RTRecordClosure);
-}
-
-/*
- * Return a ValueProfData instance that stores the data collected
- * from runtime. If \c DstData is provided by the caller, the value
- * profile data will be store in *DstData and DstData is returned,
- * otherwise the method will allocate space for the value data and
- * return pointer to the newly allocated space.
- */
-ValueProfData *
-serializeValueProfDataFromRT(const ValueProfRuntimeRecord *Record,
-                             ValueProfData *DstData) {
-  RTRecordClosure.Record = Record;
-  return serializeValueProfDataFrom(&RTRecordClosure, DstData);
-}
-
-
 #undef INSTR_PROF_COMMON_API_IMPL
 #endif /* INSTR_PROF_COMMON_API_IMPL */
 
 /*============================================================================*/
 
-
 #ifndef INSTR_PROF_DATA_DEFINED
 
-#ifndef INSTR_PROF_DATA_INC_
-#define INSTR_PROF_DATA_INC_
+#ifndef INSTR_PROF_DATA_INC
+#define INSTR_PROF_DATA_INC
 
 /* Helper macros.  */
 #define INSTR_PROF_SIMPLE_QUOTE(x) #x
@@ -695,23 +588,33 @@ serializeValueProfDataFromRT(const ValueProfRuntimeRecord *Record,
        (uint64_t)'p' << 40 | (uint64_t)'r' << 32 | (uint64_t)'o' << 24 |  \
         (uint64_t)'f' << 16 | (uint64_t)'R' << 8 | (uint64_t)129
 
-/* Raw profile format version. */
-#define INSTR_PROF_RAW_VERSION 2
-#define INSTR_PROF_INDEX_VERSION 3
-#define INSTR_PROF_COVMAP_VERSION 0
+/* Raw profile format version (start from 1). */
+#define INSTR_PROF_RAW_VERSION 4
+/* Indexed profile format version (start from 1). */
+#define INSTR_PROF_INDEX_VERSION 4
+/* Coverage mapping format vresion (start from 0). */
+#define INSTR_PROF_COVMAP_VERSION 1
 
-/* Profile version is always of type uint_64_t. Reserve the upper 8 bits in the
+/* Profile version is always of type uint64_t. Reserve the upper 8 bits in the
  * version for other variants of profile. We set the lowest bit of the upper 8
  * bits (i.e. bit 56) to 1 to indicate if this is an IR-level instrumentaiton
  * generated profile, and 0 if this is a Clang FE generated profile.
-*/
+ */
 #define VARIANT_MASKS_ALL 0xff00000000000000ULL
 #define GET_VERSION(V) ((V) & ~VARIANT_MASKS_ALL)
+#define VARIANT_MASK_IR_PROF (0x1ULL << 56)
+#define IR_LEVEL_PROF_VERSION_VAR __llvm_profile_raw_version
 
 /* Runtime section names and name strings.  */
 #define INSTR_PROF_DATA_SECT_NAME __llvm_prf_data
 #define INSTR_PROF_NAME_SECT_NAME __llvm_prf_names
 #define INSTR_PROF_CNTS_SECT_NAME __llvm_prf_cnts
+/* Array of pointers. Each pointer points to a list
+ * of value nodes associated with one value site.
+ */
+#define INSTR_PROF_VALS_SECT_NAME __llvm_prf_vals
+/* Value profile nodes section. */
+#define INSTR_PROF_VNODES_SECT_NAME __llvm_prf_vnds
 #define INSTR_PROF_COVMAP_SECT_NAME __llvm_covmap
 
 #define INSTR_PROF_DATA_SECT_NAME_STR                                          \
@@ -722,6 +625,10 @@ serializeValueProfDataFromRT(const ValueProfRuntimeRecord *Record,
   INSTR_PROF_QUOTE(INSTR_PROF_CNTS_SECT_NAME)
 #define INSTR_PROF_COVMAP_SECT_NAME_STR                                        \
   INSTR_PROF_QUOTE(INSTR_PROF_COVMAP_SECT_NAME)
+#define INSTR_PROF_VALS_SECT_NAME_STR                                          \
+  INSTR_PROF_QUOTE(INSTR_PROF_VALS_SECT_NAME)
+#define INSTR_PROF_VNODES_SECT_NAME_STR                                        \
+  INSTR_PROF_QUOTE(INSTR_PROF_VNODES_SECT_NAME)
 
 /* Macros to define start/stop section symbol for a given
  * section on Linux. For instance
@@ -751,16 +658,7 @@ typedef struct InstrProfValueData {
   uint64_t Count;
 } InstrProfValueData;
 
-/* This is an internal data structure used by value profiler. It
- * is defined here to allow serialization code sharing by LLVM
- * to be used in unit test.
- */
-typedef struct ValueProfNode {
-  InstrProfValueData VData;
-  struct ValueProfNode *Next;
-} ValueProfNode;
-
-#endif /* INSTR_PROF_DATA_INC_ */
+#endif /* INSTR_PROF_DATA_INC */
 
 #else
 #undef INSTR_PROF_DATA_DEFINED
diff --git a/contrib/compiler-rt/lib/profile/InstrProfiling.c b/contrib/compiler-rt/lib/profile/InstrProfiling.c
index 711f2b608a5f..c763a44233a0 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfiling.c
+++ b/contrib/compiler-rt/lib/profile/InstrProfiling.c
@@ -16,7 +16,7 @@
 #define INSTR_PROF_VALUE_PROF_DATA
 #include "InstrProfData.inc"
 
-char *(*GetEnvHook)(const char *) = 0;
+COMPILER_RT_VISIBILITY char *(*GetEnvHook)(const char *) = 0;
 
 COMPILER_RT_WEAK uint64_t __llvm_profile_raw_version = INSTR_PROF_RAW_VERSION;
 
@@ -46,7 +46,7 @@ COMPILER_RT_VISIBILITY void __llvm_profile_reset_counters(void) {
   const __llvm_profile_data *DataBegin = __llvm_profile_begin_data();
   const __llvm_profile_data *DataEnd = __llvm_profile_end_data();
   const __llvm_profile_data *DI;
-  for (DI = DataBegin; DI != DataEnd; ++DI) {
+  for (DI = DataBegin; DI < DataEnd; ++DI) {
     uint64_t CurrentVSiteCount = 0;
     uint32_t VKI, i;
     if (!DI->Values)
@@ -61,7 +61,7 @@ COMPILER_RT_VISIBILITY void __llvm_profile_reset_counters(void) {
       ValueProfNode *CurrentVNode = ValueCounters[i];
 
       while (CurrentVNode) {
-        CurrentVNode->VData.Count = 0;
+        CurrentVNode->Count = 0;
         CurrentVNode = CurrentVNode->Next;
       }
     }
diff --git a/contrib/compiler-rt/lib/profile/InstrProfiling.h b/contrib/compiler-rt/lib/profile/InstrProfiling.h
index d27ca569d535..b23bed8ea3a8 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfiling.h
+++ b/contrib/compiler-rt/lib/profile/InstrProfiling.h
@@ -11,6 +11,8 @@
 #define PROFILE_INSTRPROFILING_H_
 
 #include "InstrProfilingPort.h"
+
+#define INSTR_PROF_VISIBILITY COMPILER_RT_VISIBILITY
 #include "InstrProfData.inc"
 
 enum ValueKind {
@@ -30,6 +32,12 @@ typedef struct __llvm_profile_header {
 #include "InstrProfData.inc"
 } __llvm_profile_header;
 
+typedef struct ValueProfNode * PtrToNodeT;
+typedef struct ValueProfNode {
+#define INSTR_PROF_VALUE_NODE(Type, LLVMType, Name, Initializer) Type Name;
+#include "InstrProfData.inc"
+} ValueProfNode;
+
 /*!
  * \brief Get number of bytes necessary to pad the argument to eight
  * byte boundary.
@@ -55,6 +63,8 @@ const char *__llvm_profile_begin_names(void);
 const char *__llvm_profile_end_names(void);
 uint64_t *__llvm_profile_begin_counters(void);
 uint64_t *__llvm_profile_end_counters(void);
+ValueProfNode *__llvm_profile_begin_vnodes();
+ValueProfNode *__llvm_profile_end_vnodes();
 
 /*!
  * \brief Clear profile counters to zero.
@@ -63,6 +73,27 @@ uint64_t *__llvm_profile_end_counters(void);
 void __llvm_profile_reset_counters(void);
 
 /*!
+ * \brief Merge profile data from buffer.
+ *
+ * Read profile data form buffer \p Profile  and merge with
+ * in-process profile counters. The client is expected to
+ * have checked or already knows the profile data in the
+ * buffer matches the in-process counter structure before
+ * calling it.
+ */
+void __llvm_profile_merge_from_buffer(const char *Profile, uint64_t Size);
+
+/*! \brief Check if profile in buffer matches the current binary.
+ *
+ *  Returns 0 (success) if the profile data in buffer \p Profile with size
+ *  \p Size was generated by the same binary and therefore matches
+ *  structurally the in-process counters. If the profile data in buffer is
+ *  not compatible, the interface returns 1 (failure).
+ */
+int __llvm_profile_check_compatibility(const char *Profile,
+                                       uint64_t Size);
+
+/*!
  * \brief Counts the number of times a target value is seen.
  *
  * Records the target value for the CounterIndex if not seen before. Otherwise,
@@ -73,15 +104,7 @@ void __llvm_profile_reset_counters(void);
 void INSTR_PROF_VALUE_PROF_FUNC(
 #define VALUE_PROF_FUNC_PARAM(ArgType, ArgName, ArgLLVMType) ArgType ArgName
 #include "InstrProfData.inc"
-);
-
-/*!
- * \brief Prepares the value profiling data for output.
- *
- * Returns an array of pointers to value profile data.
- */
-struct ValueProfData;
-struct ValueProfData **__llvm_profile_gather_value_data(uint64_t *Size);
+    );
 
 /*!
  * \brief Write instrumentation data to the current file.
@@ -131,4 +154,31 @@ uint64_t __llvm_profile_get_magic(void);
 /*! \brief Get the version of the file format. */
 uint64_t __llvm_profile_get_version(void);
 
+/*! \brief Get the number of entries in the profile data section. */
+uint64_t __llvm_profile_get_data_size(const __llvm_profile_data *Begin,
+                                      const __llvm_profile_data *End);
+
+/*!
+ * This variable is defined in InstrProfilingRuntime.cc as a hidden
+ * symbol. Its main purpose is to enable profile runtime user to
+ * bypass runtime initialization code -- if the client code explicitly
+ * define this variable, then InstProfileRuntime.o won't be linked in.
+ * Note that this variable's visibility needs to be hidden so that the
+ * definition of this variable in an instrumented shared library won't
+ * affect runtime initialization decision of the main program.
+ */
+COMPILER_RT_VISIBILITY extern int __llvm_profile_runtime;
+
+/*!
+ * This variable is defined in InstrProfiling.c. Its main purpose is to
+ * encode the raw profile version value and other format related information
+ * such as whether the profile is from IR based instrumentation. The variable
+ * is defined as weak so that compiler can emit an overriding definition
+ * depending on user option.  Since we don't support mixing FE and IR based
+ * data in the same raw profile data file (in other words, shared libs and
+ * main program are expected to be instrumented in the same way), there is
+ * no need for this variable to be hidden.
+ */
+extern uint64_t __llvm_profile_raw_version;
+
 #endif /* PROFILE_INSTRPROFILING_H_ */
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingBuffer.c b/contrib/compiler-rt/lib/profile/InstrProfilingBuffer.c
index 4227ca6b66ea..ac259e83cbd2 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfilingBuffer.c
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingBuffer.c
@@ -23,7 +23,13 @@ uint64_t __llvm_profile_get_size_for_buffer(void) {
       DataBegin, DataEnd, CountersBegin, CountersEnd, NamesBegin, NamesEnd);
 }
 
-#define PROFILE_RANGE_SIZE(Range) (Range##End - Range##Begin)
+COMPILER_RT_VISIBILITY
+uint64_t __llvm_profile_get_data_size(const __llvm_profile_data *Begin,
+                                      const __llvm_profile_data *End) {
+  intptr_t BeginI = (intptr_t)Begin, EndI = (intptr_t)End;
+  return ((EndI + sizeof(__llvm_profile_data) - 1) - BeginI) /
+         sizeof(__llvm_profile_data);
+}
 
 COMPILER_RT_VISIBILITY
 uint64_t __llvm_profile_get_size_for_buffer_internal(
@@ -31,22 +37,23 @@ uint64_t __llvm_profile_get_size_for_buffer_internal(
     const uint64_t *CountersBegin, const uint64_t *CountersEnd,
     const char *NamesBegin, const char *NamesEnd) {
   /* Match logic in __llvm_profile_write_buffer(). */
-  const uint64_t NamesSize = PROFILE_RANGE_SIZE(Names) * sizeof(char);
+  const uint64_t NamesSize = (NamesEnd - NamesBegin) * sizeof(char);
   const uint8_t Padding = __llvm_profile_get_num_padding_bytes(NamesSize);
   return sizeof(__llvm_profile_header) +
-         PROFILE_RANGE_SIZE(Data) * sizeof(__llvm_profile_data) +
-         PROFILE_RANGE_SIZE(Counters) * sizeof(uint64_t) + NamesSize + Padding;
+         (__llvm_profile_get_data_size(DataBegin, DataEnd) *
+          sizeof(__llvm_profile_data)) +
+         (CountersEnd - CountersBegin) * sizeof(uint64_t) + NamesSize + Padding;
 }
 
 COMPILER_RT_VISIBILITY int __llvm_profile_write_buffer(char *Buffer) {
-  return llvmWriteProfData(llvmBufferWriter, Buffer, 0, 0);
+  return lprofWriteData(lprofBufferWriter, Buffer, 0);
 }
 
 COMPILER_RT_VISIBILITY int __llvm_profile_write_buffer_internal(
     char *Buffer, const __llvm_profile_data *DataBegin,
     const __llvm_profile_data *DataEnd, const uint64_t *CountersBegin,
     const uint64_t *CountersEnd, const char *NamesBegin, const char *NamesEnd) {
-  return llvmWriteProfDataImpl(llvmBufferWriter, Buffer, DataBegin, DataEnd,
-                               CountersBegin, CountersEnd, 0, 0, NamesBegin,
-                               NamesEnd);
+  return lprofWriteDataImpl(lprofBufferWriter, Buffer, DataBegin, DataEnd,
+                            CountersBegin, CountersEnd, 0, NamesBegin,
+                            NamesEnd);
 }
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingFile.c b/contrib/compiler-rt/lib/profile/InstrProfilingFile.c
index 68d088a1956a..32762d14eef2 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfilingFile.c
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingFile.c
@@ -14,12 +14,76 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
-
-#define UNCONST(ptr) ((void *)(uintptr_t)(ptr))
-
 #ifdef _MSC_VER
-#define snprintf _snprintf
+/* For _alloca. */
+#include <malloc.h>
+#endif
+#if defined(_WIN32)
+#include "WindowsMMap.h"
+/* For _chsize_s */
+#include <io.h>
+#else
+#include <sys/file.h>
+#include <sys/mman.h>
+#include <unistd.h>
+#if defined(__linux__)
+#include <sys/types.h>
 #endif
+#endif
+
+/* From where is profile name specified.
+ * The order the enumerators define their
+ * precedence. Re-order them may lead to
+ * runtime behavior change. */ 
+typedef enum ProfileNameSpecifier {
+  PNS_unknown = 0,
+  PNS_default,
+  PNS_command_line,
+  PNS_environment,
+  PNS_runtime_api
+} ProfileNameSpecifier;
+
+static const char *getPNSStr(ProfileNameSpecifier PNS) {
+  switch (PNS) {
+  case PNS_default:
+    return "default setting";
+  case PNS_command_line:
+    return "command line";
+  case PNS_environment:
+    return "environment variable";
+  case PNS_runtime_api:
+    return "runtime API";
+  default:
+    return "Unknown";
+  }
+}
+
+#define MAX_PID_SIZE 16
+/* Data structure holding the result of parsed filename pattern.  */
+typedef struct lprofFilename {
+  /* File name string possibly with %p or %h specifiers. */
+  const char *FilenamePat;
+  char PidChars[MAX_PID_SIZE];
+  char Hostname[COMPILER_RT_MAX_HOSTLEN];
+  unsigned NumPids;
+  unsigned NumHosts;
+  /* When in-process merging is enabled, this parameter specifies
+   * the total number of profile data files shared by all the processes
+   * spawned from the same binary. By default the value is 1. If merging
+   * is not enabled, its value should be 0. This parameter is specified
+   * by the %[0-9]m specifier. For instance %2m enables merging using
+   * 2 profile data files. %1m is equivalent to %m. Also %m specifier
+   * can only appear once at the end of the name pattern. */
+  unsigned MergePoolSize;
+  ProfileNameSpecifier PNS;
+} lprofFilename;
+
+lprofFilename lprofCurFilename = {0, {0}, {0}, 0, 0, 0, PNS_unknown};
+
+int getpid(void);
+static int getCurFilenameLength();
+static const char *getCurFilename(char *FilenameBuf);
+static unsigned doMerging() { return lprofCurFilename.MergePoolSize; }
 
 /* Return 1 if there is an error, otherwise return  0.  */
 static uint32_t fileWriter(ProfDataIOVec *IOVecs, uint32_t NumIOVecs,
@@ -35,59 +99,144 @@ static uint32_t fileWriter(ProfDataIOVec *IOVecs, uint32_t NumIOVecs,
 }
 
 COMPILER_RT_VISIBILITY ProfBufferIO *
-llvmCreateBufferIOInternal(void *File, uint32_t BufferSz) {
-  CallocHook = calloc;
-  FreeHook = free;
-  return llvmCreateBufferIO(fileWriter, File, BufferSz);
+lprofCreateBufferIOInternal(void *File, uint32_t BufferSz) {
+  FreeHook = &free;
+  DynamicBufferIOBuffer = (uint8_t *)calloc(BufferSz, 1);
+  VPBufferSize = BufferSz;
+  return lprofCreateBufferIO(fileWriter, File);
 }
 
-static int writeFile(FILE *File) {
+static void setupIOBuffer() {
   const char *BufferSzStr = 0;
-  uint64_t ValueDataSize = 0;
-  struct ValueProfData **ValueDataArray =
-      __llvm_profile_gather_value_data(&ValueDataSize);
-  FreeHook = &free;
-  CallocHook = &calloc;
   BufferSzStr = getenv("LLVM_VP_BUFFER_SIZE");
-  if (BufferSzStr && BufferSzStr[0])
+  if (BufferSzStr && BufferSzStr[0]) {
     VPBufferSize = atoi(BufferSzStr);
-  return llvmWriteProfData(fileWriter, File, ValueDataArray, ValueDataSize);
+    DynamicBufferIOBuffer = (uint8_t *)calloc(VPBufferSize, 1);
+  }
+}
+
+/* Read profile data in \c ProfileFile and merge with in-memory
+   profile counters. Returns -1 if there is fatal error, otheriwse
+   0 is returned.
+*/
+static int doProfileMerging(FILE *ProfileFile) {
+  uint64_t ProfileFileSize;
+  char *ProfileBuffer;
+
+  if (fseek(ProfileFile, 0L, SEEK_END) == -1) {
+    PROF_ERR("Unable to merge profile data, unable to get size: %s\n",
+             strerror(errno));
+    return -1;
+  }
+  ProfileFileSize = ftell(ProfileFile);
+
+  /* Restore file offset.  */
+  if (fseek(ProfileFile, 0L, SEEK_SET) == -1) {
+    PROF_ERR("Unable to merge profile data, unable to rewind: %s\n",
+             strerror(errno));
+    return -1;
+  }
+
+  /* Nothing to merge.  */
+  if (ProfileFileSize < sizeof(__llvm_profile_header)) {
+    if (ProfileFileSize)
+      PROF_WARN("Unable to merge profile data: %s\n",
+                "source profile file is too small.");
+    return 0;
+  }
+
+  ProfileBuffer = mmap(NULL, ProfileFileSize, PROT_READ, MAP_SHARED | MAP_FILE,
+                       fileno(ProfileFile), 0);
+  if (ProfileBuffer == MAP_FAILED) {
+    PROF_ERR("Unable to merge profile data, mmap failed: %s\n",
+             strerror(errno));
+    return -1;
+  }
+
+  if (__llvm_profile_check_compatibility(ProfileBuffer, ProfileFileSize)) {
+    (void)munmap(ProfileBuffer, ProfileFileSize);
+    PROF_WARN("Unable to merge profile data: %s\n",
+              "source profile file is not compatible.");
+    return 0;
+  }
+
+  /* Now start merging */
+  __llvm_profile_merge_from_buffer(ProfileBuffer, ProfileFileSize);
+  (void)munmap(ProfileBuffer, ProfileFileSize);
+
+  return 0;
 }
 
-static int writeFileWithName(const char *OutputName) {
+/* Open the profile data for merging. It opens the file in r+b mode with
+ * file locking.  If the file has content which is compatible with the
+ * current process, it also reads in the profile data in the file and merge
+ * it with in-memory counters. After the profile data is merged in memory,
+ * the original profile data is truncated and gets ready for the profile
+ * dumper. With profile merging enabled, each executable as well as any of
+ * its instrumented shared libraries dump profile data into their own data file.
+*/
+static FILE *openFileForMerging(const char *ProfileFileName) {
+  FILE *ProfileFile;
+  int rc;
+
+  ProfileFile = lprofOpenFileEx(ProfileFileName);
+  if (!ProfileFile)
+    return NULL;
+
+  rc = doProfileMerging(ProfileFile);
+  if (rc || COMPILER_RT_FTRUNCATE(ProfileFile, 0L) ||
+      fseek(ProfileFile, 0L, SEEK_SET) == -1) {
+    PROF_ERR("Profile Merging of file %s failed: %s\n", ProfileFileName,
+             strerror(errno));
+    fclose(ProfileFile);
+    return NULL;
+  }
+  fseek(ProfileFile, 0L, SEEK_SET);
+  return ProfileFile;
+}
+
+/* Write profile data to file \c OutputName.  */
+static int writeFile(const char *OutputName) {
   int RetVal;
   FILE *OutputFile;
-  if (!OutputName || !OutputName[0])
-    return -1;
 
-  /* Append to the file to support profiling multiple shared objects. */
-  OutputFile = fopen(OutputName, "ab");
+  if (!doMerging())
+    OutputFile = fopen(OutputName, "ab");
+  else
+    OutputFile = openFileForMerging(OutputName);
+
   if (!OutputFile)
     return -1;
 
-  RetVal = writeFile(OutputFile);
+  FreeHook = &free;
+  setupIOBuffer();
+  RetVal = lprofWriteData(fileWriter, OutputFile, lprofGetVPDataReader());
 
   fclose(OutputFile);
   return RetVal;
 }
 
-COMPILER_RT_WEAK int __llvm_profile_OwnsFilename = 0;
-COMPILER_RT_WEAK const char *__llvm_profile_CurrentFilename = NULL;
-
 static void truncateCurrentFile(void) {
   const char *Filename;
+  char *FilenameBuf;
   FILE *File;
+  int Length;
 
-  Filename = __llvm_profile_CurrentFilename;
-  if (!Filename || !Filename[0])
+  Length = getCurFilenameLength();
+  FilenameBuf = (char *)COMPILER_RT_ALLOCA(Length + 1);
+  Filename = getCurFilename(FilenameBuf);
+  if (!Filename)
     return;
 
   /* Create the directory holding the file, if needed. */
-  if (strchr(Filename, '/')) {
-    char *Copy = malloc(strlen(Filename) + 1);
-    strcpy(Copy, Filename);
+  if (strchr(Filename, DIR_SEPARATOR)
+#if defined(DIR_SEPARATOR_2)
+      || strchr(Filename, DIR_SEPARATOR_2)
+#endif
+          ) {
+    char *Copy = (char *)COMPILER_RT_ALLOCA(Length + 1);
+    strncpy(Copy, Filename, Length + 1);
     __llvm_profile_recursive_mkdir(Copy);
-    free(Copy);
   }
 
   /* Truncate the file.  Later we'll reopen and append. */
@@ -97,137 +246,254 @@ static void truncateCurrentFile(void) {
   fclose(File);
 }
 
-static void setFilename(const char *Filename, int OwnsFilename) {
-  /* Check if this is a new filename and therefore needs truncation. */
-  int NewFile = !__llvm_profile_CurrentFilename ||
-      (Filename && strcmp(Filename, __llvm_profile_CurrentFilename));
-  if (__llvm_profile_OwnsFilename)
-    free(UNCONST(__llvm_profile_CurrentFilename));
+static const char *DefaultProfileName = "default.profraw";
+static void resetFilenameToDefault(void) {
+  memset(&lprofCurFilename, 0, sizeof(lprofCurFilename));
+  lprofCurFilename.FilenamePat = DefaultProfileName;
+  lprofCurFilename.PNS = PNS_default;
+}
+
+static int containsMergeSpecifier(const char *FilenamePat, int I) {
+  return (FilenamePat[I] == 'm' ||
+          (FilenamePat[I] >= '1' && FilenamePat[I] <= '9' &&
+           /* If FilenamePat[I] is not '\0', the next byte is guaranteed
+            * to be in-bound as the string is null terminated. */
+           FilenamePat[I + 1] == 'm'));
+}
 
-  __llvm_profile_CurrentFilename = Filename;
-  __llvm_profile_OwnsFilename = OwnsFilename;
+/* Parses the pattern string \p FilenamePat and stores the result to
+ * lprofcurFilename structure. */
+static int parseFilenamePattern(const char *FilenamePat) {
+  int NumPids = 0, NumHosts = 0, I;
+  char *PidChars = &lprofCurFilename.PidChars[0];
+  char *Hostname = &lprofCurFilename.Hostname[0];
+  int MergingEnabled = 0;
 
-  /* If not a new file, append to support profiling multiple shared objects. */
-  if (NewFile)
-    truncateCurrentFile();
+  lprofCurFilename.FilenamePat = FilenamePat;
+  /* Check the filename for "%p", which indicates a pid-substitution. */
+  for (I = 0; FilenamePat[I]; ++I)
+    if (FilenamePat[I] == '%') {
+      if (FilenamePat[++I] == 'p') {
+        if (!NumPids++) {
+          if (snprintf(PidChars, MAX_PID_SIZE, "%d", getpid()) <= 0) {
+            PROF_WARN(
+                "Unable to parse filename pattern %s. Using the default name.",
+                FilenamePat);
+            return -1;
+          }
+        }
+      } else if (FilenamePat[I] == 'h') {
+        if (!NumHosts++)
+          if (COMPILER_RT_GETHOSTNAME(Hostname, COMPILER_RT_MAX_HOSTLEN)) {
+            PROF_WARN(
+                "Unable to parse filename pattern %s. Using the default name.",
+                FilenamePat);
+            return -1;
+          }
+      } else if (containsMergeSpecifier(FilenamePat, I)) {
+        if (MergingEnabled) {
+          PROF_WARN("%%m specifier can only be specified once in %s.\n",
+                    FilenamePat);
+          return -1;
+        }
+        MergingEnabled = 1;
+        if (FilenamePat[I] == 'm')
+          lprofCurFilename.MergePoolSize = 1;
+        else {
+          lprofCurFilename.MergePoolSize = FilenamePat[I] - '0';
+          I++; /* advance to 'm' */
+        }
+      }
+    }
+
+  lprofCurFilename.NumPids = NumPids;
+  lprofCurFilename.NumHosts = NumHosts;
+  return 0;
 }
 
-static void resetFilenameToDefault(void) { setFilename("default.profraw", 0); }
+static void parseAndSetFilename(const char *FilenamePat,
+                                ProfileNameSpecifier PNS) {
 
-int getpid(void);
-static int setFilenamePossiblyWithPid(const char *Filename) {
-#define MAX_PID_SIZE 16
-  char PidChars[MAX_PID_SIZE] = {0};
-  int NumPids = 0, PidLength = 0;
-  char *Allocated;
-  int I, J;
+  const char *OldFilenamePat = lprofCurFilename.FilenamePat;
+  ProfileNameSpecifier OldPNS = lprofCurFilename.PNS;
 
-  /* Reset filename on NULL, except with env var which is checked by caller. */
-  if (!Filename) {
+  if (PNS < OldPNS)
+    return;
+
+  if (!FilenamePat)
+    FilenamePat = DefaultProfileName;
+
+  /* When -fprofile-instr-generate=<path> is specified on the
+   * command line, each module will be instrumented with runtime
+   * init call to __llvm_profile_init function which calls
+   * __llvm_profile_override_default_filename. In most of the cases,
+   * the path will be identical, so bypass the parsing completely.
+   */
+  if (OldFilenamePat && !strcmp(OldFilenamePat, FilenamePat)) {
+    lprofCurFilename.PNS = PNS;
+    return;
+  }
+
+  /* When PNS >= OldPNS, the last one wins. */
+  if (!FilenamePat || parseFilenamePattern(FilenamePat))
     resetFilenameToDefault();
-    return 0;
+  lprofCurFilename.PNS = PNS;
+
+  if (!OldFilenamePat) {
+    PROF_NOTE("Set profile file path to \"%s\" via %s.\n",
+              lprofCurFilename.FilenamePat, getPNSStr(PNS));
+  } else {
+    PROF_NOTE("Override old profile path \"%s\" via %s to \"%s\" via %s.\n",
+              OldFilenamePat, getPNSStr(OldPNS), lprofCurFilename.FilenamePat,
+              getPNSStr(PNS));
   }
 
-  /* Check the filename for "%p", which indicates a pid-substitution. */
-  for (I = 0; Filename[I]; ++I)
-    if (Filename[I] == '%' && Filename[++I] == 'p')
-      if (!NumPids++) {
-        PidLength = snprintf(PidChars, MAX_PID_SIZE, "%d", getpid());
-        if (PidLength <= 0)
-          return -1;
-      }
-  if (!NumPids) {
-    setFilename(Filename, 0);
+  if (!lprofCurFilename.MergePoolSize)
+    truncateCurrentFile();
+}
+
+/* Return buffer length that is required to store the current profile
+ * filename with PID and hostname substitutions. */
+/* The length to hold uint64_t followed by 2 digit pool id including '_' */
+#define SIGLEN 24
+static int getCurFilenameLength() {
+  int Len;
+  if (!lprofCurFilename.FilenamePat || !lprofCurFilename.FilenamePat[0])
     return 0;
-  }
 
-  /* Allocate enough space for the substituted filename. */
-  Allocated = malloc(I + NumPids*(PidLength - 2) + 1);
-  if (!Allocated)
-    return -1;
+  if (!(lprofCurFilename.NumPids || lprofCurFilename.NumHosts ||
+        lprofCurFilename.MergePoolSize))
+    return strlen(lprofCurFilename.FilenamePat);
+
+  Len = strlen(lprofCurFilename.FilenamePat) +
+        lprofCurFilename.NumPids * (strlen(lprofCurFilename.PidChars) - 2) +
+        lprofCurFilename.NumHosts * (strlen(lprofCurFilename.Hostname) - 2);
+  if (lprofCurFilename.MergePoolSize)
+    Len += SIGLEN;
+  return Len;
+}
+
+/* Return the pointer to the current profile file name (after substituting
+ * PIDs and Hostnames in filename pattern. \p FilenameBuf is the buffer
+ * to store the resulting filename. If no substitution is needed, the
+ * current filename pattern string is directly returned. */
+static const char *getCurFilename(char *FilenameBuf) {
+  int I, J, PidLength, HostNameLength;
+  const char *FilenamePat = lprofCurFilename.FilenamePat;
+
+  if (!lprofCurFilename.FilenamePat || !lprofCurFilename.FilenamePat[0])
+    return 0;
 
+  if (!(lprofCurFilename.NumPids || lprofCurFilename.NumHosts ||
+        lprofCurFilename.MergePoolSize))
+    return lprofCurFilename.FilenamePat;
+
+  PidLength = strlen(lprofCurFilename.PidChars);
+  HostNameLength = strlen(lprofCurFilename.Hostname);
   /* Construct the new filename. */
-  for (I = 0, J = 0; Filename[I]; ++I)
-    if (Filename[I] == '%') {
-      if (Filename[++I] == 'p') {
-        memcpy(Allocated + J, PidChars, PidLength);
+  for (I = 0, J = 0; FilenamePat[I]; ++I)
+    if (FilenamePat[I] == '%') {
+      if (FilenamePat[++I] == 'p') {
+        memcpy(FilenameBuf + J, lprofCurFilename.PidChars, PidLength);
         J += PidLength;
+      } else if (FilenamePat[I] == 'h') {
+        memcpy(FilenameBuf + J, lprofCurFilename.Hostname, HostNameLength);
+        J += HostNameLength;
+      } else if (containsMergeSpecifier(FilenamePat, I)) {
+        char LoadModuleSignature[SIGLEN];
+        int S;
+        int ProfilePoolId = getpid() % lprofCurFilename.MergePoolSize;
+        S = snprintf(LoadModuleSignature, SIGLEN, "%" PRIu64 "_%d",
+                     lprofGetLoadModuleSignature(), ProfilePoolId);
+        if (S == -1 || S > SIGLEN)
+          S = SIGLEN;
+        memcpy(FilenameBuf + J, LoadModuleSignature, S);
+        J += S;
+        if (FilenamePat[I] != 'm')
+          I++;
       }
       /* Drop any unknown substitutions. */
     } else
-      Allocated[J++] = Filename[I];
-  Allocated[J] = 0;
+      FilenameBuf[J++] = FilenamePat[I];
+  FilenameBuf[J] = 0;
 
-  /* Use the computed name. */
-  setFilename(Allocated, 1);
-  return 0;
+  return FilenameBuf;
 }
 
-static int setFilenameFromEnvironment(void) {
+/* Returns the pointer to the environment variable
+ * string. Returns null if the env var is not set. */
+static const char *getFilenamePatFromEnv(void) {
   const char *Filename = getenv("LLVM_PROFILE_FILE");
-
   if (!Filename || !Filename[0])
-    return -1;
-
-  return setFilenamePossiblyWithPid(Filename);
-}
-
-static void setFilenameAutomatically(void) {
-  if (!setFilenameFromEnvironment())
-    return;
-
-  resetFilenameToDefault();
+    return 0;
+  return Filename;
 }
 
+/* This method is invoked by the runtime initialization hook
+ * InstrProfilingRuntime.o if it is linked in. Both user specified
+ * profile path via -fprofile-instr-generate= and LLVM_PROFILE_FILE
+ * environment variable can override this default value. */
 COMPILER_RT_VISIBILITY
 void __llvm_profile_initialize_file(void) {
-  /* Check if the filename has been initialized. */
-  if (__llvm_profile_CurrentFilename)
-    return;
+  const char *FilenamePat;
 
-  /* Detect the filename and truncate. */
-  setFilenameAutomatically();
+  FilenamePat = getFilenamePatFromEnv();
+  parseAndSetFilename(FilenamePat, FilenamePat ? PNS_environment : PNS_default);
 }
 
+/* This API is directly called by the user application code. It has the
+ * highest precedence compared with LLVM_PROFILE_FILE environment variable
+ * and command line option -fprofile-instr-generate=<profile_name>.
+ */
 COMPILER_RT_VISIBILITY
-void __llvm_profile_set_filename(const char *Filename) {
-  setFilenamePossiblyWithPid(Filename);
+void __llvm_profile_set_filename(const char *FilenamePat) {
+  parseAndSetFilename(FilenamePat, PNS_runtime_api);
 }
 
+/*
+ * This API is invoked by the global initializers emitted by Clang/LLVM when
+ * -fprofile-instr-generate=<..> is specified (vs -fprofile-instr-generate
+ *  without an argument). This option has lower precedence than the
+ *  LLVM_PROFILE_FILE environment variable.
+ */
 COMPILER_RT_VISIBILITY
-void __llvm_profile_override_default_filename(const char *Filename) {
-  /* If the env var is set, skip setting filename from argument. */
-  const char *Env_Filename = getenv("LLVM_PROFILE_FILE");
-  if (Env_Filename && Env_Filename[0])
-    return;
-  setFilenamePossiblyWithPid(Filename);
+void __llvm_profile_override_default_filename(const char *FilenamePat) {
+  parseAndSetFilename(FilenamePat, PNS_command_line);
 }
 
+/* The public API for writing profile data into the file with name
+ * set by previous calls to __llvm_profile_set_filename or
+ * __llvm_profile_override_default_filename or
+ * __llvm_profile_initialize_file. */
 COMPILER_RT_VISIBILITY
 int __llvm_profile_write_file(void) {
-  int rc;
+  int rc, Length;
+  const char *Filename;
+  char *FilenameBuf;
+
+  Length = getCurFilenameLength();
+  FilenameBuf = (char *)COMPILER_RT_ALLOCA(Length + 1);
+  Filename = getCurFilename(FilenameBuf);
 
-  GetEnvHook = &getenv;
   /* Check the filename. */
-  if (!__llvm_profile_CurrentFilename) {
-    PROF_ERR("LLVM Profile: Failed to write file : %s\n", "Filename not set");
+  if (!Filename) {
+    PROF_ERR("Failed to write file : %s\n", "Filename not set");
     return -1;
   }
 
   /* Check if there is llvm/runtime version mismatch.  */
   if (GET_VERSION(__llvm_profile_get_version()) != INSTR_PROF_RAW_VERSION) {
-    PROF_ERR("LLVM Profile: runtime and instrumentation version mismatch : "
+    PROF_ERR("Runtime and instrumentation version mismatch : "
              "expected %d, but get %d\n",
              INSTR_PROF_RAW_VERSION,
              (int)GET_VERSION(__llvm_profile_get_version()));
     return -1;
   }
 
-  /* Write the file. */
-  rc = writeFileWithName(__llvm_profile_CurrentFilename);
+  /* Write profile data to the file. */
+  rc = writeFile(Filename);
   if (rc)
-    PROF_ERR("LLVM Profile: Failed to write file \"%s\": %s\n",
-            __llvm_profile_CurrentFilename, strerror(errno));
+    PROF_ERR("Failed to write file \"%s\": %s\n", Filename, strerror(errno));
   return rc;
 }
 
@@ -240,6 +506,8 @@ int __llvm_profile_register_write_file_atexit(void) {
   if (HasBeenRegistered)
     return 0;
 
+  lprofSetupValueProfiler();
+
   HasBeenRegistered = 1;
   return atexit(writeFileWithoutReturn);
 }
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingInternal.h b/contrib/compiler-rt/lib/profile/InstrProfilingInternal.h
index 4aab78ea509c..44f308206ca8 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfilingInternal.h
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingInternal.h
@@ -68,51 +68,123 @@ typedef struct ProfBufferIO {
 } ProfBufferIO;
 
 /* The creator interface used by testing.  */
-ProfBufferIO *llvmCreateBufferIOInternal(void *File, uint32_t DefaultBufferSz);
+ProfBufferIO *lprofCreateBufferIOInternal(void *File, uint32_t BufferSz);
+
 /*!
  * This is the interface to create a handle for buffered IO.
  */
-ProfBufferIO *llvmCreateBufferIO(WriterCallback FileWriter, void *File,
-                                 uint32_t DefaultBufferSz);
+ProfBufferIO *lprofCreateBufferIO(WriterCallback FileWriter, void *File);
+
 /*!
  * The interface to destroy the bufferIO handle and reclaim
  * the memory.
  */
-void llvmDeleteBufferIO(ProfBufferIO *BufferIO);
+void lprofDeleteBufferIO(ProfBufferIO *BufferIO);
 
 /*!
  * This is the interface to write \c Data of \c Size bytes through
  * \c BufferIO. Returns 0 if successful, otherwise return -1.
  */
-int llvmBufferIOWrite(ProfBufferIO *BufferIO, const uint8_t *Data,
-                      uint32_t Size);
+int lprofBufferIOWrite(ProfBufferIO *BufferIO, const uint8_t *Data,
+                       uint32_t Size);
 /*!
  * The interface to flush the remaining data in the buffer.
  * through the low level writer callback.
  */
-int llvmBufferIOFlush(ProfBufferIO *BufferIO);
+int lprofBufferIOFlush(ProfBufferIO *BufferIO);
 
 /* The low level interface to write data into a buffer. It is used as the
  * callback by other high level writer methods such as buffered IO writer
  * and profile data writer.  */
-uint32_t llvmBufferWriter(ProfDataIOVec *IOVecs, uint32_t NumIOVecs,
-                          void **WriterCtx);
-
-int llvmWriteProfData(WriterCallback Writer, void *WriterCtx,
-                      struct ValueProfData **ValueDataArray,
-                      const uint64_t ValueDataSize);
-int llvmWriteProfDataImpl(WriterCallback Writer, void *WriterCtx,
-                          const __llvm_profile_data *DataBegin,
-                          const __llvm_profile_data *DataEnd,
-                          const uint64_t *CountersBegin,
-                          const uint64_t *CountersEnd,
-                          struct ValueProfData **ValueDataBeginArray,
-                          const uint64_t ValueDataSize, const char *NamesBegin,
-                          const char *NamesEnd);
-
-extern char *(*GetEnvHook)(const char *);
-extern void (*FreeHook)(void *);
-extern void* (*CallocHook)(size_t, size_t);
-extern uint32_t VPBufferSize;
+uint32_t lprofBufferWriter(ProfDataIOVec *IOVecs, uint32_t NumIOVecs,
+                           void **WriterCtx);
+
+struct ValueProfData;
+struct ValueProfRecord;
+struct InstrProfValueData;
+struct ValueProfNode;
+
+/*!
+ * The class that defines a set of methods to read value profile
+ * data for streaming/serialization from the instrumentation runtime.
+ */
+typedef struct VPDataReaderType {
+  uint32_t (*InitRTRecord)(const __llvm_profile_data *Data,
+                           uint8_t *SiteCountArray[]);
+  /* Function pointer to getValueProfRecordHeader method. */
+  uint32_t (*GetValueProfRecordHeaderSize)(uint32_t NumSites);
+  /* Function pointer to getFristValueProfRecord method. */  
+  struct ValueProfRecord *(*GetFirstValueProfRecord)(struct ValueProfData *);
+  /* Return the number of value data for site \p Site.  */
+  uint32_t (*GetNumValueDataForSite)(uint32_t VK, uint32_t Site);
+  /* Return the total size of the value profile data of the 
+   * current function.  */
+  uint32_t (*GetValueProfDataSize)(void);
+  /*! 
+   * Read the next \p N value data for site \p Site and store the data
+   * in \p Dst. \p StartNode is the first value node to start with if
+   * it is not null. The function returns the pointer to the value
+   * node pointer to be used as the \p StartNode of the next batch reading.
+   * If there is nothing left, it returns NULL.
+   */
+  struct ValueProfNode *(*GetValueData)(uint32_t ValueKind, uint32_t Site,
+                                        struct InstrProfValueData *Dst,
+                                        struct ValueProfNode *StartNode,
+                                        uint32_t N);
+} VPDataReaderType;
+
+int lprofWriteData(WriterCallback Writer, void *WriterCtx,
+                   VPDataReaderType *VPDataReader);
+int lprofWriteDataImpl(WriterCallback Writer, void *WriterCtx,
+                       const __llvm_profile_data *DataBegin,
+                       const __llvm_profile_data *DataEnd,
+                       const uint64_t *CountersBegin,
+                       const uint64_t *CountersEnd,
+                       VPDataReaderType *VPDataReader, const char *NamesBegin,
+                       const char *NamesEnd);
+
+/* Merge value profile data pointed to by SrcValueProfData into
+ * in-memory profile counters pointed by to DstData.  */
+void lprofMergeValueProfData(struct ValueProfData *SrcValueProfData,
+                             __llvm_profile_data *DstData);
+
+VPDataReaderType *lprofGetVPDataReader();
+
+/* Internal interface used by test to reset the max number of 
+ * tracked values per value site to be \p MaxVals.
+ */
+void lprofSetMaxValsPerSite(uint32_t MaxVals);
+void lprofSetupValueProfiler();
+
+/* Return the profile header 'signature' value associated with the current
+ * executable or shared library. The signature value can be used to for
+ * a profile name that is unique to this load module so that it does not
+ * collide with profiles from other binaries. It also allows shared libraries
+ * to dump merged profile data into its own profile file. */
+uint64_t lprofGetLoadModuleSignature();
+
+/* GCOV_PREFIX and GCOV_PREFIX_STRIP support */
+/* Return the path prefix specified by GCOV_PREFIX environment variable.
+ * If GCOV_PREFIX_STRIP is also specified, the strip level (integer value)
+ * is returned via \c *PrefixStrip. The prefix length is stored in *PrefixLen.
+ */
+const char *lprofGetPathPrefix(int *PrefixStrip, size_t *PrefixLen);
+/* Apply the path prefix specified in \c Prefix to path string in \c PathStr,
+ * and store the result to buffer pointed to by \c Buffer. If \c PrefixStrip
+ * is not zero, path prefixes are stripped from \c PathStr (the level of
+ * stripping is specified by \c PrefixStrip) before \c Prefix is added.
+ */
+void lprofApplyPathPrefix(char *Dest, const char *PathStr, const char *Prefix,
+                          size_t PrefixLen, int PrefixStrip);
+
+COMPILER_RT_VISIBILITY extern char *(*GetEnvHook)(const char *);
+COMPILER_RT_VISIBILITY extern void (*FreeHook)(void *);
+COMPILER_RT_VISIBILITY extern uint8_t *DynamicBufferIOBuffer;
+COMPILER_RT_VISIBILITY extern uint32_t VPBufferSize;
+COMPILER_RT_VISIBILITY extern uint32_t VPMaxNumValsPerSite;
+/* Pointer to the start of static value counters to be allocted. */
+COMPILER_RT_VISIBILITY extern ValueProfNode *CurrentVNode;
+COMPILER_RT_VISIBILITY extern ValueProfNode *EndVNode;
+extern void (*VPMergeHook)(struct ValueProfData *, __llvm_profile_data *);
 
 #endif
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingMerge.c b/contrib/compiler-rt/lib/profile/InstrProfilingMerge.c
new file mode 100644
index 000000000000..a2021154df73
--- /dev/null
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingMerge.c
@@ -0,0 +1,132 @@
+/*===- InstrProfilingMerge.c - Profile in-process Merging  ---------------===*\
+|*
+|*                     The LLVM Compiler Infrastructure
+|*
+|* This file is distributed under the University of Illinois Open Source
+|* License. See LICENSE.TXT for details.
+|*
+|*===----------------------------------------------------------------------===*
+|* This file defines the API needed for in-process merging of profile data
+|* stored in memory buffer.
+\*===---------------------------------------------------------------------===*/
+
+#include "InstrProfiling.h"
+#include "InstrProfilingInternal.h"
+#include "InstrProfilingUtil.h"
+
+#define INSTR_PROF_VALUE_PROF_DATA
+#include "InstrProfData.inc"
+
+COMPILER_RT_WEAK void (*VPMergeHook)(ValueProfData *,
+                                     __llvm_profile_data *) = NULL;
+COMPILER_RT_VISIBILITY
+uint64_t lprofGetLoadModuleSignature() {
+  /* A very fast way to compute a module signature.  */
+  uint64_t CounterSize = (uint64_t)(__llvm_profile_end_counters() -
+                                    __llvm_profile_begin_counters());
+  uint64_t DataSize = __llvm_profile_get_data_size(__llvm_profile_begin_data(),
+                                                   __llvm_profile_end_data());
+  uint64_t NamesSize =
+      (uint64_t)(__llvm_profile_end_names() - __llvm_profile_begin_names());
+  uint64_t NumVnodes =
+      (uint64_t)(__llvm_profile_end_vnodes() - __llvm_profile_begin_vnodes());
+  const __llvm_profile_data *FirstD = __llvm_profile_begin_data();
+
+  return (NamesSize << 40) + (CounterSize << 30) + (DataSize << 20) +
+         (NumVnodes << 10) + (DataSize > 0 ? FirstD->NameRef : 0);
+}
+
+/* Returns 1 if profile is not structurally compatible.  */
+COMPILER_RT_VISIBILITY
+int __llvm_profile_check_compatibility(const char *ProfileData,
+                                       uint64_t ProfileSize) {
+  /* Check profile header only for now  */
+  __llvm_profile_header *Header = (__llvm_profile_header *)ProfileData;
+  __llvm_profile_data *SrcDataStart, *SrcDataEnd, *SrcData, *DstData;
+  SrcDataStart =
+      (__llvm_profile_data *)(ProfileData + sizeof(__llvm_profile_header));
+  SrcDataEnd = SrcDataStart + Header->DataSize;
+
+  if (ProfileSize < sizeof(__llvm_profile_header))
+    return 1;
+
+  /* Check the header first.  */
+  if (Header->Magic != __llvm_profile_get_magic() ||
+      Header->Version != __llvm_profile_get_version() ||
+      Header->DataSize !=
+          __llvm_profile_get_data_size(__llvm_profile_begin_data(),
+                                       __llvm_profile_end_data()) ||
+      Header->CountersSize != (uint64_t)(__llvm_profile_end_counters() -
+                                         __llvm_profile_begin_counters()) ||
+      Header->NamesSize != (uint64_t)(__llvm_profile_end_names() -
+                                      __llvm_profile_begin_names()) ||
+      Header->ValueKindLast != IPVK_Last)
+    return 1;
+
+  if (ProfileSize < sizeof(__llvm_profile_header) +
+                        Header->DataSize * sizeof(__llvm_profile_data) +
+                        Header->NamesSize + Header->CountersSize)
+    return 1;
+
+  for (SrcData = SrcDataStart,
+       DstData = (__llvm_profile_data *)__llvm_profile_begin_data();
+       SrcData < SrcDataEnd; ++SrcData, ++DstData) {
+    if (SrcData->NameRef != DstData->NameRef ||
+        SrcData->FuncHash != DstData->FuncHash ||
+        SrcData->NumCounters != DstData->NumCounters)
+      return 1;
+  }
+
+  /* Matched! */
+  return 0;
+}
+
+COMPILER_RT_VISIBILITY
+void __llvm_profile_merge_from_buffer(const char *ProfileData,
+                                      uint64_t ProfileSize) {
+  __llvm_profile_data *SrcDataStart, *SrcDataEnd, *SrcData, *DstData;
+  __llvm_profile_header *Header = (__llvm_profile_header *)ProfileData;
+  uint64_t *SrcCountersStart;
+  const char *SrcNameStart;
+  ValueProfData *SrcValueProfDataStart, *SrcValueProfData;
+
+  SrcDataStart =
+      (__llvm_profile_data *)(ProfileData + sizeof(__llvm_profile_header));
+  SrcDataEnd = SrcDataStart + Header->DataSize;
+  SrcCountersStart = (uint64_t *)SrcDataEnd;
+  SrcNameStart = (const char *)(SrcCountersStart + Header->CountersSize);
+  SrcValueProfDataStart =
+      (ValueProfData *)(SrcNameStart + Header->NamesSize +
+                        __llvm_profile_get_num_padding_bytes(
+                            Header->NamesSize));
+
+  for (SrcData = SrcDataStart,
+      DstData = (__llvm_profile_data *)__llvm_profile_begin_data(),
+      SrcValueProfData = SrcValueProfDataStart;
+       SrcData < SrcDataEnd; ++SrcData, ++DstData) {
+    uint64_t *SrcCounters;
+    uint64_t *DstCounters = (uint64_t *)DstData->CounterPtr;
+    unsigned I, NC, NVK = 0;
+
+    NC = SrcData->NumCounters;
+    SrcCounters = SrcCountersStart +
+                  ((size_t)SrcData->CounterPtr - Header->CountersDelta) /
+                      sizeof(uint64_t);
+    for (I = 0; I < NC; I++)
+      DstCounters[I] += SrcCounters[I];
+
+    /* Now merge value profile data.  */
+    if (!VPMergeHook)
+      continue;
+
+    for (I = 0; I <= IPVK_Last; I++)
+      NVK += (SrcData->NumValueSites[I] != 0);
+
+    if (!NVK)
+      continue;
+
+    VPMergeHook(SrcValueProfData, DstData);
+    SrcValueProfData = (ValueProfData *)((char *)SrcValueProfData +
+                                         SrcValueProfData->TotalSize);
+  }
+}
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingMergeFile.c b/contrib/compiler-rt/lib/profile/InstrProfilingMergeFile.c
new file mode 100644
index 000000000000..ac5ee9fbedc6
--- /dev/null
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingMergeFile.c
@@ -0,0 +1,41 @@
+/*===- InstrProfilingMergeFile.c - Profile in-process Merging  ------------===*\
+|*
+|*                     The LLVM Compiler Infrastructure
+|*
+|* This file is distributed under the University of Illinois Open Source
+|* License. See LICENSE.TXT for details.
+|*
+|*===----------------------------------------------------------------------===
+|* This file defines APIs needed to support in-process merging for profile data
+|* stored in files.
+\*===----------------------------------------------------------------------===*/
+
+#include "InstrProfiling.h"
+#include "InstrProfilingInternal.h"
+#include "InstrProfilingUtil.h"
+
+#define INSTR_PROF_VALUE_PROF_DATA
+#include "InstrProfData.inc"
+
+void (*VPMergeHook)(ValueProfData *,
+                    __llvm_profile_data *) = &lprofMergeValueProfData;
+
+/* Merge value profile data pointed to by SrcValueProfData into
+ * in-memory profile counters pointed by to DstData.  */
+void lprofMergeValueProfData(ValueProfData *SrcValueProfData,
+                             __llvm_profile_data *DstData) {
+  unsigned I, S, V, C;
+  InstrProfValueData *VData;
+  ValueProfRecord *VR = getFirstValueProfRecord(SrcValueProfData);
+  for (I = 0; I < SrcValueProfData->NumValueKinds; I++) {
+    VData = getValueProfRecordValueData(VR);
+    for (S = 0; S < VR->NumValueSites; S++) {
+      uint8_t NV = VR->SiteCountArray[S];
+      for (V = 0; V < NV; V++) {
+        for (C = 0; C < VData[V].Count; C++)
+          __llvm_profile_instrument_target(VData[V].Value, DstData, S);
+      }
+    }
+    VR = getValueProfRecordNext(VR);
+  }
+}
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingPlatformDarwin.c b/contrib/compiler-rt/lib/profile/InstrProfilingPlatformDarwin.c
index 30ddbd2e4982..8931abaddf7f 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfilingPlatformDarwin.c
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingPlatformDarwin.c
@@ -30,6 +30,13 @@ extern uint64_t
     CountersEnd __asm("section$end$__DATA$" INSTR_PROF_CNTS_SECT_NAME_STR);
 
 COMPILER_RT_VISIBILITY
+extern ValueProfNode
+    VNodesStart __asm("section$start$__DATA$" INSTR_PROF_VNODES_SECT_NAME_STR);
+COMPILER_RT_VISIBILITY
+extern ValueProfNode
+    VNodesEnd __asm("section$end$__DATA$" INSTR_PROF_VNODES_SECT_NAME_STR);
+
+COMPILER_RT_VISIBILITY
 const __llvm_profile_data *__llvm_profile_begin_data(void) {
   return &DataStart;
 }
@@ -43,4 +50,14 @@ COMPILER_RT_VISIBILITY
 uint64_t *__llvm_profile_begin_counters(void) { return &CountersStart; }
 COMPILER_RT_VISIBILITY
 uint64_t *__llvm_profile_end_counters(void) { return &CountersEnd; }
+
+COMPILER_RT_VISIBILITY
+ValueProfNode *__llvm_profile_begin_vnodes(void) {
+  return &VNodesStart;
+}
+COMPILER_RT_VISIBILITY
+ValueProfNode *__llvm_profile_end_vnodes(void) { return &VNodesEnd; }
+
+COMPILER_RT_VISIBILITY ValueProfNode *CurrentVNode = &VNodesStart;
+COMPILER_RT_VISIBILITY ValueProfNode *EndVNode = &VNodesEnd;
 #endif
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingPlatformLinux.c b/contrib/compiler-rt/lib/profile/InstrProfilingPlatformLinux.c
index 7843f47caa1b..b6c780ff514f 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfilingPlatformLinux.c
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingPlatformLinux.c
@@ -18,6 +18,8 @@
 #define PROF_NAME_STOP INSTR_PROF_SECT_STOP(INSTR_PROF_NAME_SECT_NAME)
 #define PROF_CNTS_START INSTR_PROF_SECT_START(INSTR_PROF_CNTS_SECT_NAME)
 #define PROF_CNTS_STOP INSTR_PROF_SECT_STOP(INSTR_PROF_CNTS_SECT_NAME)
+#define PROF_VNODES_START INSTR_PROF_SECT_START(INSTR_PROF_VNODES_SECT_NAME)
+#define PROF_VNODES_STOP INSTR_PROF_SECT_STOP(INSTR_PROF_VNODES_SECT_NAME)
 
 /* Declare section start and stop symbols for various sections
  * generated by compiler instrumentation.
@@ -28,6 +30,8 @@ extern uint64_t PROF_CNTS_START COMPILER_RT_VISIBILITY;
 extern uint64_t PROF_CNTS_STOP COMPILER_RT_VISIBILITY;
 extern char PROF_NAME_START COMPILER_RT_VISIBILITY;
 extern char PROF_NAME_STOP COMPILER_RT_VISIBILITY;
+extern ValueProfNode PROF_VNODES_START COMPILER_RT_VISIBILITY;
+extern ValueProfNode PROF_VNODES_STOP COMPILER_RT_VISIBILITY;
 
 /* Add dummy data to ensure the section is always created. */
 __llvm_profile_data
@@ -35,6 +39,7 @@ __llvm_profile_data
 uint64_t
     __prof_cnts_sect_data[0] COMPILER_RT_SECTION(INSTR_PROF_CNTS_SECT_NAME_STR);
 char __prof_nms_sect_data[0] COMPILER_RT_SECTION(INSTR_PROF_NAME_SECT_NAME_STR);
+ValueProfNode __prof_vnodes_sect_data[0] COMPILER_RT_SECTION(INSTR_PROF_VNODES_SECT_NAME_STR);
 
 COMPILER_RT_VISIBILITY const __llvm_profile_data *
 __llvm_profile_begin_data(void) {
@@ -56,4 +61,15 @@ COMPILER_RT_VISIBILITY uint64_t *__llvm_profile_begin_counters(void) {
 COMPILER_RT_VISIBILITY uint64_t *__llvm_profile_end_counters(void) {
   return &PROF_CNTS_STOP;
 }
+
+COMPILER_RT_VISIBILITY ValueProfNode *
+__llvm_profile_begin_vnodes(void) {
+  return &PROF_VNODES_START;
+}
+COMPILER_RT_VISIBILITY ValueProfNode *__llvm_profile_end_vnodes(void) {
+  return &PROF_VNODES_STOP;
+}
+COMPILER_RT_VISIBILITY ValueProfNode *CurrentVNode = &PROF_VNODES_START;
+COMPILER_RT_VISIBILITY ValueProfNode *EndVNode = &PROF_VNODES_STOP;
+
 #endif
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingPlatformOther.c b/contrib/compiler-rt/lib/profile/InstrProfilingPlatformOther.c
index 58ceb3458a0a..b25966487e91 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfilingPlatformOther.c
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingPlatformOther.c
@@ -10,6 +10,7 @@
 #include "InstrProfiling.h"
 
 #if !defined(__APPLE__) && !defined(__linux__) && !defined(__FreeBSD__)
+
 #include <stdlib.h>
 
 static const __llvm_profile_data *DataFirst = NULL;
@@ -19,6 +20,14 @@ static const char *NamesLast = NULL;
 static uint64_t *CountersFirst = NULL;
 static uint64_t *CountersLast = NULL;
 
+static const void *getMinAddr(const void *A1, const void *A2) {
+  return A1 < A2 ? A1 : A2;
+}
+
+static const void *getMaxAddr(const void *A1, const void *A2) {
+  return A1 > A2 ? A1 : A2;
+}
+
 /*!
  * \brief Register an instrumented function.
  *
@@ -33,24 +42,30 @@ void __llvm_profile_register_function(void *Data_) {
   if (!DataFirst) {
     DataFirst = Data;
     DataLast = Data + 1;
-    NamesFirst = Data->NamePtr;
-    NamesLast = (const char *)Data->NamePtr + Data->NameSize;
     CountersFirst = Data->CounterPtr;
     CountersLast = (uint64_t *)Data->CounterPtr + Data->NumCounters;
     return;
   }
 
-#define UPDATE_FIRST(First, New) First = New < First ? New : First
-  UPDATE_FIRST(DataFirst, Data);
-  UPDATE_FIRST(NamesFirst, (const char *)Data->NamePtr);
-  UPDATE_FIRST(CountersFirst, (uint64_t *)Data->CounterPtr);
-#undef UPDATE_FIRST
+  DataFirst = (const __llvm_profile_data *)getMinAddr(DataFirst, Data);
+  CountersFirst = (uint64_t *)getMinAddr(CountersFirst, Data->CounterPtr);
+
+  DataLast = (const __llvm_profile_data *)getMaxAddr(DataLast, Data + 1);
+  CountersLast = (uint64_t *)getMaxAddr(
+      CountersLast, (uint64_t *)Data->CounterPtr + Data->NumCounters);
+}
 
-#define UPDATE_LAST(Last, New) Last = New > Last ? New : Last
-  UPDATE_LAST(DataLast, Data + 1);
-  UPDATE_LAST(NamesLast, (const char *)Data->NamePtr + Data->NameSize);
-  UPDATE_LAST(CountersLast, (uint64_t *)Data->CounterPtr + Data->NumCounters);
-#undef UPDATE_LAST
+COMPILER_RT_VISIBILITY
+void __llvm_profile_register_names_function(void *NamesStart,
+                                            uint64_t NamesSize) {
+  if (!NamesFirst) {
+    NamesFirst = (const char *)NamesStart;
+    NamesLast = (const char *)NamesStart + NamesSize;
+    return;
+  }
+  NamesFirst = (const char *)getMinAddr(NamesFirst, NamesStart);
+  NamesLast =
+      (const char *)getMaxAddr(NamesLast, (const char *)NamesStart + NamesSize);
 }
 
 COMPILER_RT_VISIBILITY
@@ -65,4 +80,15 @@ COMPILER_RT_VISIBILITY
 uint64_t *__llvm_profile_begin_counters(void) { return CountersFirst; }
 COMPILER_RT_VISIBILITY
 uint64_t *__llvm_profile_end_counters(void) { return CountersLast; }
+
+COMPILER_RT_VISIBILITY
+ValueProfNode *__llvm_profile_begin_vnodes(void) {
+  return 0;
+}
+COMPILER_RT_VISIBILITY
+ValueProfNode *__llvm_profile_end_vnodes(void) { return 0; }
+
+COMPILER_RT_VISIBILITY ValueProfNode *CurrentVNode = 0;
+COMPILER_RT_VISIBILITY ValueProfNode *EndVNode = 0;
+
 #endif
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingPort.h b/contrib/compiler-rt/lib/profile/InstrProfilingPort.h
index e07f59878730..c947153e2517 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfilingPort.h
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingPort.h
@@ -13,39 +13,99 @@
 #ifdef _MSC_VER
 #define COMPILER_RT_ALIGNAS(x) __declspec(align(x))
 #define COMPILER_RT_VISIBILITY
+/* FIXME: selectany does not have the same semantics as weak. */
 #define COMPILER_RT_WEAK __declspec(selectany)
+/* Need to include <windows.h> */
+#define COMPILER_RT_ALLOCA _alloca
+/* Need to include <stdio.h> and <io.h> */
+#define COMPILER_RT_FTRUNCATE(f,l) _chsize(_fileno(f),l)
 #elif __GNUC__
 #define COMPILER_RT_ALIGNAS(x) __attribute__((aligned(x)))
 #define COMPILER_RT_VISIBILITY __attribute__((visibility("hidden")))
 #define COMPILER_RT_WEAK __attribute__((weak))
+#define COMPILER_RT_ALLOCA __builtin_alloca
+#define COMPILER_RT_FTRUNCATE(f,l) ftruncate(fileno(f),l)
 #endif
 
+#if defined(__APPLE__)
+#define COMPILER_RT_SEG "__DATA,"
+#else
+#define COMPILER_RT_SEG ""
+#endif
+
+#ifdef _MSC_VER
+#define COMPILER_RT_SECTION(Sect) __declspec(allocate(Sect))
+#else
 #define COMPILER_RT_SECTION(Sect) __attribute__((section(Sect)))
+#endif
+
+#define COMPILER_RT_MAX_HOSTLEN 128
+#ifdef _MSC_VER
+#define COMPILER_RT_GETHOSTNAME(Name, Len) gethostname(Name, Len)
+#elif defined(__ORBIS__)
+#define COMPILER_RT_GETHOSTNAME(Name, Len) ((void)(Name), (void)(Len), (-1))
+#else
+#define COMPILER_RT_GETHOSTNAME(Name, Len) lprofGetHostName(Name, Len)
+#define COMPILER_RT_HAS_UNAME 1
+#endif
 
 #if COMPILER_RT_HAS_ATOMICS == 1
 #ifdef _MSC_VER
 #include <windows.h>
+#if _MSC_VER < 1900
+#define snprintf _snprintf
+#endif
 #if defined(_WIN64)
 #define COMPILER_RT_BOOL_CMPXCHG(Ptr, OldV, NewV)                              \
   (InterlockedCompareExchange64((LONGLONG volatile *)Ptr, (LONGLONG)NewV,      \
                                 (LONGLONG)OldV) == (LONGLONG)OldV)
+#define COMPILER_RT_PTR_FETCH_ADD(DomType, PtrVar, PtrIncr)                    \
+  (DomType *)InterlockedExchangeAdd64((LONGLONG volatile *)&PtrVar,            \
+                                      (LONGLONG)sizeof(DomType) * PtrIncr)
 #else /* !defined(_WIN64) */
 #define COMPILER_RT_BOOL_CMPXCHG(Ptr, OldV, NewV)                              \
   (InterlockedCompareExchange((LONG volatile *)Ptr, (LONG)NewV, (LONG)OldV) == \
    (LONG)OldV)
+#define COMPILER_RT_PTR_FETCH_ADD(DomType, PtrVar, PtrIncr)                    \
+  (DomType *)InterlockedExchangeAdd((LONG volatile *)&PtrVar,                  \
+                                    (LONG)sizeof(DomType) * PtrIncr)
 #endif
 #else /* !defined(_MSC_VER) */
 #define COMPILER_RT_BOOL_CMPXCHG(Ptr, OldV, NewV)                              \
   __sync_bool_compare_and_swap(Ptr, OldV, NewV)
+#define COMPILER_RT_PTR_FETCH_ADD(DomType, PtrVar, PtrIncr)                    \
+  (DomType *)__sync_fetch_and_add((long *)&PtrVar, sizeof(DomType) * PtrIncr)
 #endif
 #else /* COMPILER_RT_HAS_ATOMICS != 1 */
+#include "InstrProfilingUtil.h"
 #define COMPILER_RT_BOOL_CMPXCHG(Ptr, OldV, NewV)                              \
-  BoolCmpXchg((void **)Ptr, OldV, NewV)
+  lprofBoolCmpXchg((void **)Ptr, OldV, NewV)
+#define COMPILER_RT_PTR_FETCH_ADD(DomType, PtrVar, PtrIncr)                    \
+  (DomType *)lprofPtrFetchAdd((void **)&PtrVar, sizeof(DomType) * PtrIncr)
+#endif
+
+#if defined(_WIN32)
+#define DIR_SEPARATOR '\\'
+#define DIR_SEPARATOR_2 '/'
+#else
+#define DIR_SEPARATOR '/'
 #endif
 
+#ifndef DIR_SEPARATOR_2
+#define IS_DIR_SEPARATOR(ch) ((ch) == DIR_SEPARATOR)
+#else /* DIR_SEPARATOR_2 */
+#define IS_DIR_SEPARATOR(ch)                                                   \
+  (((ch) == DIR_SEPARATOR) || ((ch) == DIR_SEPARATOR_2))
+#endif /* DIR_SEPARATOR_2 */
+
 #define PROF_ERR(Format, ...)                                                  \
-  if (GetEnvHook && GetEnvHook("LLVM_PROFILE_VERBOSE_ERRORS"))                 \
-    fprintf(stderr, Format, __VA_ARGS__);
+  fprintf(stderr, "LLVM Profile Error: " Format, __VA_ARGS__);
+
+#define PROF_WARN(Format, ...)                                                 \
+  fprintf(stderr, "LLVM Profile Warning: " Format, __VA_ARGS__);
+
+#define PROF_NOTE(Format, ...)                                                 \
+  fprintf(stderr, "LLVM Profile Note: " Format, __VA_ARGS__);
 
 #if defined(__FreeBSD__)
 
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingUtil.c b/contrib/compiler-rt/lib/profile/InstrProfilingUtil.c
index 6f0443d3bb5d..5c66933bc1af 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfilingUtil.c
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingUtil.c
@@ -12,25 +12,175 @@
 
 #ifdef _WIN32
 #include <direct.h>
-#elif I386_FREEBSD
-int mkdir(const char*, unsigned short);
+#include <io.h>
+#include <windows.h>
 #else
 #include <sys/stat.h>
 #include <sys/types.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <errno.h>
 #endif
 
+#ifdef COMPILER_RT_HAS_UNAME
+#include <sys/utsname.h>
+#endif
+
+#include <stdlib.h>
+#include <string.h>
+
 COMPILER_RT_VISIBILITY
 void __llvm_profile_recursive_mkdir(char *path) {
   int i;
 
   for (i = 1; path[i] != '\0'; ++i) {
-    if (path[i] != '/') continue;
+    char save = path[i];
+    if (!(path[i] == '/' || path[i] == '\\'))
+      continue;
     path[i] = '\0';
 #ifdef _WIN32
     _mkdir(path);
 #else
-    mkdir(path, 0755);  /* Some of these will fail, ignore it. */
+    mkdir(path, 0755); /* Some of these will fail, ignore it. */
 #endif
-    path[i] = '/';
+    path[i] = save;
+  }
+}
+
+#if COMPILER_RT_HAS_ATOMICS != 1
+COMPILER_RT_VISIBILITY
+uint32_t lprofBoolCmpXchg(void **Ptr, void *OldV, void *NewV) {
+  void *R = *Ptr;
+  if (R == OldV) {
+    *Ptr = NewV;
+    return 1;
+  }
+  return 0;
+}
+COMPILER_RT_VISIBILITY
+void *lprofPtrFetchAdd(void **Mem, long ByteIncr) {
+  void *Old = *Mem;
+  *((char **)Mem) += ByteIncr;
+  return Old;
+}
+
+#endif
+
+#ifdef COMPILER_RT_HAS_UNAME
+COMPILER_RT_VISIBILITY int lprofGetHostName(char *Name, int Len) {
+  struct utsname N;
+  int R;
+  if (!(R = uname(&N)))
+    strncpy(Name, N.nodename, Len);
+  return R;
+}
+#endif
+
+COMPILER_RT_VISIBILITY FILE *lprofOpenFileEx(const char *ProfileName) {
+  FILE *f;
+  int fd;
+#ifdef COMPILER_RT_HAS_FCNTL_LCK
+  struct flock s_flock;
+
+  s_flock.l_whence = SEEK_SET;
+  s_flock.l_start = 0;
+  s_flock.l_len = 0; /* Until EOF.  */
+  s_flock.l_pid = getpid();
+
+  s_flock.l_type = F_WRLCK;
+  fd = open(ProfileName, O_RDWR | O_CREAT, 0666);
+  if (fd < 0)
+    return NULL;
+
+  while (fcntl(fd, F_SETLKW, &s_flock) == -1) {
+    if (errno != EINTR) {
+      if (errno == ENOLCK) {
+        PROF_WARN("Data may be corrupted during profile merging : %s\n",
+                  "Fail to obtain file lock due to system limit.");
+      }
+      break;
+    }
+  }
+
+  f = fdopen(fd, "r+b");
+#elif defined(_WIN32)
+  // FIXME: Use the wide variants to handle Unicode filenames.
+  HANDLE h = CreateFileA(ProfileName, GENERIC_READ | GENERIC_WRITE, 0, 0,
+                         OPEN_ALWAYS, FILE_ATTRIBUTE_NORMAL, 0);
+  if (h == INVALID_HANDLE_VALUE)
+    return NULL;
+
+  fd = _open_osfhandle((intptr_t)h, 0);
+  if (fd == -1) {
+    CloseHandle(h);
+    return NULL;
   }
+
+  f = _fdopen(fd, "r+b");
+  if (f == 0) {
+    CloseHandle(h);
+    return NULL;
+  }
+#else
+  /* Worst case no locking applied.  */
+  PROF_WARN("Concurrent file access is not supported : %s\n",
+            "lack file locking");
+  fd = open(ProfileName, O_RDWR | O_CREAT, 0666);
+  if (fd < 0)
+    return NULL;
+  f = fdopen(fd, "r+b");
+#endif
+
+  return f;
+}
+
+COMPILER_RT_VISIBILITY const char *lprofGetPathPrefix(int *PrefixStrip,
+                                                      size_t *PrefixLen) {
+  const char *Prefix = getenv("GCOV_PREFIX");
+  const char *PrefixStripStr = getenv("GCOV_PREFIX_STRIP");
+
+  *PrefixLen = 0;
+  *PrefixStrip = 0;
+  if (Prefix == NULL || Prefix[0] == '\0')
+    return NULL;
+
+  if (PrefixStripStr) {
+    *PrefixStrip = atoi(PrefixStripStr);
+
+    /* Negative GCOV_PREFIX_STRIP values are ignored */
+    if (*PrefixStrip < 0)
+      *PrefixStrip = 0;
+  } else {
+    *PrefixStrip = 0;
+  }
+  *PrefixLen = strlen(Prefix);
+
+  return Prefix;
+}
+
+COMPILER_RT_VISIBILITY void
+lprofApplyPathPrefix(char *Dest, const char *PathStr, const char *Prefix,
+                     size_t PrefixLen, int PrefixStrip) {
+
+  const char *Ptr;
+  int Level;
+  const char *StrippedPathStr = PathStr;
+
+  for (Level = 0, Ptr = PathStr + 1; Level < PrefixStrip; ++Ptr) {
+    if (*Ptr == '\0')
+      break;
+
+    if (!IS_DIR_SEPARATOR(*Ptr))
+      continue;
+
+    StrippedPathStr = Ptr;
+    ++Level;
+  }
+
+  memcpy(Dest, Prefix, PrefixLen);
+
+  if (!IS_DIR_SEPARATOR(Prefix[PrefixLen - 1]))
+    Dest[PrefixLen++] = DIR_SEPARATOR;
+
+  memcpy(Dest + PrefixLen, StrippedPathStr, strlen(StrippedPathStr) + 1);
 }
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingUtil.h b/contrib/compiler-rt/lib/profile/InstrProfilingUtil.h
index 756b18e7c56d..16d3fbf420f2 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfilingUtil.h
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingUtil.h
@@ -10,7 +10,24 @@
 #ifndef PROFILE_INSTRPROFILINGUTIL_H
 #define PROFILE_INSTRPROFILINGUTIL_H
 
+#include <stddef.h>
+#include <stdio.h>
+
 /*! \brief Create a directory tree. */
 void __llvm_profile_recursive_mkdir(char *Pathname);
 
-#endif  /* PROFILE_INSTRPROFILINGUTIL_H */
+/*! Open file \c Filename for read+write with write
+ * lock for exclusive access. The caller will block
+ * if the lock is already held by another process. */
+FILE *lprofOpenFileEx(const char *Filename);
+/* PS4 doesn't have getenv. Define a shim. */
+#if __ORBIS__
+static inline char *getenv(const char *name) { return NULL; }
+#endif /* #if __ORBIS__ */
+
+int lprofGetHostName(char *Name, int Len);
+
+unsigned lprofBoolCmpXchg(void **Ptr, void *OldV, void *NewV);
+void *lprofPtrFetchAdd(void **Mem, long ByteIncr);
+
+#endif /* PROFILE_INSTRPROFILINGUTIL_H */
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingValue.c b/contrib/compiler-rt/lib/profile/InstrProfilingValue.c
index 68e16cff9cbc..93957e323762 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfilingValue.c
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingValue.c
@@ -9,6 +9,7 @@
 
 #include "InstrProfiling.h"
 #include "InstrProfilingInternal.h"
+#include "InstrProfilingUtil.h" /* For PS4 getenv shim. */
 #include <limits.h>
 #include <stdio.h>
 #include <stdlib.h>
@@ -17,25 +18,39 @@
 #define INSTR_PROF_COMMON_API_IMPL
 #include "InstrProfData.inc"
 
-#define PROF_OOM(Msg) PROF_ERR(Msg ":%s\n", "Out of memory");
-#define PROF_OOM_RETURN(Msg)                                                   \
-  {                                                                            \
-    PROF_OOM(Msg)                                                              \
-    free(ValueDataArray);                                                      \
-    return NULL;                                                               \
-  }
+static int hasStaticCounters = 1;
+static int OutOfNodesWarnings = 0;
+static int hasNonDefaultValsPerSite = 0;
+#define INSTR_PROF_MAX_VP_WARNS 10
+#define INSTR_PROF_DEFAULT_NUM_VAL_PER_SITE 8
+#define INSTR_PROF_VNODE_POOL_SIZE 1024
+
+#ifndef _MSC_VER
+/* A shared static pool in addition to the vnodes statically
+ * allocated by the compiler.  */
+COMPILER_RT_VISIBILITY ValueProfNode
+    lprofValueProfNodes[INSTR_PROF_VNODE_POOL_SIZE] COMPILER_RT_SECTION(
+       COMPILER_RT_SEG INSTR_PROF_VNODES_SECT_NAME_STR);
+#endif
 
-#if COMPILER_RT_HAS_ATOMICS != 1
-COMPILER_RT_VISIBILITY
-uint32_t BoolCmpXchg(void **Ptr, void *OldV, void *NewV) {
-  void *R = *Ptr;
-  if (R == OldV) {
-    *Ptr = NewV;
-    return 1;
+COMPILER_RT_VISIBILITY uint32_t VPMaxNumValsPerSite =
+    INSTR_PROF_DEFAULT_NUM_VAL_PER_SITE;
+
+COMPILER_RT_VISIBILITY void lprofSetupValueProfiler() {
+  const char *Str = 0;
+  Str = getenv("LLVM_VP_MAX_NUM_VALS_PER_SITE");
+  if (Str && Str[0]) {
+    VPMaxNumValsPerSite = atoi(Str);
+    hasNonDefaultValsPerSite = 1;
   }
-  return 0;
+  if (VPMaxNumValsPerSite > INSTR_PROF_MAX_NUM_VAL_PER_SITE)
+    VPMaxNumValsPerSite = INSTR_PROF_MAX_NUM_VAL_PER_SITE;
+}
+
+COMPILER_RT_VISIBILITY void lprofSetMaxValsPerSite(uint32_t MaxVals) {
+  VPMaxNumValsPerSite = MaxVals;
+  hasNonDefaultValsPerSite = 1;
 }
-#endif
 
 /* This method is only used in value profiler mock testing.  */
 COMPILER_RT_VISIBILITY void
@@ -65,6 +80,15 @@ __llvm_get_function_addr(const __llvm_profile_data *Data) {
 static int allocateValueProfileCounters(__llvm_profile_data *Data) {
   uint64_t NumVSites = 0;
   uint32_t VKI;
+
+  /* This function will never be called when value site array is allocated
+     statically at compile time.  */
+  hasStaticCounters = 0;
+  /* When dynamic allocation is enabled, allow tracking the max number of
+   * values allowd.  */
+  if (!hasNonDefaultValsPerSite)
+    VPMaxNumValsPerSite = INSTR_PROF_MAX_NUM_VAL_PER_SITE;
+
   for (VKI = IPVK_First; VKI <= IPVK_Last; ++VKI)
     NumVSites += Data->NumValueSites[VKI];
 
@@ -79,10 +103,36 @@ static int allocateValueProfileCounters(__llvm_profile_data *Data) {
   return 1;
 }
 
+static ValueProfNode *allocateOneNode(__llvm_profile_data *Data, uint32_t Index,
+                                      uint64_t Value) {
+  ValueProfNode *Node;
+
+  if (!hasStaticCounters)
+    return (ValueProfNode *)calloc(1, sizeof(ValueProfNode));
+
+  /* Early check to avoid value wrapping around.  */
+  if (CurrentVNode + 1 > EndVNode) {
+    if (OutOfNodesWarnings++ < INSTR_PROF_MAX_VP_WARNS) {
+      PROF_WARN("Unable to track new values: %s. "
+                " Consider using option -mllvm -vp-counters-per-site=<n> to "
+                "allocate more"
+                " value profile counters at compile time. \n",
+                "Running out of static counters");
+    }
+    return 0;
+  }
+  Node = COMPILER_RT_PTR_FETCH_ADD(ValueProfNode, CurrentVNode, 1);
+  /* Due to section padding, EndVNode point to a byte which is one pass
+   * an incomplete VNode, so we need to skip the last incomplete node. */
+  if (Node + 1 > EndVNode)
+    return 0;
+
+  return Node;
+}
+
 COMPILER_RT_VISIBILITY void
 __llvm_profile_instrument_target(uint64_t TargetValue, void *Data,
                                  uint32_t CounterIndex) {
-
   __llvm_profile_data *PData = (__llvm_profile_data *)Data;
   if (!PData)
     return;
@@ -94,87 +144,184 @@ __llvm_profile_instrument_target(uint64_t TargetValue, void *Data,
 
   ValueProfNode **ValueCounters = (ValueProfNode **)PData->Values;
   ValueProfNode *PrevVNode = NULL;
-  ValueProfNode *CurrentVNode = ValueCounters[CounterIndex];
+  ValueProfNode *MinCountVNode = NULL;
+  ValueProfNode *CurVNode = ValueCounters[CounterIndex];
+  uint64_t MinCount = UINT64_MAX;
 
   uint8_t VDataCount = 0;
-  while (CurrentVNode) {
-    if (TargetValue == CurrentVNode->VData.Value) {
-      CurrentVNode->VData.Count++;
+  while (CurVNode) {
+    if (TargetValue == CurVNode->Value) {
+      CurVNode->Count++;
       return;
     }
-    PrevVNode = CurrentVNode;
-    CurrentVNode = CurrentVNode->Next;
+    if (CurVNode->Count < MinCount) {
+      MinCount = CurVNode->Count;
+      MinCountVNode = CurVNode;
+    }
+    PrevVNode = CurVNode;
+    CurVNode = CurVNode->Next;
     ++VDataCount;
   }
 
-  if (VDataCount >= INSTR_PROF_MAX_NUM_VAL_PER_SITE)
+  if (VDataCount >= VPMaxNumValsPerSite) {
+    /* Bump down the min count node's count. If it reaches 0,
+     * evict it. This eviction/replacement policy makes hot
+     * targets more sticky while cold targets less so. In other
+     * words, it makes it less likely for the hot targets to be
+     * prematurally evicted during warmup/establishment period,
+     * when their counts are still low. In a special case when
+     * the number of values tracked is reduced to only one, this
+     * policy will guarantee that the dominating target with >50%
+     * total count will survive in the end. Note that this scheme
+     * allows the runtime to track the min count node in an adaptive
+     * manner. It can correct previous mistakes and eventually
+     * lock on a cold target that is alread in stable state.
+     *
+     * In very rare cases,  this replacement scheme may still lead
+     * to target loss. For instance, out of \c N value slots, \c N-1
+     * slots are occupied by luke warm targets during the warmup
+     * period and the remaining one slot is competed by two or more
+     * very hot targets. If those hot targets occur in an interleaved
+     * way, none of them will survive (gain enough weight to throw out
+     * other established entries) due to the ping-pong effect.
+     * To handle this situation, user can choose to increase the max
+     * number of tracked values per value site. Alternatively, a more
+     * expensive eviction mechanism can be implemented. It requires
+     * the runtime to track the total number of evictions per-site.
+     * When the total number of evictions reaches certain threshold,
+     * the runtime can wipe out more than one lowest count entries
+     * to give space for hot targets.
+     */
+    if (!(--MinCountVNode->Count)) {
+      CurVNode = MinCountVNode;
+      CurVNode->Value = TargetValue;
+      CurVNode->Count++;
+    }
     return;
+  }
 
-  CurrentVNode = (ValueProfNode *)calloc(1, sizeof(ValueProfNode));
-  if (!CurrentVNode)
+  CurVNode = allocateOneNode(PData, CounterIndex, TargetValue);
+  if (!CurVNode)
     return;
-
-  CurrentVNode->VData.Value = TargetValue;
-  CurrentVNode->VData.Count++;
+  CurVNode->Value = TargetValue;
+  CurVNode->Count++;
 
   uint32_t Success = 0;
   if (!ValueCounters[CounterIndex])
     Success =
-        COMPILER_RT_BOOL_CMPXCHG(&ValueCounters[CounterIndex], 0, CurrentVNode);
+        COMPILER_RT_BOOL_CMPXCHG(&ValueCounters[CounterIndex], 0, CurVNode);
   else if (PrevVNode && !PrevVNode->Next)
-    Success = COMPILER_RT_BOOL_CMPXCHG(&(PrevVNode->Next), 0, CurrentVNode);
+    Success = COMPILER_RT_BOOL_CMPXCHG(&(PrevVNode->Next), 0, CurVNode);
 
-  if (!Success) {
-    free(CurrentVNode);
+  if (!Success && !hasStaticCounters) {
+    free(CurVNode);
     return;
   }
 }
 
-COMPILER_RT_VISIBILITY ValueProfData **
-__llvm_profile_gather_value_data(uint64_t *ValueDataSize) {
-  size_t S = 0;
-  __llvm_profile_data *I;
-  ValueProfData **ValueDataArray;
-
-  const __llvm_profile_data *DataEnd = __llvm_profile_end_data();
-  const __llvm_profile_data *DataBegin = __llvm_profile_begin_data();
-
-  if (!ValueDataSize)
-    return NULL;
-
-  ValueDataArray =
-      (ValueProfData **)calloc(DataEnd - DataBegin, sizeof(void *));
-  if (!ValueDataArray)
-    PROF_OOM_RETURN("Failed to write value profile data ");
-
-  /*
-   * Compute the total Size of the buffer to hold ValueProfData
-   * structures for functions with value profile data.
-   */
-  for (I = (__llvm_profile_data *)DataBegin; I != DataEnd; ++I) {
-    ValueProfRuntimeRecord R;
-    if (initializeValueProfRuntimeRecord(&R, I->NumValueSites, I->Values))
-      PROF_OOM_RETURN("Failed to write value profile data ");
-
-    /* Compute the size of ValueProfData from this runtime record.  */
-    if (getNumValueKindsRT(&R) != 0) {
-      ValueProfData *VD = NULL;
-      uint32_t VS = getValueProfDataSizeRT(&R);
-      VD = (ValueProfData *)calloc(VS, sizeof(uint8_t));
-      if (!VD)
-        PROF_OOM_RETURN("Failed to write value profile data ");
-      serializeValueProfDataFromRT(&R, VD);
-      ValueDataArray[I - DataBegin] = VD;
-      S += VS;
+/*
+ * A wrapper struct that represents value profile runtime data.
+ * Like InstrProfRecord class which is used by profiling host tools,
+ * ValueProfRuntimeRecord also implements the abstract intefaces defined in
+ * ValueProfRecordClosure so that the runtime data can be serialized using
+ * shared C implementation.
+ */
+typedef struct ValueProfRuntimeRecord {
+  const __llvm_profile_data *Data;
+  ValueProfNode **NodesKind[IPVK_Last + 1];
+  uint8_t **SiteCountArray;
+} ValueProfRuntimeRecord;
+
+/* ValueProfRecordClosure Interface implementation. */
+
+static uint32_t getNumValueSitesRT(const void *R, uint32_t VK) {
+  return ((const ValueProfRuntimeRecord *)R)->Data->NumValueSites[VK];
+}
+
+static uint32_t getNumValueDataRT(const void *R, uint32_t VK) {
+  uint32_t S = 0, I;
+  const ValueProfRuntimeRecord *Record = (const ValueProfRuntimeRecord *)R;
+  if (Record->SiteCountArray[VK] == INSTR_PROF_NULLPTR)
+    return 0;
+  for (I = 0; I < Record->Data->NumValueSites[VK]; I++)
+    S += Record->SiteCountArray[VK][I];
+  return S;
+}
+
+static uint32_t getNumValueDataForSiteRT(const void *R, uint32_t VK,
+                                         uint32_t S) {
+  const ValueProfRuntimeRecord *Record = (const ValueProfRuntimeRecord *)R;
+  return Record->SiteCountArray[VK][S];
+}
+
+static ValueProfRuntimeRecord RTRecord;
+static ValueProfRecordClosure RTRecordClosure = {
+    &RTRecord,          INSTR_PROF_NULLPTR, /* GetNumValueKinds */
+    getNumValueSitesRT, getNumValueDataRT,  getNumValueDataForSiteRT,
+    INSTR_PROF_NULLPTR, /* RemapValueData */
+    INSTR_PROF_NULLPTR, /* GetValueForSite, */
+    INSTR_PROF_NULLPTR  /* AllocValueProfData */
+};
+
+static uint32_t
+initializeValueProfRuntimeRecord(const __llvm_profile_data *Data,
+                                 uint8_t *SiteCountArray[]) {
+  unsigned I, J, S = 0, NumValueKinds = 0;
+  ValueProfNode **Nodes = (ValueProfNode **)Data->Values;
+  RTRecord.Data = Data;
+  RTRecord.SiteCountArray = SiteCountArray;
+  for (I = 0; I <= IPVK_Last; I++) {
+    uint16_t N = Data->NumValueSites[I];
+    if (!N)
+      continue;
+
+    NumValueKinds++;
+
+    RTRecord.NodesKind[I] = Nodes ? &Nodes[S] : INSTR_PROF_NULLPTR;
+    for (J = 0; J < N; J++) {
+      /* Compute value count for each site. */
+      uint32_t C = 0;
+      ValueProfNode *Site =
+          Nodes ? RTRecord.NodesKind[I][J] : INSTR_PROF_NULLPTR;
+      while (Site) {
+        C++;
+        Site = Site->Next;
+      }
+      if (C > UCHAR_MAX)
+        C = UCHAR_MAX;
+      RTRecord.SiteCountArray[I][J] = C;
     }
-    finalizeValueProfRuntimeRecord(&R);
+    S += N;
   }
+  return NumValueKinds;
+}
 
-  if (!S) {
-    free(ValueDataArray);
-    ValueDataArray = NULL;
+static ValueProfNode *getNextNValueData(uint32_t VK, uint32_t Site,
+                                        InstrProfValueData *Dst,
+                                        ValueProfNode *StartNode, uint32_t N) {
+  unsigned I;
+  ValueProfNode *VNode = StartNode ? StartNode : RTRecord.NodesKind[VK][Site];
+  for (I = 0; I < N; I++) {
+    Dst[I].Value = VNode->Value;
+    Dst[I].Count = VNode->Count;
+    VNode = VNode->Next;
   }
+  return VNode;
+}
+
+static uint32_t getValueProfDataSizeWrapper(void) {
+  return getValueProfDataSize(&RTRecordClosure);
+}
+
+static uint32_t getNumValueDataForSiteWrapper(uint32_t VK, uint32_t S) {
+  return getNumValueDataForSiteRT(&RTRecord, VK, S);
+}
+
+static VPDataReaderType TheVPDataReader = {
+    initializeValueProfRuntimeRecord, getValueProfRecordHeaderSize,
+    getFirstValueProfRecord,          getNumValueDataForSiteWrapper,
+    getValueProfDataSizeWrapper,      getNextNValueData};
 
-  *ValueDataSize = S;
-  return ValueDataArray;
+COMPILER_RT_VISIBILITY VPDataReaderType *lprofGetVPDataReader() {
+  return &TheVPDataReader;
 }
diff --git a/contrib/compiler-rt/lib/profile/InstrProfilingWriter.c b/contrib/compiler-rt/lib/profile/InstrProfilingWriter.c
index a07bc538ed4b..95f37e8e9b4f 100644
--- a/contrib/compiler-rt/lib/profile/InstrProfilingWriter.c
+++ b/contrib/compiler-rt/lib/profile/InstrProfilingWriter.c
@@ -9,20 +9,31 @@
 
 #include "InstrProfiling.h"
 #include "InstrProfilingInternal.h"
+#ifdef _MSC_VER
+/* For _alloca */
+#include <malloc.h>
+#endif
 #include <string.h>
 
 #define INSTR_PROF_VALUE_PROF_DATA
 #include "InstrProfData.inc"
-void (*FreeHook)(void *) = NULL;
-void* (*CallocHook)(size_t, size_t) = NULL;
-uint32_t VPBufferSize = 0;
+
+COMPILER_RT_VISIBILITY void (*FreeHook)(void *) = NULL;
+static ProfBufferIO TheBufferIO;
+#define VP_BUFFER_SIZE 8 * 1024
+static uint8_t BufferIOBuffer[VP_BUFFER_SIZE];
+static InstrProfValueData VPDataArray[16];
+static uint32_t VPDataArraySize = sizeof(VPDataArray) / sizeof(*VPDataArray);
+
+COMPILER_RT_VISIBILITY uint8_t *DynamicBufferIOBuffer = 0;
+COMPILER_RT_VISIBILITY uint32_t VPBufferSize = 0;
 
 /* The buffer writer is reponsponsible in keeping writer state
  * across the call.
  */
-COMPILER_RT_VISIBILITY uint32_t llvmBufferWriter(ProfDataIOVec *IOVecs,
-                                                 uint32_t NumIOVecs,
-                                                 void **WriterCtx) {
+COMPILER_RT_VISIBILITY uint32_t lprofBufferWriter(ProfDataIOVec *IOVecs,
+                                                  uint32_t NumIOVecs,
+                                                  void **WriterCtx) {
   uint32_t I;
   char **Buffer = (char **)WriterCtx;
   for (I = 0; I < NumIOVecs; I++) {
@@ -43,28 +54,31 @@ static void llvmInitBufferIO(ProfBufferIO *BufferIO, WriterCallback FileWriter,
 }
 
 COMPILER_RT_VISIBILITY ProfBufferIO *
-llvmCreateBufferIO(WriterCallback FileWriter, void *File, uint32_t BufferSz) {
-  ProfBufferIO *BufferIO = (ProfBufferIO *)CallocHook(1, sizeof(ProfBufferIO));
-  uint8_t *Buffer = (uint8_t *)CallocHook(1, BufferSz);
+lprofCreateBufferIO(WriterCallback FileWriter, void *File) {
+  uint8_t *Buffer = DynamicBufferIOBuffer;
+  uint32_t BufferSize = VPBufferSize;
   if (!Buffer) {
-    FreeHook(BufferIO);
-    return 0;
+    Buffer = &BufferIOBuffer[0];
+    BufferSize = sizeof(BufferIOBuffer);
   }
-  llvmInitBufferIO(BufferIO, FileWriter, File, Buffer, BufferSz);
-  return BufferIO;
+  llvmInitBufferIO(&TheBufferIO, FileWriter, File, Buffer, BufferSize);
+  return &TheBufferIO;
 }
 
-COMPILER_RT_VISIBILITY void llvmDeleteBufferIO(ProfBufferIO *BufferIO) {
-  FreeHook(BufferIO->BufferStart);
-  FreeHook(BufferIO);
+COMPILER_RT_VISIBILITY void lprofDeleteBufferIO(ProfBufferIO *BufferIO) {
+  if (DynamicBufferIOBuffer) {
+    FreeHook(DynamicBufferIOBuffer);
+    DynamicBufferIOBuffer = 0;
+    VPBufferSize = 0;
+  }
 }
 
 COMPILER_RT_VISIBILITY int
-llvmBufferIOWrite(ProfBufferIO *BufferIO, const uint8_t *Data, uint32_t Size) {
+lprofBufferIOWrite(ProfBufferIO *BufferIO, const uint8_t *Data, uint32_t Size) {
   /* Buffer is not large enough, it is time to flush.  */
   if (Size + BufferIO->CurOffset > BufferIO->BufferSz) {
-     if (llvmBufferIOFlush(BufferIO) != 0)
-       return -1;
+    if (lprofBufferIOFlush(BufferIO) != 0)
+      return -1;
   }
   /* Special case, bypass the buffer completely. */
   ProfDataIOVec IO[] = {{Data, sizeof(uint8_t), Size}};
@@ -74,13 +88,13 @@ llvmBufferIOWrite(ProfBufferIO *BufferIO, const uint8_t *Data, uint32_t Size) {
   } else {
     /* Write the data to buffer */
     uint8_t *Buffer = BufferIO->BufferStart + BufferIO->CurOffset;
-    llvmBufferWriter(IO, 1, (void **)&Buffer);
+    lprofBufferWriter(IO, 1, (void **)&Buffer);
     BufferIO->CurOffset = Buffer - BufferIO->BufferStart;
   }
   return 0;
 }
 
-COMPILER_RT_VISIBILITY int llvmBufferIOFlush(ProfBufferIO *BufferIO) {
+COMPILER_RT_VISIBILITY int lprofBufferIOFlush(ProfBufferIO *BufferIO) {
   if (BufferIO->CurOffset) {
     ProfDataIOVec IO[] = {
         {BufferIO->BufferStart, sizeof(uint8_t), BufferIO->CurOffset}};
@@ -91,60 +105,151 @@ COMPILER_RT_VISIBILITY int llvmBufferIOFlush(ProfBufferIO *BufferIO) {
   return 0;
 }
 
-COMPILER_RT_VISIBILITY int llvmWriteProfData(WriterCallback Writer,
-                                             void *WriterCtx,
-                                             ValueProfData **ValueDataArray,
-                                             const uint64_t ValueDataSize) {
-  /* Match logic in __llvm_profile_write_buffer(). */
-  const __llvm_profile_data *DataBegin = __llvm_profile_begin_data();
-  const __llvm_profile_data *DataEnd = __llvm_profile_end_data();
-  const uint64_t *CountersBegin = __llvm_profile_begin_counters();
-  const uint64_t *CountersEnd = __llvm_profile_end_counters();
-  const char *NamesBegin = __llvm_profile_begin_names();
-  const char *NamesEnd = __llvm_profile_end_names();
-  return llvmWriteProfDataImpl(Writer, WriterCtx, DataBegin, DataEnd,
-                               CountersBegin, CountersEnd, ValueDataArray,
-                               ValueDataSize, NamesBegin, NamesEnd);
+/* Write out value profile data for function specified with \c Data.
+ * The implementation does not use the method \c serializeValueProfData
+ * which depends on dynamic memory allocation. In this implementation,
+ * value profile data is written out to \c BufferIO piecemeal.
+ */
+static int writeOneValueProfData(ProfBufferIO *BufferIO,
+                                 VPDataReaderType *VPDataReader,
+                                 const __llvm_profile_data *Data) {
+  unsigned I, NumValueKinds = 0;
+  ValueProfData VPHeader;
+  uint8_t *SiteCountArray[IPVK_Last + 1];
+
+  for (I = 0; I <= IPVK_Last; I++) {
+    if (!Data->NumValueSites[I])
+      SiteCountArray[I] = 0;
+    else {
+      uint32_t Sz =
+          VPDataReader->GetValueProfRecordHeaderSize(Data->NumValueSites[I]) -
+          offsetof(ValueProfRecord, SiteCountArray);
+      /* Only use alloca for this small byte array to avoid excessive
+       * stack growth.  */
+      SiteCountArray[I] = (uint8_t *)COMPILER_RT_ALLOCA(Sz);
+      memset(SiteCountArray[I], 0, Sz);
+    }
+  }
+
+  /* If NumValueKinds returned is 0, there is nothing to write, report
+     success and return. This should match the raw profile reader's behavior. */
+  if (!(NumValueKinds = VPDataReader->InitRTRecord(Data, SiteCountArray)))
+    return 0;
+
+  /* First write the header structure. */
+  VPHeader.TotalSize = VPDataReader->GetValueProfDataSize();
+  VPHeader.NumValueKinds = NumValueKinds;
+  if (lprofBufferIOWrite(BufferIO, (const uint8_t *)&VPHeader,
+                         sizeof(ValueProfData)))
+    return -1;
+
+  /* Make sure nothing else needs to be written before value profile
+   * records. */
+  if ((void *)VPDataReader->GetFirstValueProfRecord(&VPHeader) !=
+      (void *)(&VPHeader + 1))
+    return -1;
+
+  /* Write out the value profile record for each value kind
+   * one by one. */
+  for (I = 0; I <= IPVK_Last; I++) {
+    uint32_t J;
+    ValueProfRecord RecordHeader;
+    /* The size of the value prof record header without counting the
+     * site count array .*/
+    uint32_t RecordHeaderSize = offsetof(ValueProfRecord, SiteCountArray);
+    uint32_t SiteCountArraySize;
+
+    if (!Data->NumValueSites[I])
+      continue;
+
+    /* Write out the record header.  */
+    RecordHeader.Kind = I;
+    RecordHeader.NumValueSites = Data->NumValueSites[I];
+    if (lprofBufferIOWrite(BufferIO, (const uint8_t *)&RecordHeader,
+                           RecordHeaderSize))
+      return -1;
+
+    /* Write out the site value count array including padding space. */
+    SiteCountArraySize =
+        VPDataReader->GetValueProfRecordHeaderSize(Data->NumValueSites[I]) -
+        RecordHeaderSize;
+    if (lprofBufferIOWrite(BufferIO, SiteCountArray[I], SiteCountArraySize))
+      return -1;
+
+    /* Write out the value profile data for each value site.  */
+    for (J = 0; J < Data->NumValueSites[I]; J++) {
+      uint32_t NRead, NRemain;
+      ValueProfNode *NextStartNode = 0;
+      NRemain = VPDataReader->GetNumValueDataForSite(I, J);
+      if (!NRemain)
+        continue;
+      /* Read and write out value data in small chunks till it is done. */
+      do {
+        NRead = (NRemain > VPDataArraySize ? VPDataArraySize : NRemain);
+        NextStartNode =
+            VPDataReader->GetValueData(I, /* ValueKind */
+                                       J, /* Site */
+                                       &VPDataArray[0], NextStartNode, NRead);
+        if (lprofBufferIOWrite(BufferIO, (const uint8_t *)&VPDataArray[0],
+                               NRead * sizeof(InstrProfValueData)))
+          return -1;
+        NRemain -= NRead;
+      } while (NRemain != 0);
+    }
+  }
+  /* All done report success.  */
+  return 0;
 }
 
-#define VP_BUFFER_SIZE 8 * 1024
 static int writeValueProfData(WriterCallback Writer, void *WriterCtx,
-                              ValueProfData **ValueDataBegin,
-                              uint64_t NumVData) {
+                              VPDataReaderType *VPDataReader,
+                              const __llvm_profile_data *DataBegin,
+                              const __llvm_profile_data *DataEnd) {
   ProfBufferIO *BufferIO;
-  uint32_t I = 0, BufferSz;
+  const __llvm_profile_data *DI = 0;
 
-  if (!ValueDataBegin)
+  if (!VPDataReader)
     return 0;
 
-  BufferSz = VPBufferSize ? VPBufferSize : VP_BUFFER_SIZE;
-  BufferIO = llvmCreateBufferIO(Writer, WriterCtx, BufferSz);
+  BufferIO = lprofCreateBufferIO(Writer, WriterCtx);
 
-  for (I = 0; I < NumVData; I++) {
-    ValueProfData *CurVData = ValueDataBegin[I];
-    if (!CurVData)
-      continue;
-    if (llvmBufferIOWrite(BufferIO, (const uint8_t *)CurVData,
-                          CurVData->TotalSize) != 0)
+  for (DI = DataBegin; DI < DataEnd; DI++) {
+    if (writeOneValueProfData(BufferIO, VPDataReader, DI))
       return -1;
   }
 
-  if (llvmBufferIOFlush(BufferIO) != 0)
+  if (lprofBufferIOFlush(BufferIO) != 0)
     return -1;
-  llvmDeleteBufferIO(BufferIO);
+  lprofDeleteBufferIO(BufferIO);
 
   return 0;
 }
 
-COMPILER_RT_VISIBILITY int llvmWriteProfDataImpl(
-    WriterCallback Writer, void *WriterCtx,
-    const __llvm_profile_data *DataBegin, const __llvm_profile_data *DataEnd,
-    const uint64_t *CountersBegin, const uint64_t *CountersEnd,
-    ValueProfData **ValueDataBegin, const uint64_t ValueDataSize,
-    const char *NamesBegin, const char *NamesEnd) {
+COMPILER_RT_VISIBILITY int lprofWriteData(WriterCallback Writer,
+                                          void *WriterCtx,
+                                          VPDataReaderType *VPDataReader) {
+  /* Match logic in __llvm_profile_write_buffer(). */
+  const __llvm_profile_data *DataBegin = __llvm_profile_begin_data();
+  const __llvm_profile_data *DataEnd = __llvm_profile_end_data();
+  const uint64_t *CountersBegin = __llvm_profile_begin_counters();
+  const uint64_t *CountersEnd = __llvm_profile_end_counters();
+  const char *NamesBegin = __llvm_profile_begin_names();
+  const char *NamesEnd = __llvm_profile_end_names();
+  return lprofWriteDataImpl(Writer, WriterCtx, DataBegin, DataEnd,
+                            CountersBegin, CountersEnd, VPDataReader,
+                            NamesBegin, NamesEnd);
+}
+
+COMPILER_RT_VISIBILITY int
+lprofWriteDataImpl(WriterCallback Writer, void *WriterCtx,
+                   const __llvm_profile_data *DataBegin,
+                   const __llvm_profile_data *DataEnd,
+                   const uint64_t *CountersBegin, const uint64_t *CountersEnd,
+                   VPDataReaderType *VPDataReader, const char *NamesBegin,
+                   const char *NamesEnd) {
 
   /* Calculate size of sections. */
-  const uint64_t DataSize = DataEnd - DataBegin;
+  const uint64_t DataSize = __llvm_profile_get_data_size(DataBegin, DataEnd);
   const uint64_t CountersSize = CountersEnd - CountersBegin;
   const uint64_t NamesSize = NamesEnd - NamesBegin;
   const uint64_t Padding = __llvm_profile_get_num_padding_bytes(NamesSize);
@@ -158,7 +263,7 @@ COMPILER_RT_VISIBILITY int llvmWriteProfDataImpl(
   if (!DataSize)
     return 0;
 
-  /* Initialize header struture.  */
+/* Initialize header structure.  */
 #define INSTR_PROF_RAW_HEADER(Type, Name, Init) Header.Name = Init;
 #include "InstrProfData.inc"
 
@@ -171,5 +276,6 @@ COMPILER_RT_VISIBILITY int llvmWriteProfDataImpl(
   if (Writer(IOVec, sizeof(IOVec) / sizeof(*IOVec), &WriterCtx))
     return -1;
 
-  return writeValueProfData(Writer, WriterCtx, ValueDataBegin, DataSize);
+  return writeValueProfData(Writer, WriterCtx, VPDataReader, DataBegin,
+                            DataEnd);
 }
diff --git a/contrib/compiler-rt/lib/profile/WindowsMMap.h b/contrib/compiler-rt/lib/profile/WindowsMMap.h
index 7b94eb28230c..271619aea09a 100644
--- a/contrib/compiler-rt/lib/profile/WindowsMMap.h
+++ b/contrib/compiler-rt/lib/profile/WindowsMMap.h
@@ -21,13 +21,7 @@
  */
 #define PROT_READ     0x1
 #define PROT_WRITE    0x2
-/* This flag is only available in WinXP+ */
-#ifdef FILE_MAP_EXECUTE
-#define PROT_EXEC     0x4
-#else
-#define PROT_EXEC        0x0
-#define FILE_MAP_EXECUTE 0
-#endif
+#define PROT_EXEC     0x0
 
 #define MAP_FILE      0x00
 #define MAP_SHARED    0x01