[Profile] Add new public API to get profile path prefix

[davidxl]

Summary

This API is intended to be used by user program to fetch all profiles dumped in same dir (the name of individual raw profile file is not important).

[vsk]

Looks good! I just have one concern.

lib/profile/InstrProfiling.h

155

than -> and?

lib/profile/InstrProfilingFile.c

435

Can this give the wrong answer if it comes after a call to __llvm_profile_set_filename()?

[davidxl]

Address Vedant's review comments and added more test.

[vsk]

LGTM.

lib/profile/InstrProfiling.h

 void __llvm_profile_override_default_filename(const char *Name);
 
 /*! \brief Register to write instrumentation data to file at exit. */
 int __llvm_profile_register_write_file_atexit(void);
 
 /*! \brief Initialize file handling. */
 void __llvm_profile_initialize_file(void);
 
+/*!
+ * \brief Return path prefix (excluding the base filename) of the profile data.
+ * This is useful for users using \c -fprofile-generate=./path_prefix who do
+ * not care about the default raw profile name. It is also useful to collect
+ * more than more profile data files dumped in the same directory (Online
+ * merge mode is turned on for instrumented programs with shared libs).
+ * Side-effect: this API call will invoke malloc with dynamic memory allocation.
+ */
+const char *__llvm_profile_get_path_prefix();
+
 /*! \brief Get the magic token for the file format. */
 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,

lib/profile/InstrProfilingFile.c

static const char *getPNSStr(ProfileNameSpecifier PNS) {
   }
 }
 
 #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;
+  const char *ProfilePathPrefix;
   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};
+lprofFilename lprofCurFilename = {0, 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,

  * 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 0;
   return Filename;
 }
 
+COMPILER_RT_VISIBILITY
+const char *__llvm_profile_get_path_prefix(void) {
+  int Length;
+  char *FilenameBuf, *Prefix;
+  const char *Filename, *PrefixEnd;
+
+  if (lprofCurFilename.ProfilePathPrefix)
+    return lprofCurFilename.ProfilePathPrefix;
+
+  Length = getCurFilenameLength();
+  FilenameBuf = (char *)COMPILER_RT_ALLOCA(Length + 1);
+  Filename = getCurFilename(FilenameBuf);
+  if (!Filename) {
+    lprofCurFilename.ProfilePathPrefix = "\0";
+    return "\0";
+  }
+
+  PrefixEnd = lprofFindLastDirSeparator(Filename);
+  if (!PrefixEnd) {
+    lprofCurFilename.ProfilePathPrefix = "\0";
+    return "\0";
+  }
+  Length = PrefixEnd - Filename + 1;
+  Prefix = (char *)malloc(Length + 1);
+  if (!Prefix) {
+    PROF_ERR("Failed to %s\n", "allocate memory.");
+    return "\0";
+  }
+  memcpy(Prefix, Filename, Length);
+  Prefix[Length] = '\0';
+  lprofCurFilename.ProfilePathPrefix = Prefix;
+  return Prefix;
+}
+
 /* 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) {
   const char *FilenamePat;
 

test/profile/instrprof-path.c

+// RUN: %clang_pgogen -O2 -o %t.0 %s
+// RUN: %clang_pgogen=%t.d1 -O2 -o %t.1 %s
+// RUN: %clang_pgogen=%t.d1/%t.d2 -O2 -o %t.2 %s
+//
+// RUN: %run %t.0  ""
+// RUN: env LLVM_PROFILE_FILE=%t.d1/default.profraw %run %t.0  %t.d1/
+// RUN: env LLVM_PROFILE_FILE=%t.d1/%t.d2/default.profraw %run %t.0 %t.d1/%t.d2/
+// RUN: %run %t.1 %t.d1/
+// RUN: %run %t.2 %t.d1/%t.d2/
+
+#include <string.h>
+
+const char *__llvm_profile_get_path_prefix();
+
+int main(int argc, const char *argv[]) {
+  int i;
+  const char *expected;
+  const char *prefix;
+  if (argc < 2)
+    return 1;
+
+  expected = argv[1];
+  prefix = __llvm_profile_get_path_prefix();
+
+  if (strcmp(prefix, expected))
+    return 1;
+
+  return 0;
+}