[PGO][Offload] Fix offload coverage mapping (#143490)

This pull request fixes coverage mapping on GPU targets. 

- It adds an address space cast to the coverage mapping generation pass.
- It reads the profiled function names from the ELF directly. Reading it
from public globals was causing issues in cases where multiple
device-code object files are linked together.

offload/plugins-nextgen/common/include/GlobalHandler.h

@@ -80,6 +80,7 @@ struct GPUProfGlobals {
 
   void dump() const;
   Error write() const;
+  bool empty() const;
 };
 
 /// Subclass of GlobalTy that holds the memory for a global of \p Ty.
@@ -192,9 +193,6 @@ public:
                                           /*D2H=*/false);
   }
 
-  /// Checks whether a given image contains profiling globals.
-  bool hasProfilingGlobals(GenericDeviceTy &Device, DeviceImageTy &Image);
-
   /// Reads profiling data from a GPU image to supplied profdata struct.
   /// Iterates through the image symbol table and stores global values
   /// with profiling prefixes.

offload/plugins-nextgen/common/src/GlobalHandler.cpp

@@ -173,16 +173,6 @@ Error GenericGlobalHandlerTy::readGlobalFromImage(GenericDeviceTy &Device,
   return Plugin::success();
 }
 
-bool GenericGlobalHandlerTy::hasProfilingGlobals(GenericDeviceTy &Device,
-                                                 DeviceImageTy &Image) {
-  GlobalTy global(getInstrProfNamesVarName().str(), 0);
-  if (auto Err = getGlobalMetadataFromImage(Device, Image, global)) {
-    consumeError(std::move(Err));
-    return false;
-  }
-  return true;
-}
-
 Expected<GPUProfGlobals>
 GenericGlobalHandlerTy::readProfilingGlobals(GenericDeviceTy &Device,
                                              DeviceImageTy &Image) {
@@ -204,12 +194,17 @@ GenericGlobalHandlerTy::readProfilingGlobals(GenericDeviceTy &Device,
     // Check if given current global is a profiling global based
     // on name
     if (*NameOrErr == getInstrProfNamesVarName()) {
-      // Read in profiled function names
-      DeviceProfileData.NamesData = SmallVector<uint8_t>(Sym.getSize(), 0);
-      GlobalTy NamesGlobal(NameOrErr->str(), Sym.getSize(),
-                           DeviceProfileData.NamesData.data());
-      if (auto Err = readGlobalFromDevice(Device, Image, NamesGlobal))
-        return Err;
+      // Read in profiled function names from ELF
+      auto SectionOrErr = Sym.getSection();
+      if (!SectionOrErr)
+        return SectionOrErr.takeError();
+
+      auto ContentsOrErr = (*SectionOrErr)->getContents();
+      if (!ContentsOrErr)
+        return ContentsOrErr.takeError();
+
+      SmallVector<uint8_t> NameBytes(ContentsOrErr->bytes());
+      DeviceProfileData.NamesData = NameBytes;
     } else if (NameOrErr->starts_with(getInstrProfCountersVarPrefix())) {
       // Read global variable profiling counts
       SmallVector<int64_t> Counts(Sym.getSize() / sizeof(int64_t), 0);
@@ -322,3 +317,7 @@ Error GPUProfGlobals::write() const {
 
   return Plugin::success();
 }
+
+bool GPUProfGlobals::empty() const {
+  return Counts.empty() && Data.empty() && NamesData.empty();
+}

offload/plugins-nextgen/common/src/PluginInterface.cpp

@@ -858,14 +858,13 @@ Error GenericDeviceTy::deinit(GenericPluginTy &Plugin) {
 
   for (auto *Image : LoadedImages) {
     GenericGlobalHandlerTy &Handler = Plugin.getGlobalHandler();
-    if (!Handler.hasProfilingGlobals(*this, *Image))
-      continue;
-
-    GPUProfGlobals profdata;
     auto ProfOrErr = Handler.readProfilingGlobals(*this, *Image);
     if (!ProfOrErr)
       return ProfOrErr.takeError();
 
+    if (ProfOrErr->empty())
+      continue;
+
     // Dump out profdata
     if ((OMPX_DebugKind.get() & uint32_t(DeviceDebugKind::PGODump)) ==
         uint32_t(DeviceDebugKind::PGODump))