[Clang][NFC] Move GPU include directory to proper place (llvm#160608)

Summary:
This should be handled in the toolchain, not in the middle of clang.

Author: jhuber6

clang/lib/Driver/ToolChains/AMDGPU.cpp

@@ -882,6 +882,16 @@ void AMDGPUToolChain::addClangWarningOptions(ArgStringList &CC1Args) const {
   CC1Args.push_back("-Werror=atomic-alignment");
 }
 
+void AMDGPUToolChain::AddClangSystemIncludeArgs(const ArgList &DriverArgs,
+                                                ArgStringList &CC1Args) const {
+  if (DriverArgs.hasArg(options::OPT_nostdinc) ||
+      DriverArgs.hasArg(options::OPT_nostdlibinc))
+    return;
+
+  if (std::optional<std::string> Path = getStdlibIncludePath())
+    addSystemInclude(DriverArgs, CC1Args, *Path);
+}
+
 StringRef
 AMDGPUToolChain::getGPUArch(const llvm::opt::ArgList &DriverArgs) const {
   return getProcessorFromTargetID(

clang/lib/Driver/ToolChains/AMDGPU.h

@@ -79,6 +79,9 @@ class LLVM_LIBRARY_VISIBILITY AMDGPUToolChain : public Generic_ELF {
   void addClangTargetOptions(const llvm::opt::ArgList &DriverArgs,
                              llvm::opt::ArgStringList &CC1Args,
                              Action::OffloadKind DeviceOffloadKind) const override;
+  void
+  AddClangSystemIncludeArgs(const llvm::opt::ArgList &DriverArgs,
+                            llvm::opt::ArgStringList &CC1Args) const override;
 
   /// Return whether denormals should be flushed, and treated as 0 by default
   /// for the subtarget.

clang/lib/Driver/ToolChains/Clang.cpp

@@ -1109,26 +1109,15 @@ void Clang::AddPreprocessingOptions(Compilation &C, const JobAction &JA,
   if (!Args.hasArg(options::OPT_nostdinc) &&
       Args.hasFlag(options::OPT_offload_inc, options::OPT_no_offload_inc,
                    true) &&
-      !Args.hasArg(options::OPT_nobuiltininc)) {
-    // Without an offloading language we will include these headers directly.
-    // Offloading languages will instead only use the declarations stored in
-    // the resource directory at clang/lib/Headers/llvm_libc_wrappers.
-    if (getToolChain().getTriple().isGPU() &&
-        C.getActiveOffloadKinds() == Action::OFK_None) {
-      SmallString<128> P(llvm::sys::path::parent_path(D.Dir));
-      llvm::sys::path::append(P, "include");
-      llvm::sys::path::append(P, getToolChain().getTripleString());
-      CmdArgs.push_back("-internal-isystem");
-      CmdArgs.push_back(Args.MakeArgString(P));
-    } else if (C.getActiveOffloadKinds() == Action::OFK_OpenMP) {
-      // TODO: CUDA / HIP include their own headers for some common functions
-      // implemented here. We'll need to clean those up so they do not conflict.
-      SmallString<128> P(D.ResourceDir);
-      llvm::sys::path::append(P, "include");
-      llvm::sys::path::append(P, "llvm_libc_wrappers");
-      CmdArgs.push_back("-internal-isystem");
-      CmdArgs.push_back(Args.MakeArgString(P));
-    }
+      !Args.hasArg(options::OPT_nobuiltininc) &&
+      (C.getActiveOffloadKinds() == Action::OFK_OpenMP)) {
+    // TODO: CUDA / HIP include their own headers for some common functions
+    // implemented here. We'll need to clean those up so they do not conflict.
+    SmallString<128> P(D.ResourceDir);
+    llvm::sys::path::append(P, "include");
+    llvm::sys::path::append(P, "llvm_libc_wrappers");
+    CmdArgs.push_back("-internal-isystem");
+    CmdArgs.push_back(Args.MakeArgString(P));
   }
 
   // Add system include arguments for all targets but IAMCU.

clang/lib/Driver/ToolChains/Cuda.cpp

@@ -778,6 +778,16 @@ void NVPTXToolChain::addClangTargetOptions(
     const llvm::opt::ArgList &DriverArgs, llvm::opt::ArgStringList &CC1Args,
     Action::OffloadKind DeviceOffloadingKind) const {}
 
+void NVPTXToolChain::AddClangSystemIncludeArgs(const ArgList &DriverArgs,
+                                               ArgStringList &CC1Args) const {
+  if (DriverArgs.hasArg(options::OPT_nostdinc) ||
+      DriverArgs.hasArg(options::OPT_nostdlibinc))
+    return;
+
+  if (std::optional<std::string> Path = getStdlibIncludePath())
+    addSystemInclude(DriverArgs, CC1Args, *Path);
+}
+
 bool NVPTXToolChain::supportsDebugInfoOption(const llvm::opt::Arg *A) const {
   const Option &O = A->getOption();
   return (O.matches(options::OPT_gN_Group) &&

clang/lib/Driver/ToolChains/Cuda.h

@@ -92,6 +92,9 @@ class LLVM_LIBRARY_VISIBILITY NVPTXToolChain : public ToolChain {
   addClangTargetOptions(const llvm::opt::ArgList &DriverArgs,
                         llvm::opt::ArgStringList &CC1Args,
                         Action::OffloadKind DeviceOffloadKind) const override;
+  void
+  AddClangSystemIncludeArgs(const llvm::opt::ArgList &DriverArgs,
+                            llvm::opt::ArgStringList &CC1Args) const override;
 
   // Never try to use the integrated assembler with CUDA; always fork out to
   // ptxas.