[CUDA][OpenMP] Add a generic offload action builder

[sfantao]

Summary

This patch proposes a new class to generate and record action dependences related with offloading. The builder provides three main functionalities:

• Add device dependences to host actions.
• Add host dependence to device actions.
• Register device top-level actions.

The constructor of the builder detect the programming models that should be supported, and generates a specialized builder for each. If a new programming model is to be added in the future, only a new specialized builder has to be implemented.

When the specialized builder is generated, it produces programming-model-specific diagnostics.

A CUDA specialized builder is proposed in the patch that mostly consists of the partition of the current buildCudaAction by the three different functionalities.

[sfantao]

Rebase.

[sfantao]

• Rebase.

[sfantao]

• Rebase.

[ABataev]

No '\brief's

lib/Driver/Driver.cpp

1393 ↗

(On Diff #62231)

1. 'final'
2. default initializers for fields

1437 ↗

(On Diff #62231)

default initializer

1486 ↗

(On Diff #62231)

'final'

[sfantao]

• Fix comments, annotate classes with final, and add default initializers.

[sfantao]

Hi Alexey,

Thanks for the review! Addressed your comments in the new diff.

I'll wait for your response on my question in http://reviews.llvm.org/D18172 to address the issue with \brief properly.

Thanks again,
Samuel

[jlebar]

Alexey, it seems that you're asking for "final" on all classes that are not inherited from. Forgive my ignorance, but would you mind pointing me to the document that talks about our position on "final" in LLVM source? I don't see it in the style guide, but I may be missing something.

The style guide does talk a good bit about writing concise and generally not-misleading code. My concern is that adding "final" everywhere paints an inaccurate picture and will mislead readers. Specifically, "final" is useful as a signal to readers that a class cannot safely be inherited from. "Don't even think about it, buster." But here we're adding "final" to a lot of classes that, as far as I can tell, aren't distinctive except in that they have no subclasses today. The problem with this is that, if we use "final" in this way, it dilutes the first "don't even try" meaning. Now when I see a class with "final" that I want to subclass, I'm just going to rip the "final" off, because chances are, I can do so safely. Now "final" does not serve as a warning to me that I shouldn't do this.

Sorry to focus on a superficial issue, but I think this really does matter for usability.

[sfantao]

• Remove \brief.

[ABataev]

In D18172#471861, @jlebar wrote:

Alexey, it seems that you're asking for "final" on all classes that are not inherited from. Forgive my ignorance, but would you mind pointing me to the document that talks about our position on "final" in LLVM source? I don't see it in the style guide, but I may be missing something.

The style guide does talk a good bit about writing concise and generally not-misleading code. My concern is that adding "final" everywhere paints an inaccurate picture and will mislead readers. Specifically, "final" is useful as a signal to readers that a class cannot safely be inherited from. "Don't even think about it, buster." But here we're adding "final" to a lot of classes that, as far as I can tell, aren't distinctive except in that they have no subclasses today. The problem with this is that, if we use "final" in this way, it dilutes the first "don't even try" meaning. Now when I see a class with "final" that I want to subclass, I'm just going to rip the "final" off, because chances are, I can do so safely. Now "final" does not serve as a warning to me that I shouldn't do this.

Sorry to focus on a superficial issue, but I think this really does matter for usability.

Hi Justin,
There are no strict rules on using 'final'. But I believe that it is better to protect developers from unintended use of code and if language allows to do it we should do it. If some class is not intended to be used as a base class, it is better to mark it as 'final' to be sure that nobody will inherit from it for sure. Consider it as an additional contract of the class.

[sfantao]

In D18172#471861, @jlebar wrote:

Alexey, it seems that you're asking for "final" on all classes that are not inherited from. Forgive my ignorance, but would you mind pointing me to the document that talks about our position on "final" in LLVM source? I don't see it in the style guide, but I may be missing something.

The style guide does talk a good bit about writing concise and generally not-misleading code. My concern is that adding "final" everywhere paints an inaccurate picture and will mislead readers. Specifically, "final" is useful as a signal to readers that a class cannot safely be inherited from. "Don't even think about it, buster." But here we're adding "final" to a lot of classes that, as far as I can tell, aren't distinctive except in that they have no subclasses today. The problem with this is that, if we use "final" in this way, it dilutes the first "don't even try" meaning. Now when I see a class with "final" that I want to subclass, I'm just going to rip the "final" off, because chances are, I can do so safely. Now "final" does not serve as a warning to me that I shouldn't do this.

Sorry to focus on a superficial issue, but I think this really does matter for usability.

Hi Justin,

Thanks for the comment!

I understand both interpretations can be useful for different purposes, i.e. "don't inherit, it is not safe" vs "feel free to change this class without worrying with subclasses". I tend to think that the latter (which I believe is what Alexey has in mind) can, in general, be more useful in the sense that a class is usually safe to extend - you can always add features that are somewhat orthogonal to what the parent class is doing. I can, of course, imagine cases where your argument is valid - you can have an implementation that allocates memory based on the type of the parent therefore it does not observe the extra storage required by derived classes. However, I think these cases are less common.

I also look at the programming guidelines and couldn't find a clear answer to the issue you raised.

I don't have a strong opinion about this, just giving my two cents.

Thanks again!
Samuel

[jlebar]

Yeah, I'd say that in the absence of a rule, consistency with surrounding
code is king. Otherwise we're sending a message when we don't mean to be.

I'm not at my machine, but my recollection is that most of the driver uses
final sparingly. But whatever the convention is we should do that, I think.

[sfantao]

• Use double instead of triple slash in one comment.

[sfantao]

In D18172#472152, @jlebar wrote:

Yeah, I'd say that in the absence of a rule, consistency with surrounding
code is king. Otherwise we're sending a message when we don't mean to be.

I'm not at my machine, but my recollection is that most of the driver uses
final sparingly. But whatever the convention is we should do that, I think.

True, final is not widely used in the driver. Again, I don't feel strongly about doing this one way or the other. For the moment, I'll update the diffs according to what Alexey requested. If you guys decide that I should do that differently, I'm happy to go to back to each review and update it accordingly.

Thanks again!
Samuel

[sfantao]

• Rebase

[sfantao]

• Rebase.

[sfantao]

• Remove redundant phases from cuda-phases.cu and use DAG check.
• Rebase.

[sfantao]

Any more comments on this patch or depending ones?

Thanks!
Samuel

[Hahnfeld]

In D18172#500029, @sfantao wrote:

Any more comments on this patch or depending ones?

Thanks!
Samuel

I can report that the latest patches are working for me and that they fix all points that I previously raised.
That said they look good from a user perspective but @jlebar and @ABataev have been driving the reviews so far.

Regards,
Jonas

[sfantao]

• Rebase.

[ABataev]

LG for me.

[hfinkel]

A nice abstraction and cleanup. LGTM.

lib/Driver/Driver.cpp

1625 ↗

(On Diff #72117)

use -> uses

[jlebar]

Hi, Samuel.

This change introduced a new crash / assertion failure in the driver.

$ echo | llvm-run clang -emit-llvm -c -x cuda - -o /dev/null

Before this patch, we get an error (perhaps not as helpful as we might want, but still):

clang-3.8: error: cannot specify -o when generating multiple output files

After this patch, we get an assertion failure:

clang: ../clang/lib/Driver/Driver.cpp:1610: virtual (anonymous namespace)::OffloadingActionBuilder::DeviceActionBuilder::ActionBuilderReturnCode (anonymous namespace)::OffloadingActionBuilder::CudaActionBuilder::getDeviceDepences(OffloadAction::DeviceDependences &, phases::ID, phases::ID, PhasesTy &): Assertion `CurPhase < phases::Backend && "Generating single CUDA " "instructions should only occur " "before the backend phase!"' failed.
#0 0x0000000001b07e28 llvm::sys::PrintStackTrace(llvm::raw_ostream&) (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x1b07e28)
#1 0x0000000001b08566 SignalHandler(int) (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x1b08566)
#2 0x00007f4bb7f89330 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x10330)
#3 0x00007f4bb6b7cc37 gsignal /build/eglibc-oGUzwX/eglibc-2.19/signal/../nptl/sysdeps/unix/sysv/linux/raise.c:56:0
#4 0x00007f4bb6b80028 abort /build/eglibc-oGUzwX/eglibc-2.19/stdlib/abort.c:91:0
#5 0x00007f4bb6b75bf6 __assert_fail_base /build/eglibc-oGUzwX/eglibc-2.19/assert/assert.c:92:0
#6 0x00007f4bb6b75ca2 (/lib/x86_64-linux-gnu/libc.so.6+0x2fca2)
#7 0x0000000001fa809b (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x1fa809b)
#8 0x0000000001f99981 clang::driver::Driver::BuildActions(clang::driver::Compilation&, llvm::opt::DerivedArgList&, llvm::SmallVector<std::pair<clang::driver::types::ID, llvm::opt::Arg const*>, 16u> const&, llvm::SmallVector<clang::driver::Action*, 3u>&) const (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x1f99981)
#9 0x0000000001f9431c clang::driver::Driver::BuildCompilation(llvm::ArrayRef<char const*>) (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x1f9431c)
#10 0x00000000007c0254 main (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x7c0254)
#11 0x00007f4bb6b67f45 __libc_start_main /build/eglibc-oGUzwX/eglibc-2.19/csu/libc-start.c:321:0
#12 0x00000000007bd9a2 _start (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x7bd9a2)
Stack dump:
0.	Program arguments: /usr/local/google/home/jlebar/llvm/release/bin/clang -emit-llvm -c -x cuda - -o /dev/null 
1.	Compilation construction
2.	Building compilation actions
Aborted (core dumped)

This was reported two weeks ago by Gurunath Kadam on cfe-dev, but I just got around to bisecting it.

Would you mind spinning a fix for this?

[sfantao]

Hi Justin

Thanks for letting me know. I'm looking into it.

Thanks again,
Samuel

[sfantao]

In D18172#580276, @jlebar wrote:

Hi, Samuel.

This change introduced a new crash / assertion failure in the driver.

$ echo | llvm-run clang -emit-llvm -c -x cuda - -o /dev/null

Before this patch, we get an error (perhaps not as helpful as we might want, but still):

clang-3.8: error: cannot specify -o when generating multiple output files

After this patch, we get an assertion failure:

clang: ../clang/lib/Driver/Driver.cpp:1610: virtual (anonymous namespace)::OffloadingActionBuilder::DeviceActionBuilder::ActionBuilderReturnCode (anonymous namespace)::OffloadingActionBuilder::CudaActionBuilder::getDeviceDepences(OffloadAction::DeviceDependences &, phases::ID, phases::ID, PhasesTy &): Assertion `CurPhase < phases::Backend && "Generating single CUDA " "instructions should only occur " "before the backend phase!"' failed.
#0 0x0000000001b07e28 llvm::sys::PrintStackTrace(llvm::raw_ostream&) (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x1b07e28)
#1 0x0000000001b08566 SignalHandler(int) (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x1b08566)
#2 0x00007f4bb7f89330 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x10330)
#3 0x00007f4bb6b7cc37 gsignal /build/eglibc-oGUzwX/eglibc-2.19/signal/../nptl/sysdeps/unix/sysv/linux/raise.c:56:0
#4 0x00007f4bb6b80028 abort /build/eglibc-oGUzwX/eglibc-2.19/stdlib/abort.c:91:0
#5 0x00007f4bb6b75bf6 __assert_fail_base /build/eglibc-oGUzwX/eglibc-2.19/assert/assert.c:92:0
#6 0x00007f4bb6b75ca2 (/lib/x86_64-linux-gnu/libc.so.6+0x2fca2)
#7 0x0000000001fa809b (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x1fa809b)
#8 0x0000000001f99981 clang::driver::Driver::BuildActions(clang::driver::Compilation&, llvm::opt::DerivedArgList&, llvm::SmallVector<std::pair<clang::driver::types::ID, llvm::opt::Arg const*>, 16u> const&, llvm::SmallVector<clang::driver::Action*, 3u>&) const (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x1f99981)
#9 0x0000000001f9431c clang::driver::Driver::BuildCompilation(llvm::ArrayRef<char const*>) (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x1f9431c)
#10 0x00000000007c0254 main (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x7c0254)
#11 0x00007f4bb6b67f45 __libc_start_main /build/eglibc-oGUzwX/eglibc-2.19/csu/libc-start.c:321:0
#12 0x00000000007bd9a2 _start (/usr/local/google/home/jlebar/llvm/release/bin/clang+0x7bd9a2)
Stack dump:
0.	Program arguments: /usr/local/google/home/jlebar/llvm/release/bin/clang -emit-llvm -c -x cuda - -o /dev/null 
1.	Compilation construction
2.	Building compilation actions
Aborted (core dumped)

This was reported two weeks ago by Gurunath Kadam on cfe-dev, but I just got around to bisecting it.

Would you mind spinning a fix for this?

This should be fixed in r285263.

You should now get the same error message as before.

cfe/trunk/include/clang/Driver/Compilation.h

typedef std::pair<const_offload_toolchains_iterator,
                     const_offload_toolchains_iterator>
       const_offload_toolchains_range;
 
   template <Action::OffloadKind Kind>
   const_offload_toolchains_range getOffloadToolChains() const {
     return OrderedOffloadingToolchains.equal_range(Kind);
   }
 
+  /// Return true if an offloading tool chain of a given kind exists.
+  template <Action::OffloadKind Kind> bool hasOffloadToolChain() const {
+    return OrderedOffloadingToolchains.find(Kind) !=
+           OrderedOffloadingToolchains.end();
+  }
+
   /// Return an offload toolchain of the provided kind. Only one is expected to
   /// exist.
   template <Action::OffloadKind Kind>
   const ToolChain *getSingleOffloadToolChain() const {
     auto TCs = getOffloadToolChains<Kind>();
 
     assert(TCs.first != TCs.second &&
            "No tool chains of the selected kind exist!");

cfe/trunk/lib/Driver/Driver.cpp

void Driver::BuildInputs(const ToolChain &TC, DerivedArgList &Args,
   if (CCCIsCPP() && Inputs.empty()) {
     // If called as standalone preprocessor, stdin is processed
     // if no other input is present.
     Arg *A = MakeInputArg(Args, Opts, "-");
     Inputs.push_back(std::make_pair(types::TY_C, A));
   }
 }
 
-// For each unique --cuda-gpu-arch= argument creates a TY_CUDA_DEVICE
-// input action and then wraps each in CudaDeviceAction paired with
-// appropriate GPU arch name. In case of partial (i.e preprocessing
-// only) or device-only compilation, each device action is added to /p
-// Actions and /p Current is released. Otherwise the function creates
-// and returns a new CudaHostAction which wraps /p Current and device
-// side actions.
-static Action *buildCudaActions(Compilation &C, DerivedArgList &Args,
-                                const Arg *InputArg, Action *HostAction,
-                                ActionList &Actions) {
-  Arg *PartialCompilationArg = Args.getLastArg(
-      options::OPT_cuda_host_only, options::OPT_cuda_device_only,
-      options::OPT_cuda_compile_host_device);
-  bool CompileHostOnly =
-      PartialCompilationArg &&
-      PartialCompilationArg->getOption().matches(options::OPT_cuda_host_only);
-  bool CompileDeviceOnly =
-      PartialCompilationArg &&
-      PartialCompilationArg->getOption().matches(options::OPT_cuda_device_only);
+namespace {
+/// Provides a convenient interface for different programming models to generate
+/// the required device actions.
+class OffloadingActionBuilder final {
+  /// Flag used to trace errors in the builder.
+  bool IsValid = false;
+
+  /// The compilation that is using this builder.
+  Compilation &C;
+
+  /// The derived arguments associated with this builder.
+  DerivedArgList &Args;
+
+  /// Map between an input argument and the offload kinds used to process it.
+  std::map<const Arg *, unsigned> InputArgToOffloadKindMap;
+
+  /// Builder interface. It doesn't build anything or keep any state.
+  class DeviceActionBuilder {
+  public:
+    typedef llvm::SmallVector<phases::ID, phases::MaxNumberOfPhases> PhasesTy;
+
+    enum ActionBuilderReturnCode {
+      // The builder acted successfully on the current action.
+      ABRT_Success,
+      // The builder didn't have to act on the current action.
+      ABRT_Inactive,
+      // The builder was successful and requested the host action to not be
+      // generated.
+      ABRT_Ignore_Host,
+    };
+
+  protected:
+    /// Compilation associated with this builder.
+    Compilation &C;
+
+    /// Tool chains associated with this builder. The same programming
+    /// model may have associated one or more tool chains.
+    SmallVector<const ToolChain *, 2> ToolChains;
+
+    /// The derived arguments associated with this builder.
+    DerivedArgList &Args;
+
+    /// The inputs associated with this builder.
+    const Driver::InputList &Inputs;
+
+    /// The associated offload kind.
+    Action::OffloadKind AssociatedOffloadKind = Action::OFK_None;
+
+  public:
+    DeviceActionBuilder(Compilation &C, DerivedArgList &Args,
+                        const Driver::InputList &Inputs,
+                        Action::OffloadKind AssociatedOffloadKind)
+        : C(C), Args(Args), Inputs(Inputs),
+          AssociatedOffloadKind(AssociatedOffloadKind) {}
+    virtual ~DeviceActionBuilder() {}
+
+    /// Fill up the array \a DA with all the device dependences that should be
+    /// added to the provided host action \a HostAction. By default it is
+    /// inactive.
+    virtual ActionBuilderReturnCode
+    getDeviceDepences(OffloadAction::DeviceDependences &DA, phases::ID CurPhase,
+                      phases::ID FinalPhase, PhasesTy &Phases) {
+      return ABRT_Inactive;
+    }
+
+    /// Update the state to include the provided host action \a HostAction as a
+    /// dependency of the current device action. By default it is inactive.
+    virtual ActionBuilderReturnCode addDeviceDepences(Action *HostAction) {
+      return ABRT_Inactive;
+    }
+
+    /// Append top level actions generated by the builder. Return true if errors
+    /// were found.
+    virtual void appendTopLevelActions(ActionList &AL) {}
+
+    /// Append linker actions generated by the builder. Return true if errors
+    /// were found.
+    virtual void appendLinkDependences(OffloadAction::DeviceDependences &DA) {}
+
+    /// Initialize the builder. Return true if any initialization errors are
+    /// found.
+    virtual bool initialize() { return false; }
+
+    /// Return true if this builder is valid. We have a valid builder if we have
+    /// associated device tool chains.
+    bool isValid() { return !ToolChains.empty(); }
+
+    /// Return the associated offload kind.
+    Action::OffloadKind getAssociatedOffloadKind() {
+      return AssociatedOffloadKind;
+    }
+  };
+
+  /// \brief CUDA action builder. It injects device code in the host backend
+  /// action.
+  class CudaActionBuilder final : public DeviceActionBuilder {
+    /// Flags to signal if the user requested host-only or device-only
+    /// compilation.
+    bool CompileHostOnly = false;
+    bool CompileDeviceOnly = false;
+
+    /// List of GPU architectures to use in this compilation.
+    SmallVector<CudaArch, 4> GpuArchList;
+
+    /// The CUDA actions for the current input.
+    ActionList CudaDeviceActions;
+
+    /// The CUDA fat binary if it was generated for the current input.
+    Action *CudaFatBinary = nullptr;
+
+    /// Flag that is set to true if this builder acted on the current input.
+    bool IsActive = false;
+
+  public:
+    CudaActionBuilder(Compilation &C, DerivedArgList &Args,
+                      const Driver::InputList &Inputs)
+        : DeviceActionBuilder(C, Args, Inputs, Action::OFK_Cuda) {}
+
+    ActionBuilderReturnCode
+    getDeviceDepences(OffloadAction::DeviceDependences &DA, phases::ID CurPhase,
+                      phases::ID FinalPhase, PhasesTy &Phases) override {
+      if (!IsActive)
+        return ABRT_Inactive;
+
+      // If we don't have more CUDA actions, we don't have any dependences to
+      // create for the host.
+      if (CudaDeviceActions.empty())
+        return ABRT_Success;
+
+      assert(CudaDeviceActions.size() == GpuArchList.size() &&
+             "Expecting one action per GPU architecture.");
+      assert(!CompileHostOnly &&
+             "Not expecting CUDA actions in host-only compilation.");
+
+      // If we are generating code for the device or we are in a backend phase,
+      // we attempt to generate the fat binary. We compile each arch to ptx and
+      // assemble to cubin, then feed the cubin *and* the ptx into a device
+      // "link" action, which uses fatbinary to combine these cubins into one
+      // fatbin.  The fatbin is then an input to the host action if not in
+      // device-only mode.
+      if (CompileDeviceOnly || CurPhase == phases::Backend) {
+        ActionList DeviceActions;
+        for (unsigned I = 0, E = GpuArchList.size(); I != E; ++I) {
+          // Produce the device action from the current phase up to the assemble
+          // phase.
+          for (auto Ph : Phases) {
+            // Skip the phases that were already dealt with.
+            if (Ph < CurPhase)
+              continue;
+            // We have to be consistent with the host final phase.
+            if (Ph > FinalPhase)
+              break;
+
+            CudaDeviceActions[I] = C.getDriver().ConstructPhaseAction(
+                C, Args, Ph, CudaDeviceActions[I]);
+
+            if (Ph == phases::Assemble)
+              break;
+          }
+
+          // If we didn't reach the assemble phase, we can't generate the fat
+          // binary. We don't need to generate the fat binary if we are not in
+          // device-only mode.
+          if (!isa<AssembleJobAction>(CudaDeviceActions[I]) ||
+              CompileDeviceOnly)
+            continue;
+
+          Action *AssembleAction = CudaDeviceActions[I];
+          assert(AssembleAction->getType() == types::TY_Object);
+          assert(AssembleAction->getInputs().size() == 1);
+
+          Action *BackendAction = AssembleAction->getInputs()[0];
+          assert(BackendAction->getType() == types::TY_PP_Asm);
+
+          for (auto &A : {AssembleAction, BackendAction}) {
+            OffloadAction::DeviceDependences DDep;
+            DDep.add(*A, *ToolChains.front(), CudaArchToString(GpuArchList[I]),
+                     Action::OFK_Cuda);
+            DeviceActions.push_back(
+                C.MakeAction<OffloadAction>(DDep, A->getType()));
+          }
+        }
+
+        // We generate the fat binary if we have device input actions.
+        if (!DeviceActions.empty()) {
+          CudaFatBinary =
+              C.MakeAction<LinkJobAction>(DeviceActions, types::TY_CUDA_FATBIN);
+
+          if (!CompileDeviceOnly) {
+            DA.add(*CudaFatBinary, *ToolChains.front(), /*BoundArch=*/nullptr,
+                   Action::OFK_Cuda);
+            // Clear the fat binary, it is already a dependence to an host
+            // action.
+            CudaFatBinary = nullptr;
+          }
+
+          // Remove the CUDA actions as they are already connected to an host
+          // action or fat binary.
+          CudaDeviceActions.clear();
+        }
+
+        // We avoid creating host action in device-only mode.
+        return CompileDeviceOnly ? ABRT_Ignore_Host : ABRT_Success;
+      }
+
+      assert(CurPhase < phases::Backend && "Generating single CUDA "
+                                           "instructions should only occur "
+                                           "before the backend phase!");
+
+      // By default, we produce an action for each device arch.
+      for (Action *&A : CudaDeviceActions)
+        A = C.getDriver().ConstructPhaseAction(C, Args, CurPhase, A);
+
+      return ABRT_Success;
+    }
+
+    ActionBuilderReturnCode addDeviceDepences(Action *HostAction) override {
+      // While generating code for CUDA, we only depend on the host input action
+      // to trigger the creation of all the CUDA device actions.
+
+      // If we are dealing with an input action, replicate it for each GPU
+      // architecture. If we are in host-only mode we return 'success' so that
+      // the host uses the CUDA offload kind.
+      if (auto *IA = dyn_cast<InputAction>(HostAction)) {
+        assert(!GpuArchList.empty() &&
+               "We should have at least one GPU architecture.");
+
+        // If the host input is not CUDA, we don't need to bother about this
+        // input.
+        if (IA->getType() != types::TY_CUDA) {
+          // The builder will ignore this input.
+          IsActive = false;
+          return ABRT_Inactive;
+        }
+
+        // Set the flag to true, so that the builder acts on the current input.
+        IsActive = true;
+
+        if (CompileHostOnly)
+          return ABRT_Success;
+
+        // Replicate inputs for each GPU architecture.
+        for (unsigned I = 0, E = GpuArchList.size(); I != E; ++I)
+          CudaDeviceActions.push_back(C.MakeAction<InputAction>(
+              IA->getInputArg(), types::TY_CUDA_DEVICE));
+
+        return ABRT_Success;
+      }
+
+      return IsActive ? ABRT_Success : ABRT_Inactive;
+    }
+
+    void appendTopLevelActions(ActionList &AL) override {
+      // Utility to append actions to the top level list.
+      auto AddTopLevel = [&](Action *A, CudaArch BoundArch) {
+        OffloadAction::DeviceDependences Dep;
+        Dep.add(*A, *ToolChains.front(), CudaArchToString(BoundArch),
+                Action::OFK_Cuda);
+        AL.push_back(C.MakeAction<OffloadAction>(Dep, A->getType()));
+      };
+
+      // If we have a fat binary, add it to the list.
+      if (CudaFatBinary) {
+        AddTopLevel(CudaFatBinary, CudaArch::UNKNOWN);
+        CudaDeviceActions.clear();
+        CudaFatBinary = nullptr;
+        return;
+      }
+
+      if (CudaDeviceActions.empty())
+        return;
+
+      // If we have CUDA actions at this point, that's because we have a have
+      // partial compilation, so we should have an action for each GPU
+      // architecture.
+      assert(CudaDeviceActions.size() == GpuArchList.size() &&
+             "Expecting one action per GPU architecture.");
+      assert(ToolChains.size() == 1 &&
+             "Expecting to have a sing CUDA toolchain.");
+      for (unsigned I = 0, E = GpuArchList.size(); I != E; ++I)
+        AddTopLevel(CudaDeviceActions[I], GpuArchList[I]);
+
+      CudaDeviceActions.clear();
+    }
+
+    bool initialize() override {
+      // We don't need to support CUDA.
+      if (!C.hasOffloadToolChain<Action::OFK_Cuda>())
+        return false;
+
       const ToolChain *HostTC = C.getSingleOffloadToolChain<Action::OFK_Host>();
       assert(HostTC && "No toolchain for host compilation.");
       if (HostTC->getTriple().isNVPTX()) {
         // We do not support targeting NVPTX for host compilation. Throw
         // an error and abort pipeline construction early so we don't trip
         // asserts that assume device-side compilation.
         C.getDriver().Diag(diag::err_drv_cuda_nvptx_host);
-    return nullptr;
+        return true;
       }
 
-  if (CompileHostOnly) {
-    OffloadAction::HostDependence HDep(*HostAction, *HostTC,
-                                       /*BoundArch=*/nullptr, Action::OFK_Cuda);
-    return C.MakeAction<OffloadAction>(HDep);
-  }
+      ToolChains.push_back(C.getSingleOffloadToolChain<Action::OFK_Cuda>());
+
+      Arg *PartialCompilationArg = Args.getLastArg(
+          options::OPT_cuda_host_only, options::OPT_cuda_device_only,
+          options::OPT_cuda_compile_host_device);
+      CompileHostOnly = PartialCompilationArg &&
+                        PartialCompilationArg->getOption().matches(
+                            options::OPT_cuda_host_only);
+      CompileDeviceOnly = PartialCompilationArg &&
+                          PartialCompilationArg->getOption().matches(
+                              options::OPT_cuda_device_only);
 
       // Collect all cuda_gpu_arch parameters, removing duplicates.
-  SmallVector<CudaArch, 4> GpuArchList;
       llvm::SmallSet<CudaArch, 4> GpuArchs;
+      bool Error = false;
       for (Arg *A : Args) {
         if (!A->getOption().matches(options::OPT_cuda_gpu_arch_EQ))
           continue;
         A->claim();
 
         const auto &ArchStr = A->getValue();
         CudaArch Arch = StringToCudaArch(ArchStr);
-    if (Arch == CudaArch::UNKNOWN)
+        if (Arch == CudaArch::UNKNOWN) {
           C.getDriver().Diag(clang::diag::err_drv_cuda_bad_gpu_arch) << ArchStr;
-    else if (GpuArchs.insert(Arch).second)
+          Error = true;
+        } else if (GpuArchs.insert(Arch).second)
           GpuArchList.push_back(Arch);
       }
 
-  // Default to sm_20 which is the lowest common denominator for supported GPUs.
+      // Default to sm_20 which is the lowest common denominator for supported
+      // GPUs.
       // sm_20 code should work correctly, if suboptimally, on all newer GPUs.
       if (GpuArchList.empty())
         GpuArchList.push_back(CudaArch::SM_20);
 
-  // Replicate inputs for each GPU architecture.
-  Driver::InputList CudaDeviceInputs;
-  for (unsigned I = 0, E = GpuArchList.size(); I != E; ++I)
-    CudaDeviceInputs.push_back(std::make_pair(types::TY_CUDA_DEVICE, InputArg));
+      return Error;
+    }
+  };
 
-  // Build actions for all device inputs.
-  ActionList CudaDeviceActions;
-  C.getDriver().BuildActions(C, Args, CudaDeviceInputs, CudaDeviceActions);
-  assert(GpuArchList.size() == CudaDeviceActions.size() &&
-         "Failed to create actions for all devices");
-
-  // Check whether any of device actions stopped before they could generate PTX.
-  bool PartialCompilation =
-      llvm::any_of(CudaDeviceActions, [](const Action *a) {
-        return a->getKind() != Action::AssembleJobClass;
-      });
+  /// Add the implementation for other specialized builders here.
 
-  const ToolChain *CudaTC = C.getSingleOffloadToolChain<Action::OFK_Cuda>();
+  /// Specialized builders being used by this offloading action builder.
+  SmallVector<DeviceActionBuilder *, 4> SpecializedBuilders;
 
-  // Figure out what to do with device actions -- pass them as inputs to the
-  // host action or run each of them independently.
-  if (PartialCompilation || CompileDeviceOnly) {
-    // In case of partial or device-only compilation results of device actions
-    // are not consumed by the host action device actions have to be added to
-    // top-level actions list with AtTopLevel=true and run independently.
-
-    // -o is ambiguous if we have more than one top-level action.
-    if (Args.hasArg(options::OPT_o) &&
-        (!CompileDeviceOnly || GpuArchList.size() > 1)) {
-      C.getDriver().Diag(
-          clang::diag::err_drv_output_argument_with_multiple_files);
+public:
+  OffloadingActionBuilder(Compilation &C, DerivedArgList &Args,
+                          const Driver::InputList &Inputs)
+      : C(C), Args(Args) {
+    // Create a specialized builder for each device toolchain.
+
+    IsValid = true;
+
+    // Create a specialized builder for CUDA.
+    SpecializedBuilders.push_back(new CudaActionBuilder(C, Args, Inputs));
+
+    //
+    // TODO: Build other specialized builders here.
+    //
+
+    // Initialize all the builders, keeping track of errors.
+    for (auto *SB : SpecializedBuilders)
+      IsValid = IsValid && !SB->initialize();
+  }
+
+  ~OffloadingActionBuilder() {
+    for (auto *SB : SpecializedBuilders)
+      delete SB;
+  }
+
+  /// Generate an action that adds device dependences (if any) to a host action.
+  /// If no device dependence actions exist, just return the host action \a
+  /// HostAction. If an error is found or if no builder requires the host action
+  /// to be generated, return nullptr.
+  Action *
+  addDeviceDependencesToHostAction(Action *HostAction, const Arg *InputArg,
+                                   phases::ID CurPhase, phases::ID FinalPhase,
+                                   DeviceActionBuilder::PhasesTy &Phases) {
+    if (!IsValid)
       return nullptr;
-    }
 
-    for (unsigned I = 0, E = GpuArchList.size(); I != E; ++I) {
-      OffloadAction::DeviceDependences DDep;
-      DDep.add(*CudaDeviceActions[I], *CudaTC, CudaArchToString(GpuArchList[I]),
-               Action::OFK_Cuda);
-      Actions.push_back(
-          C.MakeAction<OffloadAction>(DDep, CudaDeviceActions[I]->getType()));
+    if (SpecializedBuilders.empty())
+      return HostAction;
+
+    assert(HostAction && "Invalid host action!");
+
+    OffloadAction::DeviceDependences DDeps;
+    // Check if all the programming models agree we should not emit the host
+    // action. Also, keep track of the offloading kinds employed.
+    auto &OffloadKind = InputArgToOffloadKindMap[InputArg];
+    unsigned InactiveBuilders = 0u;
+    unsigned IgnoringBuilders = 0u;
+    for (auto *SB : SpecializedBuilders) {
+      if (!SB->isValid()) {
+        ++InactiveBuilders;
+        continue;
       }
-    // Kill host action in case of device-only compilation.
-    if (CompileDeviceOnly)
+
+      auto RetCode = SB->getDeviceDepences(DDeps, CurPhase, FinalPhase, Phases);
+
+      // If the builder explicitly says the host action should be ignored,
+      // we need to increment the variable that tracks the builders that request
+      // the host object to be ignored.
+      if (RetCode == DeviceActionBuilder::ABRT_Ignore_Host)
+        ++IgnoringBuilders;
+
+      // Unless the builder was inactive for this action, we have to record the
+      // offload kind because the host will have to use it.
+      if (RetCode != DeviceActionBuilder::ABRT_Inactive)
+        OffloadKind |= SB->getAssociatedOffloadKind();
+    }
+
+    // If all builders agree that the host object should be ignored, just return
+    // nullptr.
+    if (IgnoringBuilders &&
+        SpecializedBuilders.size() == (InactiveBuilders + IgnoringBuilders))
       return nullptr;
+
+    if (DDeps.getActions().empty())
       return HostAction;
+
+    // We have dependences we need to bundle together. We use an offload action
+    // for that.
+    OffloadAction::HostDependence HDep(
+        *HostAction, *C.getSingleOffloadToolChain<Action::OFK_Host>(),
+        /*BoundArch=*/nullptr, DDeps);
+    return C.MakeAction<OffloadAction>(HDep, DDeps);
   }
 
-  // If we're not a partial or device-only compilation, we compile each arch to
-  // ptx and assemble to cubin, then feed the cubin *and* the ptx into a device
-  // "link" action, which uses fatbinary to combine these cubins into one
-  // fatbin.  The fatbin is then an input to the host compilation.
-  ActionList DeviceActions;
-  for (unsigned I = 0, E = GpuArchList.size(); I != E; ++I) {
-    Action* AssembleAction = CudaDeviceActions[I];
-    assert(AssembleAction->getType() == types::TY_Object);
-    assert(AssembleAction->getInputs().size() == 1);
+  /// Generate an action that adds a host dependence to a device action. The
+  /// results will be kept in this action builder. Return true if an error was
+  /// found.
+  bool addHostDependenceToDeviceActions(Action *HostAction,
+                                        const Arg *InputArg) {
+    if (!IsValid)
+      return true;
 
-    Action* BackendAction = AssembleAction->getInputs()[0];
-    assert(BackendAction->getType() == types::TY_PP_Asm);
+    assert(HostAction && "Invalid host action!");
 
-    for (auto &A : {AssembleAction, BackendAction}) {
-      OffloadAction::DeviceDependences DDep;
-      DDep.add(*A, *CudaTC, CudaArchToString(GpuArchList[I]), Action::OFK_Cuda);
-      DeviceActions.push_back(C.MakeAction<OffloadAction>(DDep, A->getType()));
+    // Register the offload kinds that are used.
+    auto &OffloadKind = InputArgToOffloadKindMap[InputArg];
+    for (auto *SB : SpecializedBuilders) {
+      if (!SB->isValid())
+        continue;
+
+      auto RetCode = SB->addDeviceDepences(HostAction);
+
+      // Host dependences for device actions are not compatible with that same
+      // action being ignored.
+      assert(RetCode != DeviceActionBuilder::ABRT_Ignore_Host &&
+             "Host dependence not expected to be ignored.!");
+
+      // Unless the builder was inactive for this action, we have to record the
+      // offload kind because the host will have to use it.
+      if (RetCode != DeviceActionBuilder::ABRT_Inactive)
+        OffloadKind |= SB->getAssociatedOffloadKind();
     }
+
+    return false;
   }
-  auto FatbinAction =
-      C.MakeAction<LinkJobAction>(DeviceActions, types::TY_CUDA_FATBIN);
 
-  // Return a new host action that incorporates original host action and all
-  // device actions.
-  OffloadAction::HostDependence HDep(*HostAction, *HostTC,
-                                     /*BoundArch=*/nullptr, Action::OFK_Cuda);
-  OffloadAction::DeviceDependences DDep;
-  DDep.add(*FatbinAction, *CudaTC, /*BoundArch=*/nullptr, Action::OFK_Cuda);
-  return C.MakeAction<OffloadAction>(HDep, DDep);
+  /// Add the offloading top level actions to the provided action list.
+  bool appendTopLevelActions(ActionList &AL, Action *HostAction,
+                             const Arg *InputArg) {
+    auto NumActions = AL.size();
+
+    for (auto *SB : SpecializedBuilders) {
+      if (!SB->isValid())
+        continue;
+      SB->appendTopLevelActions(AL);
+    }
+
+    assert(NumActions <= AL.size() && "Expecting more actions, not less!");
+
+    // Propagate to the current host action (if any) the offload information
+    // associated with the current input.
+    if (HostAction)
+      HostAction->propagateHostOffloadInfo(InputArgToOffloadKindMap[InputArg],
+                                           /*BoundArch=*/nullptr);
+
+    // If any action is added by the builders, -o is ambiguous if we have more
+    // than one top-level action.
+    if (NumActions < AL.size() && Args.hasArg(options::OPT_o) &&
+        AL.size() > 1) {
+      C.getDriver().Diag(
+          clang::diag::err_drv_output_argument_with_multiple_files);
+      return true;
+    }
+
+    return false;
+  }
+
+  /// Processes the host linker action. This currently consists of replacing it
+  /// with an offload action if there are device link objects and propagate to
+  /// the host action all the offload kinds used in the current compilation. The
+  /// resulting action is returned.
+  Action *processHostLinkAction(Action *HostAction) {
+    // Add all the dependences from the device linking actions.
+    OffloadAction::DeviceDependences DDeps;
+    for (auto *SB : SpecializedBuilders) {
+      if (!SB->isValid())
+        continue;
+
+      SB->appendLinkDependences(DDeps);
     }
 
+    // Calculate all the offload kinds used in the current compilation.
+    unsigned ActiveOffloadKinds = 0u;
+    for (auto &I : InputArgToOffloadKindMap)
+      ActiveOffloadKinds |= I.second;
+
+    // If we don't have device dependencies, we don't have to create an offload
+    // action.
+    if (DDeps.getActions().empty()) {
+      // Propagate all the active kinds to host action. Given that it is a link
+      // action it is assumed to depend on all actions generated so far.
+      HostAction->propagateHostOffloadInfo(ActiveOffloadKinds,
+                                           /*BoundArch=*/nullptr);
+      return HostAction;
+    }
+
+    // Create the offload action with all dependences. When an offload action
+    // is created the kinds are propagated to the host action, so we don't have
+    // to do that explicitely here.
+    OffloadAction::HostDependence HDep(
+        *HostAction, *C.getSingleOffloadToolChain<Action::OFK_Host>(),
+        /*BoundArch*/ nullptr, ActiveOffloadKinds);
+    return C.MakeAction<OffloadAction>(HDep, DDeps);
+  }
+};
+} // anonymous namespace.
+
 void Driver::BuildActions(Compilation &C, DerivedArgList &Args,
                           const InputList &Inputs, ActionList &Actions) const {
   llvm::PrettyStackTraceString CrashInfo("Building compilation actions");
 
   if (!SuppressMissingInputWarning && Inputs.empty()) {
     Diag(clang::diag::err_drv_no_input_files);
     return;
   }
if (Args.hasArg(options::OPT__SLASH_Y_)) {
     // /Y- disables all pch handling.  Rather than check for it everywhere,
     // just remove clang-cl pch-related flags here.
     Args.eraseArg(options::OPT__SLASH_Fp);
     Args.eraseArg(options::OPT__SLASH_Yc);
     Args.eraseArg(options::OPT__SLASH_Yu);
     YcArg = YuArg = nullptr;
   }
 
-  // Track the host offload kinds used on this compilation.
-  unsigned CompilationActiveOffloadHostKinds = 0u;
+  // Builder to be used to build offloading actions.
+  OffloadingActionBuilder OffloadBuilder(C, Args, Inputs);
 
   // Construct the actions to perform.
   ActionList LinkerInputs;
 
   llvm::SmallVector<phases::ID, phases::MaxNumberOfPhases> PL;
   for (auto &I : Inputs) {
     types::ID InputType = I.first;
     const Arg *InputArg = I.second;
if (YcArg) {
         assert(ClangClPch);
         Actions.push_back(ClangClPch);
         // The driver currently exits after the first failed command.  This
         // relies on that behavior, to make sure if the pch generation fails,
         // the main compilation won't run.
       }
     }
 
-    phases::ID CudaInjectionPhase =
-        (phases::Compile < FinalPhase &&
-         llvm::find(PL, phases::Compile) != PL.end())
-            ? phases::Compile
-            : FinalPhase;
-
-    // Track the host offload kinds used on this input.
-    unsigned InputActiveOffloadHostKinds = 0u;
-
     // Build the pipeline for this file.
     Action *Current = C.MakeAction<InputAction>(*InputArg, InputType);
+
+    // Use the current host action in any of the offloading actions, if
+    // required.
+    if (OffloadBuilder.addHostDependenceToDeviceActions(Current, InputArg))
+      break;
+
     for (SmallVectorImpl<phases::ID>::iterator i = PL.begin(), e = PL.end();
          i != e; ++i) {
       phases::ID Phase = *i;
 
       // We are done if this step is past what the user requested.
       if (Phase > FinalPhase)
         break;
 
+      // Add any offload action the host action depends on.
+      Current = OffloadBuilder.addDeviceDependencesToHostAction(
+          Current, InputArg, Phase, FinalPhase, PL);
+      if (!Current)
+        break;
+
       // Queue linker inputs.
       if (Phase == phases::Link) {
         assert((i + 1) == e && "linking must be final compilation step.");
         LinkerInputs.push_back(Current);
         Current = nullptr;
         break;
       }
 
-      // Some types skip the assembler phase (e.g., llvm-bc), but we can't
-      // encode this in the steps because the intermediate type depends on
-      // arguments. Just special case here.
-      if (Phase == phases::Assemble && Current->getType() != types::TY_PP_Asm)
+      // Otherwise construct the appropriate action.
+      auto *NewCurrent = ConstructPhaseAction(C, Args, Phase, Current);
+
+      // We didn't create a new action, so we will just move to the next phase.
+      if (NewCurrent == Current)
         continue;
 
-      // Otherwise construct the appropriate action.
-      Current = ConstructPhaseAction(C, Args, Phase, Current);
+      Current = NewCurrent;
 
-      if (InputType == types::TY_CUDA && Phase == CudaInjectionPhase) {
-        Current = buildCudaActions(C, Args, InputArg, Current, Actions);
-        if (!Current)
-          break;
+      // Use the current host action in any of the offloading actions, if
+      // required.
+      if (OffloadBuilder.addHostDependenceToDeviceActions(Current, InputArg))
 
-        // We produced a CUDA action for this input, so the host has to support
-        // CUDA.
-        InputActiveOffloadHostKinds |= Action::OFK_Cuda;
-        CompilationActiveOffloadHostKinds |= Action::OFK_Cuda;
-      }
-
       if (Current->getType() == types::TY_Nothing)
         break;
     }
 
-    // If we ended with something, add to the output list. Also, propagate the
-    // offload information to the top-level host action related with the current
-    // input.
-    if (Current) {
-      if (InputActiveOffloadHostKinds)
-        Current->propagateHostOffloadInfo(InputActiveOffloadHostKinds,
-                                          /*BoundArch=*/nullptr);
+    // If we ended with something, add to the output list.
+    if (Current)
       Actions.push_back(Current);
-    }
+
+    // Add any top level actions generated for offloading.
+    OffloadBuilder.appendTopLevelActions(Actions, Current, InputArg);
   }
 
-  // Add a link action if necessary and propagate the offload information for
-  // the current compilation.
+  // Add a link action if necessary.
   if (!LinkerInputs.empty()) {
-    Actions.push_back(
-        C.MakeAction<LinkJobAction>(LinkerInputs, types::TY_Image));
-    Actions.back()->propagateHostOffloadInfo(CompilationActiveOffloadHostKinds,
-                                             /*BoundArch=*/nullptr);
+    Action *LA = C.MakeAction<LinkJobAction>(LinkerInputs, types::TY_Image);
+    LA = OffloadBuilder.processHostLinkAction(LA);
+    Actions.push_back(LA);
   }
 
   // If we are linking, claim any options which are obviously only used for
   // compilation.
   if (FinalPhase == phases::Link && PL.size() == 1) {
     Args.ClaimAllArgs(options::OPT_CompileOnly_Group);
     Args.ClaimAllArgs(options::OPT_cl_compile_Group);
   }
 
   // Claim ignored clang-cl options.
   Args.ClaimAllArgs(options::OPT_cl_ignored_Group);
 
   // Claim --cuda-host-only and --cuda-compile-host-device, which may be passed
   // to non-CUDA compilations and should not trigger warnings there.
   Args.ClaimAllArgs(options::OPT_cuda_host_only);
   Args.ClaimAllArgs(options::OPT_cuda_compile_host_device);
 }
 
 Action *Driver::ConstructPhaseAction(Compilation &C, const ArgList &Args,
                                      phases::ID Phase, Action *Input) const {
   llvm::PrettyStackTraceString CrashInfo("Constructing phase actions");
+
+  // Some types skip the assembler phase (e.g., llvm-bc), but we can't
+  // encode this in the steps because the intermediate type depends on
+  // arguments. Just special case here.
+  if (Phase == phases::Assemble && Input->getType() != types::TY_PP_Asm)
+    return Input;
+
   // Build the appropriate action.
   switch (Phase) {
   case phases::Link:
     llvm_unreachable("link action invalid here.");
   case phases::Preprocess: {
     types::ID OutputTy;
     // -{M, MM} alter the output type.
     if (Args.hasArg(options::OPT_M, options::OPT_MM)) {

cfe/trunk/lib/Driver/Types.cpp

if (!onlyPrecompileType(Id)) {
       if (!onlyAssembleType(Id)) {
         P.push_back(phases::Compile);
         P.push_back(phases::Backend);
       }
       P.push_back(phases::Assemble);
     }
   }
 
-  if (!onlyPrecompileType(Id) && Id != TY_CUDA_DEVICE) {
+  if (!onlyPrecompileType(Id)) {
     P.push_back(phases::Link);
   }
   assert(0 < P.size() && "Not enough phases in list");
   assert(P.size() <= phases::MaxNumberOfPhases && "Too many phases in list");
 }
 
 ID types::lookupCXXTypeForCType(ID Id) {
   switch (Id) {

cfe/trunk/test/Driver/cuda-bindings.cu

 // BIN-NOT: cuda-bindings-device-cuda-nvptx64
 // BIN: # "powerpc64le-ibm-linux-gnu" - "GNU::Linker", inputs:{{.*}}, output: "a.out"
 
 //
 // Test single gpu architecture up to the assemble phase.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-bindings --cuda-gpu-arch=sm_30 %s -S 2>&1 \
 // RUN: | FileCheck -check-prefix=ASM %s
-// ASM: # "nvptx64-nvidia-cuda" - "clang",{{.*}} output: "cuda-bindings-device-cuda-nvptx64-nvidia-cuda-sm_30.s"
-// ASM: # "powerpc64le-ibm-linux-gnu" - "clang",{{.*}} output: "cuda-bindings.s"
+// ASM-DAG: # "nvptx64-nvidia-cuda" - "clang",{{.*}} output: "cuda-bindings-device-cuda-nvptx64-nvidia-cuda-sm_30.s"
+// ASM-DAG: # "powerpc64le-ibm-linux-gnu" - "clang",{{.*}} output: "cuda-bindings.s"
 
 //
 // Test two gpu architectures with complete compilation.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-bindings --cuda-gpu-arch=sm_30 --cuda-gpu-arch=sm_35 %s 2>&1 \
 // RUN: | FileCheck -check-prefix=BIN2 %s
 // BIN2: # "nvptx64-nvidia-cuda" - "clang",{{.*}} output:
 // BIN2-NOT: cuda-bindings-device-cuda-nvptx64

 // BIN2: # "powerpc64le-ibm-linux-gnu" - "GNU::Linker", inputs:{{.*}}, output: "a.out"
 
 //
 // Test two gpu architectures up to the assemble phase.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-bindings \
 // RUN:        --cuda-gpu-arch=sm_30 --cuda-gpu-arch=sm_35 %s -S 2>&1 \
 // RUN: | FileCheck -check-prefix=ASM2 %s
-// ASM2: # "nvptx64-nvidia-cuda" - "clang",{{.*}} output: "cuda-bindings-device-cuda-nvptx64-nvidia-cuda-sm_30.s"
-// ASM2: # "nvptx64-nvidia-cuda" - "clang",{{.*}} output: "cuda-bindings-device-cuda-nvptx64-nvidia-cuda-sm_35.s"
-// ASM2: # "powerpc64le-ibm-linux-gnu" - "clang",{{.*}} output: "cuda-bindings.s"
+// ASM2-DAG: # "nvptx64-nvidia-cuda" - "clang",{{.*}} output: "cuda-bindings-device-cuda-nvptx64-nvidia-cuda-sm_30.s"
+// ASM2-DAG: # "nvptx64-nvidia-cuda" - "clang",{{.*}} output: "cuda-bindings-device-cuda-nvptx64-nvidia-cuda-sm_35.s"
+// ASM2-DAG: # "powerpc64le-ibm-linux-gnu" - "clang",{{.*}} output: "cuda-bindings.s"
 
 //
 // Test one or more gpu architecture with complete compilation in host-only
 // compilation mode.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-bindings \
 // RUN:        --cuda-gpu-arch=sm_30 %s --cuda-host-only 2>&1 \
 // RUN: | FileCheck -check-prefix=HBIN %s

cfe/trunk/test/Driver/cuda-phases.cu

 // Tests the phases generated for a CUDA offloading target for different
 // combinations of:
 // - Number of gpu architectures;
 // - Host/device-only compilation;
 // - User-requested final phase - binary or assembly.
 
 // REQUIRES: clang-driver
 // REQUIRES: powerpc-registered-target
 // REQUIRES: nvptx-registered-target
 
 //
 // Test single gpu architecture with complete compilation.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 %s 2>&1 \
 // RUN: | FileCheck -check-prefix=BIN %s
-// BIN: 0: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
-// BIN: 1: preprocessor, {0}, cuda-cpp-output, (host-cuda)
-// BIN: 2: compiler, {1}, ir, (host-cuda)
-// BIN: 3: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
-// BIN: 4: preprocessor, {3}, cuda-cpp-output, (device-cuda, sm_30)
-// BIN: 5: compiler, {4}, ir, (device-cuda, sm_30)
-// BIN: 6: backend, {5}, assembler, (device-cuda, sm_30)
-// BIN: 7: assembler, {6}, object, (device-cuda, sm_30)
-// BIN: 8: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {7}, object
-// BIN: 9: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {6}, assembler
-// BIN: 10: linker, {8, 9}, cuda-fatbin, (device-cuda)
-// BIN: 11: offload, "host-cuda (powerpc64le-ibm-linux-gnu)" {2}, "device-cuda (nvptx64-nvidia-cuda)" {10}, ir
-// BIN: 12: backend, {11}, assembler, (host-cuda)
-// BIN: 13: assembler, {12}, object, (host-cuda)
-// BIN: 14: linker, {13}, image, (host-cuda)
+// BIN-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
+// BIN-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (host-cuda)
+// BIN-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (host-cuda)
+// BIN-DAG: [[P3:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
+// BIN-DAG: [[P4:[0-9]+]]: preprocessor, {[[P3]]}, cuda-cpp-output, (device-cuda, sm_30)
+// BIN-DAG: [[P5:[0-9]+]]: compiler, {[[P4]]}, ir, (device-cuda, sm_30)
+// BIN-DAG: [[P6:[0-9]+]]: backend, {[[P5]]}, assembler, (device-cuda, sm_30)
+// BIN-DAG: [[P7:[0-9]+]]: assembler, {[[P6]]}, object, (device-cuda, sm_30)
+// BIN-DAG: [[P8:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {[[P7]]}, object
+// BIN-DAG: [[P9:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {[[P6]]}, assembler
+// BIN-DAG: [[P10:[0-9]+]]: linker, {[[P8]], [[P9]]}, cuda-fatbin, (device-cuda)
+// BIN-DAG: [[P11:[0-9]+]]: offload, "host-cuda (powerpc64le-ibm-linux-gnu)" {[[P2]]}, "device-cuda (nvptx64-nvidia-cuda)" {[[P10]]}, ir
+// BIN-DAG: [[P12:[0-9]+]]: backend, {[[P11]]}, assembler, (host-cuda)
+// BIN-DAG: [[P13:[0-9]+]]: assembler, {[[P12]]}, object, (host-cuda)
+// BIN-DAG: [[P14:[0-9]+]]: linker, {[[P13]]}, image, (host-cuda)
 
 //
 // Test single gpu architecture up to the assemble phase.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 %s -S 2>&1 \
 // RUN: | FileCheck -check-prefix=ASM %s
-// ASM: 0: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
-// ASM: 1: preprocessor, {0}, cuda-cpp-output, (device-cuda, sm_30)
-// ASM: 2: compiler, {1}, ir, (device-cuda, sm_30)
-// ASM: 3: backend, {2}, assembler, (device-cuda, sm_30)
-// ASM: 4: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {3}, assembler
-// ASM: 5: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
-// ASM: 6: preprocessor, {5}, cuda-cpp-output, (host-cuda)
-// ASM: 7: compiler, {6}, ir, (host-cuda)
-// ASM: 8: backend, {7}, assembler, (host-cuda)
+// ASM-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
+// ASM-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (device-cuda, sm_30)
+// ASM-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (device-cuda, sm_30)
+// ASM-DAG: [[P3:[0-9]+]]: backend, {[[P2]]}, assembler, (device-cuda, sm_30)
+// ASM-DAG: [[P4:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {[[P3]]}, assembler
+// ASM-DAG: [[P5:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
+// ASM-DAG: [[P6:[0-9]+]]: preprocessor, {[[P5]]}, cuda-cpp-output, (host-cuda)
+// ASM-DAG: [[P7:[0-9]+]]: compiler, {[[P6]]}, ir, (host-cuda)
+// ASM-DAG: [[P8:[0-9]+]]: backend, {[[P7]]}, assembler, (host-cuda)
 
 //
 // Test two gpu architectures with complete compilation.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 --cuda-gpu-arch=sm_35 %s 2>&1 \
 // RUN: | FileCheck -check-prefix=BIN2 %s
-// BIN2: 0: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
-// BIN2: 1: preprocessor, {0}, cuda-cpp-output, (host-cuda)
-// BIN2: 2: compiler, {1}, ir, (host-cuda)
-// BIN2: 3: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
-// BIN2: 4: preprocessor, {3}, cuda-cpp-output, (device-cuda, sm_30)
-// BIN2: 5: compiler, {4}, ir, (device-cuda, sm_30)
-// BIN2: 6: backend, {5}, assembler, (device-cuda, sm_30)
-// BIN2: 7: assembler, {6}, object, (device-cuda, sm_30)
-// BIN2: 8: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {7}, object
-// BIN2: 9: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {6}, assembler
-// BIN2: 10: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_35)
-// BIN2: 11: preprocessor, {10}, cuda-cpp-output, (device-cuda, sm_35)
-// BIN2: 12: compiler, {11}, ir, (device-cuda, sm_35)
-// BIN2: 13: backend, {12}, assembler, (device-cuda, sm_35)
-// BIN2: 14: assembler, {13}, object, (device-cuda, sm_35)
-// BIN2: 15: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {14}, object
-// BIN2: 16: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {13}, assembler
-// BIN2: 17: linker, {8, 9, 15, 16}, cuda-fatbin, (device-cuda)
-// BIN2: 18: offload, "host-cuda (powerpc64le-ibm-linux-gnu)" {2}, "device-cuda (nvptx64-nvidia-cuda)" {17}, ir
-// BIN2: 19: backend, {18}, assembler, (host-cuda)
-// BIN2: 20: assembler, {19}, object, (host-cuda)
-// BIN2: 21: linker, {20}, image, (host-cuda)
+// BIN2-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
+// BIN2-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (host-cuda)
+// BIN2-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (host-cuda)
+// BIN2-DAG: [[P3:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
+// BIN2-DAG: [[P4:[0-9]+]]: preprocessor, {[[P3]]}, cuda-cpp-output, (device-cuda, sm_30)
+// BIN2-DAG: [[P5:[0-9]+]]: compiler, {[[P4]]}, ir, (device-cuda, sm_30)
+// BIN2-DAG: [[P6:[0-9]+]]: backend, {[[P5]]}, assembler, (device-cuda, sm_30)
+// BIN2-DAG: [[P7:[0-9]+]]: assembler, {[[P6]]}, object, (device-cuda, sm_30)
+// BIN2-DAG: [[P8:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {[[P7]]}, object
+// BIN2-DAG: [[P9:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {[[P6]]}, assembler
+// BIN2-DAG: [[P10:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_35)
+// BIN2-DAG: [[P11:[0-9]+]]: preprocessor, {[[P10]]}, cuda-cpp-output, (device-cuda, sm_35)
+// BIN2-DAG: [[P12:[0-9]+]]: compiler, {[[P11]]}, ir, (device-cuda, sm_35)
+// BIN2-DAG: [[P13:[0-9]+]]: backend, {[[P12]]}, assembler, (device-cuda, sm_35)
+// BIN2-DAG: [[P14:[0-9]+]]: assembler, {[[P13]]}, object, (device-cuda, sm_35)
+// BIN2-DAG: [[P15:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {[[P14]]}, object
+// BIN2-DAG: [[P16:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {[[P13]]}, assembler
+// BIN2-DAG: [[P17:[0-9]+]]: linker, {[[P8]], [[P9]], [[P15]], [[P16]]}, cuda-fatbin, (device-cuda)
+// BIN2-DAG: [[P18:[0-9]+]]: offload, "host-cuda (powerpc64le-ibm-linux-gnu)" {[[P2]]}, "device-cuda (nvptx64-nvidia-cuda)" {[[P17]]}, ir
+// BIN2-DAG: [[P19:[0-9]+]]: backend, {[[P18]]}, assembler, (host-cuda)
+// BIN2-DAG: [[P20:[0-9]+]]: assembler, {[[P19]]}, object, (host-cuda)
+// BIN2-DAG: [[P21:[0-9]+]]: linker, {[[P20]]}, image, (host-cuda)
 
 //
 // Test two gpu architecturess up to the assemble phase.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 --cuda-gpu-arch=sm_35 %s -S 2>&1 \
 // RUN: | FileCheck -check-prefix=ASM2 %s
-// ASM2: 0: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
-// ASM2: 1: preprocessor, {0}, cuda-cpp-output, (device-cuda, sm_30)
-// ASM2: 2: compiler, {1}, ir, (device-cuda, sm_30)
-// ASM2: 3: backend, {2}, assembler, (device-cuda, sm_30)
-// ASM2: 4: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {3}, assembler
-// ASM2: 5: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_35)
-// ASM2: 6: preprocessor, {5}, cuda-cpp-output, (device-cuda, sm_35)
-// ASM2: 7: compiler, {6}, ir, (device-cuda, sm_35)
-// ASM2: 8: backend, {7}, assembler, (device-cuda, sm_35)
-// ASM2: 9: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {8}, assembler
-// ASM2: 10: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
-// ASM2: 11: preprocessor, {10}, cuda-cpp-output, (host-cuda)
-// ASM2: 12: compiler, {11}, ir, (host-cuda)
-// ASM2: 13: backend, {12}, assembler, (host-cuda)
+// ASM2-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
+// ASM2-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (device-cuda, sm_30)
+// ASM2-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (device-cuda, sm_30)
+// ASM2-DAG: [[P3:[0-9]+]]: backend, {[[P2]]}, assembler, (device-cuda, sm_30)
+// ASM2-DAG: [[P4:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {[[P3]]}, assembler
+// ASM2-DAG: [[P5:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_35)
+// ASM2-DAG: [[P6:[0-9]+]]: preprocessor, {[[P5]]}, cuda-cpp-output, (device-cuda, sm_35)
+// ASM2-DAG: [[P7:[0-9]+]]: compiler, {[[P6]]}, ir, (device-cuda, sm_35)
+// ASM2-DAG: [[P8:[0-9]+]]: backend, {[[P7]]}, assembler, (device-cuda, sm_35)
+// ASM2-DAG: [[P9:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {[[P8]]}, assembler
+// ASM2-DAG: [[P10:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
+// ASM2-DAG: [[P11:[0-9]+]]: preprocessor, {[[P10]]}, cuda-cpp-output, (host-cuda)
+// ASM2-DAG: [[P12:[0-9]+]]: compiler, {[[P11]]}, ir, (host-cuda)
+// ASM2-DAG: [[P13:[0-9]+]]: backend, {[[P12]]}, assembler, (host-cuda)
 
 //
 // Test single gpu architecture with complete compilation in host-only
 // compilation mode.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 %s --cuda-host-only 2>&1 \
 // RUN: | FileCheck -check-prefix=HBIN %s
-// HBIN: 0: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
-// HBIN: 1: preprocessor, {0}, cuda-cpp-output, (host-cuda)
-// HBIN: 2: compiler, {1}, ir, (host-cuda)
-// HBIN: 3: offload, "host-cuda (powerpc64le-ibm-linux-gnu)" {2}, ir
-// HBIN: 4: backend, {3}, assembler, (host-cuda)
-// HBIN: 5: assembler, {4}, object, (host-cuda)
-// HBIN: 6: linker, {5}, image, (host-cuda)
-
+// HBIN-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
+// HBIN-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (host-cuda)
+// HBIN-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (host-cuda)
+// HBIN-DAG: [[P3:[0-9]+]]: backend, {[[P2]]}, assembler, (host-cuda)
+// HBIN-DAG: [[P4:[0-9]+]]: assembler, {[[P3]]}, object, (host-cuda)
+// HBIN-DAG: [[P5:[0-9]+]]: linker, {[[P4]]}, image, (host-cuda)
 //
 // Test single gpu architecture up to the assemble phase in host-only
 // compilation mode.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 %s --cuda-host-only -S 2>&1 \
 // RUN: | FileCheck -check-prefix=HASM %s
-// HASM: 0: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
-// HASM: 1: preprocessor, {0}, cuda-cpp-output, (host-cuda)
-// HASM: 2: compiler, {1}, ir, (host-cuda)
-// HASM: 3: offload, "host-cuda (powerpc64le-ibm-linux-gnu)" {2}, ir
-// HASM: 4: backend, {3}, assembler, (host-cuda)
+// HASM-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
+// HASM-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (host-cuda)
+// HASM-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (host-cuda)
+// HASM-DAG: [[P3:[0-9]+]]: backend, {[[P2]]}, assembler, (host-cuda)
 
 //
 // Test two gpu architectures with complete compilation in host-only
 // compilation mode.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 --cuda-gpu-arch=sm_35 %s --cuda-host-only 2>&1 \
 // RUN: | FileCheck -check-prefix=HBIN2 %s
-// HBIN2: 0: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
-// HBIN2: 1: preprocessor, {0}, cuda-cpp-output, (host-cuda)
-// HBIN2: 2: compiler, {1}, ir, (host-cuda)
-// HBIN2: 3: offload, "host-cuda (powerpc64le-ibm-linux-gnu)" {2}, ir
-// HBIN2: 4: backend, {3}, assembler, (host-cuda)
-// HBIN2: 5: assembler, {4}, object, (host-cuda)
-// HBIN2: 6: linker, {5}, image, (host-cuda)
+// HBIN2-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
+// HBIN2-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (host-cuda)
+// HBIN2-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (host-cuda)
+// HBIN2-DAG: [[P3:[0-9]+]]: backend, {[[P2]]}, assembler, (host-cuda)
+// HBIN2-DAG: [[P4:[0-9]+]]: assembler, {[[P3]]}, object, (host-cuda)
+// HBIN2-DAG: [[P5:[0-9]+]]: linker, {[[P4]]}, image, (host-cuda)
 
 //
 // Test two gpu architectures up to the assemble phase in host-only
 // compilation mode.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 --cuda-gpu-arch=sm_35 %s --cuda-host-only -S 2>&1 \
 // RUN: | FileCheck -check-prefix=HASM2 %s
-// HASM2: 0: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
-// HASM2: 1: preprocessor, {0}, cuda-cpp-output, (host-cuda)
-// HASM2: 2: compiler, {1}, ir, (host-cuda)
-// HASM2: 3: offload, "host-cuda (powerpc64le-ibm-linux-gnu)" {2}, ir
-// HASM2: 4: backend, {3}, assembler, (host-cuda)
+// HASM2-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (host-cuda)
+// HASM2-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (host-cuda)
+// HASM2-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (host-cuda)
+// HASM2-DAG: [[P3:[0-9]+]]: backend, {[[P2]]}, assembler, (host-cuda)
 
 //
 // Test single gpu architecture with complete compilation in device-only
 // compilation mode.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 %s --cuda-device-only 2>&1 \
 // RUN: | FileCheck -check-prefix=DBIN %s
-// DBIN: 0: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
-// DBIN: 1: preprocessor, {0}, cuda-cpp-output, (device-cuda, sm_30)
-// DBIN: 2: compiler, {1}, ir, (device-cuda, sm_30)
-// DBIN: 3: backend, {2}, assembler, (device-cuda, sm_30)
-// DBIN: 4: assembler, {3}, object, (device-cuda, sm_30)
-// DBIN: 5: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {4}, object
+// DBIN-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
+// DBIN-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (device-cuda, sm_30)
+// DBIN-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (device-cuda, sm_30)
+// DBIN-DAG: [[P3:[0-9]+]]: backend, {[[P2]]}, assembler, (device-cuda, sm_30)
+// DBIN-DAG: [[P4:[0-9]+]]: assembler, {[[P3]]}, object, (device-cuda, sm_30)
+// DBIN-DAG: [[P5:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {[[P4]]}, object
 
 //
 // Test single gpu architecture up to the assemble phase in device-only
 // compilation mode.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 %s --cuda-device-only -S 2>&1 \
 // RUN: | FileCheck -check-prefix=DASM %s
-// DASM: 0: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
-// DASM: 1: preprocessor, {0}, cuda-cpp-output, (device-cuda, sm_30)
-// DASM: 2: compiler, {1}, ir, (device-cuda, sm_30)
-// DASM: 3: backend, {2}, assembler, (device-cuda, sm_30)
-// DASM: 4: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {3}, assembler
+// DASM-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
+// DASM-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (device-cuda, sm_30)
+// DASM-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (device-cuda, sm_30)
+// DASM-DAG: [[P3:[0-9]+]]: backend, {[[P2]]}, assembler, (device-cuda, sm_30)
+// DASM-DAG: [[P4:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {[[P3]]}, assembler
 
 //
 // Test two gpu architectures with complete compilation in device-only
 // compilation mode.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 --cuda-gpu-arch=sm_35 %s --cuda-device-only 2>&1 \
 // RUN: | FileCheck -check-prefix=DBIN2 %s
-// DBIN2: 0: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
-// DBIN2: 1: preprocessor, {0}, cuda-cpp-output, (device-cuda, sm_30)
-// DBIN2: 2: compiler, {1}, ir, (device-cuda, sm_30)
-// DBIN2: 3: backend, {2}, assembler, (device-cuda, sm_30)
-// DBIN2: 4: assembler, {3}, object, (device-cuda, sm_30)
-// DBIN2: 5: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {4}, object
-// DBIN2: 6: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_35)
-// DBIN2: 7: preprocessor, {6}, cuda-cpp-output, (device-cuda, sm_35)
-// DBIN2: 8: compiler, {7}, ir, (device-cuda, sm_35)
-// DBIN2: 9: backend, {8}, assembler, (device-cuda, sm_35)
-// DBIN2: 10: assembler, {9}, object, (device-cuda, sm_35)
-// DBIN2: 11: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {10}, object
+// DBIN2-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
+// DBIN2-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (device-cuda, sm_30)
+// DBIN2-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (device-cuda, sm_30)
+// DBIN2-DAG: [[P3:[0-9]+]]: backend, {[[P2]]}, assembler, (device-cuda, sm_30)
+// DBIN2-DAG: [[P4:[0-9]+]]: assembler, {[[P3]]}, object, (device-cuda, sm_30)
+// DBIN2-DAG: [[P5:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {[[P4]]}, object
+// DBIN2-DAG: [[P6:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_35)
+// DBIN2-DAG: [[P7:[0-9]+]]: preprocessor, {[[P6]]}, cuda-cpp-output, (device-cuda, sm_35)
+// DBIN2-DAG: [[P8:[0-9]+]]: compiler, {[[P7]]}, ir, (device-cuda, sm_35)
+// DBIN2-DAG: [[P9:[0-9]+]]: backend, {[[P8]]}, assembler, (device-cuda, sm_35)
+// DBIN2-DAG: [[P10:[0-9]+]]: assembler, {[[P9]]}, object, (device-cuda, sm_35)
+// DBIN2-DAG: [[P11:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {[[P10]]}, object
 
 //
 // Test two gpu architectures up to the assemble phase in device-only
 // compilation mode.
 //
 // RUN: %clang -target powerpc64le-ibm-linux-gnu -ccc-print-phases --cuda-gpu-arch=sm_30 --cuda-gpu-arch=sm_35 %s --cuda-device-only -S 2>&1 \
 // RUN: | FileCheck -check-prefix=DASM2 %s
-// DASM2: 0: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
-// DASM2: 1: preprocessor, {0}, cuda-cpp-output, (device-cuda, sm_30)
-// DASM2: 2: compiler, {1}, ir, (device-cuda, sm_30)
-// DASM2: 3: backend, {2}, assembler, (device-cuda, sm_30)
-// DASM2: 4: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {3}, assembler
-// DASM2: 5: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_35)
-// DASM2: 6: preprocessor, {5}, cuda-cpp-output, (device-cuda, sm_35)
-// DASM2: 7: compiler, {6}, ir, (device-cuda, sm_35)
-// DASM2: 8: backend, {7}, assembler, (device-cuda, sm_35)
-// DASM2: 9: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {8}, assembler
+// DASM2-DAG: [[P0:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_30)
+// DASM2-DAG: [[P1:[0-9]+]]: preprocessor, {[[P0]]}, cuda-cpp-output, (device-cuda, sm_30)
+// DASM2-DAG: [[P2:[0-9]+]]: compiler, {[[P1]]}, ir, (device-cuda, sm_30)
+// DASM2-DAG: [[P3:[0-9]+]]: backend, {[[P2]]}, assembler, (device-cuda, sm_30)
+// DASM2-DAG: [[P4:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_30)" {[[P3]]}, assembler
+// DASM2-DAG: [[P5:[0-9]+]]: input, "{{.*}}cuda-phases.cu", cuda, (device-cuda, sm_35)
+// DASM2-DAG: [[P6:[0-9]+]]: preprocessor, {[[P5]]}, cuda-cpp-output, (device-cuda, sm_35)
+// DASM2-DAG: [[P7:[0-9]+]]: compiler, {[[P6]]}, ir, (device-cuda, sm_35)
+// DASM2-DAG: [[P8:[0-9]+]]: backend, {[[P7]]}, assembler, (device-cuda, sm_35)
+// DASM2-DAG: [[P9:[0-9]+]]: offload, "device-cuda (nvptx64-nvidia-cuda:sm_35)" {[[P8]]}, assembler