[hip] Fix device-only relocatable code compilation.

[hliao]

Summary

• In HIP, just as the regular device-only compilation, the device-only relocatable code compilation should not involve offload bundle.
• In addition, that device-only relocatable code compilation should have the similar 3 steps, namely preprocessor, compile, and backend, to the regular code generation with -emit-llvm.

[tra]

LGTM in general, but I'll let Sam stamp it.

clang/lib/Driver/Driver.cpp

4594–4599

This is rather hard to understand.
Would it be simpler to specify when we shouldn't add .tmp?
At the very least I'd extract the newly added clause into a temporary variable and would add some comments explaining why -fgpu-rdc gets special treatment.

clang/test/Driver/hip-device-only.hip

6

We appear to test -fgpu-rdc only and the test case seems to be fairly specific to the way that option is handled by HIP.
Perhaps the test file should reflect that. hip-rdc-device-only ?

[hliao]

Revise following reviewer's comment.

[yaxunl]

LGTM. thanks

[thakis]

This doesn't pass tests: http://45.33.8.238/linux/19977/step_7.txt

Please take a look, and please revert for now if fixing takes a while.

[hliao]

In D81427#2085739, @thakis wrote:

This doesn't pass tests: http://45.33.8.238/linux/19977/step_7.txt

Please take a look, and please revert for now if fixing takes a while.

Thanks. I am building with PowerPC enabled to reproduce this issue. There should be a minor fix on that test case.

[hliao]

In D81427#2085739, @thakis wrote:

This doesn't pass tests: http://45.33.8.238/linux/19977/step_7.txt

Please take a look, and please revert for now if fixing takes a while.

Fixed in https://github.com/llvm/llvm-project/commit/6dd058083208d58c6a7005bfb092cee085fd9a48

clang/lib/Driver/Driver.cpp

public:
     ActionBuilderReturnCode
     getDeviceDependences(OffloadAction::DeviceDependences &DA,
                          phases::ID CurPhase, phases::ID FinalPhase,
                          PhasesTy &Phases) override {
       // amdgcn does not support linking of object files, therefore we skip
       // backend and assemble phases to output LLVM IR. Except for generating
       // non-relocatable device coee, where we generate fat binary for device
       // code and pass to host in Backend phase.
-      if (CudaDeviceActions.empty() ||
-          (CurPhase == phases::Backend && Relocatable) ||
-          CurPhase == phases::Assemble)
+      if (CudaDeviceActions.empty())
         return ABRT_Success;
 
       assert(((CurPhase == phases::Link && Relocatable) ||
               CudaDeviceActions.size() == GpuArchList.size()) &&
              "Expecting one action per GPU architecture.");
       assert(!CompileHostOnly &&
              "Not expecting CUDA actions in host-only compilation.");
 
getDeviceDependences(OffloadAction::DeviceDependences &DA,
 
         // We will pass the device action as a host dependence, so we don't
         // need to do anything else with them.
         CudaDeviceActions.clear();
         return ABRT_Success;
       }
 
       // By default, we produce an action for each device arch.
+      if (!Relocatable || CurPhase <= phases::Backend) {
         for (Action *&A : CudaDeviceActions)
           A = C.getDriver().ConstructPhaseAction(C, Args, CurPhase, A,
                                                  AssociatedOffloadKind);
+      }
 
       return (CompileDeviceOnly && CurPhase == FinalPhase) ? ABRT_Ignore_Host
                                                            : ABRT_Success;
     }
 
     void appendLinkDependences(OffloadAction::DeviceDependences &DA) override {
       // Append a new link action for each device.
       unsigned I = 0;
case phases::Compile: {
     return C.MakeAction<CompileJobAction>(Input, types::TY_LLVM_BC);
   }
   case phases::Backend: {
     if (isUsingLTO() && TargetDeviceOffloadKind == Action::OFK_None) {
       types::ID Output =
           Args.hasArg(options::OPT_S) ? types::TY_LTO_IR : types::TY_LTO_BC;
       return C.MakeAction<BackendJobAction>(Input, Output);
     }
-    if (Args.hasArg(options::OPT_emit_llvm)) {
+    if (Args.hasArg(options::OPT_emit_llvm) ||
+        (TargetDeviceOffloadKind == Action::OFK_HIP &&
+         Args.hasFlag(options::OPT_fgpu_rdc, options::OPT_fno_gpu_rdc,
+                      false))) {
       types::ID Output =
           Args.hasArg(options::OPT_S) ? types::TY_LLVM_IR : types::TY_LLVM_BC;
       return C.MakeAction<BackendJobAction>(Input, Output);
     }
     return C.MakeAction<BackendJobAction>(Input, types::TY_PP_Asm);
   }
   case phases::Assemble:
     return C.MakeAction<AssembleJobAction>(std::move(Input), types::TY_Object);
if ((JA.getType() == types::TY_Object || JA.getType() == types::TY_LTO_BC) &&
     Suffixed += OffloadingPrefix;
     if (MultipleArchs && !BoundArch.empty()) {
       Suffixed += "-";
       Suffixed.append(BoundArch);
     }
     // When using both -save-temps and -emit-llvm, use a ".tmp.bc" suffix for
     // the unoptimized bitcode so that it does not get overwritten by the ".bc"
     // optimized bitcode output.
-    if (!AtTopLevel && C.getArgs().hasArg(options::OPT_emit_llvm) &&
-        JA.getType() == types::TY_LLVM_BC)
+    if (!AtTopLevel && JA.getType() == types::TY_LLVM_BC &&
+        (C.getArgs().hasArg(options::OPT_emit_llvm) ||
+         (JA.getOffloadingDeviceKind() == Action::OFK_HIP &&
+          isa<CompileJobAction>(JA) &&
+          C.getArgs().hasFlag(options::OPT_fgpu_rdc, options::OPT_fno_gpu_rdc,
+                              false))))
       Suffixed += ".tmp";
     Suffixed += '.';
     Suffixed += Suffix;
     NamedOutput = C.getArgs().MakeArgString(Suffixed.c_str());
   }
 
   // Prepend object file path if -save-temps=obj
   if (!AtTopLevel && isSaveTempsObj() && C.getArgs().hasArg(options::OPT_o) &&

clang/test/Driver/hip-device-only.hip

+// REQUIRES: clang-driver
+// REQUIRES: x86-registered-target
+// REQUIRES: amdgpu-registered-target
+
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   -x hip --cuda-gpu-arch=gfx803 --cuda-gpu-arch=gfx900 \
+// RUN:   -c -nogpuinc -nogpulib --cuda-device-only -fgpu-rdc \
+// RUN:   %S/Inputs/hip_multiple_inputs/a.cu \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck -check-prefixes=COMMON,EMITBC %s
+
+// With `-emit-llvm`, the output should be the same as the aforementioned line
+// as `-fgpu-rdc` in HIP implies `-emit-llvm`.
+
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   -x hip --cuda-gpu-arch=gfx803 --cuda-gpu-arch=gfx900 \
+// RUN:   -c -emit-llvm -nogpuinc -nogpulib --cuda-device-only -fgpu-rdc \
+// RUN:   %S/Inputs/hip_multiple_inputs/a.cu \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck -check-prefixes=COMMON,EMITBC %s
+
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   -x hip --cuda-gpu-arch=gfx803 --cuda-gpu-arch=gfx900 \
+// RUN:   -S -nogpuinc -nogpulib --cuda-device-only -fgpu-rdc \
+// RUN:   %S/Inputs/hip_multiple_inputs/a.cu \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck -check-prefixes=COMMON,EMITLL %s
+
+// With `-emit-llvm`, the output should be the same as the aforementioned line
+// as `-fgpu-rdc` in HIP implies `-emit-llvm`.
+
+// RUN: %clang -### -target x86_64-linux-gnu \
+// RUN:   -x hip --cuda-gpu-arch=gfx803 --cuda-gpu-arch=gfx900 \
+// RUN:   -S -emit-llvm -nogpuinc -nogpulib --cuda-device-only -fgpu-rdc \
+// RUN:   %S/Inputs/hip_multiple_inputs/a.cu \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck -check-prefixes=COMMON,EMITLL %s
+
+// With `-save-temps`, commane lines for each steps are dumped. For assembly
+// output, there should 3 steps (preprocessor, compile, and backend) per source
+// and per target, totally 12 steps.
+
+// RUN: %clang -### -save-temps -target x86_64-linux-gnu \
+// RUN:   -x hip --cuda-gpu-arch=gfx803 --cuda-gpu-arch=gfx900 \
+// RUN:   -S -nogpuinc -nogpulib --cuda-device-only -fgpu-rdc \
+// RUN:   %S/Inputs/hip_multiple_inputs/a.cu \
+// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck -check-prefix=SAVETEMP %s
+
+// COMMON: [[CLANG:".*clang.*"]] "-cc1" "-triple" "amdgcn-amd-amdhsa"
+// COMMON-SAME: "-aux-triple" "x86_64-unknown-linux-gnu"
+// EMITBC-SAME: "-emit-llvm-bc"
+// EMITLL-SAME: "-emit-llvm"
+// COMMON-SAME: {{.*}} "-main-file-name" "a.cu" {{.*}} "-target-cpu" "gfx803"
+// COMMON-SAME: "-fcuda-is-device" "-fgpu-rdc" "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
+// COMMON-SAME: "-fapply-global-visibility-to-externs"
+// EMITBC-SAME: {{.*}} "-o" {{"a.*bc"}} "-x" "hip"
+// EMITLL-SAME: {{.*}} "-o" {{"a.*ll"}} "-x" "hip"
+// CHECK-SAME: {{.*}} {{".*a.cu"}}
+
+// COMMON: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa"
+// COMMON-SAME: "-aux-triple" "x86_64-unknown-linux-gnu"
+// EMITBC-SAME: "-emit-llvm-bc"
+// EMITLL-SAME: "-emit-llvm"
+// COMMON-SAME: {{.*}} "-main-file-name" "a.cu" {{.*}} "-target-cpu" "gfx900"
+// COMMON-SAME: "-fcuda-is-device" "-fgpu-rdc" "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
+// COMMON-SAME: "-fapply-global-visibility-to-externs"
+// EMITBC-SAME: {{.*}} "-o" {{"a.*bc"}} "-x" "hip"
+// EMITLL-SAME: {{.*}} "-o" {{"a.*ll"}} "-x" "hip"
+// COMMON-SAME: {{.*}} {{".*a.cu"}}
+
+// COMMON: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa"
+// COMMON-SAME: "-aux-triple" "x86_64-unknown-linux-gnu"
+// EMITBC-SAME: "-emit-llvm-bc"
+// EMITLL-SAME: "-emit-llvm"
+// COMMON-SAME: {{.*}} "-main-file-name" "b.hip" {{.*}} "-target-cpu" "gfx803"
+// COMMON-SAME: "-fcuda-is-device" "-fgpu-rdc" "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
+// COMMON-SAME: "-fapply-global-visibility-to-externs"
+// EMITBC-SAME: {{.*}} "-o" {{"b.*bc"}} "-x" "hip"
+// EMITLL-SAME: {{.*}} "-o" {{"b.*ll"}} "-x" "hip"
+// COMMON-SAME: {{.*}} {{".*b.hip"}}
+
+// COMMON: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa"
+// COMMON-SAME: "-aux-triple" "x86_64-unknown-linux-gnu"
+// EMITBC-SAME: "-emit-llvm-bc"
+// EMITLL-SAME: "-emit-llvm"
+// COMMON-SAME: {{.*}} "-main-file-name" "b.hip" {{.*}} "-target-cpu" "gfx900"
+// COMMON-SAME: "-fcuda-is-device" "-fgpu-rdc" "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
+// COMMON-SAME: "-fapply-global-visibility-to-externs"
+// EMITBC-SAME: {{.*}} "-o" {{"b.*bc"}} "-x" "hip"
+// EMITLL-SAME: {{.*}} "-o" {{"b.*ll"}} "-x" "hip"
+// COMMON-SAME: {{.*}} {{".*b.hip"}}
+
+// SAVETEMP: [[CLANG:".*clang.*"]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-E"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "a.cu" {{.*}} "-target-cpu" "gfx803"
+// SAVETEMP-SAME: {{.*}} "-o" [[A_GFX803_CUI:"a.*cui"]] "-x" "hip" {{".*a.cu"}}
+// SAVETEMP-NEXT: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-emit-llvm-bc"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "a.cu" {{.*}} "-target-cpu" "gfx803"
+// SAVETEMP-SAME: {{.*}} "-o" [[A_GFX803_TMP_BC:"a.*tmp.bc"]] "-x" "hip-cpp-output" [[A_GFX803_CUI]]
+// SAVETEMP-NEXT: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-emit-llvm"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "a.cu" {{.*}} "-target-cpu" "gfx803"
+// SAVETEMP-SAME: {{.*}} "-o" {{"a.*.ll"}} "-x" "ir" [[A_GFX803_TMP_BC]]
+
+// SAVETEMP: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-E"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "a.cu" {{.*}} "-target-cpu" "gfx900"
+// SAVETEMP-SAME: {{.*}} "-o" [[A_GFX900_CUI:"a.*cui"]] "-x" "hip" {{".*a.cu"}}
+// SAVETEMP-NEXT: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-emit-llvm-bc"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "a.cu" {{.*}} "-target-cpu" "gfx900"
+// SAVETEMP-SAME: {{.*}} "-o" [[A_GFX900_TMP_BC:"a.*tmp.bc"]] "-x" "hip-cpp-output" [[A_GFX900_CUI]]
+// SAVETEMP-NEXT: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-emit-llvm"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "a.cu" {{.*}} "-target-cpu" "gfx900"
+// SAVETEMP-SAME: {{.*}} "-o" {{"a.*.ll"}} "-x" "ir" [[A_GFX900_TMP_BC]]
+
+// SAVETEMP: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-E"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "b.hip" {{.*}} "-target-cpu" "gfx803"
+// SAVETEMP-SAME: {{.*}} "-o" [[B_GFX803_CUI:"b.*cui"]] "-x" "hip" {{".*b.hip"}}
+// SAVETEMP-NEXT: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-emit-llvm-bc"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "b.hip" {{.*}} "-target-cpu" "gfx803"
+// SAVETEMP-SAME: {{.*}} "-o" [[B_GFX803_TMP_BC:"b.*tmp.bc"]] "-x" "hip-cpp-output" [[B_GFX803_CUI]]
+// SAVETEMP-NEXT: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-emit-llvm"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "b.hip" {{.*}} "-target-cpu" "gfx803"
+// SAVETEMP-SAME: {{.*}} "-o" {{"b.*.ll"}} "-x" "ir" [[B_GFX803_TMP_BC]]
+
+// SAVETEMP: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-E"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "b.hip" {{.*}} "-target-cpu" "gfx900"
+// SAVETEMP-SAME: {{.*}} "-o" [[B_GFX900_CUI:"b.*cui"]] "-x" "hip" {{".*b.hip"}}
+// SAVETEMP-NEXT: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-emit-llvm-bc"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "b.hip" {{.*}} "-target-cpu" "gfx900"
+// SAVETEMP-SAME: {{.*}} "-o" [[B_GFX900_TMP_BC:"b.*tmp.bc"]] "-x" "hip-cpp-output" [[B_GFX900_CUI]]
+// SAVETEMP-NEXT: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa" "-aux-triple" "x86_64-unknown-linux-gnu"
+// SAVETEMP-SAME: "-emit-llvm"
+// SAVETEMP-SAME: {{.*}} "-main-file-name" "b.hip" {{.*}} "-target-cpu" "gfx900"
+// SAVETEMP-SAME: {{.*}} "-o" {{"b.*.ll"}} "-x" "ir" [[B_GFX900_TMP_BC]]