Reapply "[AMDGPU] Use the AMDGPUToolChain when targeting C/C++ directly"

[jhuber6]

Summary:
This reverts commit 740e6ae.

After discussions it was determined that the behavior for IR inputs
needs to be maintained at least for now. In the future we should put a
deprecation notice on this behavior. This patch keeps the old behavior
for OpenCL and IR inputs, while others will be standalone. This is good
enough for standard compile flows.

[llvmbot]

@llvm/pr-subscribers-clang-driver
@llvm/pr-subscribers-backend-amdgpu

@llvm/pr-subscribers-clang

Author: Joseph Huber (jhuber6)

Changes

Summary:
This reverts commit 740e6ae.

After discussions it was determined that the behavior for IR inputs
needs to be maintained at least for now. In the future we should put a
deprecation notice on this behavior. This patch keeps the old beahvior
for OpenCL and IR inputs, while others will be standalone. This is good
enough for standard compile flows.

---

Full diff: https://github.com/llvm/llvm-project/pull/125744.diff

3 Files Affected:

• (modified) clang/lib/Driver/Driver.cpp (+20-2)
• (modified) clang/test/Driver/amdgpu-toolchain-opencl.cl (+5)
• (modified) clang/test/Driver/amdgpu-toolchain.c (+8)

diff --git a/clang/lib/Driver/Driver.cpp b/clang/lib/Driver/Driver.cpp
index 912777a9808b4be..7420ebb0fe50ed6 100644
--- a/clang/lib/Driver/Driver.cpp
+++ b/clang/lib/Driver/Driver.cpp
@@ -146,6 +146,7 @@ getNVIDIAOffloadTargetTriple(const Driver &D, const ArgList &Args,
   D.Diag(diag::err_drv_invalid_or_unsupported_offload_target) << TT->str();
   return std::nullopt;
 }
+
 static std::optional<llvm::Triple>
 getHIPOffloadTargetTriple(const Driver &D, const ArgList &Args) {
   if (!Args.hasArg(options::OPT_offload_EQ)) {
@@ -168,6 +169,18 @@ getHIPOffloadTargetTriple(const Driver &D, const ArgList &Args) {
   return std::nullopt;
 }
 
+template <typename F>
+static bool usesInput(const ArgList &Args, F &&Fn) {
+  return llvm::any_of(Args, [&](Arg *A) {
+    return (A->getOption().matches(options::OPT_x) &&
+            Fn(types::lookupTypeForTypeSpecifier(A->getValue()))) ||
+           (A->getOption().getKind() == Option::InputClass &&
+            StringRef(A->getValue()).rfind('.') != StringRef::npos &&
+            Fn(types::lookupTypeForExtension(
+                &A->getValue()[StringRef(A->getValue()).rfind('.') + 1])));
+  });
+}
+
 // static
 std::string Driver::GetResourcesPath(StringRef BinaryPath) {
   // Since the resource directory is embedded in the module hash, it's important
@@ -6672,9 +6685,14 @@ const ToolChain &Driver::getToolChain(const ArgList &Args,
     case llvm::Triple::CUDA:
       TC = std::make_unique<toolchains::NVPTXToolChain>(*this, Target, Args);
       break;
-    case llvm::Triple::AMDHSA:
-      TC = std::make_unique<toolchains::ROCMToolChain>(*this, Target, Args);
+    case llvm::Triple::AMDHSA: {
+      bool DL =
+          usesInput(Args, types::isOpenCL) || usesInput(Args, types::isLLVMIR);
+      TC = DL ? std::make_unique<toolchains::ROCMToolChain>(*this, Target, Args)
+              : std::make_unique<toolchains::AMDGPUToolChain>(*this, Target,
+                                                              Args);
       break;
+    }
     case llvm::Triple::AMDPAL:
     case llvm::Triple::Mesa3D:
       TC = std::make_unique<toolchains::AMDGPUToolChain>(*this, Target, Args);
diff --git a/clang/test/Driver/amdgpu-toolchain-opencl.cl b/clang/test/Driver/amdgpu-toolchain-opencl.cl
index 755d5e0ad8c502f..ffcaa1d7a232f10 100644
--- a/clang/test/Driver/amdgpu-toolchain-opencl.cl
+++ b/clang/test/Driver/amdgpu-toolchain-opencl.cl
@@ -35,3 +35,8 @@
 
 // RUN: %clang -### --target=amdgcn-amd-amdhsa-opencl -x cl -c -emit-llvm -mcpu=fiji -nogpulib %s 2>&1 | FileCheck -check-prefix=CHECK-WARN-ATOMIC %s
 // CHECK-WARN-ATOMIC: "-cc1"{{.*}} "-Werror=atomic-alignment"
+
+// RUN: %clang -### --target=amdgcn-amd-amdhsa -x cl -c -emit-llvm -mcpu=fiji %s 2>&1 \
+// RUN:  -mcode-object-version=5 --rocm-device-lib-path=%S/Inputs/rocm/amdgcn/bitcode \
+// RUN: | FileCheck -check-prefix=DEVICE-LIBS %s
+// DEVICE-LIBS: "-mlink-builtin-bitcode" "[[ROCM_PATH:.+]]/opencl.bc" "-mlink-builtin-bitcode" "[[ROCM_PATH]]/ocml.bc" "-mlink-builtin-bitcode" "[[ROCM_PATH]]/ockl.bc"
diff --git a/clang/test/Driver/amdgpu-toolchain.c b/clang/test/Driver/amdgpu-toolchain.c
index c1c5aa8e90e6868..cfe9ea1d4c7a99a 100644
--- a/clang/test/Driver/amdgpu-toolchain.c
+++ b/clang/test/Driver/amdgpu-toolchain.c
@@ -36,3 +36,11 @@
 // RUN: %clang -target amdgcn-amd-amdhsa -march=gfx90a -stdlib -startfiles \
 // RUN:   -nogpulib -nogpuinc -### %s 2>&1 | FileCheck -check-prefix=STARTUP %s
 // STARTUP: ld.lld{{.*}}"-lc" "-lm" "{{.*}}crt1.o"
+
+// RUN: %clang -### --target=amdgcn-amd-amdhsa -mcpu=gfx906 %s 2>&1 | FileCheck -check-prefix=ROCM %s
+// ROCM-NOT: -mlink-builtin-bitcode
+
+// RUN: %clang -### --target=amdgcn-amd-amdhsa -mcpu=fiji -x ir %s \
+// RUN:  --rocm-device-lib-path=%S/Inputs/rocm/amdgcn/bitcode 2>&1 \
+// RUN: | FileCheck -check-prefix=DEVICE-LIBS %s
+// DEVICE-LIBS: "-mlink-builtin-bitcode" "[[ROCM_PATH:.+]]/opencl.bc" "-mlink-builtin-bitcode" "[[ROCM_PATH]]/ocml.bc" "-mlink-builtin-bitcode" "[[ROCM_PATH]]/ockl.bc"

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.