[clang] [AMDGPU] Use the AMDGPUToolChain when targeting C/C++ directly (PR #99687)

[Joseph Huber via cfe-commits]

https://github.com/jhuber6 updated https://github.com/llvm/llvm-project/pull/99687

>From a6df42bc9a7a9f3779f598f77db45e5334910bc2 Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn at outlook.com>
Date: Fri, 24 Jan 2025 16:15:57 -0600
Subject: [PATCH] [AMDGPU] Use the AMDGPUToolChain when targeting C/C++
 directly

Summary:
The getToolChain pass uses the triple to determine which toolchain to
create. Currently the amdgcn-amd-amdhsa triple maps to the
ROCmToolChain which uses things expected to be provided by ROCm.
This is neded for OpenCL, but directly targeting C++ does not want this
since it's primarily being used for creating GPU runtime code. As far as
I know I'm the only user of this, so this shouldn't change anything.

Unfortunately, there's no good logic for detercting this, so I simply
checked ahead of time if the input is either foo.cl or -x cl foo.c
to choose between the two. This allows us to use the AMDGPU target
normally, as otherwise it will error without passing -nogpulib.
---
 clang/lib/Driver/Driver.cpp          | 16 ++++++++++++++--
 clang/test/Driver/amdgpu-toolchain.c |  3 +++
 2 files changed, 17 insertions(+), 2 deletions(-)

diff --git a/clang/lib/Driver/Driver.cpp b/clang/lib/Driver/Driver.cpp
index 87855fdb799710..a9bd52920d823f 100644
--- a/clang/lib/Driver/Driver.cpp
+++ b/clang/lib/Driver/Driver.cpp
@@ -6669,9 +6669,21 @@ 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 CL = llvm::any_of(Args, [&](Arg *A) {
+        return (A->getOption().matches(options::OPT_x) &&
+                types::isOpenCL(
+                    types::lookupTypeForExtension(A->getValue()))) ||
+               (A->getOption().getKind() == Option::InputClass &&
+                StringRef(A->getValue()).rfind('.') != StringRef::npos &&
+                types::isOpenCL(types::lookupTypeForExtension(
+                    &A->getValue()[StringRef(A->getValue()).rfind('.') + 1])));
+      });
+      TC = CL ? 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.c b/clang/test/Driver/amdgpu-toolchain.c
index c1c5aa8e90e686..3299930ee40dbb 100644
--- a/clang/test/Driver/amdgpu-toolchain.c
+++ b/clang/test/Driver/amdgpu-toolchain.c
@@ -36,3 +36,6 @@
 // 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