[clang] [NVPTX] Correctly forward the PTX feature to the nvlink wrapper (PR #100607)

[Joseph Huber via cfe-commits]

https://github.com/jhuber6 updated https://github.com/llvm/llvm-project/pull/100607

>From 0cc36f8aa42c1b673b1d239bf1518990582c5479 Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn at outlook.com>
Date: Thu, 25 Jul 2024 12:31:09 -0500
Subject: [PATCH] [NVPTX] Correctly forward the PTX feature to the nvlink
 wrapper

Summary:
This is necessary for LTO when the user specifies it or has a CUDA
version that supports a sufficiently high version. Previously it would
default.
---
 clang/lib/Driver/ToolChains/Cuda.cpp     | 7 +++++++
 clang/test/Driver/cuda-cross-compiling.c | 8 ++++++++
 2 files changed, 15 insertions(+)

diff --git a/clang/lib/Driver/ToolChains/Cuda.cpp b/clang/lib/Driver/ToolChains/Cuda.cpp
index ffe5f4a6f4a25..e98e574d6cc2b 100644
--- a/clang/lib/Driver/ToolChains/Cuda.cpp
+++ b/clang/lib/Driver/ToolChains/Cuda.cpp
@@ -625,6 +625,13 @@ void NVPTX::Linker::ConstructJob(Compilation &C, const JobAction &JA,
     addLTOOptions(getToolChain(), Args, CmdArgs, Output, Inputs[0],
                   C.getDriver().getLTOMode() == LTOK_Thin);
 
+  // Forward the PTX features if the nvlink-wrapper needs it.
+  std::vector<StringRef> Features;
+  getNVPTXTargetFeatures(C.getDriver(), getToolChain().getTriple(), Args,
+                         Features);
+  for (StringRef Feature : Features)
+    CmdArgs.append({"--feature", Args.MakeArgString(Feature)});
+
   addGPULibraries(getToolChain(), Args, CmdArgs);
 
   // Add paths for the default clang library path.
diff --git a/clang/test/Driver/cuda-cross-compiling.c b/clang/test/Driver/cuda-cross-compiling.c
index 42d56cbfcc321..c2e538c25329e 100644
--- a/clang/test/Driver/cuda-cross-compiling.c
+++ b/clang/test/Driver/cuda-cross-compiling.c
@@ -90,3 +90,11 @@
 // RUN:   | FileCheck -check-prefix=GENERIC %s
 
 // GENERIC-NOT: -cc1" "-triple" "nvptx64-nvidia-cuda" {{.*}} "-target-cpu"
+
+//
+// Test forwarding the necessary +ptx feature.
+//
+// RUN: %clang -target nvptx64-nvidia-cuda --cuda-feature=+ptx63 -march=sm_52 -### %s 2>&1 \
+// RUN:   | FileCheck -check-prefix=FEATURE %s
+
+// FEATURE: clang-nvlink-wrapper{{.*}}"--feature" "+ptx63"