[clang] [CLANG][DRIVER] Resolve Argument Visibility and Duplication Issue for Cuda ToolChain (PR #86807)

[Jefferson Le Quellec via cfe-commits]

https://github.com/jle-quel created https://github.com/llvm/llvm-project/pull/86807

# Description

This PR makes the argument `-Xcuda-ptxas` visible to the driver in cl-mode.

Furthermore, it has been noticed that the arguments are being passed twice to `ptxas`.
This also has been fixed by filtering out the arguments before appending them to the new `DAL` created by `CudaToolChain::TranslateArgs`.

>From e28fab463b54192b8e8440aa8c349c2d42bb949a Mon Sep 17 00:00:00 2001
From: Jefferson Le Quellec <jefferson.lequellec at codeplay.com>
Date: Wed, 27 Mar 2024 15:25:16 +0100
Subject: [PATCH 1/3] Make the argument -Xcuda-ptxas visible to the driver in
 cl-mode

---
 clang/include/clang/Driver/Options.td | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index 29066ea14280c2..a5e132dc48a3ef 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -1001,7 +1001,8 @@ def : Joined<["-"], "Xclang=">, Group<CompileOnly_Group>,
 def Xcuda_fatbinary : Separate<["-"], "Xcuda-fatbinary">,
   HelpText<"Pass <arg> to fatbinary invocation">, MetaVarName<"<arg>">;
 def Xcuda_ptxas : Separate<["-"], "Xcuda-ptxas">,
-  HelpText<"Pass <arg> to the ptxas assembler">, MetaVarName<"<arg>">;
+  HelpText<"Pass <arg> to the ptxas assembler">, MetaVarName<"<arg>">,
+  Visibility<[ClangOption, CLOption]>;
 def Xopenmp_target : Separate<["-"], "Xopenmp-target">, Group<CompileOnly_Group>,
   HelpText<"Pass <arg> to the target offloading toolchain.">, MetaVarName<"<arg>">;
 def Xopenmp_target_EQ : JoinedAndSeparate<["-"], "Xopenmp-target=">, Group<CompileOnly_Group>,

>From be1fd093d9c3f2553cefd061d476d052848b853a Mon Sep 17 00:00:00 2001
From: Jefferson Le Quellec <jefferson.lequellec at codeplay.com>
Date: Wed, 27 Mar 2024 15:26:52 +0100
Subject: [PATCH 2/3] Make sure arguments are not duplicated in new DAL

---
 clang/lib/Driver/ToolChains/Cuda.cpp | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/clang/lib/Driver/ToolChains/Cuda.cpp b/clang/lib/Driver/ToolChains/Cuda.cpp
index 5f0b516e1a1a08..6634e6d818b33e 100644
--- a/clang/lib/Driver/ToolChains/Cuda.cpp
+++ b/clang/lib/Driver/ToolChains/Cuda.cpp
@@ -990,7 +990,10 @@ CudaToolChain::TranslateArgs(const llvm::opt::DerivedArgList &Args,
   }
 
   for (Arg *A : Args) {
-    DAL->append(A);
+    // Make sure flags are not duplicated.
+    if (!llvm::is_contained(*DAL, A)) {
+      DAL->append(A);
+    }
   }
 
   if (!BoundArch.empty()) {

>From fe2ab25ac7efc10851ca3aaa51caecc80f9000d4 Mon Sep 17 00:00:00 2001
From: Jefferson Le Quellec <jefferson.lequellec at codeplay.com>
Date: Wed, 27 Mar 2024 15:27:33 +0100
Subject: [PATCH 3/3] Add visibility and duplication tests

---
 clang/test/Driver/cuda-external-tools.cu | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/clang/test/Driver/cuda-external-tools.cu b/clang/test/Driver/cuda-external-tools.cu
index 946e144fce38fb..d9564d026b4faa 100644
--- a/clang/test/Driver/cuda-external-tools.cu
+++ b/clang/test/Driver/cuda-external-tools.cu
@@ -86,6 +86,12 @@
 // RUN:   -Xcuda-fatbinary -bar1 -Xcuda-ptxas -foo2 -Xcuda-fatbinary -bar2 %s 2>&1 \
 // RUN: | FileCheck -check-prefixes=CHECK,SM35,PTXAS-EXTRA,FATBINARY-EXTRA %s
 
+// Check -Xcuda-ptxas with clang-cl
+// RUN: %clang_cl -### -c -Xcuda-ptxas -foo1 \
+// RUN:   --offload-arch=sm_35 --cuda-path=%S/Inputs/CUDA/usr/local/cuda \
+// RUN:   -Xcuda-ptxas -foo2 %s 2>&1 \
+// RUN: | FileCheck -check-prefixes=CHECK,SM35,PTXAS-EXTRA %s
+
 // MacOS spot-checks
 // RUN: %clang -### --target=x86_64-apple-macosx -O0 -c %s 2>&1 \
 // RUN:   --offload-arch=sm_35 --cuda-path=%S/Inputs/CUDA/usr/local/cuda \
@@ -140,6 +146,8 @@
 // CHECK-SAME: "[[PTXFILE]]"
 // PTXAS-EXTRA-SAME: "-foo1"
 // PTXAS-EXTRA-SAME: "-foo2"
+// CHECK-NOT: "-foo1"
+// CHECK-NOT: "-foo2"
 // RDC-SAME: "-c"
 // CHECK-NOT: "-c"