[SYCL][NativeCPU] Remove -fsycl-is-native-cpu. (#19446)

hvdijk · web-flow · commit 566fc15d8a83 · 2025-07-15T14:43:24.000Z
Now that we are able to check whether the triple is native_cpu, we no
longer need this additional compiler option.
diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
@@ -9032,10 +9032,6 @@ def fenable_sycl_dae : Flag<["-"], "fenable-sycl-dae">,
 def fsycl_enable_int_header_diags: Flag<["-"], "fsycl-enable-int-header-diags">,
   HelpText<"Enable diagnostics that require the SYCL integration header.">,
   MarshallingInfoFlag<LangOpts<"SYCLEnableIntHeaderDiags">>;
-def fsycl_is_native_cpu : Flag<["-"], "fsycl-is-native-cpu">,
-  HelpText<"Perform device compilation for Native CPU.">,
-  Visibility<[CC1Option]>,
-  MarshallingInfoFlag<LangOpts<"SYCLIsNativeCPU">>;
 // TODO: This option can be removed once a fix goes in that can
 // work with the community changes for using the alloca address space.
 defm offload_use_alloca_addrspace_for_srets : BoolFOption<"offload-use-alloca-addrspace-for-srets",
diff --git a/clang/lib/AST/MicrosoftMangle.cpp b/clang/lib/AST/MicrosoftMangle.cpp
@@ -3219,12 +3219,12 @@ void MicrosoftCXXNameMangler::mangleCallingConvention(CallingConv CC,
         Out << "w";
       return;
     case CC_OpenCLKernel:
-      // This can occur on the SYCl NativeCPU device
+      // This can occur on the SYCL NativeCPU device
       // where device code is compiled with the same
       // target triple (eg for Windows) as host code.
       // FIXME: 1.) provide mangling if needed
       //        2.) check if other conventions need to be handled.
-      if (!getASTContext().getLangOpts().SYCLIsNativeCPU)
+      if (!getASTContext().getTargetInfo().getTriple().isNativeCPU())
         // Currently we only allow this convention in
         // SYCLNativeCPU and raise the usual error otherwise.
         llvm_unreachable("Unsupported CC for mangling");
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -1396,8 +1396,6 @@ void CodeGenModule::Release() {
 
   if (LangOpts.SYCLIsDevice) {
     getModule().addModuleFlag(llvm::Module::Error, "sycl-device", 1);
-    if (LangOpts.SYCLIsNativeCPU)
-      getModule().addModuleFlag(llvm::Module::Error, "is-native-cpu", 1);
   }
 
   if (LangOpts.EHAsynch)
@@ -6955,8 +6953,7 @@ void CodeGenModule::EmitGlobalFunctionDefinition(GlobalDecl GD,
     if (GD.getKernelReferenceKind() == KernelReferenceKind::Stub &&
         !D->hasAttr<NoInlineAttr>() &&
         !Fn->hasFnAttribute(llvm::Attribute::NoInline) &&
-        !D->hasAttr<OptimizeNoneAttr>() && !LangOpts.SYCLIsNativeCPU &&
-        !LangOpts.SYCLIsDevice &&
+        !D->hasAttr<OptimizeNoneAttr>() && !LangOpts.SYCLIsDevice &&
         !Fn->hasFnAttribute(llvm::Attribute::OptimizeNone) &&
         !ShouldAddOptNone) {
       Fn->addFnAttr(llvm::Attribute::AlwaysInline);
diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -5696,7 +5696,6 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
         CmdArgs.push_back("-sycl-opt");
       }
       if (RawTriple.isNativeCPU()) {
-        CmdArgs.push_back("-fsycl-is-native-cpu");
         CmdArgs.push_back("-D");
         CmdArgs.push_back("__SYCL_NATIVE_CPU__");
         CmdArgs.push_back("-fno-autolink");
diff --git a/clang/lib/Frontend/InitPreprocessor.cpp b/clang/lib/Frontend/InitPreprocessor.cpp
@@ -1549,7 +1549,7 @@ static void InitializePredefinedMacros(const TargetInfo &TI,
     const llvm::Triple::SubArchType SubArch = Triple.getSubArch();
     if (Triple.isNVPTX() || Triple.isAMDGPU() ||
         (Triple.isSPIR() && SubArch != llvm::Triple::SPIRSubArch_fpga) ||
-        LangOpts.SYCLIsNativeCPU)
+        Triple.isNativeCPU())
       Builder.defineMacro("SYCL_USE_NATIVE_FP_ATOMICS");
     // Enable generation of USM address spaces for FPGA.
     if (SubArch == llvm::Triple::SPIRSubArch_fpga) {
diff --git a/clang/lib/Sema/SemaSYCL.cpp b/clang/lib/Sema/SemaSYCL.cpp
@@ -1303,7 +1303,7 @@ constructKernelName(SemaSYCL &S, const FunctionDecl *KernelCallerFunc,
   // instead to always compile the NativeCPU device code in GNU mode which
   // may cause issues when compiling headers with non-standard extensions
   // written for compilers with different C++ ABIs (like MS VS).
-  if (S.getLangOpts().SYCLIsNativeCPU) {
+  if (S.getASTContext().getTargetInfo().getTriple().isNativeCPU()) {
     MangledName = StableName;
   }
 
diff --git a/clang/test/CodeGenSYCL/native_cpu_as.cpp b/clang/test/CodeGenSYCL/native_cpu_as.cpp
@@ -1,10 +1,10 @@
 // This test is temporarily disabled for SYCL Native CPU on Windows
 // UNSUPPORTED: system-windows
-// Checks that name mangling matches between SYCL Native CPU and OpenCL when -fsycl-is-native-cpu is set
-// RUN: %clang_cc1 -triple=native_cpu -DCPP -fsycl-is-device -emit-llvm -internal-isystem %S/Inputs -fsycl-is-native-cpu -o %t_sycl.ll %s
+// Checks that name mangling matches between SYCL Native CPU and OpenCL
+// RUN: %clang_cc1 -triple=native_cpu -DCPP -fsycl-is-device -emit-llvm -internal-isystem %S/Inputs -o %t_sycl.ll %s
 // RUN: FileCheck -input-file=%t_sycl.ll %s
 
-// RUN: %clang_cc1 -triple=native_cpu -x cl -DOCL -emit-llvm -internal-isystem %S/Inputs -fsycl-is-native-cpu -o %t_ocl.ll %s
+// RUN: %clang_cc1 -triple=native_cpu -x cl -DOCL -emit-llvm -internal-isystem %S/Inputs -o %t_ocl.ll %s
 // RUN: FileCheck -input-file=%t_ocl.ll %s
 
 #ifdef CPP
diff --git a/clang/test/CodeGenSYCL/native_cpu_basic.cpp b/clang/test/CodeGenSYCL/native_cpu_basic.cpp
diff --git a/clang/test/CodeGenSYCL/native_cpu_mangling.cpp b/clang/test/CodeGenSYCL/native_cpu_mangling.cpp
@@ -1,10 +1,10 @@
 // This test ensures the native-cpu device generates the expected kernel names,
 // and that the MS mangler doesn't assert on the code below.
 
-// RUN: %clang_cc1 -triple x86_64-pc-windows-msvc   -aux-triple x86_64-pc-windows-msvc   -I %S/Inputs -fsycl-is-device -fsycl-is-native-cpu -emit-llvm -o - -x c++ %s | FileCheck %s
-// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -aux-triple x86_64-unknown-linux-gnu -I %S/Inputs -fsycl-is-device -fsycl-is-native-cpu -emit-llvm -o - -x c++ %s | FileCheck %s
-// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -aux-triple x86_64-pc-windows-msvc   -I %S/Inputs -fsycl-is-device -fsycl-is-native-cpu -emit-llvm -o - -x c++ %s | FileCheck %s
-// RUN: %clang_cc1 -triple x86_64-pc-windows-msvc   -aux-triple x86_64-unknown-linux-gnu -I %S/Inputs -fsycl-is-device -fsycl-is-native-cpu -emit-llvm -o - -x c++ %s | FileCheck %s
+// RUN: %clang_cc1 -triple native_cpu -aux-triple x86_64-pc-windows-msvc   -I %S/Inputs -fsycl-is-device -emit-llvm -o - -x c++ %s | FileCheck %s
+// RUN: %clang_cc1 -triple native_cpu -aux-triple x86_64-unknown-linux-gnu -I %S/Inputs -fsycl-is-device -emit-llvm -o - -x c++ %s | FileCheck %s
+// RUN: %clang_cc1 -triple native_cpu -aux-triple x86_64-pc-windows-msvc   -I %S/Inputs -fsycl-is-device -emit-llvm -o - -x c++ %s | FileCheck %s
+// RUN: %clang_cc1 -triple native_cpu -aux-triple x86_64-unknown-linux-gnu -I %S/Inputs -fsycl-is-device -emit-llvm -o - -x c++ %s | FileCheck %s
 // Todo: check other cpus
 
 #include "sycl.hpp"
diff --git a/clang/test/CodeGenSYCL/native_cpu_target_features.cpp b/clang/test/CodeGenSYCL/native_cpu_target_features.cpp
@@ -1,9 +1,9 @@
-// RUN: %clang_cc1 -triple native_cpu -aux-triple x86_64-unknown-linux-gnu -fsycl-is-device -emit-llvm -fsycl-is-native-cpu -o - %s | FileCheck %s --check-prefixes=CHECK,NOAVX
-// RUN: %clang_cc1 -triple native_cpu -aux-triple x86_64-unknown-linux-gnu -aux-target-cpu skylake -fsycl-is-device -emit-llvm -fsycl-is-native-cpu -o - %s | FileCheck %s --check-prefixes=CHECK,AVX
-// RUN: %clang_cc1 -triple native_cpu -aux-triple x86_64-unknown-linux-gnu -aux-target-feature +avx -fsycl-is-device -emit-llvm -fsycl-is-native-cpu -o - %s | FileCheck %s --check-prefixes=CHECK,AVX
+// RUN: %clang_cc1 -triple native_cpu -aux-triple x86_64-unknown-linux-gnu -fsycl-is-device -emit-llvm -o - %s | FileCheck %s --check-prefixes=CHECK,NOAVX
+// RUN: %clang_cc1 -triple native_cpu -aux-triple x86_64-unknown-linux-gnu -aux-target-cpu skylake -fsycl-is-device -emit-llvm -o - %s | FileCheck %s --check-prefixes=CHECK,AVX
+// RUN: %clang_cc1 -triple native_cpu -aux-triple x86_64-unknown-linux-gnu -aux-target-feature +avx -fsycl-is-device -emit-llvm -o - %s | FileCheck %s --check-prefixes=CHECK,AVX
 //
 // This is not sensible but check that we do not crash.
-// RUN: %clang_cc1 -triple native_cpu -aux-triple native_cpu -fsycl-is-device -emit-llvm -fsycl-is-native-cpu -o - %s | FileCheck %s --check-prefixes=CHECK,NOAVX
+// RUN: %clang_cc1 -triple native_cpu -aux-triple native_cpu -fsycl-is-device -emit-llvm -o - %s | FileCheck %s --check-prefixes=CHECK,NOAVX
 
 #include "Inputs/sycl.hpp"
 using namespace sycl;
diff --git a/clang/test/Driver/sycl-native-cpu-fsycl.cpp b/clang/test/Driver/sycl-native-cpu-fsycl.cpp
@@ -42,7 +42,7 @@
 //CHECK_BINDINGS:# "{{.*}}" - "offload wrapper", inputs: ["[[TABLEFILE]].table"], output: "[[WRAPPEROBJ:.*]].o"
 //CHECK_BINDINGS:# "{{.*}}" - "{{.*}}::Linker", inputs: ["[[HOSTOBJ]].o", "[[KERNELOBJ]].o", "[[WRAPPEROBJ]].o"], output: "a.{{.*}}"
 
-//CHECK_INVO:{{.*}}clang{{.*}}-fsycl-is-device{{.*}}"-fsycl-is-native-cpu" "-D" "__SYCL_NATIVE_CPU__"
+//CHECK_INVO:{{.*}}clang{{.*}}"-fsycl-is-device"{{.*}} "-D" "__SYCL_NATIVE_CPU__"
 //CHECK_INVO:{{.*}}clang{{.*}}"-fsycl-is-host"{{.*}}
 //CHECK_INVO:{{.*}}clang{{.*}}"-x" "ir"
 //CHECK_INVO:{{.*}}sycl-post-link{{.*}}"-emit-program-metadata"
diff --git a/clang/test/Driver/sycl-native-cpu.cpp b/clang/test/Driver/sycl-native-cpu.cpp
@@ -3,10 +3,10 @@
 
 
 // checks that the host triple is native_cpu, the device triple is set, and that the sycl-native-cpu LLVM option is set
-// CHECK: clang{{.*}}"-triple" "native_cpu"{{.*}}"-aux-triple" "{{.*}}" "-fsycl-is-native-cpu"{{.*}}"-D" "__SYCL_NATIVE_CPU__"
+// CHECK: clang{{.*}}"-triple" "native_cpu"{{.*}}"-aux-triple" "{{.*}}" {{.*}}"-D" "__SYCL_NATIVE_CPU__"
 
 // checks that the target triples are set correctly when the target is set explicitly
-// CHECK-AARCH64: clang{{.*}}"-triple" "native_cpu"{{.*}}"-aux-triple" "aarch64-unknown-linux-gnu"{{.*}}"-fsycl-is-native-cpu"{{.*}}"-D" "__SYCL_NATIVE_CPU__"
+// CHECK-AARCH64: clang{{.*}}"-triple" "native_cpu"{{.*}}"-aux-triple" "aarch64-unknown-linux-gnu" {{.*}}"-D" "__SYCL_NATIVE_CPU__"
 
 // RUN:   %clang -### --target=x86_64-unknown-linux-gnu -fsycl -fsycl-targets=native_cpu -g %s 2>&1 | FileCheck -check-prefix=CHECK-LINUX %s
 // CHECK-LINUX: {{.*}}"-fsycl-is-device"{{.*}}"-dwarf-version=[[DVERSION:.*]]" "-debugger-tuning=gdb"
@@ -32,6 +32,5 @@
 
 // Checking that coverage testing options are accepted by native_cpu, and that device and host compilation invocations receive the same options
 // RUN: %clangxx -fsycl -fsycl-targets=native_cpu -Werror -fno-profile-instr-generate -fprofile-instr-generate -fno-coverage-mapping -fcoverage-mapping -### %s 2>&1 | FileCheck %s --check-prefix=CHECK_COV_INVO
-// CHECK_COV_INVO:{{.*}}clang{{.*}}-fsycl-is-device{{.*}}"-fsycl-is-native-cpu" "-D" "__SYCL_NATIVE_CPU__"{{.*}}"-fprofile-instrument=clang"{{.*}}"-fcoverage-mapping" "-fcoverage-compilation-dir={{.*}}"
+// CHECK_COV_INVO:{{.*}}clang{{.*}}"-fsycl-is-device"{{.*}} "-D" "__SYCL_NATIVE_CPU__"{{.*}}"-fprofile-instrument=clang"{{.*}}"-fcoverage-mapping" "-fcoverage-compilation-dir={{.*}}"
 // CHECK_COV_INVO:{{.*}}clang{{.*}}"-fsycl-is-host"{{.*}}"-fprofile-instrument=clang"{{.*}}"-fcoverage-mapping" "-fcoverage-compilation-dir={{.*}}"
-
diff --git a/clang/test/Preprocessor/sycl-macro-target-specific.cpp b/clang/test/Preprocessor/sycl-macro-target-specific.cpp
@@ -46,7 +46,7 @@
 // RUN: | FileCheck --check-prefix=CHECK-SYCL-FP-ATOMICS %s
 // RUN: %clang_cc1 %s -fsycl-is-device -triple amdgcn-amdhsa-amdhsa -E -dM \
 // RUN: | FileCheck --check-prefix=CHECK-SYCL-FP-ATOMICS %s
-// RUN: %clang_cc1 %s -fsycl-is-device -triple x86_64-unknown-linux-gnu -fsycl-is-native-cpu  \
+// RUN: %clang_cc1 %s -fsycl-is-device -triple native_cpu \
 // RUN: -E -dM | FileCheck --check-prefix=CHECK-SYCL-FP-ATOMICS %s
 // CHECK-SYCL-FP-ATOMICS: #define SYCL_USE_NATIVE_FP_ATOMICS
 // CHECK-SYCL-FP-ATOMICS-NEG-NOT: #define SYCL_USE_NATIVE_FP_ATOMICS
diff --git a/libclc/CMakeLists.txt b/libclc/CMakeLists.txt
@@ -464,9 +464,6 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
     elseif( ARCH STREQUAL amdgcn )
       set( build_flags )
       set( opt_flags -O3 --amdgpu-oclc-reflect-enable=false )
-    elseif( ARCH STREQUAL native_cpu )
-      set( build_flags -Xclang -fsycl-is-native-cpu )
-      set( opt_flags -O3 )
     else()
       set( build_flags )
       set( opt_flags -O3 )
diff --git a/llvm/include/llvm/SYCLLowerIR/UtilsSYCLNativeCPU.h b/llvm/include/llvm/SYCLLowerIR/UtilsSYCLNativeCPU.h
@@ -31,10 +31,6 @@ inline llvm::Twine addSYCLNativeCPUSuffix(StringRef S) {
   return llvm::Twine(S, SYCLNATIVECPUSUFFIX);
 }
 
-inline bool isSYCLNativeCPU(const Module &M) {
-  return M.getModuleFlag("is-native-cpu") != nullptr;
-}
-
 } // namespace utils
 } // namespace sycl
 } // namespace llvm
diff --git a/llvm/lib/SYCLLowerIR/LowerWGScope.cpp b/llvm/lib/SYCLLowerIR/LowerWGScope.cpp
@@ -933,7 +933,7 @@ GlobalVariable *spirv::createWGLocalVariable(Module &M, Type *T,
 // Return a value equals to 0 if and only if the local linear id is 0.
 Value *spirv::genPseudoLocalID(Instruction &Before, const Triple &TT) {
   Module &M = *Before.getModule();
-  if (TT.isNVPTX() || TT.isAMDGCN() || sycl::utils::isSYCLNativeCPU(M)) {
+  if (TT.isNVPTX() || TT.isAMDGCN() || TT.isNativeCPU()) {
     LLVMContext &Ctx = Before.getContext();
     Type *RetTy = getSizeTTy(M);
 
diff --git a/sycl/test/check_device_code/native_cpu/device_builtins.cpp b/sycl/test/check_device_code/native_cpu/device_builtins.cpp
@@ -32,5 +32,3 @@ DefTestCastForSpace(volatile int*)
 DefTestCastForSpace(const volatile int*)
 
 int main(){}
-// check that the generated module has the is-native-cpu module flag set
-// CHECK: !{{[0-9]*}} = !{i32 1, !"is-native-cpu", i32 1}
diff --git a/sycl/test/check_device_code/native_cpu/native_cpu_builtins.cpp b/sycl/test/check_device_code/native_cpu/native_cpu_builtins.cpp
@@ -88,6 +88,3 @@ int main() {
     });
   });
 }
-
-// check that the generated module has the is-native-cpu module flag set
-// CHECK: !{{[0-9]*}} = !{i32 1, !"is-native-cpu", i32 1}

Original file line number	Diff line number	Diff line change
`@@ -5696,7 +5696,6 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,`
`5696`	`5696`	`CmdArgs.push_back("-sycl-opt");`
`5697`	`5697`	`}`
`5698`	`5698`	`if (RawTriple.isNativeCPU()) {`
`5699`		`- CmdArgs.push_back("-fsycl-is-native-cpu");`
`5700`	`5699`	`CmdArgs.push_back("-D");`
`5701`	`5700`	`CmdArgs.push_back("__SYCL_NATIVE_CPU__");`
`5702`	`5701`	`CmdArgs.push_back("-fno-autolink");`
Original file line number	Diff line number	Diff line change
`@@ -1303,7 +1303,7 @@ constructKernelName(SemaSYCL &S, const FunctionDecl *KernelCallerFunc,`
`1303`	`1303`	`// instead to always compile the NativeCPU device code in GNU mode which`
`1304`	`1304`	`// may cause issues when compiling headers with non-standard extensions`
`1305`	`1305`	`// written for compilers with different C++ ABIs (like MS VS).`
`1306`		`- if (S.getLangOpts().SYCLIsNativeCPU) {`
	`1306`	`+ if (S.getASTContext().getTargetInfo().getTriple().isNativeCPU()) {`
`1307`	`1307`	`MangledName = StableName;`
`1308`	`1308`	`}`
`1309`	`1309`