llvm · jchlanda · Nov 25, 2024
diff --git a/clang/include/clang/Basic/TargetInfo.h b/clang/include/clang/Basic/TargetInfo.h
@@ -271,6 +271,9 @@ class TargetInfo : public TransferrableTargetInfo,
   LLVM_PREFERRED_TYPE(bool)
   unsigned AllowAMDGPUUnsafeFPAtomics : 1;
 
+  LLVM_PREFERRED_TYPE(bool)
+  unsigned OpenCLDefIsGenericAddrSpace : 1;
+
   LLVM_PREFERRED_TYPE(bool)
   unsigned HasUnalignedAccess : 1;
 

diff --git a/clang/include/clang/Basic/TargetOptions.h b/clang/include/clang/Basic/TargetOptions.h
@@ -78,6 +78,10 @@ class TargetOptions {
   /// \brief If enabled, allow AMDGPU unsafe floating point atomics.
   bool AllowAMDGPUUnsafeFPAtomics = false;
 
+  /// \brief If enabled, allow overriding of the default address space (from
+  /// private to generic).
+  bool OpenCLDefIsGenericAddrSpace = false;
+
   /// \brief Code object version for AMDGPU.
   llvm::CodeObjectVersionKind CodeObjectVersion =
       llvm::CodeObjectVersionKind::COV_None;

diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
@@ -5164,6 +5164,13 @@ defm unsafe_fp_atomics : BoolMOption<"unsafe-fp-atomics",
           "for certain memory destinations. (AMDGPU only)">,
   NegFlag<SetFalse>>;
 
+defm opencl_def_is_generic_addrspace: BoolMOption<"opencl-def-is-generic-addrspace",
+  TargetOpts<"OpenCLDefIsGenericAddrSpace">, DefaultFalse,
+  PosFlag<SetTrue, [], [ClangOption, CC1Option],
+          "Enable overriding of OpenCL's default address space: from private "
+          "to generic (AMDGPU only).">,
+  NegFlag<SetFalse>>;
+
 def faltivec : Flag<["-"], "faltivec">, Group<f_Group>;
 def fno_altivec : Flag<["-"], "fno-altivec">, Group<f_Group>;
 let Flags = [TargetSpecific] in {

diff --git a/clang/lib/Basic/TargetInfo.cpp b/clang/lib/Basic/TargetInfo.cpp
@@ -157,6 +157,7 @@ TargetInfo::TargetInfo(const llvm::Triple &T) : Triple(T) {
   HasAArch64SVETypes = false;
   HasRISCVVTypes = false;
   AllowAMDGPUUnsafeFPAtomics = false;
+  OpenCLDefIsGenericAddrSpace = false;
   HasUnalignedAccess = false;
   ARMCDECoprocMask = 0;
 

diff --git a/clang/lib/Basic/Targets/AMDGPU.cpp b/clang/lib/Basic/Targets/AMDGPU.cpp
@@ -240,6 +240,7 @@ AMDGPUTargetInfo::AMDGPUTargetInfo(const llvm::Triple &Triple,
   HasFloat16 = true;
   WavefrontSize = (GPUFeatures & llvm::AMDGPU::FEATURE_WAVE32) ? 32 : 64;
   AllowAMDGPUUnsafeFPAtomics = Opts.AllowAMDGPUUnsafeFPAtomics;
+  OpenCLDefIsGenericAddrSpace = Opts.OpenCLDefIsGenericAddrSpace;
 
   // Set pointer width and alignment for the generic address space.
   PointerWidth = PointerAlign = getPointerWidthV(LangAS::Default);
@@ -262,8 +263,13 @@ void AMDGPUTargetInfo::adjust(DiagnosticsEngine &Diags, LangOptions &Opts) {
   // ToDo: There are still a few places using default address space as private
   // address space in OpenCL, which needs to be cleaned up, then the references
   // to OpenCL can be removed from the following line.
-  setAddressSpaceMap((Opts.OpenCL && !Opts.OpenCLGenericAddressSpace) ||
-                     !isAMDGCN(getTriple()));
+  bool DefaultIsPrivate = (Opts.OpenCL && !Opts.OpenCLGenericAddressSpace) ||
+                          !isAMDGCN(getTriple());
+  // Allow for overriding of the default address space in OpenCL (from private
+  // to generic).
+  if (Opts.OpenCL && OpenCLDefIsGenericAddrSpace)
+    DefaultIsPrivate = false;
+  setAddressSpaceMap(DefaultIsPrivate);
 }
 
 ArrayRef<Builtin::Info> AMDGPUTargetInfo::getTargetBuiltins() const {

diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -7888,6 +7888,8 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
 
     Args.addOptInFlag(CmdArgs, options::OPT_munsafe_fp_atomics,
                       options::OPT_mno_unsafe_fp_atomics);
+    Args.addOptInFlag(CmdArgs, options::OPT_mopencl_def_is_generic_addrspace,
+                      options::OPT_mno_opencl_def_is_generic_addrspace);
     Args.addOptOutFlag(CmdArgs, options::OPT_mamdgpu_ieee,
                        options::OPT_mno_amdgpu_ieee);
   }

diff --git a/clang/test/CodeGen/AMDGPU/opencl_def_is_generic_addrspace.cl b/clang/test/CodeGen/AMDGPU/opencl_def_is_generic_addrspace.cl
@@ -0,0 +1,18 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
+// RUN: %clang_cc1 -triple amdgcn-unknown-amdhsa -emit-llvm -target-cpu gfx900 -mopencl-def-is-generic-addrspace -cl-std=CL3.0 -cl-ext=+__opencl_c_generic_address_space -o - %s | FileCheck -check-prefix=CHECK-OPENCL-DEF-IS-GENERIC %s
+// RUN: %clang_cc1 -triple amdgcn-unknown-amdhsa -emit-llvm -target-cpu gfx900 -cl-std=CL3.0 -cl-ext=+__opencl_c_generic_address_space -o - %s | FileCheck -check-prefix=CHECK-NO-OPENCL-DEF-IS-GENERIC %s
+
+// CHECK-OPENCL-DEF-IS-GENERIC-LABEL: define dso_local void @_Z14funcGenericPtrfPf(
+// CHECK-OPENCL-DEF-IS-GENERIC-SAME: float noundef [[ARGS_0:%.*]], ptr nocapture noundef readnone [[ARGS_1:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+// CHECK-OPENCL-DEF-IS-GENERIC-NEXT:  [[ENTRY:.*:]]
+// CHECK-OPENCL-DEF-IS-GENERIC-NEXT:    ret void
+//
+// CHECK-NO-OPENCL-DEF-IS-GENERIC-LABEL: define dso_local void @_Z14funcGenericPtrfPU3AS0f(
+// CHECK-NO-OPENCL-DEF-IS-GENERIC-SAME: float noundef [[ARGS_0:%.*]], ptr nocapture noundef readnone [[ARGS_1:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+// CHECK-NO-OPENCL-DEF-IS-GENERIC-NEXT:  [[ENTRY:.*:]]
+// CHECK-NO-OPENCL-DEF-IS-GENERIC-NEXT:    ret void
+//
+__attribute__((overloadable)) void funcGenericPtr(float args_0,
+                                                  float __generic *args_1) {
+  return;
+}