[clang][OMPIRBuilder] Fix two missed function pointer type issues (llvm#162914)

Two small issues, when storing function pointers we need to use the
program address space.

With this change there are no asserts if I run all OpenMP tests with the
offload target manually changed to SPIR-V, so we are getting somewhere.

About 10 test fails though.

Signed-off-by: Sarnie, Nick <[email protected]>

Author: sarnex

clang/lib/CodeGen/CGOpenMPRuntime.cpp

@@ -1762,8 +1762,11 @@ void CGOpenMPRuntime::emitDeclareTargetFunction(const FunctionDecl *FD,
   // access its value.
   llvm::GlobalValue *Addr = GV;
   if (CGM.getLangOpts().OpenMPIsTargetDevice) {
+    llvm::PointerType *FnPtrTy = llvm::PointerType::get(
+        CGM.getLLVMContext(),
+        CGM.getModule().getDataLayout().getProgramAddressSpace());
     Addr = new llvm::GlobalVariable(
-        CGM.getModule(), CGM.VoidPtrTy,
+        CGM.getModule(), FnPtrTy,
         /*isConstant=*/true, llvm::GlobalValue::ExternalLinkage, GV, Name,
         nullptr, llvm::GlobalValue::NotThreadLocal,
         CGM.getModule().getDataLayout().getDefaultGlobalsAddressSpace());

clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp

@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 //
 // This provides a generalized class for OpenMP runtime code generation
-// specialized by GPU targets NVPTX and AMDGCN.
+// specialized by GPU targets NVPTX, AMDGCN and SPIR-V.
 //
 //===----------------------------------------------------------------------===//
 
@@ -1242,12 +1242,13 @@ void CGOpenMPRuntimeGPU::emitParallelCall(
     CGBuilderTy &Bld = CGF.Builder;
     llvm::Value *NumThreadsVal = NumThreads;
     llvm::Function *WFn = WrapperFunctionsMap[OutlinedFn];
-    llvm::Value *ID = llvm::ConstantPointerNull::get(CGM.Int8PtrTy);
-    if (WFn)
-      ID = Bld.CreateBitOrPointerCast(WFn, CGM.Int8PtrTy);
-    llvm::Type *FnPtrTy = llvm::PointerType::get(
+    llvm::PointerType *FnPtrTy = llvm::PointerType::get(
         CGF.getLLVMContext(), CGM.getDataLayout().getProgramAddressSpace());
 
+    llvm::Value *ID = llvm::ConstantPointerNull::get(FnPtrTy);
+    if (WFn)
+      ID = Bld.CreateBitOrPointerCast(WFn, FnPtrTy);
+
     llvm::Value *FnPtr = Bld.CreateBitOrPointerCast(OutlinedFn, FnPtrTy);
 
     // Create a private scope that will globalize the arguments

clang/test/OpenMP/target_indirect_codegen.cpp

@@ -4,6 +4,12 @@
 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=51 -x c++ -triple amdgcn-amd-amdhsa %s -fopenmp-is-target-device -fvisibility=protected -fopenmp-host-ir-file-path %t-host.bc -emit-pch -o %t
 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=51 -x c++ -triple amdgcn-amd-amdhsa -emit-llvm %s -fopenmp-is-target-device -fvisibility=protected -fopenmp-host-ir-file-path %t-host.bc -include-pch %t -o - | FileCheck %s --check-prefix=DEVICE
 
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=51 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=spirv64-intel -emit-llvm %s -o - | FileCheck %s --check-prefix=HOST
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=51 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=spirv64-intel -emit-llvm-bc %s -o %t-spirv-host.bc
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=51 -x c++ -triple spirv64-intel -emit-llvm %s -fopenmp-is-target-device -fvisibility=protected -fopenmp-host-ir-file-path %t-spirv-host.bc -o - | FileCheck %s --check-prefix=DEVICE
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=51 -x c++ -triple spirv64-intel %s -fopenmp-is-target-device -fvisibility=protected -fopenmp-host-ir-file-path %t-spirv-host.bc -emit-pch -o %t
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=51 -x c++ -triple spirv64-intel -emit-llvm %s -fopenmp-is-target-device -fvisibility=protected -fopenmp-host-ir-file-path %t-spirv-host.bc -include-pch %t -o - | FileCheck %s --check-prefix=DEVICE
+
 // expected-no-diagnostics
 #ifndef HEADER
 #define HEADER
@@ -19,10 +25,10 @@
 // HOST: @[[BAR_ENTRY_NAME:.+]] = internal unnamed_addr constant [{{[0-9]+}} x i8] c"[[BAR_NAME:__omp_offloading_[0-9a-z]+_[0-9a-z]+_bar_l[0-9]+]]\00"
 // HOST: @.offloading.entry.[[BAR_NAME]] = weak constant %struct.__tgt_offload_entry { i64 0, i16 1, i16 1, i32 8, ptr @_ZL3barv, ptr @[[BAR_ENTRY_NAME]], i64 8, i64 0, ptr null }
 //.
-// DEVICE: @[[FOO_NAME:__omp_offloading_[0-9a-z]+_[0-9a-z]+_foo_l[0-9]+]] = protected addrspace(1) constant ptr @_Z3foov
-// DEVICE: @[[BAZ_NAME:__omp_offloading_[0-9a-z]+_[0-9a-z]+_baz_l[0-9]+]] = protected addrspace(1) constant ptr @_Z3bazv
+// DEVICE: @[[FOO_NAME:__omp_offloading_[0-9a-z]+_[0-9a-z]+_foo_l[0-9]+]] = protected addrspace(1) constant {{ptr|ptr addrspace\(9\)}} @_Z3foov
+// DEVICE: @[[BAZ_NAME:__omp_offloading_[0-9a-z]+_[0-9a-z]+_baz_l[0-9]+]] = protected addrspace(1) constant {{ptr|ptr addrspace\(9\)}} @_Z3bazv
 // DEVICE: @var = protected addrspace(1) global i8 0, align 1
-// DEVICE: @[[BAR_NAME:__omp_offloading_[0-9a-z]+_[0-9a-z]+_bar_l[0-9]+]] = protected addrspace(1) constant ptr @_ZL3barv
+// DEVICE: @[[BAR_NAME:__omp_offloading_[0-9a-z]+_[0-9a-z]+_bar_l[0-9]+]] = protected addrspace(1) constant {{ptr|ptr addrspace\(9\)}} @_ZL3barv
 //.
 void foo() { }
 #pragma omp declare target to(foo) indirect

clang/test/OpenMP/amdgcn_parallel_num_threads_strict_messages.cpp renamed to clang/test/OpenMP/target_parallel_num_threads_strict_messages.cpp

@@ -5,6 +5,13 @@
 // RUN: %clang_cc1 -DF3 -verify -fopenmp -fopenmp-version=60 -triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm-bc %s -o %t-ppc-host-ppc.bc
 // RUN: %clang_cc1 -DF3 -DTARGET -verify -fopenmp -fopenmp-version=60 -triple amdgcn-amd-amdhsa -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host-ppc.bc -o /dev/null
 
+// RUN: %clang_cc1 -DF1 -verify -fopenmp -fopenmp-version=60 -triple x86_64-unknown-unknown -fopenmp-targets=spirv64-intel -emit-llvm-bc %s -o %t-spirv-ppc-host-ppc.bc
+// RUN: %clang_cc1 -DF1 -DTARGET -verify -fopenmp -fopenmp-version=60 -triple spirv64-intel -fopenmp-targets=spirv64-intel -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-spirv-ppc-host-ppc.bc -o /dev/null
+// RUN: %clang_cc1 -DF2 -verify -fopenmp -fopenmp-version=60 -triple x86_64-unknown-unknown -fopenmp-targets=spirv64-intel -emit-llvm-bc %s -o %t-spirv-ppc-host-ppc.bc
+// RUN: %clang_cc1 -DF2 -DTARGET -verify -fopenmp -fopenmp-version=60 -triple spirv64-intel -fopenmp-targets=spirv64-intel -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-spirv-ppc-host-ppc.bc -o /dev/null
+// RUN: %clang_cc1 -DF3 -verify -fopenmp -fopenmp-version=60 -triple x86_64-unknown-unknown -fopenmp-targets=spirv64-intel -emit-llvm-bc %s -o %t-spirv-ppc-host-ppc.bc
+// RUN: %clang_cc1 -DF3 -DTARGET -verify -fopenmp -fopenmp-version=60 -triple spirv64-intel -fopenmp-targets=spirv64-intel -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-spirv-ppc-host-ppc.bc -o /dev/null
+
 #ifndef TARGET
 // expected-no-diagnostics
 #endif

llvm/include/llvm/Frontend/OpenMP/OMPKinds.def

@@ -472,7 +472,7 @@ __OMP_RTL(__kmpc_target_init, false, Int32, KernelEnvironmentPtr, KernelLaunchEn
 __OMP_RTL(__kmpc_target_deinit, false, Void,)
 __OMP_RTL(__kmpc_kernel_prepare_parallel, false, Void, VoidPtr)
 __OMP_RTL(__kmpc_parallel_51, false, Void, IdentPtr, Int32, Int32, Int32, Int32,
-          FuncPtrTy, VoidPtr, VoidPtrPtr, SizeTy)
+          FuncPtrTy, FuncPtrTy, VoidPtrPtr, SizeTy)
 __OMP_RTL(__kmpc_for_static_loop_4, false, Void, IdentPtr, VoidPtr, VoidPtr, Int32, Int32, Int32, Int8)
 __OMP_RTL(__kmpc_for_static_loop_4u, false, Void, IdentPtr, VoidPtr, VoidPtr, Int32, Int32, Int32, Int8)
 __OMP_RTL(__kmpc_for_static_loop_8, false, Void, IdentPtr, VoidPtr, VoidPtr, Int64, Int64, Int64, Int8)