[OpenMP] Fix incorrect interop type for number of dependencies

[jhuber6]

Summary

The interop types use the number of dependencies in the function
interface. Every other function uses an i32 to count the number of
dependencies except for the initialization function. This leads to
codegen issues when the rest of the compiler passes in an i32 that
then creates an invalid call. Fix this to be consistent with the other
uses.

[jhuber6]

Fix test

[tianshilei1992]

Does this cause the IR issue?

[jhuber6]

In D150156#4328360, @tianshilei1992 wrote:

Does this cause the IR issue?

Not sure, I just get an undefined symbol error in the linker now. Not sure if that means it's resolved or I just can't reproduce it.

[jhuber6]

Fix add_attributes.ll

clang/test/OpenMP/interop_irbuilder.cpp

 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --include-generated-funcs
 // RUN: %clang_cc1 -verify -triple x86_64-unknown-linux -fopenmp -x c++ -emit-llvm %s -o - | FileCheck %s
 
 // expected-no-diagnostics
 typedef void *omp_interop_t;
 
 void test1() {
 
   int device_id = 4;
   int D0, D1;
   omp_interop_t interop;
 
-#pragma omp interop init(target \
-                         : interop)
+#pragma omp interop init(target : interop)
 
-#pragma omp interop init(targetsync \
-                         : interop)
+#pragma omp interop init(targetsync : interop)
 
-#pragma omp interop init(target \
-                         : interop) device(device_id)
+#pragma omp interop init(target : interop) device(device_id)
 
-#pragma omp interop init(targetsync \
-                         : interop) device(device_id)
+#pragma omp interop init(targetsync : interop) device(device_id)
 
-#pragma omp interop use(interop) depend(in \
-                                        : D0, D1) nowait
+#pragma omp interop use(interop) depend(in : D0, D1) nowait
 
-#pragma omp interop destroy(interop) depend(in \
-                                            : D0, D1)
+#pragma omp interop destroy(interop) depend(in : D0, D1)
 }
 
 struct S {
   omp_interop_t interop;
   void member_test();
 };
 
 void S::member_test() {
 
   int device_id = 4;
   int D0, D1;
 
-#pragma omp interop init(target \
-                         : interop)
+#pragma omp interop init(target : interop)
 
-#pragma omp interop init(targetsync \
-                         : interop)
+#pragma omp interop init(targetsync : interop)
 
-#pragma omp interop init(target \
-                         : interop) device(device_id)
+#pragma omp interop init(target : interop) device(device_id)
 
-#pragma omp interop init(targetsync \
-                         : interop) device(device_id)
+#pragma omp interop init(targetsync : interop) device(device_id)
 
-#pragma omp interop use(interop) depend(in \
-                                        : D0, D1) nowait
+#pragma omp interop use(interop) depend(in : D0, D1) nowait
 
-#pragma omp interop destroy(interop) depend(in \
-                                            : D0, D1)
+#pragma omp interop destroy(interop) depend(in : D0, D1)
 }
 // CHECK-LABEL: @_Z5test1v(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[DEVICE_ID:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    [[D0:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    [[D1:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    [[INTEROP:%.*]] = alloca ptr, align 8
 // CHECK-NEXT:    [[DOTDEP_ARR_ADDR:%.*]] = alloca [2 x %struct.kmp_depend_info], align 8
 // CHECK-NEXT:    [[DEP_COUNTER_ADDR:%.*]] = alloca i64, align 8
 // CHECK-NEXT:    [[DOTDEP_ARR_ADDR5:%.*]] = alloca [2 x %struct.kmp_depend_info], align 8
 // CHECK-NEXT:    [[DEP_COUNTER_ADDR6:%.*]] = alloca i64, align 8
 // CHECK-NEXT:    store i32 4, ptr [[DEVICE_ID]], align 4
 // CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1:[0-9]+]])
-// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], ptr [[INTEROP]], i32 1, i32 -1, i64 0, ptr null, i32 0)
+// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], ptr [[INTEROP]], i32 1, i32 -1, i32 0, ptr null, i32 0)
 // CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
-// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], ptr [[INTEROP]], i32 2, i32 -1, i64 0, ptr null, i32 0)
+// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], ptr [[INTEROP]], i32 2, i32 -1, i32 0, ptr null, i32 0)
 // CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DEVICE_ID]], align 4
 // CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
-// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM2]], ptr [[INTEROP]], i32 1, i32 [[TMP0]], i64 0, ptr null, i32 0)
+// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM2]], ptr [[INTEROP]], i32 1, i32 [[TMP0]], i32 0, ptr null, i32 0)
 // CHECK-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DEVICE_ID]], align 4
 // CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
-// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], ptr [[INTEROP]], i32 2, i32 [[TMP1]], i64 0, ptr null, i32 0)
+// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], ptr [[INTEROP]], i32 2, i32 [[TMP1]], i32 0, ptr null, i32 0)
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [2 x %struct.kmp_depend_info], ptr [[DOTDEP_ARR_ADDR]], i64 0, i64 0
 // CHECK-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[D0]] to i64
 // CHECK-NEXT:    [[TMP4:%.*]] = getelementptr [[STRUCT_KMP_DEPEND_INFO:%.*]], ptr [[TMP2]], i64 0
 // CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], ptr [[TMP4]], i32 0, i32 0
 // CHECK-NEXT:    store i64 [[TMP3]], ptr [[TMP5]], align 8
 // CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], ptr [[TMP4]], i32 0, i32 1
 // CHECK-NEXT:    store i64 4, ptr [[TMP6]], align 8
 // CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], ptr [[TMP4]], i32 0, i32 2

 // CHECK-NEXT:    [[DEP_COUNTER_ADDR:%.*]] = alloca i64, align 8
 // CHECK-NEXT:    [[DOTDEP_ARR_ADDR10:%.*]] = alloca [2 x %struct.kmp_depend_info], align 8
 // CHECK-NEXT:    [[DEP_COUNTER_ADDR11:%.*]] = alloca i64, align 8
 // CHECK-NEXT:    store ptr [[THIS:%.*]], ptr [[THIS_ADDR]], align 8
 // CHECK-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK-NEXT:    store i32 4, ptr [[DEVICE_ID]], align 4
 // CHECK-NEXT:    [[INTEROP:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
-// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], ptr [[INTEROP]], i32 1, i32 -1, i64 0, ptr null, i32 0)
+// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], ptr [[INTEROP]], i32 1, i32 -1, i32 0, ptr null, i32 0)
 // CHECK-NEXT:    [[INTEROP2:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[THIS1]], i32 0, i32 0
 // CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
-// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], ptr [[INTEROP2]], i32 2, i32 -1, i64 0, ptr null, i32 0)
+// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], ptr [[INTEROP2]], i32 2, i32 -1, i32 0, ptr null, i32 0)
 // CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DEVICE_ID]], align 4
 // CHECK-NEXT:    [[INTEROP4:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[THIS1]], i32 0, i32 0
 // CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM5:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
-// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM5]], ptr [[INTEROP4]], i32 1, i32 [[TMP0]], i64 0, ptr null, i32 0)
+// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM5]], ptr [[INTEROP4]], i32 1, i32 [[TMP0]], i32 0, ptr null, i32 0)
 // CHECK-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DEVICE_ID]], align 4
 // CHECK-NEXT:    [[INTEROP6:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[THIS1]], i32 0, i32 0
 // CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM7:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
-// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], ptr [[INTEROP6]], i32 2, i32 [[TMP1]], i64 0, ptr null, i32 0)
+// CHECK-NEXT:    call void @__tgt_interop_init(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], ptr [[INTEROP6]], i32 2, i32 [[TMP1]], i32 0, ptr null, i32 0)
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [2 x %struct.kmp_depend_info], ptr [[DOTDEP_ARR_ADDR]], i64 0, i64 0
 // CHECK-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[D0]] to i64
 // CHECK-NEXT:    [[TMP4:%.*]] = getelementptr [[STRUCT_KMP_DEPEND_INFO:%.*]], ptr [[TMP2]], i64 0
 // CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], ptr [[TMP4]], i32 0, i32 0
 // CHECK-NEXT:    store i64 [[TMP3]], ptr [[TMP5]], align 8
 // CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], ptr [[TMP4]], i32 0, i32 1
 // CHECK-NEXT:    store i64 4, ptr [[TMP6]], align 8
 // CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_DEPEND_INFO]], ptr [[TMP4]], i32 0, i32 2

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

 __OMP_RTL(__kmpc_doacross_fini, false, Void, IdentPtr, Int32)
 
 __OMP_RTL(__kmpc_alloc, false, VoidPtr, /* Int */ Int32, SizeTy, VoidPtr)
 __OMP_RTL(__kmpc_aligned_alloc, false, VoidPtr, /* Int */ Int32, SizeTy, SizeTy,
           VoidPtr)
 __OMP_RTL(__kmpc_free, false, Void, /* Int */ Int32, VoidPtr, VoidPtr)
 
 __OMP_RTL(__tgt_interop_init, false, Void, IdentPtr, Int32, VoidPtrPtr, Int32,
-          Int32, Int64, VoidPtr, Int32)
+          Int32, Int32, VoidPtr, Int32)
 __OMP_RTL(__tgt_interop_destroy, false, Void, IdentPtr, Int32, VoidPtrPtr,
           Int32, Int32, VoidPtr, Int32)
 __OMP_RTL(__tgt_interop_use, false, Void, IdentPtr, Int32, VoidPtrPtr, Int32,
           Int32, VoidPtr, Int32)
 
 __OMP_RTL(__kmpc_init_allocator, false, /* omp_allocator_handle_t */ VoidPtr,
           /* Int */ Int32, /* omp_memespace_handle_t */ VoidPtr,
           /* Int */ Int32, /* omp_alloctrait_t */ VoidPtr)

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

CallInst *OpenMPIRBuilder::createOMPInteropInit(
   uint32_t SrcLocStrSize;
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
   Value *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
   Value *ThreadId = getOrCreateThreadID(Ident);
   if (Device == nullptr)
     Device = ConstantInt::get(Int32, -1);
   Constant *InteropTypeVal = ConstantInt::get(Int32, (int)InteropType);
   if (NumDependences == nullptr) {
-    NumDependences = ConstantInt::get(Int64, 0);
+    NumDependences = ConstantInt::get(Int32, 0);
     PointerType *PointerTypeVar = Type::getInt8PtrTy(M.getContext());
     DependenceAddress = ConstantPointerNull::get(PointerTypeVar);
   }
   Value *HaveNowaitClauseVal = ConstantInt::get(Int32, HaveNowaitClause);
   Value *Args[] = {
       Ident,  ThreadId,       InteropVar,        InteropTypeVal,
       Device, NumDependences, DependenceAddress, HaveNowaitClauseVal};
 

llvm/test/Transforms/OpenMP/add_attributes.ll

 declare i64 @__kmpc_shuffle_int64(i64, i16, i16);
 
 declare void @__kmpc_target_deinit(ptr, i8);
 
 declare i32 @__kmpc_target_init(ptr, i8, i1);
 
 declare void @__tgt_interop_destroy(ptr, i32, ptr, i32, i32, ptr, i32);
 
-declare void @__tgt_interop_init(ptr, i32, ptr, i32, i32, i64, ptr, i32);
+declare void @__tgt_interop_init(ptr, i32, ptr, i32, i32, i32, ptr, i32);
 
 declare void @__tgt_interop_use(ptr, i32, ptr, i32, i32, ptr, i32);
 
 declare void @__tgt_target_data_begin_mapper_issue(ptr, i64, i32, ptr, ptr, ptr, ptr, ptr, ptr, ptr);
 
 declare void @__tgt_target_data_begin_mapper_wait(i64, ptr);
 
 declare i32 @__tgt_target_kernel(ptr, i64, i32, i32, ptr, ptr);

 
 ; CHECK-NOT: Function Attrs
 ; CHECK: declare i32 @__kmpc_target_init(ptr, i8, i1)
 
 ; CHECK-NOT: Function Attrs
 ; CHECK: declare void @__tgt_interop_destroy(ptr, i32, ptr, i32, i32, ptr, i32)
 
 ; CHECK-NOT: Function Attrs
-; CHECK: declare void @__tgt_interop_init(ptr, i32, ptr, i32, i32, i64, ptr, i32)
+; CHECK: declare void @__tgt_interop_init(ptr, i32, ptr, i32, i32, i32, ptr, i32)
 
 ; CHECK-NOT: Function Attrs
 ; CHECK: declare void @__tgt_interop_use(ptr, i32, ptr, i32, i32, ptr, i32)
 
 ; CHECK-NOT: Function Attrs
 ; CHECK: declare void @__tgt_target_data_begin_mapper_issue(ptr, i64, i32, ptr, ptr, ptr, ptr, ptr, ptr, ptr)
 
 ; CHECK-NOT: Function Attrs

 
 ; OPTIMISTIC-NOT: Function Attrs
 ; OPTIMISTIC: declare i32 @__kmpc_target_init(ptr, i8, i1)
 
 ; OPTIMISTIC-NOT: Function Attrs
 ; OPTIMISTIC: declare void @__tgt_interop_destroy(ptr, i32, ptr, i32, i32, ptr, i32)
 
 ; OPTIMISTIC-NOT: Function Attrs
-; OPTIMISTIC: declare void @__tgt_interop_init(ptr, i32, ptr, i32, i32, i64, ptr, i32)
+; OPTIMISTIC: declare void @__tgt_interop_init(ptr, i32, ptr, i32, i32, i32, ptr, i32)
 
 ; OPTIMISTIC-NOT: Function Attrs
 ; OPTIMISTIC: declare void @__tgt_interop_use(ptr, i32, ptr, i32, i32, ptr, i32)
 
 ; OPTIMISTIC-NOT: Function Attrs
 ; OPTIMISTIC: declare void @__tgt_target_data_begin_mapper_issue(ptr, i64, i32, ptr, ptr, ptr, ptr, ptr, ptr, ptr)
 
 ; OPTIMISTIC-NOT: Function Attrs

 
 ; EXT-NOT: Function Attrs
 ; EXT: declare signext i32 @__kmpc_target_init(ptr, i8 signext, i1 signext)
 
 ; EXT-NOT: Function Attrs
 ; EXT: declare void @__tgt_interop_destroy(ptr, i32 signext, ptr, i32 signext, i32 signext, ptr, i32 signext)
 
 ; EXT-NOT: Function Attrs
-; EXT: declare void @__tgt_interop_init(ptr, i32 signext, ptr, i32 signext, i32 signext, i64, ptr, i32 signext)
+; EXT: declare void @__tgt_interop_init(ptr, i32 signext, ptr, i32 signext, i32 signext, i32, ptr, i32 signext)
 
 ; EXT-NOT: Function Attrs
 ; EXT: declare void @__tgt_interop_use(ptr, i32 signext, ptr, i32 signext, i32 signext, ptr, i32 signext)
 
 ; EXT-NOT: Function Attrs
 ; EXT: declare void @__tgt_target_data_begin_mapper_issue(ptr, i64, i32 signext, ptr, ptr, ptr, ptr, ptr, ptr, ptr)
 
 ; EXT-NOT: Function Attrs

openmp/libomptarget/src/interop.cpp

 typedef int64_t kmp_int64;
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 void __tgt_interop_init(ident_t *LocRef, kmp_int32 Gtid,
                         omp_interop_val_t *&InteropPtr,
                         kmp_interop_type_t InteropType, kmp_int32 DeviceId,
-                        kmp_int64 Ndeps, kmp_depend_info_t *DepList,
+                        kmp_int32 Ndeps, kmp_depend_info_t *DepList,
                         kmp_int32 HaveNowait) {
   kmp_int32 NdepsNoalias = 0;
   kmp_depend_info_t *NoaliasDepList = NULL;
   assert(InteropType != kmp_interop_type_unknown &&
          "Cannot initialize with unknown interop_type!");
   if (DeviceId == -1) {
     DeviceId = omp_get_default_device();
   }