[Flang][OpenMP][MLIR] Initial declare target to for variables implementation

While the infrastructure for declare target to/enter and link for variables exists in the MLIR dialect and at the Flang level, the current lowering from MLIR -> LLVM IR isn't in place, it's only in place for variables that have the link clause applied.

This PR aims to extend that lowering to an initial implementation that incorporates declare target to as well, which primarily requires changes in the OpenMPToLLVMIRTranslation phase. However, a minor addition to the OpenMP dialect was required to extend the declare target enumerator to include a default None field as well.

This also requires a minor change to the Flang lowering's MapInfoFinlization.cpp pass to alter the map type for descriptors to deal with cases where a variable is marked declare to. Currently, when a descriptor variable is mapped declare target to the descriptor component can become attatched, and cannot be updated, this results in issues when an unusual allocation range is specified (effectively an off-by X error). The current solution is to map the descriptor always, as we always require an up-to-date version of this data. However, this also requires an interlinked PR that adds a more intricate type of mapping of structures/record types that clang currently implements, to circumvent the overwriting of the pointer in the descriptor.

3/3 required PRs to enable declare target to mapping, this PR should pass all tests and provide an all green CI.

Co-authored-by: Raghu Maddhipatla [email protected]

Author: agozillon

flang/include/flang/Optimizer/Builder/FIRBuilder.h

@@ -970,6 +970,7 @@ void genLifetimeEnd(mlir::OpBuilder &builder, mlir::Location loc,
 mlir::Value getDescriptorWithNewBaseAddress(fir::FirOpBuilder &builder,
                                             mlir::Location loc, mlir::Value box,
                                             mlir::Value newAddr);
+
 } // namespace fir::factory
 
 #endif // FORTRAN_OPTIMIZER_BUILDER_FIRBUILDER_H

flang/test/Lower/OpenMP/derived-type-map.f90

@@ -1,5 +1,6 @@
 !RUN: %flang_fc1 -emit-hlfir -fopenmp %s -o - | FileCheck %s
 
+!CHECK: omp.declare_mapper @[[MAPPER1:_QQFmaptype_derived_implicit_allocatablescalar_and_array_omp_default_mapper]] : !fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}> {
 
 !CHECK: %[[ALLOCA:.*]] = fir.alloca !fir.type<_QFmaptype_derived_implicitTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}> {bindc_name = "scalar_arr", uniq_name = "_QFmaptype_derived_implicitEscalar_arr"}
 !CHECK: %[[DECLARE:.*]]:2 = hlfir.declare %[[ALLOCA]] {uniq_name = "_QFmaptype_derived_implicitEscalar_arr"} : (!fir.ref<!fir.type<_QFmaptype_derived_implicitTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>) -> (!fir.ref<!fir.type<_QFmaptype_derived_implicitTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>, !fir.ref<!fir.type<_QFmaptype_derived_implicitTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>)
@@ -18,6 +19,26 @@ subroutine mapType_derived_implicit
   !$omp end target
 end subroutine mapType_derived_implicit
 
+!CHECK: %[[ALLOCA:.*]] = fir.alloca !fir.box<!fir.heap<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>> {bindc_name = "scalar_arr", uniq_name = "_QFmaptype_derived_implicit_allocatableEscalar_arr"}
+!CHECK: %[[DECLARE:.*]]:2 = hlfir.declare %[[ALLOCA]] {fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFmaptype_derived_implicit_allocatableEscalar_arr"} : (!fir.ref<!fir.box<!fir.heap<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>>, !fir.ref<!fir.box<!fir.heap<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>>)
+!CHECK: %[[BOX_ADDR:.*]] = fir.box_offset %[[DECLARE]]#1 base_addr : (!fir.ref<!fir.box<!fir.heap<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>>) -> !fir.llvm_ptr<!fir.ref<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>
+!CHECK: %[[BASE_MAP:.*]] = omp.map.info var_ptr(%[[DECLARE]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>>, !fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>) map_clauses(implicit, tofrom) capture(ByRef) var_ptr_ptr(%[[BOX_ADDR]] : !fir.llvm_ptr<!fir.ref<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>) mapper(@[[MAPPER1]]) -> !fir.llvm_ptr<!fir.ref<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>> {name = ""}
+!CHECK: %[[DESC_MAP:.*]] = omp.map.info var_ptr(%[[DECLARE]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>>, !fir.box<!fir.heap<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>) map_clauses(always, implicit, descriptor, to, attach) capture(ByRef) members(%[[BASE_MAP]] : [0] : !fir.llvm_ptr<!fir.ref<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>) -> !fir.ref<!fir.box<!fir.heap<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>> {name = "scalar_arr"}
+!CHECK:     omp.target map_entries(%[[DESC_MAP]] -> %[[ARG0:.*]], %[[BASE_MAP]] -> %[[ARG1:.*]] : !fir.ref<!fir.box<!fir.heap<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>>, !fir.llvm_ptr<!fir.ref<!fir.type<_QFmaptype_derived_implicit_allocatableTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>>) {
+subroutine mapType_derived_implicit_allocatable
+    type :: scalar_and_array
+      real(4) :: real
+      integer(4) :: array(10)
+      integer(4) :: int
+    end type scalar_and_array
+    type(scalar_and_array), allocatable :: scalar_arr 
+
+    allocate (scalar_arr)
+    !$omp target
+       scalar_arr%int = 1
+    !$omp end target
+end subroutine mapType_derived_implicit_allocatable
+
 !CHECK: %[[ALLOCA:.*]] = fir.alloca !fir.type<_QFmaptype_derived_explicitTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}> {bindc_name = "scalar_arr", uniq_name = "_QFmaptype_derived_explicitEscalar_arr"}
 !CHECK: %[[DECLARE:.*]]:2 = hlfir.declare %[[ALLOCA]] {uniq_name = "_QFmaptype_derived_explicitEscalar_arr"} : (!fir.ref<!fir.type<_QFmaptype_derived_explicitTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>) -> (!fir.ref<!fir.type<_QFmaptype_derived_explicitTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>, !fir.ref<!fir.type<_QFmaptype_derived_explicitTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>)
 !CHECK: %[[MAP:.*]] = omp.map.info var_ptr(%[[DECLARE]]#1 : !fir.ref<!fir.type<_QFmaptype_derived_explicitTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>>, !fir.type<_QFmaptype_derived_explicitTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>) map_clauses(tofrom) capture(ByRef) -> !fir.ref<!fir.type<_QFmaptype_derived_explicitTscalar_and_array{real:f32,array:!fir.array<10xi32>,int:i32}>> {name = "scalar_arr"}

flang/test/Lower/OpenMP/map-character.f90

@@ -63,4 +63,3 @@ end subroutine TestOfCharacter
 !CHECK:    %[[UNBOXED_TGT_A0:.*]]:2 = fir.unboxchar %[[TGT_A0_BC_LD]] : (!fir.boxchar<1>) -> (!fir.ref<!fir.char<1,?>>, index)
 !CHECK:    %[[TGT_A0_DECL:.*]]:2 = hlfir.declare %[[TGT_A0]] typeparams %[[UNBOXED_TGT_A0]]#1 {{.*}} -> (!fir.boxchar<1>, !fir.ref<!fir.char<1,?>>)
 !CHECK:    %[[TGT_A1_DECL:.*]]:2 = hlfir.declare %[[TGT_A1]] typeparams %[[UNBOXED_TGT_A1]]#1 {{.*}} -> (!fir.boxchar<1>, !fir.ref<!fir.char<1,?>>)
-

flang/test/Lower/OpenMP/map-neg-alloca-derived-type-array.f90

@@ -24,4 +24,4 @@ subroutine map_negative_bounds_allocatable_dtype()
 ! CHECK:           %[[VAL_11:.*]] = fir.coordinate_of %[[VAL_10]], data : (!fir.ref<!fir.type<_QFmap_negative_bounds_allocatable_dtypeTderived_type{data:!fir.box<!fir.ptr<!fir.array<?x?x?xf32>>>}>>) -> !fir.ref<!fir.box<!fir.ptr<!fir.array<?x?x?xf32>>>>
 ! CHECK:           %[[VAL_12:.*]] = fir.box_offset %[[VAL_11]] base_addr : (!fir.ref<!fir.box<!fir.ptr<!fir.array<?x?x?xf32>>>>) -> !fir.llvm_ptr<!fir.ref<!fir.array<?x?x?xf32>>>
 ! CHECK:           %[[VAL_13:.*]] = omp.map.info var_ptr(%[[VAL_11]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?x?x?xf32>>>>, f32) map_clauses(tofrom) capture(ByRef) var_ptr_ptr(%[[VAL_12]] : !fir.llvm_ptr<!fir.ref<!fir.array<?x?x?xf32>>>) bounds({{.*}}) -> !fir.llvm_ptr<!fir.ref<!fir.array<?x?x?xf32>>> {name = ""}
-! CHECK:           %[[VAL_14:.*]] = omp.map.info var_ptr(%[[VAL_11]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?x?x?xf32>>>>, !fir.box<!fir.ptr<!fir.array<?x?x?xf32>>>) map_clauses(to) capture(ByRef) -> !fir.ref<!fir.box<!fir.ptr<!fir.array<?x?x?xf32>>>> {name = {{.*}}}
+! CHECK:           %[[VAL_14:.*]] = omp.map.info var_ptr(%[[VAL_11]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?x?x?xf32>>>>, !fir.box<!fir.ptr<!fir.array<?x?x?xf32>>>) map_clauses(always, to) capture(ByRef) -> !fir.ref<!fir.box<!fir.ptr<!fir.array<?x?x?xf32>>>> {name = {{.*}}}

flang/test/Lower/OpenMP/optional-argument-map-2.f90

@@ -105,7 +105,7 @@ end module mod
 ! CHECK-NO-FPRIV:           %[[VAL_21:.*]] = omp.map.bounds lower_bound(%[[VAL_17]] : index) upper_bound(%[[VAL_20]] : index) extent(%[[VAL_19]]#1 : index) stride(%[[VAL_18]] : index) start_idx(%[[VAL_17]] : index) {stride_in_bytes = true}
 ! CHECK-NO-FPRIV:           %[[VAL_22:.*]] = fir.box_offset %[[VAL_0]] base_addr : (!fir.ref<!fir.boxchar<1>>) -> !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>>
 ! CHECK-NO-FPRIV:           %[[VAL_23:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.boxchar<1>>, !fir.char<1,?>) map_clauses(implicit, to) capture(ByRef) var_ptr_ptr(%[[VAL_22]] : !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>>) bounds(%14) -> !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>> {name = ""}
-! CHECK-NO-FPRIV:           %[[VAL_24:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.boxchar<1>>, !fir.boxchar<1>) map_clauses(implicit, to) capture(ByRef) members(%[[VAL_23]] : [0] : !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>>) -> !fir.ref<!fir.boxchar<1>> {name = ""}
+! CHECK-NO-FPRIV:           %[[VAL_24:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.boxchar<1>>, !fir.boxchar<1>) map_clauses(always, implicit, to) capture(ByRef) members(%[[VAL_23]] : [0] : !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>>) -> !fir.ref<!fir.boxchar<1>> {name = ""}
 ! CHECK-NO-FPRIV:           omp.target map_entries(%[[VAL_7]] -> %[[VAL_25:.*]], %[[VAL_16]] -> %[[VAL_26:.*]], %[[VAL_24]] -> %[[VAL_27:.*]], %[[VAL_23]] -> %[[VAL_28:.*]] : !fir.ref<!fir.char<1,4>>, !fir.ref<!fir.char<1,?>>, !fir.ref<!fir.boxchar<1>>, !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>>) {
 ! CHECK-NO-FPRIV:             %[[VAL_29:.*]] = fir.load %[[VAL_27]] : !fir.ref<!fir.boxchar<1>>
 ! CHECK-NO-FPRIV:             %[[VAL_30:.*]]:2 = fir.unboxchar %[[VAL_29]] : (!fir.boxchar<1>) -> (!fir.ref<!fir.char<1,?>>, index)

flang/test/Lower/OpenMP/optional-argument-map-3.f90

@@ -33,7 +33,7 @@ end subroutine foo
 ! CHECK:             }
 ! CHECK:             %[[VAL_3:.*]] = fir.box_offset %[[VAL_0]] base_addr : (!fir.ref<!fir.box<!fir.array<?xf32>>>) -> !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>
 ! CHECK:             %[[VAL_4:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>, f32) map_clauses(implicit, tofrom) capture(ByRef) var_ptr_ptr(%[[VAL_3]] : !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>) bounds(%{{.*}}) -> !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>> {name = ""}
-! CHECK:             %[[VAL_5:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>, !fir.box<!fir.array<?xf32>>) map_clauses(implicit, to) capture(ByRef) members(%[[VAL_4]] : [0] : !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>) -> !fir.ref<!fir.array<?xf32>> {name = "dt"}
+! CHECK:             %[[VAL_5:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>, !fir.box<!fir.array<?xf32>>) map_clauses(always, implicit, to) capture(ByRef) members(%[[VAL_4]] : [0] : !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>) -> !fir.ref<!fir.array<?xf32>> {name = "dt"}
 ! CHECK:             omp.target host_eval({{.*}}) map_entries({{.*}}%[[VAL_5]] -> {{.*}}, %[[VAL_4]] -> {{.*}} : {{.*}}) {
 ! CHECK:           } else {
 ! CHECK:             %[[VAL_6:.*]] = fir.is_present %[[VAL_1]]#1 : (!fir.box<!fir.array<?xf32>>) -> i1
@@ -42,5 +42,5 @@ end subroutine foo
 ! CHECK:             }
 ! CHECK:             %[[VAL_7:.*]] = fir.box_offset %[[VAL_0]] base_addr : (!fir.ref<!fir.box<!fir.array<?xf32>>>) -> !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>
 ! CHECK:             %[[VAL_8:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>, f32) map_clauses(implicit, tofrom) capture(ByRef) var_ptr_ptr(%[[VAL_7]] : !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>) bounds(%{{.*}}) -> !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>> {name = ""}
-! CHECK:             %[[VAL_9:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>, !fir.box<!fir.array<?xf32>>) map_clauses(implicit, to) capture(ByRef) members(%[[VAL_8]] : [0] : !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>) -> !fir.ref<!fir.array<?xf32>> {name = "dt"}
+! CHECK:             %[[VAL_9:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>, !fir.box<!fir.array<?xf32>>) map_clauses(always, implicit, to) capture(ByRef) members(%[[VAL_8]] : [0] : !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>) -> !fir.ref<!fir.array<?xf32>> {name = "dt"}
 ! CHECK:             omp.target host_eval({{.*}}) map_entries({{.*}}, %[[VAL_9]] ->{{.*}}, %[[VAL_8]] -> {{.*}} : {{.*}}) {

flang/test/Lower/OpenMP/target-enter-data-default-openmp52.f90

@@ -11,7 +11,7 @@ subroutine initialize()
   allocate(A)
   !$omp target enter data map(A)
   !CHECK-52: omp.map.info var_ptr(%2 : !fir.ref<!fir.box<!fir.heap<f32>>>, f32) map_clauses(to) capture(ByRef) var_ptr_ptr(%5 : !fir.llvm_ptr<!fir.ref<f32>>) -> !fir.llvm_ptr<!fir.ref<f32>> {name = ""}
-  !CHECK-52: omp.map.info var_ptr(%2 : !fir.ref<!fir.box<!fir.heap<f32>>>, !fir.box<!fir.heap<f32>>) map_clauses(to) capture(ByRef) members(%6 : [0] : !fir.llvm_ptr<!fir.ref<f32>>) -> !fir.ref<!fir.box<!fir.heap<f32>>> {name = "a"}
+  !CHECK-52: omp.map.info var_ptr(%2 : !fir.ref<!fir.box<!fir.heap<f32>>>, !fir.box<!fir.heap<f32>>) map_clauses(always, to) capture(ByRef) members(%6 : [0] : !fir.llvm_ptr<!fir.ref<f32>>) -> !fir.ref<!fir.box<!fir.heap<f32>>> {name = "a"}
   !CHECK-51: to and alloc map types are permitted
 
   end subroutine initialize

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

@@ -7957,12 +7957,8 @@ static Expected<Function *> createOutlinedFunction(
     // preceding mapped arguments that refer to the same global that may be
     // seperate segments. To prevent this, we defer global processing until all
     // other processing has been performed.
-<<<<<<< HEAD
     if (llvm::isa<llvm::GlobalValue, llvm::GlobalObject, llvm::GlobalVariable>(
             removeASCastIfPresent(Input))) {
-=======
-    if (llvm::isa<llvm::GlobalValue, llvm::GlobalObject, llvm::GlobalVariable>(removeASCastIfPresent(Input))) {
->>>>>>> 1d1536f712a6 ([Flang][OpenMP] Additional global address space modifications for device)
       DeferredReplacement.push_back(std::make_pair(Input, InputCopy));
       continue;
     }