[flang][openacc] map data operand results to symbols inside compute region (#162306)

Variable references inside OpenACC compute and loop region were
currently always lowered to usages of the same SSA values than in the
host thread, even for variables that appear in data clauses and for
which acc data operations are created.

This makes it a non-trivial task to identify implicit data usages vs
usage of data appearing in clauses because the SSA addresses used in the
region may have a non-trivial SSA relationship with the SSA addresses
used as inputs of the data operations, especially after CSE runs that
may merge component or array element addressing operations with similar
addressing on the host thread (fir.coordinate/hlfir.designate).

This patch updates OpenACC lowering to remap the Symbol that appear in
data clauses to the related acc data operation result for the scope of
the compute or loop region.

To allow FIR passes to reason about these addresses, a new hlfir.declare
operation is created with the acc data operation result. This gives
access to the shape, contiguity, attributes, and dummy argument
relationships inside the region without having FIR extended to
understand the data operations.

Author: jeanPerier

flang/include/flang/Lower/AbstractConverter.h

@@ -123,6 +123,10 @@ class AbstractConverter {
   virtual Fortran::lower::SymMap::StorageDesc
   getSymbolStorage(SymbolRef sym) = 0;
 
+  /// Return the Symbol Map used to map semantics::Symbol to their SSA
+  /// values in the generated MLIR.
+  virtual Fortran::lower::SymMap &getSymbolMap() = 0;
+
   /// Override lowering of expression with pre-lowered values.
   /// Associate mlir::Value to evaluate::Expr. All subsequent call to
   /// genExprXXX() will replace any occurrence of an overridden

flang/lib/Lower/Bridge.cpp

@@ -643,6 +643,8 @@ class FirConverter : public Fortran::lower::AbstractConverter {
     return localSymbols.lookupStorage(sym);
   }
 
+  Fortran::lower::SymMap &getSymbolMap() override final { return localSymbols; }
+
   void
   overrideExprValues(const Fortran::lower::ExprToValueMap *map) override final {
     exprValueOverrides = map;

flang/lib/Lower/OpenACC.cpp

@@ -41,19 +41,21 @@ subroutine test(a)
 ! CHECK:           %[[VAL_5:.*]]:2 = hlfir.declare %[[VAL_4]] {uniq_name = "_QMm_firstprivate_derived_alloc_compFtestEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK:           %[[VAL_6:.*]] = acc.firstprivate varPtr(%[[VAL_1]]#0 : !fir.ref<!fir.type<_QMm_firstprivate_derived_alloc_compTpoint{x:!fir.box<!fir.heap<!fir.array<?xf32>>>}>>) -> !fir.ref<!fir.type<_QMm_firstprivate_derived_alloc_compTpoint{x:!fir.box<!fir.heap<!fir.array<?xf32>>>}>> {name = "a"}
 ! CHECK:           acc.parallel combined(loop) firstprivate(@firstprivatization_ref_rec__QMm_firstprivate_derived_alloc_compTpoint -> %[[VAL_6]] : !fir.ref<!fir.type<_QMm_firstprivate_derived_alloc_compTpoint{x:!fir.box<!fir.heap<!fir.array<?xf32>>>}>>) {
-! CHECK:             %[[VAL_7:.*]] = arith.constant 1 : i32
-! CHECK:             %[[VAL_8:.*]] = fir.load %[[VAL_5]]#0 : !fir.ref<i32>
+! CHECK:             %[[VAL_7:.*]] = fir.dummy_scope : !fir.dscope
+! CHECK:             %[[VAL_8:.*]]:2 = hlfir.declare %[[VAL_6]] dummy_scope %[[VAL_7]] {uniq_name = "_QMm_firstprivate_derived_alloc_compFtestEa"} : (!fir.ref<!fir.type<_QMm_firstprivate_derived_alloc_compTpoint{x:!fir.box<!fir.heap<!fir.array<?xf32>>>}>>, !fir.dscope) -> (!fir.ref<!fir.type<_QMm_firstprivate_derived_alloc_compTpoint{x:!fir.box<!fir.heap<!fir.array<?xf32>>>}>>, !fir.ref<!fir.type<_QMm_firstprivate_derived_alloc_compTpoint{x:!fir.box<!fir.heap<!fir.array<?xf32>>>}>>)
 ! CHECK:             %[[VAL_9:.*]] = arith.constant 1 : i32
-! CHECK:             %[[VAL_10:.*]] = acc.private varPtr(%[[VAL_3]]#0 : !fir.ref<i32>) -> !fir.ref<i32> {implicit = true, name = "i"}
-! CHECK:             %[[VAL_11:.*]]:2 = hlfir.declare %[[VAL_10]] {uniq_name = "_QMm_firstprivate_derived_alloc_compFtestEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-! CHECK:             acc.loop combined(parallel) private(@privatization_ref_i32 -> %[[VAL_10]] : !fir.ref<i32>) control(%[[VAL_12:.*]] : i32) = (%[[VAL_7]] : i32) to (%[[VAL_8]] : i32)  step (%[[VAL_9]] : i32) {
-! CHECK:               fir.store %[[VAL_12]] to %[[VAL_11]]#0 : !fir.ref<i32>
-! CHECK:               %[[VAL_13:.*]] = arith.constant 1.000000e+00 : f32
-! CHECK:               %[[VAL_14:.*]] = hlfir.designate %[[VAL_1]]#0{"x"}   {fortran_attrs = #fir.var_attrs<allocatable>} : (!fir.ref<!fir.type<_QMm_firstprivate_derived_alloc_compTpoint{x:!fir.box<!fir.heap<!fir.array<?xf32>>>}>>) -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
-! CHECK:               %[[VAL_15:.*]] = fir.load %[[VAL_14]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
-! CHECK:               %[[VAL_16:.*]] = arith.constant 10 : index
-! CHECK:               %[[VAL_17:.*]] = hlfir.designate %[[VAL_15]] (%[[VAL_16]])  : (!fir.box<!fir.heap<!fir.array<?xf32>>>, index) -> !fir.ref<f32>
-! CHECK:               hlfir.assign %[[VAL_13]] to %[[VAL_17]] : f32, !fir.ref<f32>
+! CHECK:             %[[VAL_10:.*]] = fir.load %[[VAL_5]]#0 : !fir.ref<i32>
+! CHECK:             %[[VAL_11:.*]] = arith.constant 1 : i32
+! CHECK:             %[[VAL_12:.*]] = acc.private varPtr(%[[VAL_3]]#0 : !fir.ref<i32>) -> !fir.ref<i32> {implicit = true, name = "i"}
+! CHECK:             acc.loop combined(parallel) private(@privatization_ref_i32 -> %[[VAL_12]] : !fir.ref<i32>) control(%[[VAL_14:.*]] : i32) = (%[[VAL_9]] : i32) to (%[[VAL_10]] : i32)  step (%[[VAL_11]] : i32) {
+! CHECK:               %[[VAL_13:.*]]:2 = hlfir.declare %[[VAL_12]] {uniq_name = "_QMm_firstprivate_derived_alloc_compFtestEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK:               fir.store %[[VAL_14]] to %[[VAL_13]]#0 : !fir.ref<i32>
+! CHECK:               %[[VAL_15:.*]] = arith.constant 1.000000e+00 : f32
+! CHECK:               %[[VAL_16:.*]] = hlfir.designate %[[VAL_8]]#0{"x"}   {fortran_attrs = #fir.var_attrs<allocatable>} : (!fir.ref<!fir.type<_QMm_firstprivate_derived_alloc_compTpoint{x:!fir.box<!fir.heap<!fir.array<?xf32>>>}>>) -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
+! CHECK:               %[[VAL_17:.*]] = fir.load %[[VAL_16]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
+! CHECK:               %[[VAL_18:.*]] = arith.constant 10 : index
+! CHECK:               %[[VAL_19:.*]] = hlfir.designate %[[VAL_17]] (%[[VAL_18]])  : (!fir.box<!fir.heap<!fir.array<?xf32>>>, index) -> !fir.ref<f32>
+! CHECK:               hlfir.assign %[[VAL_15]] to %[[VAL_19]] : f32, !fir.ref<f32>
 ! CHECK:               acc.yield
 ! CHECK:             } attributes {inclusiveUpperbound = array<i1: true>, independent = [#acc.device_type<none>]}
 ! CHECK:             acc.yield

flang/test/Lower/OpenACC/acc-data-operands-remapping.f90

@@ -41,19 +41,21 @@ subroutine test(a)
 ! CHECK:           %[[VAL_5:.*]]:2 = hlfir.declare %[[VAL_4]] {uniq_name = "_QMm_firstprivate_derived_ptr_compFtestEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK:           %[[VAL_6:.*]] = acc.firstprivate varPtr(%[[VAL_1]]#0 : !fir.ref<!fir.type<_QMm_firstprivate_derived_ptr_compTpoint{x:!fir.box<!fir.ptr<!fir.array<?xf32>>>}>>) -> !fir.ref<!fir.type<_QMm_firstprivate_derived_ptr_compTpoint{x:!fir.box<!fir.ptr<!fir.array<?xf32>>>}>> {name = "a"}
 ! CHECK:           acc.parallel combined(loop) firstprivate(@firstprivatization_ref_rec__QMm_firstprivate_derived_ptr_compTpoint -> %[[VAL_6]] : !fir.ref<!fir.type<_QMm_firstprivate_derived_ptr_compTpoint{x:!fir.box<!fir.ptr<!fir.array<?xf32>>>}>>) {
-! CHECK:             %[[VAL_7:.*]] = arith.constant 1 : i32
-! CHECK:             %[[VAL_8:.*]] = fir.load %[[VAL_5]]#0 : !fir.ref<i32>
+! CHECK:             %[[VAL_7:.*]] = fir.dummy_scope : !fir.dscope
+! CHECK:             %[[VAL_8:.*]]:2 = hlfir.declare %[[VAL_6]] dummy_scope %[[VAL_7]] {uniq_name = "_QMm_firstprivate_derived_ptr_compFtestEa"} : (!fir.ref<!fir.type<_QMm_firstprivate_derived_ptr_compTpoint{x:!fir.box<!fir.ptr<!fir.array<?xf32>>>}>>, !fir.dscope) -> (!fir.ref<!fir.type<_QMm_firstprivate_derived_ptr_compTpoint{x:!fir.box<!fir.ptr<!fir.array<?xf32>>>}>>, !fir.ref<!fir.type<_QMm_firstprivate_derived_ptr_compTpoint{x:!fir.box<!fir.ptr<!fir.array<?xf32>>>}>>)
 ! CHECK:             %[[VAL_9:.*]] = arith.constant 1 : i32
-! CHECK:             %[[VAL_10:.*]] = acc.private varPtr(%[[VAL_3]]#0 : !fir.ref<i32>) -> !fir.ref<i32> {implicit = true, name = "i"}
-! CHECK:             %[[VAL_11:.*]]:2 = hlfir.declare %[[VAL_10]] {uniq_name = "_QMm_firstprivate_derived_ptr_compFtestEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-! CHECK:             acc.loop combined(parallel) private(@privatization_ref_i32 -> %[[VAL_10]] : !fir.ref<i32>) control(%[[VAL_12:.*]] : i32) = (%[[VAL_7]] : i32) to (%[[VAL_8]] : i32)  step (%[[VAL_9]] : i32) {
-! CHECK:               fir.store %[[VAL_12]] to %[[VAL_11]]#0 : !fir.ref<i32>
-! CHECK:               %[[VAL_13:.*]] = arith.constant 1.000000e+00 : f32
-! CHECK:               %[[VAL_14:.*]] = hlfir.designate %[[VAL_1]]#0{"x"}   {fortran_attrs = #fir.var_attrs<pointer>} : (!fir.ref<!fir.type<_QMm_firstprivate_derived_ptr_compTpoint{x:!fir.box<!fir.ptr<!fir.array<?xf32>>>}>>) -> !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>
-! CHECK:               %[[VAL_15:.*]] = fir.load %[[VAL_14]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>
-! CHECK:               %[[VAL_16:.*]] = arith.constant 10 : index
-! CHECK:               %[[VAL_17:.*]] = hlfir.designate %[[VAL_15]] (%[[VAL_16]])  : (!fir.box<!fir.ptr<!fir.array<?xf32>>>, index) -> !fir.ref<f32>
-! CHECK:               hlfir.assign %[[VAL_13]] to %[[VAL_17]] : f32, !fir.ref<f32>
+! CHECK:             %[[VAL_10:.*]] = fir.load %[[VAL_5]]#0 : !fir.ref<i32>
+! CHECK:             %[[VAL_11:.*]] = arith.constant 1 : i32
+! CHECK:             %[[VAL_12:.*]] = acc.private varPtr(%[[VAL_3]]#0 : !fir.ref<i32>) -> !fir.ref<i32> {implicit = true, name = "i"}
+! CHECK:             acc.loop combined(parallel) private(@privatization_ref_i32 -> %[[VAL_12]] : !fir.ref<i32>) control(%[[VAL_14:.*]] : i32) = (%[[VAL_9]] : i32) to (%[[VAL_10]] : i32)  step (%[[VAL_11]] : i32) {
+! CHECK:               %[[VAL_13:.*]]:2 = hlfir.declare %[[VAL_12]] {uniq_name = "_QMm_firstprivate_derived_ptr_compFtestEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK:               fir.store %[[VAL_14]] to %[[VAL_13]]#0 : !fir.ref<i32>
+! CHECK:               %[[VAL_15:.*]] = arith.constant 1.000000e+00 : f32
+! CHECK:               %[[VAL_16:.*]] = hlfir.designate %[[VAL_8]]#0{"x"}   {fortran_attrs = #fir.var_attrs<pointer>} : (!fir.ref<!fir.type<_QMm_firstprivate_derived_ptr_compTpoint{x:!fir.box<!fir.ptr<!fir.array<?xf32>>>}>>) -> !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>
+! CHECK:               %[[VAL_17:.*]] = fir.load %[[VAL_16]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>
+! CHECK:               %[[VAL_18:.*]] = arith.constant 10 : index
+! CHECK:               %[[VAL_19:.*]] = hlfir.designate %[[VAL_17]] (%[[VAL_18]])  : (!fir.box<!fir.ptr<!fir.array<?xf32>>>, index) -> !fir.ref<f32>
+! CHECK:               hlfir.assign %[[VAL_15]] to %[[VAL_19]] : f32, !fir.ref<f32>
 ! CHECK:               acc.yield
 ! CHECK:             } attributes {inclusiveUpperbound = array<i1: true>, independent = [#acc.device_type<none>]}
 ! CHECK:             acc.yield

flang/test/Lower/OpenACC/acc-firstprivate-derived-allocatable-component.f90

@@ -55,16 +55,17 @@ subroutine test()
 ! CHECK:           %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_5]] {uniq_name = "_QMm_firstprivate_derived_user_defFtestEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK:           %[[VAL_7:.*]] = acc.firstprivate varPtr(%[[VAL_2]]#0 : !fir.ref<!fir.type<_QMm_firstprivate_derived_user_defTpoint{x:f32,y:f32,z:f32}>>) -> !fir.ref<!fir.type<_QMm_firstprivate_derived_user_defTpoint{x:f32,y:f32,z:f32}>> {name = "a"}
 ! CHECK:           acc.parallel combined(loop) firstprivate(@firstprivatization_ref_rec__QMm_firstprivate_derived_user_defTpoint -> %[[VAL_7]] : !fir.ref<!fir.type<_QMm_firstprivate_derived_user_defTpoint{x:f32,y:f32,z:f32}>>) {
-! CHECK:             %[[VAL_8:.*]] = arith.constant 1 : i32
-! CHECK:             %[[VAL_9:.*]] = fir.load %[[VAL_6]]#0 : !fir.ref<i32>
-! CHECK:             %[[VAL_10:.*]] = arith.constant 1 : i32
-! CHECK:             %[[VAL_11:.*]] = acc.private varPtr(%[[VAL_4]]#0 : !fir.ref<i32>) -> !fir.ref<i32> {implicit = true, name = "i"}
-! CHECK:             %[[VAL_12:.*]]:2 = hlfir.declare %[[VAL_11]] {uniq_name = "_QMm_firstprivate_derived_user_defFtestEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-! CHECK:             acc.loop combined(parallel) private(@privatization_ref_i32 -> %[[VAL_11]] : !fir.ref<i32>) control(%[[VAL_13:.*]] : i32) = (%[[VAL_8]] : i32) to (%[[VAL_9]] : i32)  step (%[[VAL_10]] : i32) {
-! CHECK:               fir.store %[[VAL_13]] to %[[VAL_12]]#0 : !fir.ref<i32>
-! CHECK:               %[[VAL_14:.*]] = arith.constant 1.000000e+00 : f32
-! CHECK:               %[[VAL_15:.*]] = hlfir.designate %[[VAL_2]]#0{"x"}   : (!fir.ref<!fir.type<_QMm_firstprivate_derived_user_defTpoint{x:f32,y:f32,z:f32}>>) -> !fir.ref<f32>
-! CHECK:               hlfir.assign %[[VAL_14]] to %[[VAL_15]] : f32, !fir.ref<f32>
+! CHECK:             %[[VAL_8:.*]]:2 = hlfir.declare %[[VAL_7]] {uniq_name = "_QMm_firstprivate_derived_user_defFtestEa"} : (!fir.ref<!fir.type<_QMm_firstprivate_derived_user_defTpoint{x:f32,y:f32,z:f32}>>) -> (!fir.ref<!fir.type<_QMm_firstprivate_derived_user_defTpoint{x:f32,y:f32,z:f32}>>, !fir.ref<!fir.type<_QMm_firstprivate_derived_user_defTpoint{x:f32,y:f32,z:f32}>>)
+! CHECK:             %[[VAL_9:.*]] = arith.constant 1 : i32
+! CHECK:             %[[VAL_10:.*]] = fir.load %[[VAL_6]]#0 : !fir.ref<i32>
+! CHECK:             %[[VAL_11:.*]] = arith.constant 1 : i32
+! CHECK:             %[[VAL_12:.*]] = acc.private varPtr(%[[VAL_4]]#0 : !fir.ref<i32>) -> !fir.ref<i32> {implicit = true, name = "i"}
+! CHECK:             acc.loop combined(parallel) private(@privatization_ref_i32 -> %[[VAL_12]] : !fir.ref<i32>) control(%[[VAL_14:.*]] : i32) = (%[[VAL_9]] : i32) to (%[[VAL_10]] : i32)  step (%[[VAL_11]] : i32) {
+  ! CHECK:             %[[VAL_13:.*]]:2 = hlfir.declare %[[VAL_12]] {uniq_name = "_QMm_firstprivate_derived_user_defFtestEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK:               fir.store %[[VAL_14]] to %[[VAL_13]]#0 : !fir.ref<i32>
+! CHECK:               %[[VAL_15:.*]] = arith.constant 1.000000e+00 : f32
+! CHECK:               %[[VAL_16:.*]] = hlfir.designate %[[VAL_8]]#0{"x"}   : (!fir.ref<!fir.type<_QMm_firstprivate_derived_user_defTpoint{x:f32,y:f32,z:f32}>>) -> !fir.ref<f32>
+! CHECK:               hlfir.assign %[[VAL_15]] to %[[VAL_16]] : f32, !fir.ref<f32>
 ! CHECK:               acc.yield
 ! CHECK:             } attributes {inclusiveUpperbound = array<i1: true>, independent = [#acc.device_type<none>]}
 ! CHECK:             acc.yield

flang/test/Lower/OpenACC/acc-firstprivate-derived-pointer-component.f90

@@ -41,16 +41,17 @@ subroutine test()
 ! CHECK:           %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_5]] {uniq_name = "_QMm_firstprivate_derivedFtestEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK:           %[[VAL_7:.*]] = acc.firstprivate varPtr(%[[VAL_2]]#0 : !fir.ref<!fir.type<_QMm_firstprivate_derivedTpoint{x:f32,y:f32,z:f32}>>) -> !fir.ref<!fir.type<_QMm_firstprivate_derivedTpoint{x:f32,y:f32,z:f32}>> {name = "a"}
 ! CHECK:           acc.parallel combined(loop) firstprivate(@firstprivatization_ref_rec__QMm_firstprivate_derivedTpoint -> %[[VAL_7]] : !fir.ref<!fir.type<_QMm_firstprivate_derivedTpoint{x:f32,y:f32,z:f32}>>) {
-! CHECK:             %[[VAL_8:.*]] = arith.constant 1 : i32
-! CHECK:             %[[VAL_9:.*]] = fir.load %[[VAL_6]]#0 : !fir.ref<i32>
-! CHECK:             %[[VAL_10:.*]] = arith.constant 1 : i32
-! CHECK:             %[[VAL_11:.*]] = acc.private varPtr(%[[VAL_4]]#0 : !fir.ref<i32>) -> !fir.ref<i32> {implicit = true, name = "i"}
-! CHECK:             %[[VAL_12:.*]]:2 = hlfir.declare %[[VAL_11]] {uniq_name = "_QMm_firstprivate_derivedFtestEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-! CHECK:             acc.loop combined(parallel) private(@privatization_ref_i32 -> %[[VAL_11]] : !fir.ref<i32>) control(%[[VAL_13:.*]] : i32) = (%[[VAL_8]] : i32) to (%[[VAL_9]] : i32)  step (%[[VAL_10]] : i32) {
-! CHECK:               fir.store %[[VAL_13]] to %[[VAL_12]]#0 : !fir.ref<i32>
-! CHECK:               %[[VAL_14:.*]] = arith.constant 1.000000e+00 : f32
-! CHECK:               %[[VAL_15:.*]] = hlfir.designate %[[VAL_2]]#0{"x"}   : (!fir.ref<!fir.type<_QMm_firstprivate_derivedTpoint{x:f32,y:f32,z:f32}>>) -> !fir.ref<f32>
-! CHECK:               hlfir.assign %[[VAL_14]] to %[[VAL_15]] : f32, !fir.ref<f32>
+! CHECK:             %[[VAL_8:.*]]:2 = hlfir.declare %[[VAL_7]] {uniq_name = "_QMm_firstprivate_derivedFtestEa"} : (!fir.ref<!fir.type<_QMm_firstprivate_derivedTpoint{x:f32,y:f32,z:f32}>>) -> (!fir.ref<!fir.type<_QMm_firstprivate_derivedTpoint{x:f32,y:f32,z:f32}>>, !fir.ref<!fir.type<_QMm_firstprivate_derivedTpoint{x:f32,y:f32,z:f32}>>)
+! CHECK:             %[[VAL_9:.*]] = arith.constant 1 : i32
+! CHECK:             %[[VAL_10:.*]] = fir.load %[[VAL_6]]#0 : !fir.ref<i32>
+! CHECK:             %[[VAL_11:.*]] = arith.constant 1 : i32
+! CHECK:             %[[VAL_12:.*]] = acc.private varPtr(%[[VAL_4]]#0 : !fir.ref<i32>) -> !fir.ref<i32> {implicit = true, name = "i"}
+! CHECK:             acc.loop combined(parallel) private(@privatization_ref_i32 -> %[[VAL_12]] : !fir.ref<i32>) control(%[[VAL_14:.*]] : i32) = (%[[VAL_9]] : i32) to (%[[VAL_10]] : i32)  step (%[[VAL_11]] : i32) {
+  ! CHECK:             %[[VAL_13:.*]]:2 = hlfir.declare %[[VAL_12]] {uniq_name = "_QMm_firstprivate_derivedFtestEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK:               fir.store %[[VAL_14]] to %[[VAL_13]]#0 : !fir.ref<i32>
+! CHECK:               %[[VAL_15:.*]] = arith.constant 1.000000e+00 : f32
+! CHECK:               %[[VAL_16:.*]] = hlfir.designate %[[VAL_8]]#0{"x"}   : (!fir.ref<!fir.type<_QMm_firstprivate_derivedTpoint{x:f32,y:f32,z:f32}>>) -> !fir.ref<f32>
+! CHECK:               hlfir.assign %[[VAL_15]] to %[[VAL_16]] : f32, !fir.ref<f32>
 ! CHECK:               acc.yield
 ! CHECK:             } attributes {inclusiveUpperbound = array<i1: true>, independent = [#acc.device_type<none>]}
 ! CHECK:             acc.yield

flang/test/Lower/OpenACC/acc-firstprivate-derived-user-assign.f90

@@ -16,8 +16,8 @@ subroutine sub1(x, a)
 ! CHECK-LABEL: func.func @_QPsub1
 ! CHECK: %[[A:.*]]:2 = hlfir.declare %arg1 dummy_scope %{{[0-9]+}} {uniq_name = "_QFsub1Ea"} : (!fir.ref<i32>, !fir.dscope) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK: %[[I:.*]]:2 = hlfir.declare %{{[0-9]+}} {uniq_name = "_QFsub1Ei"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-! CHECK: %[[I:.*]]:2 = hlfir.declare %{{[0-9]+}} {uniq_name = "_QFsub1Ei"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK: %[[EXIT_COND:.*]] = acc.loop
+! CHECK:   %[[I:.*]]:2 = hlfir.declare %{{[0-9]+}} {uniq_name = "_QFsub1Ei"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK: ^bb{{.*}}:
 ! CHECK: ^bb{{.*}}:
 ! CHECK:   %[[LOAD_I:.*]] = fir.load %[[I]]#0 : !fir.ref<i32> 

flang/test/Lower/OpenACC/acc-firstprivate-derived.f90

@@ -426,7 +426,7 @@ subroutine acc_private_use()
 ! CHECK: %[[DECL_I:.*]]:2 = hlfir.declare %[[I]] {uniq_name = "_QFacc_private_useEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK: acc.parallel
 ! CHECK: %[[PRIV_I:.*]] = acc.private varPtr(%[[DECL_I]]#0 : !fir.ref<i32>) -> !fir.ref<i32> {implicit = true, name = "i"}
-! CHECK: %[[DECL_PRIV_I:.*]]:2 = hlfir.declare %[[PRIV_I]] {uniq_name = "_QFacc_private_useEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK: acc.loop {{.*}} private(@privatization_ref_i32 -> %[[PRIV_I]] : !fir.ref<i32>) control(%[[IV0:.*]] : i32) = (%c1{{.*}} : i32) to (%c10{{.*}} : i32) step (%c1{{.*}} : i32)
+! CHECK:   %[[DECL_PRIV_I:.*]]:2 = hlfir.declare %[[PRIV_I]] {uniq_name = "_QFacc_private_useEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK:   fir.store %[[IV0]] to %[[DECL_PRIV_I]]#0 : !fir.ref<i32>
 ! CHECK:   %{{.*}} = fir.load %[[DECL_PRIV_I]]#0 : !fir.ref<i32>