Automerge: [Flang][OpenMP] Fix crash when loop index var is pointer or allocatable (#129717)

Use hlfir dereferencing for pointers and allocatables and use hlfir
assign. Also, change the code updating IV in lastprivate.

Note: This is a small change. Modifications in existing tests are
changes from fir.store to hlfir.assign.

Fixes #121290

Author: kiranchandramohan

flang/lib/Lower/OpenMP/DataSharingProcessor.cpp

@@ -307,8 +307,10 @@ void DataSharingProcessor::insertLastPrivateCompare(mlir::Operation *op) {
     auto ifOp = firOpBuilder.create<fir::IfOp>(loc, cmpOp, /*else*/ false);
     firOpBuilder.setInsertionPointToStart(&ifOp.getThenRegion().front());
     for (auto [v, loopIV] : llvm::zip_equal(vs, loopIVs)) {
-      assert(loopIV && "loopIV was not set");
-      firOpBuilder.createStoreWithConvert(loc, v, loopIV);
+      hlfir::Entity loopIVEntity{loopIV};
+      loopIVEntity =
+          hlfir::derefPointersAndAllocatables(loc, firOpBuilder, loopIVEntity);
+      firOpBuilder.create<hlfir::AssignOp>(loc, v, loopIVEntity);
     }
     lastPrivIP = firOpBuilder.saveInsertionPoint();
   } else if (mlir::isa<mlir::omp::SectionsOp>(op)) {

flang/lib/Lower/OpenMP/OpenMP.cpp

@@ -789,8 +789,12 @@ createAndSetPrivatizedLoopVar(lower::AbstractConverter &converter,
 
   firOpBuilder.restoreInsertionPoint(insPt);
   mlir::Value cvtVal = firOpBuilder.createConvert(loc, tempTy, indexVal);
-  mlir::Operation *storeOp = firOpBuilder.create<fir::StoreOp>(
-      loc, cvtVal, converter.getSymbolAddress(*sym));
+  hlfir::Entity lhs{converter.getSymbolAddress(*sym)};
+
+  lhs = hlfir::derefPointersAndAllocatables(loc, firOpBuilder, lhs);
+
+  mlir::Operation *storeOp =
+      firOpBuilder.create<hlfir::AssignOp>(loc, cvtVal, lhs);
   return storeOp;
 }
 

flang/test/Lower/OpenMP/DelayedPrivatization/distribute-standalone-private.f90

@@ -31,7 +31,7 @@ end subroutine standalone_distribute
 ! CHECK:               %[[VAR_PRIV_DECL:.*]]:2 = hlfir.declare %[[VAR_ARG]]
 ! CHECK:               %[[I_PRIV_DECL:.*]]:2 = hlfir.declare %[[I_ARG]]
 
-! CHECK:               fir.store %{{.*}} to %[[I_PRIV_DECL]]#1 : !fir.ref<i32>
+! CHECK:               hlfir.assign %{{.*}} to %[[I_PRIV_DECL]]#1 : i32, !fir.ref<i32>
 ! CHECK:               %{{.*}} = fir.load %[[VAR_PRIV_DECL]]#0 : !fir.ref<i32>
 ! CHECK:               %{{.*}} = fir.load %[[I_PRIV_DECL]]#0 : !fir.ref<i32>
 ! CHECK:               arith.addi %{{.*}}, %{{.*}} : i32

flang/test/Lower/OpenMP/DelayedPrivatization/wsloop.f90

@@ -28,7 +28,7 @@ end subroutine wsloop_private
 ! CHECK:       omp.loop_nest (%[[IV:.*]]) : i32 = {{.*}} {
 ! CHECK:         %[[X_PRIV_DECL:.*]]:2 = hlfir.declare %[[X_ARG]] {uniq_name = "{{.*}}x"}
 ! CHECK:         %[[I_PRIV_DECL:.*]]:2 = hlfir.declare %[[I_ARG]] {uniq_name = "{{.*}}i"}
-! CHECK:         fir.store %[[IV]] to %[[I_PRIV_DECL]]#1
+! CHECK:         hlfir.assign %[[IV]] to %[[I_PRIV_DECL]]#1
 ! CHECK:         %[[X_VAL:.*]] = fir.load %[[X_PRIV_DECL]]#0
 ! CHECK:         %[[I_VAL:.*]] = fir.load %[[I_PRIV_DECL]]#0
 ! CHECK:         %[[ADD_VAL:.*]] = arith.addi %[[X_VAL]], %[[I_VAL]]

flang/test/Lower/OpenMP/Todo/omp-default-clause-inner-loop.f90

@@ -14,7 +14,7 @@
 ! CHECK: %[[const_3:.*]] = arith.constant 1 : i32
 ! CHECK: omp.wsloop private(@{{.*}} %{{.*}} -> %[[TEMP:.*]] : !fir.ref<i32>) {
 ! CHECK-NEXT: omp.loop_nest (%[[ARG:.*]]) : i32 = (%[[const_1]]) to (%[[const_2]]) inclusive step (%[[const_3]]) {
-! CHECK: fir.store %[[ARG]] to %[[TEMP]] : !fir.ref<i32>
+! CHECK: hlfir.assign %[[ARG]] to %[[TEMP]] : i32, !fir.ref<i32>
 ! EXPECTED: %[[temp_1:.*]] = fir.load %[[PRIVATE_Z]] : !fir.ref<i32>
 ! CHECK: %[[temp_1:.*]] = fir.load %{{.*}} : !fir.ref<i32>
 ! CHECK: %[[temp_2:.*]] = fir.load %[[TEMP]] : !fir.ref<i32>

flang/test/Lower/OpenMP/copyin.f90

@@ -161,7 +161,7 @@ subroutine copyin_derived_type()
 ! CHECK:             omp.wsloop private(@{{.*}} %{{.*}} -> %[[VAL_6:.*]] : !fir.ref<i32>) {
 ! CHECK-NEXT:          omp.loop_nest (%[[VAL_14:.*]]) : i32 = (%[[VAL_11]]) to (%[[VAL_12]]) inclusive step (%[[VAL_13]]) {
 ! CHECK:                 %[[VAL_7:.*]]:2 = hlfir.declare %[[VAL_6]] {uniq_name = "_QFcombined_parallel_worksharing_loopEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-! CHECK:                 fir.store %[[VAL_14]] to %[[VAL_7]]#1 : !fir.ref<i32>
+! CHECK:                 hlfir.assign %[[VAL_14]] to %[[VAL_7]]#1 : i32, !fir.ref<i32>
 ! CHECK:                 fir.call @_QPsub4(%[[VAL_9]]#1) fastmath<contract> : (!fir.ref<i32>) -> ()
 ! CHECK:                 omp.yield
 ! CHECK:               }
@@ -326,7 +326,7 @@ subroutine common_1()
 ! CHECK:             omp.wsloop private(@{{.*}} %{{.*}} -> %[[VAL_19:.*]] : !fir.ref<i32>) {
 ! CHECK-NEXT:          omp.loop_nest (%[[VAL_37:.*]]) : i32 = (%[[VAL_34]]) to (%[[VAL_35]]) inclusive step (%[[VAL_36]]) {
 ! CHECK:             %[[VAL_20:.*]]:2 = hlfir.declare %[[VAL_19]] {uniq_name = "_QFcommon_2Ei"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-! CHECK:                 fir.store %[[VAL_37]] to %[[VAL_20]]#1 : !fir.ref<i32>
+! CHECK:                 hlfir.assign %[[VAL_37]] to %[[VAL_20]]#1 : i32, !fir.ref<i32>
 ! CHECK:                 %[[VAL_38:.*]] = fir.load %[[VAL_31]]#0 : !fir.ref<i32>
 ! CHECK:                 %[[VAL_39:.*]] = fir.load %[[VAL_20]]#0 : !fir.ref<i32>
 ! CHECK:                 %[[VAL_40:.*]] = arith.addi %[[VAL_38]], %[[VAL_39]] : i32

flang/test/Lower/OpenMP/critical.f90

@@ -42,7 +42,7 @@ subroutine predetermined_privatization()
     !CHECK: omp.wsloop private(@{{.*}} %{{.*}} -> %[[PRIV_I_ALLOC:.*]] : !fir.ref<i32>)
     !CHECK: omp.loop_nest (%[[IV:[^[:space:]]+]])
     !CHECK: %[[PRIV_I_DECL:.*]]:2 = hlfir.declare %[[PRIV_I_ALLOC]]
-    !CHECK: fir.store %[[IV]] to %[[PRIV_I_DECL]]#1
+    !CHECK: hlfir.assign %[[IV]] to %[[PRIV_I_DECL]]#1
     !CHECK: omp.critical
     !$omp critical
     a(i) = a(i-1) + 1

flang/test/Lower/OpenMP/default-clause.f90

@@ -291,7 +291,7 @@ subroutine nested_default_clause_test4
 !CHECK: omp.loop_nest (%[[ARG:.*]]) : i32 = (%[[CONST_LB]]) to (%[[CONST_UB]]) inclusive step (%[[CONST_STEP]]) {
 !CHECK: %[[X_DECLARE:.*]]:2 = hlfir.declare %[[X_ALLOCA]] {{.*}}
 !CHECK: %[[LOOP_VAR_DECLARE:.*]]:2 = hlfir.declare %[[LOOP_VAR_ALLOCA]] {{.*}}
-!CHECK: fir.store %[[ARG]] to %[[LOOP_VAR_DECLARE]]#1 : !fir.ref<i32>
+!CHECK: hlfir.assign %[[ARG]] to %[[LOOP_VAR_DECLARE]]#1 : i32, !fir.ref<i32>
 !CHECK: %[[LOADED_X:.*]] = fir.load %[[X_DECLARE]]#0 : !fir.ref<i32>
 !CHECK: %[[CONST:.*]] = arith.constant 1 : i32
 !CHECK: %[[RESULT:.*]] = arith.addi %[[LOADED_X]], %[[CONST]] : i32
@@ -324,7 +324,7 @@ subroutine nested_default_clause_test5
 ! CHECK: omp.wsloop private(@{{.*}} %{{.*}} -> %[[LOOP_VAR:.*]] : !fir.ref<i32>) {
 !CHECK: omp.loop_nest (%[[ARG:.*]]) : i32 = (%[[CONST_LB]]) to (%[[CONST_UB]]) inclusive step (%[[CONST_STEP]]) {
 !CHECK: %[[LOOP_VAR_DECLARE:.*]]:2 = hlfir.declare %[[LOOP_VAR]] {{.*}}
-!CHECK: fir.store %[[ARG]] to %[[LOOP_VAR_DECLARE]]#1 : !fir.ref<i32>
+!CHECK: hlfir.assign %[[ARG]] to %[[LOOP_VAR_DECLARE]]#1 : i32, !fir.ref<i32>
 !CHECK: %[[LOADED_X:.*]] = fir.load %[[X_VAR_DECLARE]]#0 : !fir.ref<i32>
 !CHECK: %[[CONST:.*]] = arith.constant 1 : i32
 !CHECK: %[[ADD:.*]] = arith.addi %[[LOADED_X]], %[[CONST]] : i32

flang/test/Lower/OpenMP/generic-loop-rewriting.f90

@@ -50,7 +50,7 @@ end subroutine target_teams_loop
 !CHECK:                   omp.loop_nest (%{{.*}}) : i32 = 
 !CHECK-SAME:                (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
 !CHECK:                     %[[I_PRIV_DECL:.*]]:2 = hlfir.declare %[[I_PRIV_ARG]]
-!CHECK:                     fir.store %{{.*}} to %[[I_PRIV_DECL]]#1 : !fir.ref<i32>
+!CHECK:                     hlfir.assign %{{.*}} to %[[I_PRIV_DECL]]#1 : i32, !fir.ref<i32>
 !CHECK:                   }
 !CHECK:                 }
 !CHECK:               }

flang/test/Lower/OpenMP/lastprivate-allocatable.f90

@@ -17,7 +17,7 @@
 !                        if this is the last iteration
 ! CHECK:                 fir.if %{{.*}} {
 !                          store loop IV
-! CHECK:                   fir.store %{{.*}} to %[[VAL_18]]#1 : !fir.ref<i32>
+! CHECK:                   hlfir.assign %{{.*}} to %[[VAL_18]]#1 : i32, !fir.ref<i32>
 !                          assign private variable to original copy: realloc
 ! CHECK:                   %[[VAL_23:.*]] = fir.load %[[VAL_16]]#0 : !fir.ref<!fir.box<!fir.heap<i32>>>
 ! CHECK:                   %[[VAL_24:.*]] = fir.box_addr %[[VAL_23]] : (!fir.box<!fir.heap<i32>>) -> !fir.heap<i32>