Merge branch 'main' into main

Author: YexuanXiao

flang/lib/Optimizer/HLFIR/Transforms/OptimizedBufferization.cpp

@@ -21,6 +21,7 @@
 #include "flang/Optimizer/HLFIR/HLFIROps.h"
 #include "flang/Optimizer/HLFIR/Passes.h"
 #include "flang/Optimizer/OpenMP/Passes.h"
+#include "flang/Optimizer/Support/Utils.h"
 #include "flang/Optimizer/Transforms/Utils.h"
 #include "mlir/Dialect/Func/IR/FuncOps.h"
 #include "mlir/IR/Dominance.h"
@@ -786,13 +787,55 @@ llvm::LogicalResult BroadcastAssignBufferization::matchAndRewrite(
   mlir::Value shape = hlfir::genShape(loc, builder, lhs);
   llvm::SmallVector<mlir::Value> extents =
       hlfir::getIndexExtents(loc, builder, shape);
-  hlfir::LoopNest loopNest =
-      hlfir::genLoopNest(loc, builder, extents, /*isUnordered=*/true,
-                         flangomp::shouldUseWorkshareLowering(assign));
-  builder.setInsertionPointToStart(loopNest.body);
-  auto arrayElement =
-      hlfir::getElementAt(loc, builder, lhs, loopNest.oneBasedIndices);
-  builder.create<hlfir::AssignOp>(loc, rhs, arrayElement);
+
+  if (lhs.isSimplyContiguous() && extents.size() > 1) {
+    // Flatten the array to use a single assign loop, that can be better
+    // optimized.
+    mlir::Value n = extents[0];
+    for (size_t i = 1; i < extents.size(); ++i)
+      n = builder.create<mlir::arith::MulIOp>(loc, n, extents[i]);
+    llvm::SmallVector<mlir::Value> flatExtents = {n};
+
+    mlir::Type flatArrayType;
+    mlir::Value flatArray = lhs.getBase();
+    if (mlir::isa<fir::BoxType>(lhs.getType())) {
+      shape = builder.genShape(loc, flatExtents);
+      flatArrayType = fir::BoxType::get(fir::SequenceType::get(eleTy, 1));
+      flatArray = builder.create<fir::ReboxOp>(loc, flatArrayType, flatArray,
+                                               shape, /*slice=*/mlir::Value{});
+    } else {
+      // Array references must have fixed shape, when used in assignments.
+      int64_t flatExtent = 1;
+      for (const mlir::Value &extent : extents) {
+        mlir::Operation *op = extent.getDefiningOp();
+        assert(op && "no defining operation for constant array extent");
+        flatExtent *= fir::toInt(mlir::cast<mlir::arith::ConstantOp>(*op));
+      }
+
+      flatArrayType =
+          fir::ReferenceType::get(fir::SequenceType::get({flatExtent}, eleTy));
+      flatArray = builder.createConvert(loc, flatArrayType, flatArray);
+    }
+
+    hlfir::LoopNest loopNest =
+        hlfir::genLoopNest(loc, builder, flatExtents, /*isUnordered=*/true,
+                           flangomp::shouldUseWorkshareLowering(assign));
+    builder.setInsertionPointToStart(loopNest.body);
+
+    mlir::Value arrayElement =
+        builder.create<hlfir::DesignateOp>(loc, fir::ReferenceType::get(eleTy),
+                                           flatArray, loopNest.oneBasedIndices);
+    builder.create<hlfir::AssignOp>(loc, rhs, arrayElement);
+  } else {
+    hlfir::LoopNest loopNest =
+        hlfir::genLoopNest(loc, builder, extents, /*isUnordered=*/true,
+                           flangomp::shouldUseWorkshareLowering(assign));
+    builder.setInsertionPointToStart(loopNest.body);
+    auto arrayElement =
+        hlfir::getElementAt(loc, builder, lhs, loopNest.oneBasedIndices);
+    builder.create<hlfir::AssignOp>(loc, rhs, arrayElement);
+  }
+
   rewriter.eraseOp(assign);
   return mlir::success();
 }

flang/lib/Optimizer/Transforms/DebugTypeGenerator.cpp

@@ -303,9 +303,9 @@ void DerivedTypeCache::postComponentVisitUpdate(
     return;
   ActiveLevels oldLevels;
   oldLevels.swap(activeRecursionLevels);
-  std::merge(componentActiveRecursionLevels.begin(),
-             componentActiveRecursionLevels.end(), oldLevels.begin(),
-             oldLevels.end(), std::back_inserter(activeRecursionLevels));
+  std::set_union(componentActiveRecursionLevels.begin(),
+                 componentActiveRecursionLevels.end(), oldLevels.begin(),
+                 oldLevels.end(), std::back_inserter(activeRecursionLevels));
 }
 
 void DerivedTypeCache::finalize(mlir::Type ty, mlir::LLVM::DITypeAttr attr,

flang/test/HLFIR/opt-scalar-assign.fir

@@ -12,18 +12,18 @@ func.func @_QPtest1() {
   return
 }
 // CHECK-LABEL:   func.func @_QPtest1() {
-// CHECK:           %[[VAL_0:.*]] = arith.constant 1 : index
-// CHECK:           %[[VAL_1:.*]] = arith.constant 0.000000e+00 : f32
-// CHECK:           %[[VAL_2:.*]] = arith.constant 11 : index
-// CHECK:           %[[VAL_3:.*]] = arith.constant 13 : index
-// CHECK:           %[[VAL_4:.*]] = fir.alloca !fir.array<11x13xf32> {bindc_name = "x", uniq_name = "_QFtest1Ex"}
-// CHECK:           %[[VAL_5:.*]] = fir.shape %[[VAL_2]], %[[VAL_3]] : (index, index) -> !fir.shape<2>
-// CHECK:           %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_4]](%[[VAL_5]]) {uniq_name = "_QFtest1Ex"} : (!fir.ref<!fir.array<11x13xf32>>, !fir.shape<2>) -> (!fir.ref<!fir.array<11x13xf32>>, !fir.ref<!fir.array<11x13xf32>>)
-// CHECK:           fir.do_loop %[[VAL_7:.*]] = %[[VAL_0]] to %[[VAL_3]] step %[[VAL_0]] unordered {
-// CHECK:             fir.do_loop %[[VAL_8:.*]] = %[[VAL_0]] to %[[VAL_2]] step %[[VAL_0]] unordered {
-// CHECK:               %[[VAL_9:.*]] = hlfir.designate %[[VAL_6]]#0 (%[[VAL_8]], %[[VAL_7]])  : (!fir.ref<!fir.array<11x13xf32>>, index, index) -> !fir.ref<f32>
-// CHECK:               hlfir.assign %[[VAL_1]] to %[[VAL_9]] : f32, !fir.ref<f32>
-// CHECK:             }
+// CHECK:           %[[VAL_0:.*]] = arith.constant 143 : index
+// CHECK:           %[[VAL_1:.*]] = arith.constant 1 : index
+// CHECK:           %[[VAL_2:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK:           %[[VAL_3:.*]] = arith.constant 11 : index
+// CHECK:           %[[VAL_4:.*]] = arith.constant 13 : index
+// CHECK:           %[[VAL_5:.*]] = fir.alloca !fir.array<11x13xf32> {bindc_name = "x", uniq_name = "_QFtest1Ex"}
+// CHECK:           %[[VAL_6:.*]] = fir.shape %[[VAL_3]], %[[VAL_4]] : (index, index) -> !fir.shape<2>
+// CHECK:           %[[VAL_7:.*]]:2 = hlfir.declare %[[VAL_5]](%[[VAL_6]]) {uniq_name = "_QFtest1Ex"} : (!fir.ref<!fir.array<11x13xf32>>, !fir.shape<2>) -> (!fir.ref<!fir.array<11x13xf32>>, !fir.ref<!fir.array<11x13xf32>>)
+// CHECK:           %[[VAL_8:.*]] = fir.convert %[[VAL_7]]#0 : (!fir.ref<!fir.array<11x13xf32>>) -> !fir.ref<!fir.array<143xf32>>
+// CHECK:           fir.do_loop %[[VAL_9:.*]] = %[[VAL_1]] to %[[VAL_0]] step %[[VAL_1]] unordered {
+// CHECK:             %[[VAL_10:.*]] = hlfir.designate %[[VAL_8]] (%[[VAL_9]]) : (!fir.ref<!fir.array<143xf32>>, index) -> !fir.ref<f32>
+// CHECK:             hlfir.assign %[[VAL_2]] to %[[VAL_10]] : f32, !fir.ref<f32>
 // CHECK:           }
 // CHECK:           return
 // CHECK:         }
@@ -129,3 +129,29 @@ func.func @_QPtest5(%arg0: !fir.ref<!fir.array<77xcomplex<f32>>> {fir.bindc_name
 // CHECK:           }
 // CHECK:           return
 // CHECK:         }
+
+func.func @_QPtest6(%arg0: !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>> {fir.bindc_name = "x"}) {
+  %c0_i32 = arith.constant 0 : i32
+  %0:2 = hlfir.declare %arg0 {fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFtest6Ex"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>)
+  hlfir.assign %c0_i32 to %0#0 realloc : i32, !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>
+  return
+}
+
+// CHECK-LABEL:   func.func @_QPtest6(
+// CHECK-SAME:                        %[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>> {fir.bindc_name = "x"}) {
+// CHECK:           %[[VAL_1:.*]] = arith.constant 1 : index
+// CHECK:           %[[VAL_2:.*]] = arith.constant 0 : index
+// CHECK:           %[[VAL_3:.*]] = arith.constant 0 : i32
+// CHECK:           %[[VAL_4:.*]]:2 = hlfir.declare %[[VAL_0]] {fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFtest6Ex"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>)
+// CHECK:           %[[VAL_5:.*]] = fir.load %[[VAL_4]]#0 : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>
+// CHECK:           %[[VAL_6:.*]]:3 = fir.box_dims %[[VAL_5]], %[[VAL_2]] : (!fir.box<!fir.heap<!fir.array<?x?xi32>>>, index) -> (index, index, index)
+// CHECK:           %[[VAL_7:.*]]:3 = fir.box_dims %[[VAL_5]], %[[VAL_1]] : (!fir.box<!fir.heap<!fir.array<?x?xi32>>>, index) -> (index, index, index)
+// CHECK:           %[[VAL_8:.*]] = arith.muli %[[VAL_6]]#1, %[[VAL_7]]#1 : index
+// CHECK:           %[[VAL_9:.*]] = fir.shape %[[VAL_8]] : (index) -> !fir.shape<1>
+// CHECK:           %[[VAL_10:.*]] = fir.rebox %[[VAL_5]](%[[VAL_9]]) : (!fir.box<!fir.heap<!fir.array<?x?xi32>>>, !fir.shape<1>) -> !fir.box<!fir.array<?xi32>>
+// CHECK:           fir.do_loop %[[VAL_11:.*]] = %[[VAL_1]] to %[[VAL_8]] step %[[VAL_1]] unordered {
+// CHECK:             %[[VAL_12:.*]] = hlfir.designate %[[VAL_10]] (%[[VAL_11]]) : (!fir.box<!fir.array<?xi32>>, index) -> !fir.ref<i32>
+// CHECK:             hlfir.assign %[[VAL_3]] to %[[VAL_12]] : i32, !fir.ref<i32>
+// CHECK:           }
+// CHECK:           return
+// CHECK:         }

libcxx/test/libcxx/input.output/iostreams.base/ios.base/ios.base.cons/dtor.uninitialized.pass.cpp

@@ -13,7 +13,7 @@
 
 // The fix for issue 57964 requires an updated dylib due to explicit
 // instantiations. That means Apple backdeployment targets remain broken.
-// XFAIL using-built-library-before-llvm-19
+// XFAIL: using-built-library-before-llvm-19
 
 // <ios>
 

libcxx/test/std/input.output/file.streams/fstreams/filebuf.virtuals/setbuf.pass.cpp

@@ -15,7 +15,7 @@
 
 // This test requires the fix to https://github.com/llvm/llvm-project/issues/60509 in the dylib,
 // which landed in 5afb937d8a30445642ccaf33866ee4cdd0713222.
-// XFAIL using-built-library-before-llvm-19
+// XFAIL: using-built-library-before-llvm-19
 
 #include <fstream>
 #include <cstddef>

libcxx/test/std/input.output/iostream.format/input.streams/istream.unformatted/sync.pass.cpp

@@ -16,7 +16,7 @@
 // The fix for bug 51497 and bug 51499 require and updated dylib due to
 // explicit instantiations. That means Apple backdeployment targets remain
 // broken.
-// XFAIL using-built-library-before-llvm-19
+// XFAIL: using-built-library-before-llvm-19
 
 #include <istream>
 #include <cassert>

libcxx/test/std/localization/locale.categories/category.monetary/locale.money.get/locale.money.get.members/get_long_double_fr_FR.pass.cpp

@@ -9,7 +9,7 @@
 // TODO(mordante) Investigate
 // UNSUPPORTED: apple-clang
 
-// XFAIL darwin
+// XFAIL: darwin
 
 // NetBSD does not support LC_MONETARY at the moment
 // XFAIL: netbsd

libcxx/test/std/localization/locale.categories/category.monetary/locale.money.put/locale.money.put.members/put_long_double_fr_FR.pass.cpp

@@ -9,7 +9,7 @@
 // TODO(mordante) Investigate
 // UNSUPPORTED: apple-clang
 
-// XFAIL darwin
+// XFAIL: darwin
 
 // NetBSD does not support LC_MONETARY at the moment
 // XFAIL: netbsd

libcxx/test/std/localization/locale.categories/category.monetary/locale.moneypunct.byname/grouping.pass.cpp

@@ -9,7 +9,7 @@
 // TODO(mordante) Investigate
 // UNSUPPORTED: apple-clang
 
-// XFAIL darwin
+// XFAIL: darwin
 //
 // NetBSD does not support LC_MONETARY at the moment
 // XFAIL: netbsd

libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_double.pass.cpp

@@ -11,7 +11,7 @@
 
 // The fix for LWG2381 (https://github.com/llvm/llvm-project/pull/77948) changed behavior of
 // FP parsing. This requires 3e15c97fa3812993bdc319827a5c6d867b765ae8 in the dylib.
-// XFAIL using-built-library-before-llvm-19
+// XFAIL: using-built-library-before-llvm-19
 
 // <locale>