[flang] Fixed designator codegen for contiguous boxes. #139003
+135
−9
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Contiguous variables represented with a box do not have explicit shape, but it looks like the base/shape computation was assuming that. This caused generation of raw address fir.array_coor without the shape. This patch is needed to fix failures hapenning with llvm#138797.
|
@llvm/pr-subscribers-flang-fir-hlfir Author: Slava Zakharin (vzakhari) ChangesContiguous variables represented with a box do not have Full diff: https://github.com/llvm/llvm-project/pull/139003.diff 4 Files Affected:
diff --git a/flang/include/flang/Optimizer/Builder/HLFIRTools.h b/flang/include/flang/Optimizer/Builder/HLFIRTools.h
index ac80873dc374f..bcba38ed8bd5d 100644
--- a/flang/include/flang/Optimizer/Builder/HLFIRTools.h
+++ b/flang/include/flang/Optimizer/Builder/HLFIRTools.h
@@ -533,6 +533,12 @@ Entity gen1DSection(mlir::Location loc, fir::FirOpBuilder &builder,
mlir::ArrayRef<mlir::Value> extents,
mlir::ValueRange oneBasedIndices,
mlir::ArrayRef<mlir::Value> typeParams);
+
+/// Return explicit lower bounds from a fir.shape result.
+/// Only fir.shape, fir.shift and fir.shape_shift are currently
+/// supported as \p shape.
+llvm::SmallVector<mlir::Value> getExplicitLboundsFromShape(mlir::Value shape);
+
} // namespace hlfir
#endif // FORTRAN_OPTIMIZER_BUILDER_HLFIRTOOLS_H
diff --git a/flang/lib/Optimizer/Builder/HLFIRTools.cpp b/flang/lib/Optimizer/Builder/HLFIRTools.cpp
index 51ea7305d3d26..752dc0cf86414 100644
--- a/flang/lib/Optimizer/Builder/HLFIRTools.cpp
+++ b/flang/lib/Optimizer/Builder/HLFIRTools.cpp
@@ -70,10 +70,8 @@ getExplicitExtents(fir::FortranVariableOpInterface var,
return {};
}
-// Return explicit lower bounds. For pointers and allocatables, this will not
-// read the lower bounds and instead return an empty vector.
-static llvm::SmallVector<mlir::Value>
-getExplicitLboundsFromShape(mlir::Value shape) {
+llvm::SmallVector<mlir::Value>
+hlfir::getExplicitLboundsFromShape(mlir::Value shape) {
llvm::SmallVector<mlir::Value> result;
auto *shapeOp = shape.getDefiningOp();
if (auto s = mlir::dyn_cast_or_null<fir::ShapeOp>(shapeOp)) {
@@ -89,10 +87,13 @@ getExplicitLboundsFromShape(mlir::Value shape) {
}
return result;
}
+
+// Return explicit lower bounds. For pointers and allocatables, this will not
+// read the lower bounds and instead return an empty vector.
static llvm::SmallVector<mlir::Value>
getExplicitLbounds(fir::FortranVariableOpInterface var) {
if (mlir::Value shape = var.getShape())
- return getExplicitLboundsFromShape(shape);
+ return hlfir::getExplicitLboundsFromShape(shape);
return {};
}
@@ -753,9 +754,24 @@ std::pair<mlir::Value, mlir::Value> hlfir::genVariableFirBaseShapeAndParams(
}
if (entity.isScalar())
return {fir::getBase(exv), mlir::Value{}};
+
+ // Contiguous variables that are represented with a box
+ // may require the shape to be extracted from the box (i.e. evx),
+ // because they itself may not have shape specified.
+ // This happens during late propagationg of contiguous
+ // attribute, e.g.:
+ // %9:2 = hlfir.declare %6
+ // {fortran_attrs = #fir.var_attrs<contiguous>} :
+ // (!fir.box<!fir.array<?x?x...>>) ->
+ // (!fir.box<!fir.array<?x?x...>>, !fir.box<!fir.array<?x?x...>>)
+ // The extended value is an ArrayBoxValue with base being
+ // the raw address of the array.
if (auto variableInterface = entity.getIfVariableInterface())
- return {fir::getBase(exv),
- asEmboxShape(loc, builder, exv, variableInterface.getShape())};
+ if (mlir::isa<fir::BaseBoxType>(fir::getBase(exv).getType()) ||
+ !mlir::isa<fir::BaseBoxType>(entity.getType()) ||
+ variableInterface.getShape())
+ return {fir::getBase(exv),
+ asEmboxShape(loc, builder, exv, variableInterface.getShape())};
return {fir::getBase(exv), builder.createShape(loc, exv)};
}
diff --git a/flang/lib/Optimizer/HLFIR/Transforms/ConvertToFIR.cpp b/flang/lib/Optimizer/HLFIR/Transforms/ConvertToFIR.cpp
index 8721a895b5e05..495f11a365185 100644
--- a/flang/lib/Optimizer/HLFIR/Transforms/ConvertToFIR.cpp
+++ b/flang/lib/Optimizer/HLFIR/Transforms/ConvertToFIR.cpp
@@ -412,12 +412,44 @@ class DesignateOpConversion
auto indices = designate.getIndices();
int i = 0;
auto attrs = designate.getIsTripletAttr();
+
+ // If the shape specifies a shift and the base is not a box,
+ // then we have to subtract the lower bounds, as long as
+ // fir.array_coor does not support non-default lower bounds
+ // for non-box accesses.
+ llvm::SmallVector<mlir::Value> lbounds;
+ if (shape && !mlir::isa<fir::BaseBoxType>(base.getType()))
+ lbounds = hlfir::getExplicitLboundsFromShape(shape);
+ std::size_t lboundIdx = 0;
for (auto isTriplet : attrs.asArrayRef()) {
// Coordinate of the first element are the index and triplets lower
- // bounds
- firstElementIndices.push_back(indices[i]);
+ // bounds.
+ mlir::Value index = indices[i];
+ if (!lbounds.empty()) {
+ assert(lboundIdx < lbounds.size() && "missing lbound");
+ mlir::Type indexType = builder.getIndexType();
+ mlir::Value one = builder.createIntegerConstant(loc, indexType, 1);
+ mlir::Value orig = builder.createConvert(loc, indexType, index);
+ mlir::Value lb =
+ builder.createConvert(loc, indexType, lbounds[lboundIdx]);
+ index = builder.create<mlir::arith::SubIOp>(loc, orig, lb);
+ index = builder.create<mlir::arith::AddIOp>(loc, index, one);
+ ++lboundIdx;
+ }
+ firstElementIndices.push_back(index);
i = i + (isTriplet ? 3 : 1);
}
+
+ // Remove the shift from the shape, if needed.
+ if (!lbounds.empty()) {
+ mlir::Operation *op = shape.getDefiningOp();
+ if (mlir::isa<fir::ShiftOp>(op))
+ shape = nullptr;
+ else if (auto shiftOp = mlir::dyn_cast<fir::ShapeShiftOp>(op))
+ shape = builder.create<fir::ShapeOp>(loc, shiftOp.getExtents());
+ else
+ TODO(loc, "read fir.shape to get lower bounds");
+ }
mlir::Type originalDesignateType = designate.getResult().getType();
const bool isVolatile = fir::isa_volatile_type(originalDesignateType);
mlir::Type arrayCoorType = fir::ReferenceType::get(baseEleTy, isVolatile);
diff --git a/flang/test/HLFIR/designate-codegen.fir b/flang/test/HLFIR/designate-codegen.fir
index da0a1f82b516e..d3e264941264f 100644
--- a/flang/test/HLFIR/designate-codegen.fir
+++ b/flang/test/HLFIR/designate-codegen.fir
@@ -213,3 +213,75 @@ func.func @test_polymorphic_array_elt(%arg0: !fir.class<!fir.array<?x!fir.type<_
// CHECK: %[[VAL_5:.*]] = fir.embox %[[VAL_4]] source_box %[[VAL_2]] : (!fir.ref<!fir.type<_QMtypesTt1>>, !fir.class<!fir.array<?x!fir.type<_QMtypesTt1>>>) -> !fir.class<!fir.type<_QMtypesTt1>>
// CHECK: return
// CHECK: }
+
+// Test proper generation of fir.array_coor for contiguous box with default lbounds.
+func.func @_QPtest_contiguous_derived_default(%arg0: !fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>> {fir.bindc_name = "d1", fir.contiguous, fir.optional}) {
+ %c1 = arith.constant 1 : index
+ %c16_i32 = arith.constant 16 : i32
+ %0 = fir.dummy_scope : !fir.dscope
+ %1:2 = hlfir.declare %arg0 dummy_scope %0 {fortran_attrs = #fir.var_attrs<contiguous, optional>, uniq_name = "_QFtest_contiguous_derived_defaultEd1"} : (!fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>, !fir.dscope) -> (!fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>, !fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>)
+ fir.select_type %1#1 : !fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>> [#fir.type_is<!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>, ^bb1, unit, ^bb2]
+^bb1: // pred: ^bb0
+ %2 = fir.convert %1#1 : (!fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>) -> !fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>
+ %3:2 = hlfir.declare %2 {fortran_attrs = #fir.var_attrs<contiguous>, uniq_name = "_QFtest_contiguous_derived_defaultEd1"} : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>) -> (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, !fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>)
+ %4 = hlfir.designate %3#0 (%c1, %c1) : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, index, index) -> !fir.ref<!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>
+ %5 = hlfir.designate %4{"i"} : (!fir.ref<!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>) -> !fir.ref<i32>
+ hlfir.assign %c16_i32 to %5 : i32, !fir.ref<i32>
+ cf.br ^bb3
+^bb2: // pred: ^bb0
+ %6:2 = hlfir.declare %1#1 {fortran_attrs = #fir.var_attrs<contiguous>, uniq_name = "_QFtest_contiguous_derived_defaultEd1"} : (!fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>) -> (!fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>, !fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>)
+ cf.br ^bb3
+^bb3: // 2 preds: ^bb1, ^bb2
+ return
+}
+// CHECK-LABEL: func.func @_QPtest_contiguous_derived_default(
+// CHECK: %[[VAL_0:.*]] = arith.constant 1 : index
+// CHECK: %[[VAL_9:.*]] = fir.declare %{{.*}} {fortran_attrs = #fir.var_attrs<contiguous>, uniq_name = "_QFtest_contiguous_derived_defaultEd1"} : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>) -> !fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>
+// CHECK: %[[VAL_10:.*]] = fir.rebox %[[VAL_9]] : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>) -> !fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>
+// CHECK: %[[VAL_11:.*]] = fir.box_addr %[[VAL_10]] : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>) -> !fir.ref<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>
+// CHECK: %[[VAL_12:.*]] = arith.constant 0 : index
+// CHECK: %[[VAL_13:.*]]:3 = fir.box_dims %[[VAL_10]], %[[VAL_12]] : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, index) -> (index, index, index)
+// CHECK: %[[VAL_14:.*]] = arith.constant 1 : index
+// CHECK: %[[VAL_15:.*]]:3 = fir.box_dims %[[VAL_10]], %[[VAL_14]] : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, index) -> (index, index, index)
+// CHECK: %[[VAL_16:.*]] = fir.shape %[[VAL_13]]#1, %[[VAL_15]]#1 : (index, index) -> !fir.shape<2>
+// CHECK: %[[VAL_17:.*]] = fir.array_coor %[[VAL_11]](%[[VAL_16]]) %[[VAL_0]], %[[VAL_0]] : (!fir.ref<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, !fir.shape<2>, index, index) -> !fir.ref<!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>
+
+// Test proper generation of fir.array_coor for contiguous box with non-default lbounds.
+func.func @_QPtest_contiguous_derived_lbounds(%arg0: !fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>> {fir.bindc_name = "d1", fir.contiguous}) {
+ %c3 = arith.constant 3 : index
+ %c1 = arith.constant 1 : index
+ %c16_i32 = arith.constant 16 : i32
+ %0 = fir.dummy_scope : !fir.dscope
+ %1 = fir.shift %c1, %c3 : (index, index) -> !fir.shift<2>
+ %2:2 = hlfir.declare %arg0(%1) dummy_scope %0 {fortran_attrs = #fir.var_attrs<contiguous>, uniq_name = "_QFtest_contiguous_derived_lboundsEd1"} : (!fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>, !fir.shift<2>, !fir.dscope) -> (!fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>, !fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>)
+ fir.select_type %2#1 : !fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>> [#fir.type_is<!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>, ^bb1, unit, ^bb2]
+^bb1: // pred: ^bb0
+ %3 = fir.convert %2#1 : (!fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>) -> !fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>
+ %4:2 = hlfir.declare %3(%1) {fortran_attrs = #fir.var_attrs<contiguous>, uniq_name = "_QFtest_contiguous_derived_lboundsEd1"} : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, !fir.shift<2>) -> (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, !fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>)
+ %5 = hlfir.designate %4#0 (%c1, %c3) : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, index, index) -> !fir.ref<!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>
+ %6 = hlfir.designate %5{"i"} : (!fir.ref<!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>) -> !fir.ref<i32>
+ hlfir.assign %c16_i32 to %6 : i32, !fir.ref<i32>
+ cf.br ^bb3
+^bb2: // pred: ^bb0
+ %7:2 = hlfir.declare %2#1(%1) {fortran_attrs = #fir.var_attrs<contiguous>, uniq_name = "_QFtest_contiguous_derived_lboundsEd1"} : (!fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>, !fir.shift<2>) -> (!fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>, !fir.class<!fir.array<?x?x!fir.type<_QMtypesTt1>>>)
+ cf.br ^bb3
+^bb3: // 2 preds: ^bb1, ^bb2
+ return
+}
+// CHECK-LABEL: func.func @_QPtest_contiguous_derived_lbounds(
+// CHECK: %[[VAL_0:.*]] = arith.constant 3 : index
+// CHECK: %[[VAL_1:.*]] = arith.constant 1 : index
+// CHECK: %[[VAL_8:.*]] = fir.declare %{{.*}}(%[[VAL_4:.*]]) {fortran_attrs = #fir.var_attrs<contiguous>, uniq_name = "_QFtest_contiguous_derived_lboundsEd1"} : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, !fir.shift<2>) -> !fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>
+// CHECK: %[[VAL_9:.*]] = fir.rebox %[[VAL_8]](%[[VAL_4]]) : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, !fir.shift<2>) -> !fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>
+// CHECK: %[[VAL_10:.*]] = fir.box_addr %[[VAL_9]] : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>) -> !fir.ref<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>
+// CHECK: %[[VAL_11:.*]] = arith.constant 0 : index
+// CHECK: %[[VAL_12:.*]]:3 = fir.box_dims %[[VAL_9]], %[[VAL_11]] : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, index) -> (index, index, index)
+// CHECK: %[[VAL_13:.*]] = arith.constant 1 : index
+// CHECK: %[[VAL_14:.*]]:3 = fir.box_dims %[[VAL_9]], %[[VAL_13]] : (!fir.box<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, index) -> (index, index, index)
+// CHECK: %[[VAL_15:.*]] = arith.constant 1 : index
+// CHECK: %[[VAL_16:.*]] = arith.subi %[[VAL_1]], %[[VAL_1]] : index
+// CHECK: %[[VAL_17:.*]] = arith.addi %[[VAL_16]], %[[VAL_15]] : index
+// CHECK: %[[VAL_18:.*]] = arith.constant 1 : index
+// CHECK: %[[VAL_19:.*]] = arith.subi %[[VAL_0]], %[[VAL_0]] : index
+// CHECK: %[[VAL_20:.*]] = arith.addi %[[VAL_19]], %[[VAL_18]] : index
+// CHECK: %[[VAL_21:.*]] = fir.array_coor %[[VAL_10]] %[[VAL_17]], %[[VAL_20]] : (!fir.ref<!fir.array<?x?x!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>>, index, index) -> !fir.ref<!fir.type<_QMtypesTt2{t1:!fir.type<_QMtypesTt1>,i:i32}>>
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Contiguous variables represented with a box do not have
explicit shape, but it looks like the base/shape computation
was assuming that. This caused generation of raw address
fir.array_coor without the shape. This patch is needed
to fix failures hapenning with #138797.