[flang-commits] [flang] 3064d6b - [flang][hlfir] Simplify hlfir.assign default codegen for arrays
Jean Perier via flang-commits
flang-commits at lists.llvm.org
Thu Feb 23 00:13:27 PST 2023
Author: Jean Perier
Date: 2023-02-23T09:13:09+01:00
New Revision: 3064d6b7b062cca348fd5528f3a77bff7df7e5f9
URL: https://github.com/llvm/llvm-project/commit/3064d6b7b062cca348fd5528f3a77bff7df7e5f9
DIFF: https://github.com/llvm/llvm-project/commit/3064d6b7b062cca348fd5528f3a77bff7df7e5f9.diff
LOG: [flang][hlfir] Simplify hlfir.assign default codegen for arrays
The previous code was always emitting two genAssign calls to create
a temporary copy of the RHS if it could overlap with the LHS.
This inline temporary creation is not needed anymore after:
https://github.com/llvm/llvm-project/commit/755535b5eb5f6d60e9cc347cecd9e057231b92bb
that updated the assignment runtime to detect overlap and make a
temporary copy in the runtime directly.
Note that optimized inlined assignment will still have to do the alias
analysis to skip the copy when added later.
Differential Revision: https://reviews.llvm.org/D144567
Added:
Modified:
flang/lib/Optimizer/HLFIR/Transforms/ConvertToFIR.cpp
flang/test/HLFIR/assign-codegen.fir
Removed:
################################################################################
diff --git a/flang/lib/Optimizer/HLFIR/Transforms/ConvertToFIR.cpp b/flang/lib/Optimizer/HLFIR/Transforms/ConvertToFIR.cpp
index 51cca5c4e71a3..181bc2f87b232 100644
--- a/flang/lib/Optimizer/HLFIR/Transforms/ConvertToFIR.cpp
+++ b/flang/lib/Optimizer/HLFIR/Transforms/ConvertToFIR.cpp
@@ -54,18 +54,9 @@ static mlir::Value genAllocatableTempFromSourceBox(mlir::Location loc,
return copy;
}
-static std::pair<mlir::Value, bool>
-genTempFromSourceBox(mlir::Location loc, fir::FirOpBuilder &builder,
- mlir::Value sourceBox) {
- return {genAllocatableTempFromSourceBox(loc, builder, sourceBox),
- /*cleanUpTemp=*/true};
-}
-
namespace {
/// May \p lhs alias with \p rhs?
/// TODO: implement HLFIR alias analysis.
-static bool mayAlias(hlfir::Entity lhs, hlfir::Entity rhs) { return true; }
-
class AssignOpConversion : public mlir::OpRewritePattern<hlfir::AssignOp> {
public:
explicit AssignOpConversion(mlir::MLIRContext *ctx) : OpRewritePattern{ctx} {}
@@ -92,6 +83,10 @@ class AssignOpConversion : public mlir::OpRewritePattern<hlfir::AssignOp> {
"variable to fir::ExtendedValue must not require cleanup");
if (lhs.isArray()) {
+ // There may be overlap between lhs and rhs. The runtime is able to detect
+ // and to make a copy of the rhs before modifying the lhs if needed.
+ // The code below relies on this and does not do any compile time alias
+ // analysis.
const bool rhsIsValue = fir::isa_trivial(fir::getBase(rhsExv).getType());
if (rhsIsValue) {
// createBox can only be called for fir::ExtendedValue that are
@@ -110,9 +105,6 @@ class AssignOpConversion : public mlir::OpRewritePattern<hlfir::AssignOp> {
// inline array assignment when profitable.
auto to = fir::getBase(builder.createBox(loc, lhsExv));
auto from = fir::getBase(builder.createBox(loc, rhsExv));
- bool cleanUpTemp = false;
- if (!rhsIsValue && mayAlias(rhs, lhs))
- std::tie(from, cleanUpTemp) = genTempFromSourceBox(loc, builder, from);
auto toMutableBox = builder.createTemporary(loc, to.getType());
// As per 10.2.1.2 point 1 (1) polymorphic variables must be allocatable.
@@ -120,10 +112,6 @@ class AssignOpConversion : public mlir::OpRewritePattern<hlfir::AssignOp> {
// type and that the mutableBox will not be modified.
builder.create<fir::StoreOp>(loc, to, toMutableBox);
fir::runtime::genAssign(builder, loc, toMutableBox, from);
- if (cleanUpTemp) {
- mlir::Value addr = builder.create<fir::BoxAddrOp>(loc, from);
- builder.create<fir::FreeMemOp>(loc, addr);
- }
} else {
// Assume overlap does not matter for scalar (dealt with memmove for
// characters).
diff --git a/flang/test/HLFIR/assign-codegen.fir b/flang/test/HLFIR/assign-codegen.fir
index 91ca33d8ca8df..5736ca9af98d1 100644
--- a/flang/test/HLFIR/assign-codegen.fir
+++ b/flang/test/HLFIR/assign-codegen.fir
@@ -132,7 +132,6 @@ func.func @array(%arg0: !fir.box<!fir.array<?xi32>>, %arg1: !fir.ref<!fir.array<
// CHECK-SAME: %[[VAL_0:.*]]: !fir.box<!fir.array<?xi32>>,
// CHECK-SAME: %[[VAL_1:.*]]: !fir.ref<!fir.array<100xi32>>) {
// CHECK: %[[VAL_2:.*]] = fir.alloca !fir.box<!fir.array<?xi32>>
-// CHECK: %[[VAL_3:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<100xi32>>>
// CHECK: %[[VAL_4:.*]] = arith.constant 100 : index
// CHECK: %[[VAL_5:.*]] = fir.declare %[[VAL_0]] {uniq_name = "x"} : (!fir.box<!fir.array<?xi32>>) -> !fir.box<!fir.array<?xi32>>
// CHECK: %[[VAL_6:.*]] = fir.rebox %[[VAL_5]] : (!fir.box<!fir.array<?xi32>>) -> !fir.box<!fir.array<?xi32>>
@@ -140,21 +139,10 @@ func.func @array(%arg0: !fir.box<!fir.array<?xi32>>, %arg1: !fir.ref<!fir.array<
// CHECK: %[[VAL_8:.*]] = fir.declare %[[VAL_1]](%[[VAL_7]]) {uniq_name = "y"} : (!fir.ref<!fir.array<100xi32>>, !fir.shape<1>) -> !fir.ref<!fir.array<100xi32>>
// CHECK: %[[VAL_9:.*]] = fir.shape %[[VAL_4]] : (index) -> !fir.shape<1>
// CHECK: %[[VAL_10:.*]] = fir.embox %[[VAL_8]](%[[VAL_9]]) : (!fir.ref<!fir.array<100xi32>>, !fir.shape<1>) -> !fir.box<!fir.array<100xi32>>
-// CHECK: %[[VAL_11:.*]] = fir.zero_bits !fir.heap<!fir.array<100xi32>>
-// CHECK: %[[VAL_12:.*]] = arith.constant 0 : index
-// CHECK: %[[VAL_13:.*]] = fir.shape %[[VAL_12]] : (index) -> !fir.shape<1>
-// CHECK: %[[VAL_14:.*]] = fir.embox %[[VAL_11]](%[[VAL_13]]) : (!fir.heap<!fir.array<100xi32>>, !fir.shape<1>) -> !fir.box<!fir.heap<!fir.array<100xi32>>>
-// CHECK: fir.store %[[VAL_14]] to %[[VAL_3]] : !fir.ref<!fir.box<!fir.heap<!fir.array<100xi32>>>>
-// CHECK: %[[VAL_18:.*]] = fir.convert %[[VAL_3]] : (!fir.ref<!fir.box<!fir.heap<!fir.array<100xi32>>>>) -> !fir.ref<!fir.box<none>>
-// CHECK: %[[VAL_19:.*]] = fir.convert %[[VAL_10]] : (!fir.box<!fir.array<100xi32>>) -> !fir.box<none>
-// CHECK: %[[VAL_21:.*]] = fir.call @_FortranAAssign(%[[VAL_18]], %[[VAL_19]], %{{.*}}, %{{.*}}) : (!fir.ref<!fir.box<none>>, !fir.box<none>, !fir.ref<i8>, i32) -> none
-// CHECK: %[[VAL_22:.*]] = fir.load %[[VAL_3]] : !fir.ref<!fir.box<!fir.heap<!fir.array<100xi32>>>>
// CHECK: fir.store %[[VAL_5]] to %[[VAL_2]] : !fir.ref<!fir.box<!fir.array<?xi32>>>
// CHECK: %[[VAL_26:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.box<!fir.array<?xi32>>>) -> !fir.ref<!fir.box<none>>
-// CHECK: %[[VAL_27:.*]] = fir.convert %[[VAL_22]] : (!fir.box<!fir.heap<!fir.array<100xi32>>>) -> !fir.box<none>
+// CHECK: %[[VAL_27:.*]] = fir.convert %[[VAL_10]] : (!fir.box<!fir.array<100xi32>>) -> !fir.box<none>
// CHECK: %[[VAL_29:.*]] = fir.call @_FortranAAssign(%[[VAL_26]], %[[VAL_27]], %{{.*}}, %{{.*}}) : (!fir.ref<!fir.box<none>>, !fir.box<none>, !fir.ref<i8>, i32) -> none
-// CHECK: %[[VAL_30:.*]] = fir.box_addr %[[VAL_22]] : (!fir.box<!fir.heap<!fir.array<100xi32>>>) -> !fir.heap<!fir.array<100xi32>>
-// CHECK: fir.freemem %[[VAL_30]] : !fir.heap<!fir.array<100xi32>>
func.func @test_scalar_to_array(%lhs: !fir.box<!fir.array<?xi32>>, %rhs: i32) {
More information about the flang-commits
mailing list