[flang-commits] [flang] 9b9a847 - [flang] do not generate padding/truncation code when character length are equals

[Jean Perier via flang-commits]

Author: Jean Perier
Date: 2022-12-06T09:34:13+01:00
New Revision: 9b9a8475d3ac9efad59554a25c940ef37b2056e3

URL: https://github.com/llvm/llvm-project/commit/9b9a8475d3ac9efad59554a25c940ef37b2056e3
DIFF: https://github.com/llvm/llvm-project/commit/9b9a8475d3ac9efad59554a25c940ef37b2056e3.diff

LOG: [flang] do not generate padding/truncation code when character length are equals

When generating character assignment operations, the generic code
generates some code to handle truncation and padding when the length
differ at runtime. A bypass already exists when the length are compile
time constant and match, but it was not used for the trivial case where
the RHS and LHS length is the same SSA value. In such case, even though,
the length is not know at compile time, it is known to be the same.

This will simplify the code creating character temporaries from a
variable in HLFIR that will use this assignment code.

Note that this probably has little impact on performance (llvm may be clever enough
to later catch that for us). But it makes the generated IR a lot more readable at
little cost.

Differential Revision: https://reviews.llvm.org/D139330

Added: 
    

Modified: 
    flang/lib/Optimizer/Builder/Character.cpp
    flang/test/Lower/OpenMP/omp-parallel-lastprivate-clause-scalar.f90
    flang/test/Lower/array-expression.f90

Removed: 
    


################################################################################
diff  --git a/flang/lib/Optimizer/Builder/Character.cpp b/flang/lib/Optimizer/Builder/Character.cpp
index 476187620e0fe..f3bb71efcbbb1 100644
--- a/flang/lib/Optimizer/Builder/Character.cpp
+++ b/flang/lib/Optimizer/Builder/Character.cpp
@@ -419,7 +419,8 @@ void fir::factory::CharacterExprHelper::createAssign(
   auto rhsCstLen = getCompileTimeLength(rhs);
   auto lhsCstLen = getCompileTimeLength(lhs);
   bool compileTimeSameLength =
-      lhsCstLen && rhsCstLen && *lhsCstLen == *rhsCstLen;
+      (lhsCstLen && rhsCstLen && *lhsCstLen == *rhsCstLen) ||
+      (rhs.getLen() == lhs.getLen());
 
   if (compileTimeSameLength && *lhsCstLen == 1) {
     createLengthOneAssign(lhs, rhs);

diff  --git a/flang/test/Lower/OpenMP/omp-parallel-lastprivate-clause-scalar.f90 b/flang/test/Lower/OpenMP/omp-parallel-lastprivate-clause-scalar.f90
index 09341af583f5a..cd0d8786d05f8 100644
--- a/flang/test/Lower/OpenMP/omp-parallel-lastprivate-clause-scalar.f90
+++ b/flang/test/Lower/OpenMP/omp-parallel-lastprivate-clause-scalar.f90
@@ -30,15 +30,6 @@
 !CHECK-DAG: %[[CVT:.*]] = fir.convert %[[ARG1_UNBOX]]#0 : (!fir.ref<!fir.char<1,?>>) -> !fir.ref<i8>
 !CHECK-DAG: %[[CVT1:.*]] = fir.convert %[[ARG1_PVT]] : (!fir.ref<!fir.char<1,5>>) -> !fir.ref<i8>
 !CHECK-DAG: fir.call @llvm.memmove.p0.p0.i64(%[[CVT]], %[[CVT1]]{{.*}})
-!CHECK: %[[THIRTY_TWO:.*]] = arith.constant 32 : i8
-!CHECK-DAG: %[[UNDEF:.*]] = fir.undefined !fir.char<1>
-!CHECK-DAG: %[[INSERT:.*]] = fir.insert_value %[[UNDEF]], %[[THIRTY_TWO]], [0 : index] : (!fir.char<1>, i8) -> !fir.char<1>
-!CHECK-DAG: %[[ONE_3:.*]] = arith.constant 1 : index
-!CHECK: fir.do_loop %[[ARG2:.*]] = {{.*}} {
-!CHECK-DAG: %[[CVT_2:.*]] = fir.convert %[[ARG1_UNBOX]]#0 : (!fir.ref<!fir.char<1,?>>) -> !fir.ref<!fir.array<?x!fir.char<1>>>
-!CHECK-DAG: %[[COORD:.*]] = fir.coordinate_of %[[CVT_2]], %[[ARG2]] : (!fir.ref<!fir.array<?x!fir.char<1>>>, index) -> !fir.ref<!fir.char<1>>
-!CHECK-DAG: fir.store %[[INSERT]] to %[[COORD]] : !fir.ref<!fir.char<1>>
-!CHECK-DAG: }
 !CHECK-DAG: } 
 !CHECK-DAG: omp.yield
 

diff  --git a/flang/test/Lower/array-expression.f90 b/flang/test/Lower/array-expression.f90
index 50e9f40457e63..a373548f0232a 100644
--- a/flang/test/Lower/array-expression.f90
+++ b/flang/test/Lower/array-expression.f90
@@ -998,20 +998,6 @@ end subroutine test19f
 ! CHECK:         %[[VAL_30:.*]] = fir.do_loop %[[VAL_31:.*]] = %[[VAL_28]] to %[[VAL_29]] step %[[VAL_27]] unordered iter_args(%[[VAL_32:.*]] = %[[VAL_15]]) -> (!fir.array<70x!fir.char<4,?>>) {
 ! CHECK:           %[[VAL_33:.*]] = fir.array_access %[[VAL_24]], %[[VAL_31]] : (!fir.array<140x!fir.char<2,13>>, index) -> !fir.ref<!fir.char<2,13>>
 ! CHECK:           %[[VAL_34:.*]] = fir.alloca !fir.char<4,?>(%[[VAL_4]] : index)
-! CHECK:           %[[VAL_35:.*]] = arith.cmpi slt, %[[VAL_4]], %[[VAL_4]] : index
-! CHECK:           %[[VAL_36:.*]] = arith.select %[[VAL_35]], %[[VAL_4]], %[[VAL_4]] : index
-! CHECK:           fir.char_convert %[[VAL_33]] for %[[VAL_36]] to %[[VAL_34]] : !fir.ref<!fir.char<2,13>>, index, !fir.ref<!fir.char<4,?>>
-! CHECK:           %[[VAL_37:.*]] = arith.constant 1 : index
-! CHECK:           %[[VAL_38:.*]] = arith.subi %[[VAL_4]], %[[VAL_37]] : index
-! CHECK:           %[[VAL_39:.*]] = arith.constant 32 : i32
-! CHECK:           %[[VAL_40:.*]] = fir.undefined !fir.char<4>
-! CHECK:           %[[VAL_41:.*]] = fir.insert_value %[[VAL_40]], %[[VAL_39]], [0 : index] : (!fir.char<4>, i32) -> !fir.char<4>
-! CHECK:           %[[VAL_42:.*]] = arith.constant 1 : index
-! CHECK:           fir.do_loop %[[VAL_43:.*]] = %[[VAL_36]] to %[[VAL_38]] step %[[VAL_42]] {
-! CHECK:             %[[VAL_44:.*]] = fir.convert %[[VAL_34]] : (!fir.ref<!fir.char<4,?>>) -> !fir.ref<!fir.array<?x!fir.char<4>>>
-! CHECK:             %[[VAL_45:.*]] = fir.coordinate_of %[[VAL_44]], %[[VAL_43]] : (!fir.ref<!fir.array<?x!fir.char<4>>>, index) -> !fir.ref<!fir.char<4>>
-! CHECK:             fir.store %[[VAL_41]] to %[[VAL_45]] : !fir.ref<!fir.char<4>>
-! CHECK:           }
 ! CHECK:           %[[VAL_46:.*]] = fir.array_access %[[VAL_32]], %[[VAL_31]] typeparams %[[VAL_11]] : (!fir.array<70x!fir.char<4,?>>, index, i32) -> !fir.ref<!fir.char<4,?>>
 ! CHECK:           %[[VAL_47:.*]] = fir.convert %[[VAL_11]] : (i32) -> index
 ! CHECK:           %[[VAL_48:.*]] = fir.convert %[[VAL_26]] : (i64) -> index