[flang] Use APInt to lower 128 bits integer constants

Lowering was truncating 128 bits integer to 64 bits. This
patch makes use of APInt to lower 128 bits integer correctly.

```
program bug
  print *, 170141183460469231731687303715884105727_16
end

! Before patch: 18446744073709551615
! With patch: 170141183460469231731687303715884105727
```

Reviewed By: vdonaldson

Differential Revision: https://reviews.llvm.org/D133206

Author: clementval

flang/lib/Lower/ConvertExpr.cpp

@@ -1451,6 +1451,14 @@ class ScalarExprLowering {
       const Fortran::evaluate::Scalar<Fortran::evaluate::Type<TC, KIND>>
           &value) {
     if constexpr (TC == Fortran::common::TypeCategory::Integer) {
+      if (KIND == 16) {
+        mlir::Type ty =
+            converter.genType(Fortran::common::TypeCategory::Integer, KIND);
+        auto bigInt =
+            llvm::APInt(ty.getIntOrFloatBitWidth(), value.SignedDecimal(), 10);
+        return builder.create<mlir::arith::ConstantOp>(
+            getLoc(), ty, mlir::IntegerAttr::get(ty, bigInt));
+      }
       return genIntegerConstant<KIND>(builder.getContext(), value.ToInt64());
     } else if constexpr (TC == Fortran::common::TypeCategory::Logical) {
       return genBoolConstant(value.IsTrue());

flang/test/Lower/big-integer-parameter.f90

@@ -0,0 +1,39 @@
+! Test correct lowering of 128 bit integer parameters.
+! RUN: bbc -emit-fir %s -o - | FileCheck %s
+
+program i128
+  integer(16), parameter :: maxi64 = 9223372036854775807_16
+  integer(16), parameter :: mini64 = -9223372036854775808_16
+  integer(16), parameter :: maxi128 = 170141183460469231731687303715884105727_16
+  integer(16), parameter :: mini128 = -170141183460469231731687303715884105728_16
+  integer(16), parameter :: x = 9223372036854775808_16
+  integer(16), parameter :: y = -9223372036854775809_16
+  integer(16), parameter :: z = 0_16
+  print*,x
+  print*,y
+end
+
+! CHECK-LABEL: func.func @_QQmain() {
+! CHECK-COUNT-2:  %{{.*}} = fir.call @_FortranAioOutputInteger128(%{{.*}}, %{{.*}}) : (!fir.ref<i8>, i128) -> i1
+
+
+! CHECK-LABEL: fir.global internal @_QFECmaxi128 constant : i128 {
+! CHECK-NEXT: %{{.*}} = arith.constant 170141183460469231731687303715884105727 : i128
+
+! CHECK-LABEL: fir.global internal @_QFECmaxi64 constant : i128 {
+! CHECK-NEXT:   %{{.*}} = arith.constant 9223372036854775807 : i128
+
+! CHECK-LABEL: fir.global internal @_QFECmini128 constant : i128 {
+! CHECK-NEXT: %{{.*}} = arith.constant -170141183460469231731687303715884105728 : i128
+
+! CHECK-LABEL: fir.global internal @_QFECmini64 constant : i128 {
+! CHECK-NEXT: %{{.*}} = arith.constant -9223372036854775808 : i128
+
+! CHECK-LABEL: fir.global internal @_QFECx constant : i128 {
+! CHECK-NEXT:   %{{.*}} = arith.constant 9223372036854775808 : i128
+ 
+! CHECK-LABEL: fir.global internal @_QFECy constant : i128 {
+! CHECK-NEXT: %{{.*}} = arith.constant -9223372036854775809 : i128
+
+! CHECK-LABEL: fir.global internal @_QFECz constant : i128 {
+! CHECK-NEXT: %{{.*}} = arith.constant 0 : i128