[flang][cuda] Add support to allocate scalar character types (llvm#169550)

Add support for character declared like: 

```
subroutine sub1()
  character*4, device :: b
end subroutine
```

Author: clementval

flang/lib/Optimizer/Transforms/CUFOpConversion.cpp

@@ -331,6 +331,11 @@ struct CUFAllocOpConversion : public mlir::OpRewritePattern<cuf::AllocOp> {
         std::size_t structSize = dl->getTypeSizeInBits(structTy) / 8;
         bytes = builder.createIntegerConstant(loc, builder.getIndexType(),
                                               structSize);
+      } else if (fir::isa_char(op.getInType())) {
+        mlir::Type charTy = typeConverter->convertType(op.getInType());
+        std::size_t charSize = dl->getTypeSizeInBits(charTy) / 8;
+        bytes = builder.createIntegerConstant(loc, builder.getIndexType(),
+                                              charSize);
       } else {
         mlir::emitError(loc, "unsupported type in cuf.alloc\n");
       }

flang/test/Fir/CUDA/cuda-alloc-free.fir

@@ -94,4 +94,24 @@ func.func @_QQalloc_char() attributes {fir.bindc_name = "alloc_char"} {
 // CHECK: %[[BYTES_CONV:.*]] = fir.convert %[[BYTES]] : (index) -> i64
 // CHECK: fir.call @_FortranACUFMemAlloc(%[[BYTES_CONV]], %c0{{.*}}, %{{.*}}, %{{.*}}) {cuf.data_attr = #cuf.cuda<device>} : (i64, i32, !fir.ref<i8>, i32) -> !fir.llvm_ptr<i8>
 
+
+func.func @_QQalloc_char2() {
+  %c4 = arith.constant 4 : index
+  %1 = cuf.alloc !fir.char<1,4>(%c4 : index) {bindc_name = "b", data_attr = #cuf.cuda<device>, uniq_name = "_QFsub1Eb"} -> !fir.ref<!fir.char<1,4>>
+  %2 = cuf.alloc !fir.char<2,4>(%c4 : index) {bindc_name = "c", data_attr = #cuf.cuda<device>, uniq_name = "_QFsub1Ec"} -> !fir.ref<!fir.char<2,4>>
+  %c10 = arith.constant 4 : index
+  %3 = cuf.alloc !fir.char<4,10>(%c10 : index) {bindc_name = "d", data_attr = #cuf.cuda<device>, uniq_name = "_QFsub1Ed"} -> !fir.ref<!fir.char<4,10>>
+  return
+}
+
+// CHECK-LABEL: func.func @_QQalloc_char2()
+// CHECK: %[[BYTES_4:.*]] = fir.convert %c4{{.*}} : (index) -> i64
+// CHECK: %{{.*}} = fir.call @_FortranACUFMemAlloc(%[[BYTES_4]], %{{.*}}, %{{.*}}, %{{.*}}) {cuf.data_attr = #cuf.cuda<device>} : (i64, i32, !fir.ref<i8>, i32) -> !fir.llvm_ptr<i8>
+
+// CHECK: %[[BYTES_8:.*]] = fir.convert %c8{{.*}} : (index) -> i64
+// CHECK: %{{.*}} = fir.call @_FortranACUFMemAlloc(%[[BYTES_8]], %{{.*}}, %{{.*}}, %{{.*}}) {cuf.data_attr = #cuf.cuda<device>} : (i64, i32, !fir.ref<i8>, i32) -> !fir.llvm_ptr<i8>
+
+// CHECK: %[[BYTES_40:.*]] = fir.convert %c40{{.*}} : (index) -> i64
+// CHECK: %{{.*}} = fir.call @_FortranACUFMemAlloc(%[[BYTES_40]], %{{.*}}, %{{.*}}, %{{.*}}) {cuf.data_attr = #cuf.cuda<device>} : (i64, i32, !fir.ref<i8>, i32) -> !fir.llvm_ptr<i8>
+
 } // end module