What about this

Author: wence-

cpp/benchmarks/streaming/ndsh/q09.cpp

@@ -30,9 +30,9 @@
 #include <cudf/table/table_view.hpp>
 #include <cudf/transform.hpp>
 #include <cudf/types.hpp>
-#include <rmm/mr/device/cuda_async_memory_resource.hpp>
-#include <rmm/mr/device/per_device_resource.hpp>
-#include <rmm/mr/device/pool_memory_resource.hpp>
+#include <rmm/mr/cuda_async_memory_resource.hpp>
+#include <rmm/mr/per_device_resource.hpp>
+#include <rmm/mr/pool_memory_resource.hpp>
 #include <rmm/resource_ref.hpp>
 
 #include <rapidsmpf/buffer/buffer.hpp>
@@ -476,44 +476,36 @@ static __device__ void calculate_amount(double *amount, double discount, double
     rapidsmpf::streaming::ShutdownAtExit c{ch_in, ch_out};
     co_await ctx->executor()->schedule();
     auto msg = co_await ch_in->receive();
-    ctx->comm()->logger().print("Sortby");
     // We know we only have a single chunk from the groupby
     if (msg.empty()) {
         co_return;
     }
+    ctx->comm()->logger().print("Sortby");
     auto chunk =
         rapidsmpf::ndsh::to_device(ctx, msg.release<rapidsmpf::streaming::TableChunk>());
     auto table = chunk.table_view();
-#if defined(__GNUC__) && __GNUC__ == 13
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wdeprecated-declarations"
-#endif
     auto rounded = cudf::round(
         table.column(2),
         2,
         cudf::rounding_method::HALF_EVEN,
         chunk.stream(),
         ctx->br()->device_mr()
     );
-#if defined(__GNUC__) && __GNUC__ == 13
-#pragma GCC diagnostic pop
-#endif
-    co_await ch_out->send(
-        rapidsmpf::streaming::to_message(
-            0,
-            std::make_unique<rapidsmpf::streaming::TableChunk>(
-                cudf::sort_by_key(
-                    cudf::table_view({table.column(0), table.column(1), rounded->view()}),
-                    table.select({0, 1}),
-                    {cudf::order::ASCENDING, cudf::order::DESCENDING},
-                    {cudf::null_order::BEFORE, cudf::null_order::BEFORE},
-                    chunk.stream(),
-                    ctx->br()->device_mr()
-                ),
-                chunk.stream()
-            )
+    auto result = rapidsmpf::streaming::to_message(
+        0,
+        std::make_unique<rapidsmpf::streaming::TableChunk>(
+            cudf::sort_by_key(
+                cudf::table_view({table.column(0), table.column(1), rounded->view()}),
+                table.select({0, 1}),
+                {cudf::order::ASCENDING, cudf::order::DESCENDING},
+                {cudf::null_order::BEFORE, cudf::null_order::BEFORE},
+                chunk.stream(),
+                ctx->br()->device_mr()
+            ),
+            chunk.stream()
         )
     );
+    co_await ch_out->send(std::move(result));
     co_await ch_out->drain(ctx->executor());
 }