Pinned vs cucascade pool

cpp/benchmarks/CMakeLists.txt

@@ -148,6 +148,39 @@ install(
   EXCLUDE_FROM_ALL
 )
 
+add_executable(bench_fixed_sized_pool "bench_fixed_sized_pool.cpp")
+set_target_properties(
+  bench_fixed_sized_pool
+  PROPERTIES RUNTIME_OUTPUT_DIRECTORY "$<BUILD_INTERFACE:${RAPIDSMPF_BINARY_DIR}/benchmarks>"
+             CXX_STANDARD 20
+             CXX_STANDARD_REQUIRED ON
+             CXX_EXTENSIONS ON
+             CUDA_STANDARD 20
+             CUDA_STANDARD_REQUIRED ON
+)
+target_compile_options(
+  bench_fixed_sized_pool PRIVATE "$<$<COMPILE_LANGUAGE:CXX>:${RAPIDSMPF_CXX_FLAGS}>"
+                                 "$<$<COMPILE_LANGUAGE:CUDA>:${RAPIDSMPF_CUDA_FLAGS}>"
+)
+target_link_libraries(
+  bench_fixed_sized_pool
+  PRIVATE benchmark::benchmark
+          benchmark::benchmark_main
+          rapidsmpf::rapidsmpf
+          rmm::rmm
+          $<TARGET_NAME_IF_EXISTS:rapidsmpf_cucascade_internal>
+          $<$<NOT:$<TARGET_EXISTS:rapidsmpf_cucascade_internal>>:cuCascade::cucascade>
+          $<TARGET_NAME_IF_EXISTS:conda_env>
+          maybe_asan
+          bench_utils
+)
+install(
+  TARGETS bench_fixed_sized_pool
+  COMPONENT benchmarking
+  DESTINATION bin/benchmarks/librapidsmpf
+  EXCLUDE_FROM_ALL
+)
+
 if(RAPIDSMPF_HAVE_STREAMING)
   add_subdirectory(streaming)
 endif()