I have ran into this twice now and thought it would be great if an nvbench-based benchmark could create any intermediate directories for the output JSON file.
Now, with the directory latest_json not existing, the benchmark aborts with an error:
./bin/cub.bench.transform.babelstream2.base -d 0 --stopping-criterion entropy --json latest_json/babelstream2.json
NVBench encountered an error:
[...]/cccl/build_profile/_deps/nvbench-src/nvbench/option_parser.cu:587: Error while adding json output for `latest_json/babelstream3.json`:
basic_ios::clear: iostream error
I have ran into this twice now and thought it would be great if an nvbench-based benchmark could create any intermediate directories for the output JSON file.
Now, with the directory
latest_jsonnot existing, the benchmark aborts with an error: