vuke bench currently takes a single --transform and benchmarks it in isolation. When comparing transform throughput across the board, you end up running the command a dozen times and manually collecting numbers.

An --all flag (or accepting multiple --transform values like generate/scan already do) would let you get a comparative table in one run:

$ vuke bench --all
Transform          Keys/sec
sha256             12,450,000
double_sha256       6,200,000
md5                14,100,000
direct             89,000,000
lcg:glibc           8,300,000
...

The JSON output (--json) could return an array of results instead of a single object. The existing benchmark loop in run_benchmark() already handles warmup and measurement for one transform, so this is mostly a matter of iterating over TransformType::all_variants() or a user-provided list and collecting results.

The generate and scan commands already accept num_args = 1.. for transforms, so the pattern exists.