Having benchmark metrics based on whole block metrics provides very useful insights about the overall performance of a contracts chain. However, to find and debug specific bottlenecks, we should be able to have more fine grained measurements.
Ideally, we'd use some off-chain engine like drink for that. This would make it very easy to integrate the benchmarks in ink! and Solang CI pipelines.