NLL Performance for up-to-date versions of the perf.rlo crates

Showing the time it took to cargo check each of the crates individually — simulating a non-incremental cargo check, after a touch of lib.rs or main.rs.
Each build was done 3 times and the time reported below is the average of the 3. The machine used is a dual-threaded quad-core so cargo might sometimes build more than one crate at a time.

The crate versions are generally the latest released on crates.io, and sometimes master.

These timings were done on the latest nightly: 6a1c0637c 2018-07-23.
A callgrind profile is sometimes available, linked in the column (and is probably more anecdotical evidence than our regular perf runs, but still informative). They were gathered on a local build of commit baba5007bf... from 2018-07-24.

Exceptions:

• encoding: has not seen a lot of activity since it was added to rustc-perf, and it compiles as fast with or without NLLs.
• futures: compiles as fast with or without NLLs in under 0.8s.
• script-servo: requires x11 which I don't have on this machine.
• tokio-webpush-simple: compiles as fast with or without NLLs under 0.5s.

Notes:

• cargo: the –lib profile was timed
• ripgrep: the –lib profile was timed

Timings using the master version of each crate

Results for perf.rlo versions

These are the timing results using the versions of the crates that appear on perf.rust-lang.org. These tend to be older versions where we encountered particular performance headaches before.