dive-cli
A faster TLA+ toolchain. Same specs, same results as TLC, about 5x faster.
- 5x faster than TLC on average across 169 models from tlaplus/Examples. 118 of them run at least 3x faster, and 31 at least 10x faster.
- Instant on small specs.
DieHardchecks in about 6 ms. TLC takes about half a second. - A fraction of the memory. 37–110 MiB where TLC uses 0.8–1.4 GiB.
- Drop-in. Reads your existing
.tlaand.cfgfiles. Includes atlccommand that accepts TLC's arguments. - One binary for model checking, simulation, parsing, PlusCal translation and trace validation.
Installing needs Rust 1.99 or later and a C compiler. See the
dive repository for the user guide and
how the numbers were measured. Rust callers can embed the dive-api library
instead.