Idris2/tests/idris2/perf010/run
Edwin Brady dd95a549d5
Fix performance regression #1991 (#1995)
* Normalise types fully at the REPL

It was a bit odd that we only normalised the scope of function types and
not the arguments, and I can't remember the reason for that if there
even was one.

* Better way of using nf_metavars_threshold

If a term is getting big and probably needs normalising, we now have a
sizeLimit flag in quote, so we can use that instead of checking the size
afterwards. This is a handy heuristic for speeding up unification when
there's a term with lots of suspended computation. Fixes #1991
2021-10-11 23:53:52 +01:00

4 lines
65 B
Plaintext

rm -rf build
$1 --no-color --console-width 0 --check Printf.idr