search

pepijndevos / futhark-pycffi

Python library using the Futhark C backend via CFFI
MIT License
23 stars 7 forks source link

Tests broken #16

Closed pepijndevos closed 3 years ago

pepijndevos commented 3 years ago

After merging the latest PR apparently the tests broke. https://travis-ci.org/github/pepijndevos/futhark-pycffi/jobs/760787891 It complains about this line, and I'm not sure what the correct solution is. https://github.com/pepijndevos/futhark-pycffi/blob/fc3ecc747038c6d6d88f935fee52f4b45269d0fa/test.fut#L17

I played around a bit on the repl, and iota seems to expect an i64, while the function has an i8 argument. Did i8.iota disappear, or did the syntax change?

Ping @athas

athas commented 3 years ago

Yes, i8.iota vanished in 0.17.2. Write it as map i8.i64 (iota (i64.i8 a)) instead.

pepijndevos commented 3 years ago

Ah. Any particular reason, or just maintenance burden of maintaining copies of functions for every type?

athas commented 3 years ago

Yes, mostly a general principle of minimising the prelude. With size types, they also aren't terribly useful.