Troels Henriksen

I would prefer a formatter that does not touch linebreaks. The token stream is useful because each token is associated with a start and end position in the file. We...

Philip Munksgaard writes: >> I would prefer a formatter that does not touch linebreaks. > > I am unsure what you mean by this. I would prefer a formatter that...

I'm always reluctant to add more nondeterministic constructs. I guess this one could be OK, especially if we can show that the cost of `replicate`ing an array is significant for...

For FFTs specifically, this can also be addressed by [manual reuse of arrays](https://github.com/diku-dk/fft/blob/master/lib/github.com/diku-dk/fft/stockham-radix-2.fut#L36-L42).

The `replicate n false` there is expensive enough to offset the advantage of the hypothetical `scratch`. That doesn't mean a safe `permute` is not useful, but there is no need...

Yeah, I guess with `#[unsafe]` dead code elimination would remove the inefficient replicates. That might not be a bad implementation.

Can you just print the result, like you do with a scalar? To be honest, I don't care - I am going to compile all TAIL programs that look like...

I have a system I would like to run TAIL on, but the only SML compiler available is MosML. I'll take a swing at it. ## \ Troels /\ Henriksen

I managed to find an MLton package for OpenBSD, so I won't spend any more time on this one. (Turns out that MosML does not define Real.posInf.)

Make 'make mlkit' and 'make mlton' targets, like how smackage works? ## \ Troels /\ Henriksen