mlite

Is this really possible without understanding what each pass is doing? It seems it requires each pass be both commutative and associative.

Without knowing why EnumMapSet is good for Hoopl's general use cases, I can provide the following information for you to consider. 1. What does adding the new dependencies mean to...

I support reimplementing the newtype wrappers in hoopl. On Mon, Nov 2, 2020 at 12:59 AM simonpj wrote: > If it's just a question of boilerplate newtype wrappers, I'd be...

Michal Terepeta who is also a hoopl maintainer should be the one making the call as I'm supposed to step down from the maintainer position. On Mon, Nov 2, 2020...

Is this check in arfx too late? Will adding an invariant check "all entry labels should have a fact" to analyzeAndRewriteFwd be better? A similar invariant check can be added...

It (restart) was invoked in the initial implementation, but it was, accidentally?, removed in a Dataflow rewriting to improve performance. Without CheckpointMonad, you still can write a correct analysis if...

> Do you think this could be related to issue #1 ? Using freshUniques from UniqueMonad to name new variables is not a good idea, as UniqueMonad is not restorable...

The justification of Checkpoint Monad is in the original paper. Checkpoint Monad was used in the original Dataflow implementation and removed in the performance optimized Dataflow reimplementation. Checkpoint Monad can...

Please check the revision history of Dataflow.hs. The performance optimization was needed for integrating Hoopl into GHC.

Checkout Simon Marlow’s commits. I believe he did the performance enhancement for Dataflow.hs > On Feb 11, 2019, at 12:54 PM, Siddharth wrote: > > @mlite Sorry for pinging you...