Anton Korobeynikov
Anton Korobeynikov
Checking the deps: 1. It seems that jemalloc is optional. Could simply be dropped. Or replaced by e.g. mimalloc / tcmalloc 2. Boost is only used to parse program options....
> Yes, this should be possible. jemalloc was the fastest option based on evaluations a couple of years ago but this might have changed since then. Our (= SPAdes) experience...
preset=buildbot,tools=RA,stdlib=DA @swift-ci please test with preset
Our activity analysis is just plain wrong for class arguments. In particular, the usefullness analysis is not quite prepared for such values. Currently the inouts are treated in such a...
Tagging @rxwei @BradLarson
> Issue still exists on 05/20 toolchain. We decided not to proceed with class differentiation due to many fundamental questions (classes are not values)
@fibrechannelscsi Looks like we've failed to create proper reabstraction thunk from `@escaping @callee_guaranteed (@in_guaranteed Float) -> (@out Float, @out Float)` to `(Float) -> (Float, Float)`. Apparently the inner abstraction pattern...
Ok, the problem is quite weird. Looks like we're confusing `(@in_guaranteed Float) -> (@out Float, @out Float` and `(@in_guaranteed Float) -> @out (Float, Float)` as a pullback type here. The...
@philipturner Why are you closing it?
@jkshtj Have you checked that the type of reabstraction thunk is correct now?