Gabriel Scherer

That may very well be the problem: we are in trouble if the `Effect` module is not linked. `perform` is declared as an `external` primitive in the .mli, I wonder...

To my knowledge it is an explicit design choice of the compiler that mentioning an `external` declaration of a module does not create a linking dependency on that module. The...

I personally would like to see `Load_mutable_field` and `Load_immutable_field` macros on the C side, that need not be lvalues. I could use them to express my intent / my understanding...

I agree that we should try to summarize the discussions in a form that newcomers can understand. On a ping from @OlivierNicole I asked @gadmm for more details, and I...

I wonder if the "documenting the current state" part should be placed somewhere more permanent. In memory.h? Somewhere in the user manual? I am thinking of sending a PR to...

The "documenting the current state" part is now included as a comment in memory.c as suggested by @OlivierNicole. Thanks for the feedback!

Note: it would be nice to give control to users, to let them indicate at unmarshalling time whether they believe that the data will be long-lived in their program. (@NickBarnes...

@kayceesrk : Excellent question! ## Raw data I ran my measurements again with a variation of minor heap sizes 's' (256K (the OCaml-default), 1M, 32M (the Coq default)), space overhead...

@kayceesrk I would be surprised if faster root scanning had impact on this issue, where I think the runtime is wasting time traversing the heap again and again to mark...

@damiendoligez I'm not sure what you think we should do: 1. Should we consider merging something like the current PR now, improving performance (noticeably) for Coq and similar-workload programs, at...