Ok, fair enough. Would it break anything else? (i.e. coercions) On Wed, Feb 12, 2020 at 8:34 PM Sebastian Graf wrote:

You probably couldn't do pointer tagging https://gitlab.haskell.org/ghc/ghc/wikis/commentary/rts/haskell-execution/po... anymore, which is probably a substantial performance loss.

Am Mi., 12. Feb. 2020 um 19:58 Uhr schrieb Csaba Hruska < csaba.hruska@gmail.com>:

...

Hello,

In theory could GHC codegen work if every data constructor in the whole program have a globally unique tag value instead of starting from 1 for each algebraic data type? Would this break any GHC design decision?

Regards, Csaba _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs

Ok, but what I am curious about if the tag in the info table could be globally unique? On Wed, Feb 12, 2020 at 9:01 PM Vladislav Zavialov wrote:

The globally unique tag for data constructors already exists, it’s a pointer to the StgInfoTable. You can observe it using getClosureRaw from GHC.HeapView

Example:

Prelude GHC.HeapView> getClosureRaw Nothing (0x0000000107fd8e38,[4429024840,4429024752],[])

Here, the globally unique tag is 0x0000000107fd8e38.

Note that newtype constructors do not get their own tag because newtypes guarantee that they do not change the underlying representation of data.

I’ve discovered the existence of such a tag only recently (Alexander Vershilov pointed it out to me), so I cannot say if it’s a reliable way to identify data constructors. For example, it is definitely not stable across several runs of the same binary. However, within a single run, it seems to work.

- Vlad

...

On 12 Feb 2020, at 21:57, Csaba Hruska wrote:

Hello,

In theory could GHC codegen work if every data constructor in the whole program have a globally unique tag value instead of starting from 1 for each algebraic data type? Would this break any GHC design decision?

Regards, Csaba _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs