I may be wrong, but if I remember correctly Standard Charter uses their own “dialect” called Mu[1].

1 - https://icfp21.sigplan.org/details/hiw-2021-papers/14/Haskel...

He probably means that GHC will parse and type check faster. Switching the front-end doesn't sound like it will circumvent the whole program nature of their compilation. But I'm just guessing.

From what I understand: GHC takes care of the parsing of the source file, error reporting, etc. Then the AST (or an intermediate representation) is passed to the of the Mu compiler, which performs whole-program optimisations.

Gergo Erdi gave a talk on the GHC efforts of Standard Chartered last year:

https://www.youtube.com/watch?v=fZ66Pz7015Q

I'm also interested. Just guessing, but I guess it means that while they use their own (slow compiling) 'Mu' dialect for most applications, they have switched to using 'standard' GHC for their front-end apps.

Frontend here refers to the first stage of the three-stage structure of compilers, ie. the functionalities such as parsing and type checking that come before the optimization and target code generation phases: https://en.wikipedia.org/wiki/Compiler#Three-stage_compiler_...

Didn't know that, thanks! But then I'm wondering how they can do parsing and type checking with GHC when they are using their own dialect, "Mu". How come GHC can parse that?

The differences between Mu and Haskell are not that large. They use GHC as a library in their front end. Gergo Erdi gave a talk on it last year:

https://news.ycombinator.com/item?id=35801844

GHC has a system for language extensions so they have developed a few proprietary ones for Mu. For the standard ones, see https://downloads.haskell.org/ghc/latest/docs/users_guide/ex...