Secrets of the Glasgow Haskell Compiler Inliner (1999) [pdf]

[aaronchall]

https://www.microsoft.com/en-us/research/wp-content/uploads/2002/07/inline.pdf

[houli]

Also worth checking out is this https://www.youtube.com/watch?v=uR_VzYxvbxg conference talk by Simon Peyton Jones discussing the basic constructs of GHC core and how they affect inlining and optimisation

[dualogy → houli]

SPJ must be the bounciest computing-guy-over-30 I've ever encountered. Always a refreshing pleasure to watch any of his performances =)

[akurilin → dualogy]

His enthusiasm is truly contagious, somewhat reminiscent of how excited Feynman could make you feel about science.

[watmough → akurilin]

Yes, i was lucky enough to be in the audience at St Andrews in about 1986 when he came to talk to us about parallel combinator reduction, G-machines and Alice.

Amazing enthusiasm and application. Amazing guy.

[steedsofwar → dualogy]

He was my professor at university, came in one day with shirt, tie and sandals. Took in a deep breath, and shouted it's a beautiful sunny day today.

[endgame → dualogy]

Watch something with Clifford Stoll in it. His enthusiasm is also incredible.

[amelius]

I still think it is a pity that GHC isn't much better documented. It could be a huge education and research platform, with a much wider reach than the small group currently versed in GHC internals.

[chrisseaton → amelius]

It is a huge education and research platform. There are many many papers every year developing new ideas in GHC. Look at the massive cottage industry in Haskell language extensions implemented in GHC.

[kazinator]

> A major issue for any compiler, especially for one that inlines heavily, is name capture

What? If you inline just one block of code in a million lines of code (i.e. you "inline with the dainty lightness of a frolicking fairy"), and that situation is wrong due to a name capture, you have a bug, which could stop the show.

[kr7 → kazinator]

In section 3.1 they say it is a performance issue.