#12900: Common up identical info tables -------------------------------------+------------------------------------- Reporter: dobenour | Owner: Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Runtime | Unknown/Multiple performance bug | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by simonpj): Why do you say "massive" amount of space? Do you have data? My guess would be that, except in pathological situation (e.g. thousands of identical data types) the saving would be vanishingly small. Also I think that info tables contain strings that identify the type for debuggers etc; ghci has such a debugger built in. You'd lose this entirely. So I'm a skeptical that the gain justifies the pain (implementation complexity, loss of functionality). But data might convince! If you are looking for code space saving, then I think a more profitable place might be commoning up bits of Cmm code. E.g. consider {{{ f x y = case x of True -> False; False -> y }}} We push a return address, evaluate `x`; when we come back we take a conditional jump based on the tag and, in the `True` case we just return a fixed constructor `False`. Lots of code just returns `False`! If all that code was commoned up, we might get a big space saving. Maybe there are some sequences that are SO common we can put them in a big library, always available in the RTS, so all libraries could share. We'd need automation to find these common sequences and see which ones happened a lot. I remember a PhD student trying this years ago (20 yrs?) with some success. Simon -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/12900#comment:3 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler