March 2004 - Glasgow-haskell-users

RE: [Haskell] performance tuning Data.FiniteMap
by Simon Peyton-Jones 02 Mar '04

02 Mar '04

[Moving to GHC users list] There are several things that aren't research issues: notably, faster copying, fewer intermediate lists, fewer state-monad-induced intermediate closures. These are things that would move sharply up our priority list if you had a real application that was stumbling on them. What I obviously can't promise is that improving these things would solve your problem -- if, indeed, you turn out to have one! Simon | -----Original Message----- | From: S. Alexander Jacobson [mailto:alex@i2x.com] | Sent: 26 February 2004 23:27 | To: Simon Peyton-Jones | Cc: haskell(a)haskell.org | Subject: RE: [Haskell] performance tuning Data.FiniteMap | | Is fixing GHC arrays a big research job or is it | something that someone can straightforwardly | handle if my site actually gets enough traffic to | warrant it? | | -Alex- | | On Thu, 26 Feb 2004, Simon Peyton-Jones wrote: | | > | But in managing this tradeoff, what is faster: | > | * constructing/destructing e.g. 16 trees (for a 65000 item table) | > | * 2 memcpy of 256 item arrays (perhaps after you primop?) | > | | > | If the later is not dramatically slower than I | > | will bias towards more arrayness. | > | > I doubt the latter is dramatically slower, but you'd have to experiment | > to find out. And GHC is not doing as well as it should on arrays just | > now. (One of the things on our to-do list.) Might vary between | > implementations too. | > | > Simon | > | | _________________________________________________________________ | S. Alexander Jacobson mailto:me@alexjacobson.com | tel:917-770-6565 http://alexjacobson.com

2 1

RE: config.h namespace clashes
by Simon Marlow 01 Mar '04

01 Mar '04

> The inclusion of the GHC made config.h in HsFFI.h makes it extremely > difficult to use autoconf with ones own projects without dealing with > horrible name clashes everywhere. Is there some way this can be fixed? > perhaps pre-generating the HsFFI.h. > > although I don't think it is stated explicitly by the FFI > report, I would > assume that HsFFI.h should not clash with anything that doesn't start > with Hs, hs_ or HS_. > > also, it would be good if config.h could be renamed to something like > ghc-config.h, because otherwise the conflict between the > names can hide > bugs where you get the incorrect config.h or make it tricky to get at > the GHC one if it is actually the one you want. Yes, we should definitely do that. I'm surprised this hasn't caused problems before (or maybe it has?). Cheers, Simon

2 1

RE: Happy Question
by Simon Marlow 01 Mar '04

01 Mar '04

> My hand-written S-expression parser for HScheme is very very slow. > Really quite a bit slower than it ought to be. So I'd like to > switch to > Happy instead. > > The problem is my parser should be monadic, but the monad type is > generalised (basically of the form (context m) => m). Is > there a way of > switching on monadic-ness, so I can have a monadic happyError, but > without giving type signatures to happyThen and happyReturn? Or else > allowing a class context for the type signatures? Not at the moment, but if you can describe exactly what you want to do then I can incorporate the change: either send me a patch for the .hs file generated by Happy, or (better) a patch for Happy itself. Happy's back-end is fairly easy to modify except that it has to deal with several different configurations (%monad, %monad+%lexer, no %monad etc.). Cheers, Simon

1 0

RE: Generics... no Tuples > 2 either...
by Simon Peyton-Jones 01 Mar '04

01 Mar '04

Tuples and Double are there in the HEAD build now -- but we're still thinking about how best to treat Ptr, which is essentially un-marshalable. | -----Original Message----- | From: glasgow-haskell-users-bounces(a)haskell.org [mailto:glasgow-haskell-users- | bounces(a)haskell.org] On Behalf Of MR K P SCHUPKE | Sent: 26 February 2004 18:33 | To: glasgow-haskell-users(a)haskell.org | Subject: Generics... no Tuples > 2 either... | | | Any chance of Data instances for tuples of size | greater than 2... One of the nice things about generics is | you can use them by deriving Data on your datatypes - of | course this doesn't work if you all of a sudden have to | put a load of boiler-plate in just to use tuples... | | Regards, | Keean Schupke | _______________________________________________ | Glasgow-haskell-users mailing list | Glasgow-haskell-users(a)haskell.org | http://www.haskell.org/mailman/listinfo/glasgow-haskell-users

1 0