patch applied (ghc): Improve the treatment of 'seq' (Trac #2273)

[Isaac Dupree]

Is there any particular reason to use let vs. case e.g.
    case e of !x -> ... x ...
? As far as I can tell, the main difference is the layout rule affecting 
following code...

the point being that you don't want those thunks floating around being 
used, and strict-binding makes it obvious to the compiler that you have 
no reason to duplicate that computation no matter how simple it is?  Oh, 
actually the point *is* that we *must not* duplicate it.  And the 
compiler never does, for the let! form, for some reason?

can the idiom work with deeper seq'ing (seqList, deepSeq and such) -- 
can those functions be written robustly too?

-Isaac


Don Stewart wrote:
> So would you say this affects our standard idiom for strictifying things.
> The 
> 
>     let !x = e in ... x ...
> 
> form should be used now, for robustness reasons.
> 
> -- Don
> 
> simonpj:
>> Yes that's right.  I've realised that to be fully robust you should really say
>>
>>         let !x2 = x in ...x2...
>>
>> rather than
>>
>>         x `seq` ...x....
>>
>> With the latter, if you have
>>         let x = e
>> enclosing, then substituting e for x can defeat the seq. And what humans can do, compilers may too!
>>
>> The former doesn't let you do this.
>> Simon
>>
>> | -----Original Message-----
>> | From: Don Stewart [mailto:dons at galois.com]
>> | Sent: 16 May 2008 16:49
>> | To: Simon Peyton-Jones
>> | Cc: cvs-ghc at haskell.org
>> | Subject: Re: patch applied (ghc): Improve the treatment of 'seq' (Trac #2273)
>> |
>> | simonpj:
>> | > Fri May 16 01:51:49 PDT 2008  simonpj at microsoft.com
>> | >   * Improve the treatment of 'seq' (Trac #2273)
>> | >
>> | >   Trac #2273 showed a case in which 'seq' didn't cure the space leak
>> | >   it was supposed to.  This patch does two things to help
>> | >
>> | >   a) It removes a now-redundant special case in Simplify, which
>> | >      switched off the case-binder-swap in the early stages.  This
>> | >      isn't necessary any more because FloatOut has improved since
>> | >      the Simplify code was written.  And switching off the binder-swap
>> | >      is harmful for seq.
>> | >
>> | >   However fix (a) is a bit fragile, so I did (b) too:
>> | >
>> | >   b) Desugar 'seq' specially.  See Note [Desugaring seq (2)] in DsUtils
>> | >      This isn't very robust either, since it's defeated by abstraction,
>> | >      but that's not something GHC can fix; the programmer should use
>> | >      a let! instead.
>> | >
>> | >
>> |
>> | By let! do you mean
>> |
>> |     let !x = ...
>> |
>> | ?
>> |
>> | -- Don
>>
> 
> _______________________________________________
> Cvs-ghc mailing list
> Cvs-ghc at haskell.org
> http://www.haskell.org/mailman/listinfo/cvs-ghc
>