#5237: Inefficient code generated for x^2
--------------------------------------+-------------------------------------
Reporter: scpmw | Owner:
Type: bug | Status: new
Priority: normal | Milestone: 7.4.1
Component: libraries/base | Version: 7.0.3
Resolution: | Keywords:
Testcase: | Blockedby:
Difficulty: | Os: Linux
Blocking: | Architecture: x86_64 (amd64)
Failure: Runtime performance bug |
--------------------------------------+-------------------------------------
Comment(by simonpj):
Actually I realise that we already have a good mechanism for
prioritisation: namely the simplifer's "phases".
* If you want a RULE to match on a function, you had better make sure the
function doesn't inline too early. Typically use a NOINLINE pragma:
{{{
{-# RULE forall x. f [x] = blah #-}
{-# NOINLINE [1] f #-}
f x = ...blah2...
}}}
* The simplifier works in phases, currently phase 2, then 1, then 0. The
RULE is always active. The NOINLINE stops f inlining until phase 1. So
the rule gets a chance to fire in the earlier phase 2.
* You can attach phases to rules too, which say in which phases they are
active. Ditto SPECIALISE pragmas
* GHC automatically generates specialisations for overloaded functions.
The consequent rules are active precisely when the original function's
INLINE pragma (if any) is active. So if you say `NONLINE [1] f`, then any
automatically generated specialisations of f will be active only in phase
1 and later.
Does that give you enough to go on?
Simon
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/5237#comment:15>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
