#4081: Strict constructor fields inspected in loop
---------------------------------+------------------------------------------
Reporter: rl | Owner:
Type: bug | Status: new
Priority: normal | Milestone:
Component: Compiler | Version: 6.13
Keywords: | Difficulty:
Os: Unknown/Multiple | Testcase:
Architecture: Unknown/Multiple | Failure: Runtime performance bug
---------------------------------+------------------------------------------
Comment(by simonpj):
We could arrange that w/w transformed the case to add
{{{
case m_ajy of I# m' -> ...
}}}
just inside the T.S match. As things stand today, this case would be
eliminated by the "drop redundant seqs" transformation. But perhaps that
transformation is too eager. It drops a case that looks like
{{{
case x of { C y1..yn -> body }
}}}
where the y1..yn are not mentioned in body, and x is known to be
evaluated. See
{{{
-- 2. Eliminate the case if scrutinee is evaluated
}}}
in `Simplify`.
This is correct, but it might be better to 'retain' the case if there are
any binders y1..yn, at least until the end of simplification (`CorePrep`
perhaps). Then we can drop it. That would deal with this particular case
at least.
Another opportunity that we do not exploit right now is strictness in a
free variable. Consier
{{{
f x = letrec g y = if x>0 then ..g y'....
else ...g y''...
in (g y1, g y2)
}}}
If we lambda-lifted, we'd evaluate 'x' just once; as things stand we do it
each time. A modification to the w/w transform might handle this.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/4081#comment:1>
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
