#2754: Strictness analyzer fails on an implementation of foldl
-----------------------------+----------------------------------------------
Reporter: nimnul | Owner:
Type: feature request | Status: closed
Priority: normal | Milestone:
Component: Compiler | Version: 6.8.3
Severity: normal | Resolution: invalid
Keywords: | Difficulty: Unknown
Testcase: | Architecture: x86
Os: Windows |
-----------------------------+----------------------------------------------
Comment (by simonpj):
I think you're suggesting the following: if the compiler sees `foldl`
applied to a strict function, use a {{{foldl'}}} instead. That's a
reasonable suggestion but tricky for the compiler to automate. It would
have to (a) spot the opportunity, (b) generate a specialised version
(that's {{{foldl'}}}), and (c) spot applications of `foldl` to a strict
function, and use the specialised version.
That's a big ask. GHC is not clever enough to do any of these three!
Even (c) is tricky: GHC's current rule matcher doesn't take strictness
into account. Would it matter if the function was only strict in its
first argument? Or its second?
You sensibly suggest that (a), at least, might be programmer driven. But
even then you need a language in which the programm can express the intent
(how to say "generate a specialised version for a bi-strict function
argument"?). And there's still (b) and (c) to worry about.
So while yes, we're very open to suggestions for improvements, I don't see
how to take action on this one. However two methods are already
supported:
* Write the function non-recursively, as Don suggests. Then it can be
inlined at every use site, which generates a very specialised copy!
* Use Template Haskell; this is the direct analogue of what you do with
template functions in C++
Simon
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/2754#comment:11>
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
