#2132: Optimise nested comparisons
--------------------------------------+-------------------------------------
Reporter: simonpj | Owner:
Type: bug | Status: patch
Priority: lowest | Milestone: 7.6.1
Component: Compiler | Version: 6.8.2
Resolution: | Keywords:
Os: Unknown/Multiple | Architecture: Unknown/Multiple
Failure: Runtime performance bug | Difficulty: Unknown
Testcase: | Blockedby:
Blocking: | Related:
--------------------------------------+-------------------------------------
Changes (by michalt):
* status: new => patch
Comment:
I think I've fixed the problem with shadowing (I've basically followed the
approach used in CSE). Everything validates.
However, the nofib results are not very encouraging (I've attached the
whole
output of ```nofib-analyse```):
{{{
Program Size Allocs Runtime Elapsed TotalMem
--------------------------------------------------------------------------------
...
--------------------------------------------------------------------------------
Min -0.1% -0.0% -3.5% -3.8% -4.8%
Max +0.4% +0.0% +2.0% +2.0% +0.0%
Geometric Mean -0.0% -0.0% -1.0% -1.0% -0.1%
}}}
(the results are: clean GHC and ordinary nofib run vs GHC with the patch
and
nofib run with EXTRA_HC_OPTS=-fcomparisons).
I've used the debugging output to see how often the optimization is
actually
removing some comparison: in case of GHC compile it managed to remove
around
1800 comparisons, in nofib only around 40. So I'm not sure this is worth
the
whole additional optimization. It seems that either the patch is not going
far
enough or there simply isn't that many opportunities for optimization...?
If it's
the former, then the main idea would probably be to extend the pass to be
able
to deal with arithmetic [1]. If it's the latter then maybe we should just
close
it as wontfix? What do you think?
[1] Handling, for example things like:
{{{
case x <# y + 5 of
case x <# y + 4 of
...
}}}
but this will be quite a bit more complex as we need to take into account
integer overflows, etc.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/2132#comment:25>
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
