#2252: Extreme performance degradation on minor code change
----------------------+-----------------------------------------------------
Reporter: simona | Owner:
Type: bug | Status: new
Priority: normal | Milestone:
Component: Compiler | Version: 6.8.2
Severity: normal | Resolution:
Keywords: | Difficulty: Unknown
Testcase: | Architecture: x86_64 (amd64)
Os: Linux |
----------------------+-----------------------------------------------------
Comment (by simonpj):
Thanks. More info: what command line arguments are you using when
compiling your program? In your `-ddump-simpl` I see invocations of
`GHC.Base.$`, which suggests you aren't using -O. That never occurred to
me. Do you get the same performance differences with -O?
Of course, you still should not get these huge differences even without
-O, but I'm interested because it'll give clues about where the problem
is.
The Core code looks ok to me, so I'm still v puzzled about why you would
see big perf differences. I wonder if you can make this easier to
reproduce, and at the same time give more clues. For example, suppose you
trim out most of the functions the `fixpoint` calls, and leave just some
reasonably-expensive thing (some random quadratically-expensive algorithm,
say). Does the effect still happen? Or is it somehow connected to the
code that is called (`meetAsym`, `equalityToInequalities`, etc)? And if
so can you narrow down which code is the culprit, by stubbing out other
parts?
Simon
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/2252#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
Glasgow-haskell-bugs@haskell.org
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs