[GHC] #14929: Program compiled with -O2 exhibits much worse performance

GHC ghc-devs at haskell.org
Tue Jul 10 22:11:08 UTC 2018


#14929: Program compiled with -O2 exhibits much worse performance
-------------------------------------+-------------------------------------
        Reporter:  mpickering        |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  high              |            Milestone:
       Component:  Compiler          |              Version:  8.2.2
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Runtime           |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 I have not looked at the code, but full laziness is definitely capable of
 increasing space usage.  Consider even
 {{{
 f xs = sum [x+n | n <- [1..], x <- xs]
 }}}
 Full laziness will turn this into
 {{{
 ns = [1..]
 f xs = sum [x+n | n<-ns, x<-xs]
 }}}
 which will retain a top-level CAF whose length is the longest list ever
 passed to `f`.  In this case it is probably better to re-generate the list
 `[1..]` on every call, but it's not so clear if it is `map expensive
 [1..]`.

 The OP doesn't say if the same effect happens with -O.  Is there something
 -O2 specific going on, I wonder?

 Regardless

 * Extracting the test case and uploading it here with repro instructions
 would be good
 * More specific insight into exactly what is happening would be good

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14929#comment:4>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler


More information about the ghc-tickets mailing list