[GHC] #6104: Regression: space leak in HEAD vs. 7.4

[GHC]

#6104: Regression: space leak in HEAD vs. 7.4
---------------------------------+------------------------------------------
    Reporter:  simonmar          |       Owner:                              
        Type:  bug               |      Status:  new                         
    Priority:  highest           |   Milestone:  7.6.1                       
   Component:  Compiler          |     Version:  7.5                         
    Keywords:                    |          Os:  Unknown/Multiple            
Architecture:  Unknown/Multiple  |     Failure:  Compile-time performance bug
  Difficulty:  Unknown           |    Testcase:                              
   Blockedby:                    |    Blocking:                              
     Related:                    |  
---------------------------------+------------------------------------------

Comment(by simonmar):

 I've narrowed this down to something that happened around May 8-10.  I'm
 seeing a roughly 10% increase in compile times that was in the May 10
 nightly build compared to the May 8 nightly build (I've unpacked the
 builds to check).

 e.g. compiling Cabal with -O and ghc-20120508:

 {{{
   51,777,170,264 bytes allocated in the heap
   10,213,181,272 bytes copied during GC
      174,111,968 bytes maximum residency (57 sample(s))
        6,843,024 bytes maximum slop
              496 MB total memory in use (0 MB lost due to fragmentation)

                                     Tot time (elapsed)  Avg pause  Max
 pause
   Gen  0     100232 colls,     0 par   25.37s   25.36s     0.0003s
 0.0040s
   Gen  1        57 colls,     0 par   13.46s   13.46s     0.2362s
 0.5846s

   INIT    time    0.00s  (  0.00s elapsed)
   MUT     time   58.57s  ( 61.82s elapsed)
   GC      time   38.82s  ( 38.82s elapsed)
   EXIT    time    0.00s  (  0.00s elapsed)
   Total   time   97.40s  (100.65s elapsed)
 }}}

 and with ghc-20120510:

 {{{
   55,714,073,216 bytes allocated in the heap
   11,751,752,512 bytes copied during GC
      288,316,064 bytes maximum residency (51 sample(s))
        8,407,960 bytes maximum slop
              809 MB total memory in use (0 MB lost due to fragmentation)

                                     Tot time (elapsed)  Avg pause  Max
 pause
   Gen  0     107660 colls,     0 par   28.91s   28.90s     0.0003s
 0.0044s
   Gen  1        51 colls,     0 par   16.56s   16.56s     0.3247s
 1.0106s

   INIT    time    0.00s  (  0.00s elapsed)
   MUT     time   60.13s  ( 63.20s elapsed)
   GC      time   45.47s  ( 45.46s elapsed)
   EXIT    time    0.00s  (  0.00s elapsed)
   Total   time  105.60s  (108.67s elapsed)
 }}}

 we don't have a May 9 build because the change from mtl to transformers
 broke it.

 I've had a look at the patches that went in around that time, and these
 are the only possible culprits as far as I can tell:

 {{{
 commit f7e246b50a33c329e9730296f7a342e03bc3db72
 Merge: 4b089db... ff10612...
 Author: Simon Peyton Jones <simonpj at microsoft.com>
 Date:   Wed May 9 15:50:54 2012 +0100

     Merge branch 'ghc-new-flavor'

 commit 4bbd6baee066b906ed93a0659205ecbffdbd0954
 Author: Simon Peyton Jones <simonpj at microsoft.com>
 Date:   Fri May 4 23:02:46 2012 +0100

     Simplify the SimplCont data type

 commit e08cad76e8a434aca42996f79fc8bb790f291570
 Author: Simon Peyton Jones <simonpj at microsoft.com>
 Date:   Tue May 8 10:25:22 2012 +0100

     Wibbles to 'simplify the SimplCont data type'

 commit 2112f43c466935818a371c53c706608cfa069d01
 Author: Simon Peyton Jones <simonpj at microsoft.com>
 Date:   Wed May 9 11:06:44 2012 +0100

     Be a little less aggressive about inlining (fixes Trac #5623)

 commit 980372f357667c1ba63b28acbf5798826890b7a5
 Author: Simon Peyton Jones <simonpj at microsoft.com>
 Date:   Wed May 9 16:22:49 2012 +0100

     Re-do the "function application discount" (fixes Trac #6048)
 }}}

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/6104#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler