[GHC] #15503: interpreter: sequence_ (replicate 100000000 (return ())) gobbles up memory

[GHC]

#15503: interpreter: sequence_ (replicate 100000000 (return ()))  gobbles up memory
-------------------------------------+-------------------------------------
        Reporter:  int-e             |                Owner:  osa1
            Type:  bug               |               Status:  new
        Priority:  high              |            Milestone:  8.8.1
       Component:  GHCi              |              Version:  8.5
      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 osa1):

 This regression was introduced between 7.10 and 8.0, and reproducible with
 8.4 and HEAD. I compared `+RTS -s` outputs of 7.10 and 8.4.

 7.10:

 {{{
 $ ./ghci +RTS -s
 GHCi, version 7.10.3: http://www.haskell.org/ghc/  :? for help
 λ:2> sequence_ (replicate 100000000 (return ()))
 λ:3>
 Leaving GHCi.
   12,844,310,920 bytes allocated in the heap
       26,753,232 bytes copied during GC
        7,744,552 bytes maximum residency (6 sample(s))
          138,736 bytes maximum slop
               18 MB total memory in use (0 MB lost due to fragmentation)

                                      Tot time (elapsed)  Avg pause  Max
 pause
   Gen  0      1806 colls,     0 par    0.155s   0.158s     0.0001s
 0.0055s
   Gen  1         6 colls,     0 par    0.096s   0.126s     0.0211s
 0.0415s

   TASKS: 4 (1 bound, 3 peak workers (3 total), using -N1)

   SPARKS: 0 (0 converted, 0 overflowed, 0 dud, 0 GC'd, 0 fizzled)

   INIT    time    0.000s  (  0.002s elapsed)
   MUT     time    4.065s  (  6.690s elapsed)
   GC      time    0.251s  (  0.285s elapsed)
   EXIT    time    0.020s  (  0.023s elapsed)
   Total   time    4.365s  (  7.000s elapsed)

   Alloc rate    3,159,574,299 bytes per MUT second

   Productivity  94.3% of total user, 58.8% of total elapsed

 gc_alloc_block_sync: 0
 whitehole_spin: 0
 gen[0].sync: 0
 gen[1].sync: 0
 }}}

 8.4:

 {{{
 $ ghci +RTS -s
 GHCi, version 8.4.3: http://www.haskell.org/ghc/  :? for help
 Loaded GHCi configuration from /home/omer/rcbackup/.ghci
 λ:1> sequence_ (replicate 100000000 (return ()))
 λ:2>
 Leaving GHCi.
   12,869,399,120 bytes allocated in the heap
   13,233,009,504 bytes copied during GC
    4,007,601,416 bytes maximum residency (14 sample(s))
       15,805,176 bytes maximum slop
             8652 MB total memory in use (0 MB lost due to fragmentation)

                                      Tot time (elapsed)  Avg pause  Max
 pause
   Gen  0       209 colls,     0 par    3.455s   3.486s     0.0167s
 0.4864s
   Gen  1        14 colls,     0 par    7.993s  10.629s     0.7592s
 4.4344s

   TASKS: 4 (1 bound, 3 peak workers (3 total), using -N1)

   SPARKS: 0 (0 converted, 0 overflowed, 0 dud, 0 GC'd, 0 fizzled)

   INIT    time    0.001s  (  0.001s elapsed)
   MUT     time    4.252s  (  6.710s elapsed)
   GC      time   11.448s  ( 14.115s elapsed)
   EXIT    time    0.000s  (  0.005s elapsed)
   Total   time   15.701s  ( 20.831s elapsed)

   Alloc rate    3,026,874,614 bytes per MUT second

   Productivity  27.1% of total user, 32.2% of total elapsed

 gc_alloc_block_sync: 0
 whitehole_spin: 0
 gen[0].sync: 0
 gen[1].sync: 0
 }}}

 Observations:

 - Productivity is 94% vs. 27%
 - Almost same amount of allocation in GHCs (very minor difference)
 - 500x max. residency in GHC 8.4 (7,744,552 vs. 4,007,601,416)

 So I think for some reason GHC 8.4 keeps retains more things which causes
 more GC cycles and longer pauses.

 I compared the byte codes generated by both versions and they're identical
 so I suspect this is either a problem with the interpreter code or some
 library function.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/15503#comment:4>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler