Constrained version of Probability

[chrisdone]

This concludes the test of using a constrained-normal monad wrapper over Probability.

• With RepeatedPD, I see the proper results for 53/54 iterations (fail, succeed)
• With AndGateMarkov, I see a significant speed-up with
  • The constrained monad, though I'm not sure why exactly it introduces better speed. It goes from 30s to 20s. 25G to 16G allocations.
  • Adding O.norm on the right-hand side of >>= seems to shave off a couple seconds off runtime. 20s to 17s, retains the 16G allocations.

An interesting direction could be to put a Set (p,a) instead of [(p,a)], but it'd probably only save a tiny amount (by avoiding the Data.Map.toList/Data.Map.fromList on every bind).

Note that these results are in GHCi, because we're interested in complexity more than GHC's optimizations. Here's compiled with -O:

Original

     325,369,072 bytes allocated in the heap
      33,202,640 bytes copied during GC
         233,632 bytes maximum residency (17 sample(s))
          30,328 bytes maximum slop
               0 MB total memory in use (0 MB lost due to fragmentation)

                                     Tot time (elapsed)  Avg pause  Max pause
  Gen  0       297 colls,     0 par    0.032s   0.032s     0.0001s    0.0002s
  Gen  1        17 colls,     0 par    0.004s   0.004s     0.0003s    0.0005s

  INIT    time    0.000s  (  0.000s elapsed)
  MUT     time    0.077s  (  0.077s elapsed)
  GC      time    0.036s  (  0.036s elapsed)
  EXIT    time    0.000s  (  0.000s elapsed)
  Total   time    0.113s  (  0.113s elapsed)

  %GC     time       0.0%  (0.0% elapsed)

  Alloc rate    4,237,404,076 bytes per MUT second

  Productivity  67.9% of total user, 67.9% of total elapsed

Constrained normal

     158,513,840 bytes allocated in the heap
       5,184,864 bytes copied during GC
         119,280 bytes maximum residency (3 sample(s))
          29,320 bytes maximum slop
               0 MB total memory in use (0 MB lost due to fragmentation)

                                     Tot time (elapsed)  Avg pause  Max pause
  Gen  0       150 colls,     0 par    0.007s   0.007s     0.0000s    0.0001s
  Gen  1         3 colls,     0 par    0.000s   0.000s     0.0002s    0.0002s

  INIT    time    0.000s  (  0.000s elapsed)
  MUT     time    0.044s  (  0.044s elapsed)
  GC      time    0.007s  (  0.007s elapsed)
  EXIT    time    0.000s  (  0.000s elapsed)
  Total   time    0.051s  (  0.051s elapsed)

  %GC     time       0.0%  (0.0% elapsed)

  Alloc rate    3,607,835,032 bytes per MUT second

  Productivity  85.3% of total user, 85.4% of total elapsed

norm on rhs

     151,945,968 bytes allocated in the heap
       5,815,064 bytes copied during GC
         132,128 bytes maximum residency (3 sample(s))
          29,320 bytes maximum slop
               0 MB total memory in use (0 MB lost due to fragmentation)

                                     Tot time (elapsed)  Avg pause  Max pause
  Gen  0       143 colls,     0 par    0.007s   0.008s     0.0001s    0.0002s
  Gen  1         3 colls,     0 par    0.000s   0.000s     0.0002s    0.0002s

  INIT    time    0.000s  (  0.000s elapsed)
  MUT     time    0.041s  (  0.045s elapsed)
  GC      time    0.008s  (  0.008s elapsed)
  EXIT    time    0.000s  (  0.000s elapsed)
  Total   time    0.049s  (  0.054s elapsed)

  %GC     time       0.0%  (0.0% elapsed)

  Alloc rate    3,706,812,910 bytes per MUT second

  Productivity  84.5% of total user, 84.1% of total elapsed