PoC: INLINE parser combinators

[phadej]

On my machine, trying dhall-kubernetes/examples/deployment.dhall the execution time drops significantly:

-15,31s user 0,30s system 47% cpu 33,174 total
+ 8,45s user 0,19s system 72% cpu 11,986 total

The flip side of the coin, is that Token module tortures GHC with all stuff wanting to INLINE. Yet SPJ said about a week ago

I've had a fix pending in my tree for months, but I keep getting pre-empted. Your new ticket will help incentivise me.

in https://ghc.haskell.org/trac/ghc/ticket/15630, so that issue might go away.

---

I'm not sure inlining on which (set of) functor/applicative/alternative/monad combinator(s) contributes the most, but seems that here as well, inlining composition combinators pays off in run time.

---

I hope that @f-f or someone else can continue from here. Cheers.

[Gabriella439]

@phadej: Thank you! 🙂

The test failure is weird and seems like it should not be related to this change at all. I'm digging into it

[Gabriella439]

@phadej: Oh, I know what the problem is. You also need to implement fail, too

[Gabriella439]

Thanks again!

[Gabriella439]

Before I merge, is there any reason not to merge this? This was titled a "Proof of concept" but it looks fine to me. Is it just because there are other things which could be inlined?

[phadej]

@Gabriel439 I'm not entirely sure this is optimization.

It seems that this actually doesn't really affect non-profiled build.

And with profiled, both master and this branch have very similar

COST CENTRE            MODULE                     SRC                                                 %time %alloc

noted                  Dhall.Parser.Expression    src/Dhall/Parser/Expression.hs:(31,1)-(38,58)        37.1   18.1
reserved               Dhall.Parser.Token         src/Dhall/Parser/Token.hs:402:1-56                   12.0    9.6
simpleLabel            Dhall.Parser.Token         src/Dhall/Parser/Token.hs:(160,1)-(169,76)           11.6   19.0
takeWhile              Dhall.Parser.Combinators   src/Dhall/Parser/Combinators.hs:134:1-75              5.0    6.4
splitAt                Data.Text                  Data/Text.hs:(1339,1)-(1343,63)                       4.8   15.2
makeOperatorExpression Dhall.Parser.Expression    src/Dhall/Parser/Expression.hs:(169,1)-(173,40)       2.1    1.2
naturalLiteral         Dhall.Parser.Token         src/Dhall/Parser/Token.hs:(68,1)-(70,51)              2.1    3.0
pathComponent          Dhall.Parser.Token         src/Dhall/Parser/Token.hs:(237,1)-(241,34)            1.9    2.8
doubleLiteral          Dhall.Parser.Token         src/Dhall/Parser/Token.hs:(53,1)-(57,42)              1.6    2.2
blockComment           Dhall.Parser.Token         src/Dhall/Parser/Token.hs:(131,1)-(133,24)            1.1    0.5
whitespaceChunk        Dhall.Parser.Token         src/Dhall/Parser/Token.hs:(84,1)-(92,52)              1.1    1.4
localRaw               Dhall.Parser.Expression    src/Dhall/Parser/Expression.hs:(729,1)-(758,36)       1.0    1.4
integerLiteral         Dhall.Parser.Token         src/Dhall/Parser/Token.hs:(60,1)-(65,43)              0.9    1.9
toList                 Data.HashMap.Strict.InsOrd src/Data/HashMap/Strict/InsOrd.hs:(580,1)-(584,22)    0.6    1.7
primitiveExpression    Dhall.Parser.Expression    src/Dhall/Parser/Expression.hs:(256,1)-(465,16)       0.6    1.4

(I'm using GHC-8.2.2 btw, will try 8.4.3 next)

[phadej]

The last commit is something one might want to profile in separation. On my machine it halved the time taken by noted. (EDIT: I'm quite sure about that).

[Gabriella439]

Then I think we should go ahead and merge this whenever you're ready. It can't hurt and it will make it easier to profile if we eliminate profiling false positives caused by the absence of inlining

[phadej]

Ok, I run out of low hanging fruits:

master

  INIT    time    0.000s  (  0.000s elapsed)
  MUT     time    6.256s  (  9.445s elapsed)
  GC      time    1.998s  (  1.997s elapsed)
  EXIT    time    0.003s  (  0.003s elapsed)
  Total   time    8.257s  ( 11.445s elapsed)

  %GC     time      24.2%  (17.4% elapsed)

prof

  INIT    time    0.000s  (  0.000s elapsed)
  MUT     time    4.099s  (  6.744s elapsed)
  GC      time    2.077s  (  2.076s elapsed)
  EXIT    time    0.003s  (  0.003s elapsed)
  Total   time    6.180s  (  8.824s elapsed)

  %GC     time      33.6%  (23.5% elapsed)

There is 20-25% speedup now.

[Gabriella439]

Alright, then I'll go ahead and merge this