[GHC] #9476: Implement late lambda-lifting

[GHC]

#9476: Implement late lambda-lifting
-------------------------------------+-------------------------------------
        Reporter:  simonpj           |                Owner:  sgraf
            Type:  feature request   |               Status:  closed
        Priority:  normal            |            Milestone:  8.8.1
       Component:  Compiler          |              Version:  7.8.2
      Resolution:  fixed             |             Keywords:  LateLamLift
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Runtime           |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #8763 #13286      |  Differential Rev(s):  Phab:D5224
       Wiki Page:  LateLamLift       |
-------------------------------------+-------------------------------------

Comment (by sgraf):

 Now that #15999 has a fix, I redid some of the benchmarks.

 Before I got distracted, we were discussing whether the allocation
 heuristic ("don't allow closure growth") is really effective. Here are the
 results against 8.6 as the baseline:
 https://gist.github.com/sgraf812/d3eb0b2251512a1845adcbf356248c25

 `ll-$date` is disallowing closure growth (notice absence of any regression
 in allocations), whereas `ll-c2-$date` allows closure growth and yields
 marginally better counted instructions at that.

 I expected more regressions in `ll-c2`, but there's really just
 `paraffins` regressing by over 1% wrt. `ll`. Note that `wheel-sieve1`
 still regresses by 28.7% in allocations, while not providing a real
 benefit.

 Let's see what `paraffins` is up to and what runtime measurements suggest.
 Maybe we can get rid of the closure growth heuristic (which would be a
 shame, but that's science).

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9476#comment:75>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler