[GHC] #10400: Run time increases by 40% in fractal plotter core loop

[GHC]

#10400: Run time increases by 40% in fractal plotter core loop
-------------------------------------+-------------------------------------
        Reporter:  saffroy           |                   Owner:
            Type:  bug               |                  Status:  new
        Priority:  normal            |               Milestone:
       Component:  Compiler          |                 Version:  7.8.4
      Resolution:                    |                Keywords:
Operating System:  Linux             |            Architecture:  x86_64
 Type of failure:  Runtime           |  (amd64)
  performance bug                    |               Test Case:
      Blocked By:                    |                Blocking:
 Related Tickets:                    |  Differential Revisions:
-------------------------------------+-------------------------------------

Comment (by rwbarton):

 Almost all the time is spent inside `iterations`, which is a
 20-instruction long loop in both the 7.6 and 7.8 versions. The two loops
 are very similar, but some detail causes the 7.8 loop to run about 50%
 slower, probably some pipeline stall issue. It looks hard to even track
 down the exact issue, let alone deal with it in the code generator in a
 general way.

 I note that the LLVM backend produces a smaller loop which is a bit faster
 than even the 7.6 loop. The LLVM backend is usually good at optimizing
 this kind of non-allocating code.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10400#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler