Re: [GHC] #13523: Be more explicit in generated case alternatives

4 Apr 2017

      #13523: Be more explicit in generated case alternatives
-------------------------------------+-------------------------------------
        Reporter:  bgamari           |                Owner:  (none)
            Type:  feature request   |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.0.1
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------
Description changed by bgamari:

Old description:
...
This came up during the GHC call while discussing #13397 and I thought it
would be good to record.
Often in the compiler we end up with a case analysis with a number of
alternatives and a `DEFAULT` case in place of the remaining alternative,
{{{#!hs
case x of
  True -> a
  DEFAULT -> b
}}}
Arguably we should instead generate,
{{{#!hs
case x of
  True -> a
  False -> b
}}}
instead as this may allow the code generator may be able to emit better
code when it knows the finite bounds of the switch (e.g. using a jump
table).
New description:

 This came up during the GHC call while discussing #13397 and I thought it
 would be good to record.

 Often in the compiler we end up with a case analysis with a number of
 alternatives and a `DEFAULT` case in place of the remaining alternative,

 {{{#!hs
 data T = Con1 | Con2 | Con3

 ...
   case x of
     Con1 -> a
     Con2 -> b
     DEFAULT -> c
 }}}

 Arguably we should instead generate,

 {{{#!hs
 case x of
   Con1 -> a
   Con2 -> b
   Con3 -> c
 }}}
 instead as this may allow the code generator may be able to emit better
 code when it knows the finite bounds of the switch (e.g. using a jump
 table).

--

--
Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/13523#comment:2
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler