[GHC] #14295: tagToEnum# leads to some silly closures

[GHC]

#14295: tagToEnum# leads to some silly closures
-------------------------------------+-------------------------------------
        Reporter:  dfeuer            |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:  8.4.1
       Component:  Compiler          |              Version:  8.2.1
      Resolution:                    |             Keywords:  datacon-tags
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Runtime           |  Unknown/Multiple
  performance bug                    |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 >  When we don't inline, we should still perform a bounds check to avoid
 the thunk, no?

 Good point.  I suppose so... but

 * If we did this we'd better give a different name to the primop that
 can't fail; otherwise the transformation is plain confusing.

 * I'm uncomfortable with transformations that are not readily expressed in
 Core.  E.g. we could have
 {{{
 case x of
   0# -> blah
   DEFAULT -> ...(noFailQuotInt# y x)...
 }}}
 since we know that `x` is non-zero in the default branch.  But then we
 might float the subexpression out, and the claim would no longer hold.  We
 don't have a solid way to prevent this.

   So yes one could do it as a `CorePrep` tactic, but it's delicate; and we
 risk not getting the benefits because we don't implement the follow-on
 transformations.

 * I doubt it's worth doing in practice.  i.e. the same engineering effort
 spent elsewhere might yield more benefit.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14295#comment:11>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler