#14688: Note [Lone variables] leads to missing a case-of-case opportunity -------------------------------------+------------------------------------- Reporter: mpickering | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.2 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: | -------------------------------------+------------------------------------- Comment (by simonpj): Crumbs, you are right. Looking at the `Note [Lone variables]` I see {{{ Fundamentally such contexts should not encourage inlining because the context can ``see'' the unfolding of the variable (e.g. case or a RULE) so there's no gain. If the thing is bound to a value. ... So the non-inlining of lone_variables should only apply if the unfolding is regarded as cheap; because that is when exprIsConApp_maybe looks through the unfolding. Hence the "&& is_wf" in the InlineRule branch. }}} But actually `exprIsConApp_maybe` uses `expandUnfolding_maybe` to look through the unfolding, which in turn uses the `uf_expandable` field of the unfolding, not the `uf_is_work_free` field. Conclusion: the test in `interesting_call` in `tryUnfolding` (at `Note [Lone variable]`) should not test `is_wf` but rather `is_exp`. That's an extremely simple, local change. Can you try it? I think it'll fix your problem. If it does, could you do a nofib comparison, before and after? Thanks. This looks promising. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14688#comment:1 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler