Marge Bot pushed to branch master at Glasgow Haskell Compiler / GHC Commits: 2b4f463c by Simon Peyton Jones at 2026-02-02T17:32:32+00:00 Remove exprIsCheap from doFloatFromRhs See #26854 and Note [Float when expandable] This patch simplifies the code, by removing an extra unnecessary test. - - - - - 1 changed file: - compiler/GHC/Core/Opt/Simplify/Env.hs Changes: ===================================== compiler/GHC/Core/Opt/Simplify/Env.hs ===================================== @@ -815,8 +815,9 @@ doFloatFromRhs fe lvl rec strict_bind (SimplFloats { sfLetFloats = LetFloats fs && want_to_float && can_float where - want_to_float = isTopLevel lvl || exprIsCheap rhs || exprIsExpandable rhs - -- See Note [Float when cheap or expandable] + want_to_float = isTopLevel lvl || exprIsExpandable rhs + -- See Note [Float when expandable] + can_float = case ff of FltLifted -> True FltOkSpec -> isNotTopLevel lvl && isNonRec rec @@ -828,16 +829,29 @@ doFloatFromRhs fe lvl rec strict_bind (SimplFloats { sfLetFloats = LetFloats fs floatEnabled lvl FloatNestedOnly = not (isTopLevel lvl) floatEnabled _ FloatEnabled = True -{- -Note [Float when cheap or expandable] -~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -We want to float a let from a let if the residual RHS is - a) cheap, such as (\x. blah) - b) expandable, such as (f b) if f is CONLIKE -But there are - - cheap things that are not expandable (eg \x. expensive) - - expandable things that are not cheap (eg (f b) where b is CONLIKE) -so we must take the 'or' of the two. +{- Note [Float when expandable] +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +We float when that would leave us with a RHS that is /expandable/ (#26854). +The whole purpose of the local-let-floating is to end up with + let x = rhs +where rhs has an expandable unfolding, so that `exprIsConApp_maybe` will look +inside `x`. + +Historical aside: + + In the long-distant (2011) past, we made `want_to_float` true if + EITHER exprIsExpandable OR exprIsCheap + But that seems wrong: there is no point in floating for expressions that are + cheap but not expandable: + * It costs more to test + * It gives no benefit + * Very few expressions are cheap but not expandable (e.g. error calls) + We justified this OR by saying that there may be: + - cheap things that are not expandable (eg \x. expensive) + - expandable things that are not cheap (eg (f b) where b is CONLIKE) + But that's not true: (\x.expensive) is certainly expandable. + +End of historial aside -} emptyLetFloats :: LetFloats View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/2b4f463c63bf8ff08e3ae49e4ad7b65f... -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/2b4f463c63bf8ff08e3ae49e4ad7b65f... You're receiving this email because of your account on gitlab.haskell.org.