From 51367fba96fd863ce7d3e2571bd22366b47b900a Mon Sep 17 00:00:00 2001 From: "simonpj@microsoft.com" Date: Wed, 27 Oct 2010 18:41:07 +0000 Subject: [PATCH] Don't worker-wrapper INLINABLE things See Note [Don't w/w INLINABLE things] in WorkWrap This fixes a bug that Milan found. --- compiler/stranal/WorkWrap.lhs | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) diff --git a/compiler/stranal/WorkWrap.lhs b/compiler/stranal/WorkWrap.lhs index d329b5a..a8f110c 100644 --- a/compiler/stranal/WorkWrap.lhs +++ b/compiler/stranal/WorkWrap.lhs @@ -173,6 +173,20 @@ I measured it on nofib, it didn't make much difference; just a few percent improved allocation on one benchmark (bspt/Euclid.space). But nothing got worse. +Note [Don't w/w INLINABLE things] +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +If we have + {-# INLINABLE f #-} + f x y = .... +then in principle we might get a more efficient loop by w/w'ing f. +But that would make a new unfolding which would overwrite the old +one. So we leave INLINABLE things alone too. + +This is a slight infelicity really, because it means that adding +an INLINABLE pragma could make a program a bit less efficient, +because you lose the worker/wrapper stuff. But I don't see a way +to avoid that. + Note [Wrapper activation] ~~~~~~~~~~~~~~~~~~~~~~~~~ When should the wrapper inlining be active? It must not be active @@ -260,6 +274,7 @@ tryWW is_rec fn_id rhs checkSize :: Id -> CoreExpr -> UniqSM [(Id,CoreExpr)] -> UniqSM [(Id,CoreExpr)] -- See Note [Don't w/w inline things (a) and (b)] + -- and Note [Don't w/w INLINABLE things] checkSize fn_id rhs thing_inside | isStableUnfolding unfolding -- For DFuns and INLINE things, leave their = return [ (fn_id, rhs) ] -- unfolding unchanged; but still attach @@ -271,7 +286,8 @@ checkSize fn_id rhs thing_inside | otherwise = thing_inside where - unfolding = idUnfolding fn_id + unfolding = realIdUnfolding fn_id -- We want to see the unfolding + -- for loop breakers! inline_rule = mkInlineUnfolding Nothing rhs --------------------- -- 1.7.10.4