import OccurAnal
import CoreSubst hiding( substTy )
import CoreFVs ( exprFreeVars )
+import CoreArity ( manifestArity )
import CoreUtils
import Id
import DataCon
expr 0 -- Arity of unfolding doesn't matter
(UnfWhen unSaturatedOk boringCxtOk)
-mkInlineRule :: Bool -> CoreExpr -> Arity -> Unfolding
-mkInlineRule unsat_ok expr arity
+mkInlineRule :: CoreExpr -> Maybe Arity -> Unfolding
+mkInlineRule expr mb_arity
= mkCoreUnfolding True InlineRule -- Note [Top-level flag on inline rules]
expr' arity
(UnfWhen unsat_ok boring_ok)
where
expr' = simpleOptExpr expr
+ (unsat_ok, arity) = case mb_arity of
+ Nothing -> (unSaturatedOk, manifestArity expr')
+ Just ar -> (needSaturated, ar)
+
boring_ok = case calcUnfoldingGuidance True -- Treat as cheap
False -- But not bottoming
(arity+1) expr' of
| uncondInline n_val_bndrs (iBox size)
, expr_is_cheap
-> UnfWhen unSaturatedOk boringCxtOk -- Note [INLINE for small functions]
-
| top_bot -- See Note [Do not inline top-level bottoming functions]
-> UnfNever
conSize :: DataCon -> Int -> ExprSize
conSize dc n_val_args
- | n_val_args == 0 = SizeIs (_ILIT(0)) emptyBag (_ILIT(1)) -- Like variables
+ | n_val_args == 0 = SizeIs (_ILIT(0)) emptyBag (_ILIT(1)) -- Like variables
+
+-- See Note [Constructor size]
| isUnboxedTupleCon dc = SizeIs (_ILIT(0)) emptyBag (iUnbox n_val_args +# _ILIT(1))
- | otherwise = SizeIs (_ILIT(1)) emptyBag (iUnbox n_val_args +# _ILIT(1))
- -- Treat a constructors application as size 1, regardless of how
- -- many arguments it has; we are keen to expose them
- -- (and we charge separately for their args). We can't treat
- -- them as size zero, else we find that (Just x) has size 0,
- -- which is the same as a lone variable; and hence 'v' will
- -- always be replaced by (Just x), where v is bound to Just x.
- --
- -- However, unboxed tuples count as size zero
- -- I found occasions where we had
- -- f x y z = case op# x y z of { s -> (# s, () #) }
- -- and f wasn't getting inlined
+-- See Note [Unboxed tuple result discount]
+-- | isUnboxedTupleCon dc = SizeIs (_ILIT(0)) emptyBag (_ILIT(0))
+
+-- See Note [Constructor size]
+ | otherwise = SizeIs (_ILIT(1)) emptyBag (iUnbox n_val_args +# _ILIT(1))
+\end{code}
+
+Note [Constructor size]
+~~~~~~~~~~~~~~~~~~~~~~~
+Treat a constructors application as size 1, regardless of how many
+arguments it has; we are keen to expose them (and we charge separately
+for their args). We can't treat them as size zero, else we find that
+(Just x) has size 0, which is the same as a lone variable; and hence
+'v' will always be replaced by (Just x), where v is bound to Just x.
+
+However, unboxed tuples count as size zero. I found occasions where we had
+ f x y z = case op# x y z of { s -> (# s, () #) }
+and f wasn't getting inlined.
+
+Note [Unboxed tuple result discount]
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+I tried giving unboxed tuples a *result discount* of zero (see the
+commented-out line). Why? When returned as a result they do not
+allocate, so maybe we don't want to charge so much for them If you
+have a non-zero discount here, we find that workers often get inlined
+back into wrappers, because it look like
+ f x = case $wf x of (# a,b #) -> (a,b)
+and we are keener because of the case. However while this change
+shrank binary sizes by 0.5% it also made spectral/boyer allocate 5%
+more. All other changes were very small. So it's not a big deal but I
+didn't adopt the idea.
+
+\begin{code}
primOpSize :: PrimOp -> Int -> ExprSize
primOpSize op n_val_args
| not (primOpIsDupable op) = sizeN opt_UF_DearOp
\begin{code}
couldBeSmallEnoughToInline :: Int -> CoreExpr -> Bool
couldBeSmallEnoughToInline threshold rhs
- = case calcUnfoldingGuidance False False threshold rhs of
- (_, UnfNever) -> False
- _ -> True
+ = case sizeExpr (iUnbox threshold) [] body of
+ TooBig -> False
+ _ -> True
+ where
+ (_, body) = collectBinders rhs
----------------
smallEnoughToInline :: Unfolding -> Bool
NoUnfolding -> Nothing ;
OtherCon _ -> Nothing ;
DFunUnfolding {} -> Nothing ; -- Never unfold a DFun
- CoreUnfolding { uf_tmpl = unf_template, uf_is_top = is_top, uf_is_value = is_value,
+ CoreUnfolding { uf_tmpl = unf_template, uf_is_top = is_top,
uf_is_cheap = is_cheap, uf_arity = uf_arity, uf_guidance = guidance } ->
-- uf_arity will typically be equal to (idArity id),
-- but may be less for InlineRules
interesting_saturated_call
= case cont_info of
- BoringCtxt -> not is_top && uf_arity > 0 -- Note [Nested functions]
- CaseCtxt -> not (lone_variable && is_value) -- Note [Lone variables]
- ArgCtxt {} -> uf_arity > 0 -- Note [Inlining in ArgCtxt]
- ValAppCtxt -> True -- Note [Cast then apply]
+ BoringCtxt -> not is_top && uf_arity > 0 -- Note [Nested functions]
+ CaseCtxt -> not (lone_variable && is_cheap) -- Note [Lone variables]
+ ArgCtxt {} -> uf_arity > 0 -- Note [Inlining in ArgCtxt]
+ ValAppCtxt -> True -- Note [Cast then apply]
(yes_or_no, extra_doc)
= case guidance of
text "uf arity" <+> ppr uf_arity,
text "interesting continuation" <+> ppr cont_info,
text "some_benefit" <+> ppr some_benefit,
- text "is value:" <+> ppr is_value,
text "is cheap:" <+> ppr is_cheap,
text "guidance" <+> ppr guidance,
extra_doc,
for the RHS of a 'let', we only profit from the inlining if we get a
CONLIKE thing (modulo lets).
-Note [Lone variables]
-~~~~~~~~~~~~~~~~~~~~~
+Note [Lone variables] See also Note [Interaction of exprIsCheap and lone variables]
+~~~~~~~~~~~~~~~~~~~~~ which appears below
The "lone-variable" case is important. I spent ages messing about
with unsatisfactory varaints, but this is nice. The idea is that if a
variable appears all alone
as scrutinee of a case CaseCtxt
as arg of a fn ArgCtxt
AND
- it is bound to a value
+ it is bound to a cheap expression
then we should not inline it (unless there is some other reason,
e.g. is is the sole occurrence). That is what is happening at
There's no advantage in inlining f here, and perhaps
a significant disadvantage. Hence some_val_args in the Stop case
+Note [Interaction of exprIsCheap and lone variables]
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+The lone-variable test says "don't inline if a case expression
+scrutines a lone variable whose unfolding is cheap". It's very
+important that, under these circumstances, exprIsConApp_maybe
+can spot a constructor application. So, for example, we don't
+consider
+ let x = e in (x,x)
+to be cheap, and that's good because exprIsConApp_maybe doesn't
+think that expression is a constructor application.
+
+I used to test is_value rather than is_cheap, which was utterly
+wrong, because the above expression responds True to exprIsHNF.
+
+This kind of thing can occur if you have
+
+ {-# INLINE foo #-}
+ foo = let x = e in (x,x)
+
+which Roman did.
+
\begin{code}
computeDiscount :: Int -> [Int] -> Int -> [ArgSummary] -> CallCtxt -> Int
computeDiscount n_vals_wanted arg_discounts res_discount arg_infos cont_info