#2353: GHC inliner doesn't
----------------------+-----------------------------------------------------
Reporter: guest | Owner:
Type: bug | Status: new
Priority: normal | Milestone:
Component: Compiler | Version: 6.9
Severity: normal | Resolution:
Keywords: | Difficulty: Unknown
Testcase: | Architecture: Unknown
Os: Unknown |
----------------------+-----------------------------------------------------
Comment (by guest):
If it is deliberate, why does exporting foo make a difference?
That's rather unintuitive.
I'm not sure I buy that argument that not inlining foo is the right
decision here.
Look at the difference in code without and with inlining. I think the
inlined version looks better.
{{{
-- No inline
Foo.bar =
\ (@ a_a6v) ($dC_a6F :: Foo.C a_a6v) ->
let {
foo1_s7R [Just L] :: (GHC.Base.Int -> GHC.Base.Int) -> a_a6v ->
a_a6v
[Str: DmdType]
foo1_s7R = Foo.foo @ a_a6v $dC_a6F } in
\ (f_a5Y :: GHC.Base.Int -> GHC.Base.Int) ->
let {
f1_s7T [Just L] :: a_a6v -> a_a6v
[Str: DmdType]
f1_s7T = foo1_s7R f_a5Y } in
\ (x_a6R :: a_a6v) -> f1_s7T (f1_s7T x_a6R)
-- Inline (because foo was not exported)
Foo.bar =
\ (@ a_a6v) ($dC_a6F :: Foo.C a_a6v) ->
let {
from1_s7I [Just L] :: GHC.Base.Int -> a_a6v
[Str: DmdType]
from1_s7I = Foo.from @ a_a6v $dC_a6F } in
let {
to1_s7G [Just L] :: a_a6v -> GHC.Base.Int
[Str: DmdType]
to1_s7G = Foo.to @ a_a6v $dC_a6F } in
\ (f_a5Y :: GHC.Base.Int -> GHC.Base.Int) (eta_s6T :: a_a6v) ->
from1_s7I (f_a5Y (to1_s7G (from1_s7I (f_a5Y (to1_s7G eta_s6T)))))
}}}
Also, if INLINE doesn't really inline, then I'd like a REALLYINLINE
pragma. Sometimes I know what I'm doing, and I'd like the compiler to
obey.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/2353#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
