Enable strict dicts by default at -O2.

In the common case this is a straight performance win
at a compile time cost so we enable it at -O2.

In rare cases it can lead to compile time regressions
because of changed inlining behaviour. Which can very
rarely also affect runtime performance.

Increasing the inlining threshold can help to avoid this
which is documented in the user guide.

In terms of measured results this reduced instructions executed
for nofib by 1%.
However for some cases (e.g. Cabal) enabling this
by default increases compile time by 2-3% so we enable it only
at -O2 where it's clear that a user is willing to trade compile
time for runtime.

Most of the testsuite runs without -O2 so there are few
perf changes.

Increases:
T12545/T18698: We perform more WW work because dicts are now treated strict.
T9198: Also some more work because functions are now subject to W/W

Decreases:
T14697: Compiling empty modules. Probably because of changes inside ghc.
T9203:  I can't reproduce this improvement locally. Might be spurious.

-------------------------
Metric Decrease:
     T12227
     T14697
     T9203
Metric Increase:
     T9198
     T12545
     T18698a
     T18698b
-------------------------

4 files changed, 57 insertions, 21 deletions

diff --git a/testsuite/tests/warnings/should_compile/T16282/T16282.hs b/testsuite/tests/warnings/should_compile/T16282/T16282.hs
index 0f6ab866a0..34abd24e51 100644
--- a/testsuite/tests/warnings/should_compile/T16282/T16282.hs
+++ b/testsuite/tests/warnings/should_compile/T16282/T16282.hs
@@ -1,14 +1,25 @@
-import Data.Map

-

--- If someone improves the specializer so that

--- GHC no longer misses the specialization below,

--- then this test will fail, as it expects a warning

--- to be issued.

--- Another reason this could fail is due to spelling:

--- the test checks for the "specialisation" spelling,

--- but due to changes in how the warnings are listed in DynFalgs.hs

--- the compiler may spit out the "specialization" spelling.

-main :: IO ()

-main = do

-  let m = [] :: [Map String Bool]

-  mapM_ print m

+{-# OPTIONS_GHC -Wall-missed-specialisations -O2 #-}
+
+import T16282A
+
+import Data.Map (Map)
+-- The purpose of this test is simple to trigger
+-- a missed specialization and check if GHC puts
+-- out a warning.
+-- It used to fail to specialize on the show instance
+-- for Data.Map, now that we enable -fdicts-strict by default
+-- the worker for these no longer take a dictionary (having been
+-- WWed). So instead we force it to fail to specialize on myMapM_
+
+-- If someone improves the specializer so that
+-- GHC no longer misses the specialization below,
+-- then this test will fail, as it expects a warning
+-- to be issued.
+-- Another reason this could fail is due to spelling:
+-- the test checks for the "specialisation" spelling,
+-- but due to changes in how the warnings are listed in DynFalgs.hs
+-- the compiler may spit out the "specialization" spelling.
+main :: IO ()
+main = do
+  let m = [] :: [MyMap Double]
+  myMapM_ print m
diff --git a/testsuite/tests/warnings/should_compile/T16282/T16282.stderr b/testsuite/tests/warnings/should_compile/T16282/T16282.stderr
index e9cc798546..b25c4c3563 100644
--- a/testsuite/tests/warnings/should_compile/T16282/T16282.stderr
+++ b/testsuite/tests/warnings/should_compile/T16282/T16282.stderr
@@ -1,10 +1,13 @@
+[1 of 2] Compiling T16282A          ( T16282A.hs, T16282A.o )
+[2 of 2] Compiling Main             ( T16282.hs, T16282.o )
 
 T16282.hs: warning: [-Wall-missed-specialisations]
-    Could not specialise imported function ‘Data.Foldable.$wmapM_’
-      when specialising ‘mapM_’
-    Probable fix: add INLINABLE pragma on ‘Data.Foldable.$wmapM_’
+    Could not specialise imported function ‘T16282A.$wmyMapM_’
+      when specialising ‘myMapM_’
+    Probable fix: add INLINABLE pragma on ‘T16282A.$wmyMapM_’
 
 T16282.hs: warning: [-Wall-missed-specialisations]
-    Could not specialise imported function ‘Data.Map.Internal.$w$cshowsPrec’
-      when specialising ‘Data.Map.Internal.$fShowMap_$cshowsPrec’
-    Probable fix: add INLINABLE pragma on ‘Data.Map.Internal.$w$cshowsPrec’
+    Could not specialise imported function ‘T16282A.$w$cshowsPrec’
+      when specialising ‘T16282A.$fShowMyMap_$cshowsPrec’
+    Probable fix: add INLINABLE pragma on ‘T16282A.$w$cshowsPrec’
+Linking T16282.exe ...
diff --git a/testsuite/tests/warnings/should_compile/T16282/T16282A.hs b/testsuite/tests/warnings/should_compile/T16282/T16282A.hs
new file mode 100644
index 0000000000..3bf6385fd0
--- /dev/null
+++ b/testsuite/tests/warnings/should_compile/T16282/T16282A.hs
@@ -0,0 +1,22 @@
+-- To ensure we miss the mapM specialization we:
+-- * Prevent a unfolding with NOINLINE
+-- * Turn of dicts-strict, to keep a dictionary as argument.
+
+{-# OPTIONS_GHC -fno-dicts-strict #-}
+
+module T16282A where
+
+import Data.Map as M (Map, toList)
+
+newtype MyMap v = MyMap (Map Int v)
+
+instance (Show a) => Show (MyMap a) where
+  showsPrec d (MyMap m)  = showParen (d > 10) $
+    showString "fromList " . shows (M.toList m)
+
+{-# NOINLINE myMapM_ #-}
+myMapM_ :: (Foldable t, Monad m) => (a -> m b) -> t a -> m ()
+myMapM_ f = Prelude.foldr c (return ())
+  -- See Note [List fusion and continuations in 'c']
+  where c x k = f x >> k
+        {-# INLINE c #-}
diff --git a/testsuite/tests/warnings/should_compile/T16282/all.T b/testsuite/tests/warnings/should_compile/T16282/all.T
index dfcdd0562a..c3d2bf462a 100644
--- a/testsuite/tests/warnings/should_compile/T16282/all.T
+++ b/testsuite/tests/warnings/should_compile/T16282/all.T
@@ -1 +1 @@
-test('T16282', normal, compile, ['-O2 -Wall-missed-specialisations'])
\ No newline at end of file
+test('T16282', [check_errmsg('-Wall-missed-specialisations')], multimod_compile, ['T16282', '-O2 -Wall-missed-specialisations'])