Join points

This major patch implements Join Points, as described in
https://ghc.haskell.org/trac/ghc/wiki/SequentCore.  You have
to read that page, and especially the paper it links to, to
understand what's going on; but it is very cool.

It's Luke Maurer's work, but done in close collaboration with Simon PJ.

This Phab is a squash-merge of wip/join-points branch of
http://github.com/lukemaurer/ghc. There are many, many interdependent
changes.

Reviewers: goldfire, mpickering, bgamari, simonmar, dfeuer, austin

Subscribers: simonpj, dfeuer, mpickering, Mikolaj, thomie

Differential Revision: https://phabricator.haskell.org/D2853

33 files changed, 480 insertions, 169 deletions

diff --git a/testsuite/tests/deSugar/should_compile/T2431.stderr b/testsuite/tests/deSugar/should_compile/T2431.stderr
index a8da44b73f..83826408cf 100644
--- a/testsuite/tests/deSugar/should_compile/T2431.stderr
+++ b/testsuite/tests/deSugar/should_compile/T2431.stderr
@@ -1,8 +1,9 @@
 
 ==================== Tidy Core ====================
-Result size of Tidy Core = {terms: 44, types: 34, coercions: 1}
+Result size of Tidy Core
+  = {terms: 44, types: 34, coercions: 1, joins: 0/0}
 
--- RHS size: {terms: 2, types: 4, coercions: 1}
+-- RHS size: {terms: 2, types: 4, coercions: 1, joins: 0/0}
 T2431.$WRefl [InlPrag=INLINE] :: forall a. a :~: a
 [GblId[DataConWrapper],
  Caf=NoCafRefs,
@@ -16,47 +17,47 @@ T2431.$WRefl =
   \ (@ a) ->
     T2431.Refl @ a @ a @~ (<a>_N :: (a :: *) GHC.Prim.~# (a :: *))
 
--- RHS size: {terms: 4, types: 8, coercions: 0}
+-- RHS size: {terms: 4, types: 8, coercions: 0, joins: 0/0}
 absurd :: forall a. (Int :~: Bool) -> a
 [GblId, Arity=1, Caf=NoCafRefs, Str=<L,U>x]
 absurd = \ (@ a) (x :: Int :~: Bool) -> case x of { }
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $trModule1 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs]
 $trModule1 = "main"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $trModule2 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs]
 $trModule2 = GHC.Types.TrNameS $trModule1
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $trModule3 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs]
 $trModule3 = "T2431"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $trModule4 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs]
 $trModule4 = GHC.Types.TrNameS $trModule3
 
--- RHS size: {terms: 3, types: 0, coercions: 0}
+-- RHS size: {terms: 3, types: 0, coercions: 0, joins: 0/0}
 T2431.$trModule :: GHC.Types.Module
 [GblId, Caf=NoCafRefs]
 T2431.$trModule = GHC.Types.Module $trModule2 $trModule4
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $tc'Refl1 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs]
 $tc'Refl1 = "'Refl"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $tc'Refl2 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs]
 $tc'Refl2 = GHC.Types.TrNameS $tc'Refl1
 
--- RHS size: {terms: 5, types: 0, coercions: 0}
+-- RHS size: {terms: 5, types: 0, coercions: 0, joins: 0/0}
 T2431.$tc'Refl :: GHC.Types.TyCon
 [GblId, Caf=NoCafRefs]
 T2431.$tc'Refl =
@@ -66,17 +67,17 @@ T2431.$tc'Refl =
     T2431.$trModule
     $tc'Refl2
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $tc:~:1 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs]
 $tc:~:1 = ":~:"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $tc:~:2 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs]
 $tc:~:2 = GHC.Types.TrNameS $tc:~:1
 
--- RHS size: {terms: 5, types: 0, coercions: 0}
+-- RHS size: {terms: 5, types: 0, coercions: 0, joins: 0/0}
 T2431.$tc:~: :: GHC.Types.TyCon
 [GblId, Caf=NoCafRefs]
 T2431.$tc:~: =
diff --git a/testsuite/tests/deriving/perf/all.T b/testsuite/tests/deriving/perf/all.T
index 0c3e9a4d3e..4c4bb97101 100644
--- a/testsuite/tests/deriving/perf/all.T
+++ b/testsuite/tests/deriving/perf/all.T
@@ -1,6 +1,8 @@
 test('T10858',
      [compiler_stats_num_field('bytes allocated',
-          [ (wordsize(64), 222312440, 8) ]),
+          [ (wordsize(64),  247768192, 8) ]),
+          # Initial:    222312440
+          # 2016-12-19  247768192  Join points (#19288)
       only_ways(['normal'])],
      compile,
      ['-O'])
diff --git a/testsuite/tests/numeric/should_compile/T7116.stdout b/testsuite/tests/numeric/should_compile/T7116.stdout
index 7fe4d93d87..bc2f85b85f 100644
--- a/testsuite/tests/numeric/should_compile/T7116.stdout
+++ b/testsuite/tests/numeric/should_compile/T7116.stdout
@@ -1,8 +1,9 @@
 
 ==================== Tidy Core ====================
-Result size of Tidy Core = {terms: 50, types: 25, coercions: 0}
+Result size of Tidy Core
+  = {terms: 50, types: 25, coercions: 0, joins: 0/0}
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T7116.$trModule4 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -10,7 +11,7 @@ T7116.$trModule4 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 20 0}]
 T7116.$trModule4 = "main"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T7116.$trModule3 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -19,7 +20,7 @@ T7116.$trModule3 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T7116.$trModule3 = GHC.Types.TrNameS T7116.$trModule4
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T7116.$trModule2 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -27,7 +28,7 @@ T7116.$trModule2 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 30 0}]
 T7116.$trModule2 = "T7116"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T7116.$trModule1 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -36,7 +37,7 @@ T7116.$trModule1 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T7116.$trModule1 = GHC.Types.TrNameS T7116.$trModule2
 
--- RHS size: {terms: 3, types: 0, coercions: 0}
+-- RHS size: {terms: 3, types: 0, coercions: 0, joins: 0/0}
 T7116.$trModule :: GHC.Types.Module
 [GblId,
  Caf=NoCafRefs,
@@ -46,7 +47,7 @@ T7116.$trModule :: GHC.Types.Module
 T7116.$trModule =
   GHC.Types.Module T7116.$trModule3 T7116.$trModule1
 
--- RHS size: {terms: 8, types: 3, coercions: 0}
+-- RHS size: {terms: 8, types: 3, coercions: 0, joins: 0/0}
 dr :: Double -> Double
 [GblId,
  Arity=1,
@@ -63,7 +64,7 @@ dr =
   \ (x :: Double) ->
     case x of { GHC.Types.D# x1 -> GHC.Types.D# (GHC.Prim.+## x1 x1) }
 
--- RHS size: {terms: 8, types: 3, coercions: 0}
+-- RHS size: {terms: 8, types: 3, coercions: 0, joins: 0/0}
 dl :: Double -> Double
 [GblId,
  Arity=1,
@@ -78,7 +79,7 @@ dl =
   \ (x :: Double) ->
     case x of { GHC.Types.D# y -> GHC.Types.D# (GHC.Prim.+## y y) }
 
--- RHS size: {terms: 8, types: 3, coercions: 0}
+-- RHS size: {terms: 8, types: 3, coercions: 0, joins: 0/0}
 fr :: Float -> Float
 [GblId,
  Arity=1,
@@ -97,7 +98,7 @@ fr =
     GHC.Types.F# (GHC.Prim.plusFloat# x1 x1)
     }
 
--- RHS size: {terms: 8, types: 3, coercions: 0}
+-- RHS size: {terms: 8, types: 3, coercions: 0, joins: 0/0}
 fl :: Float -> Float
 [GblId,
  Arity=1,
diff --git a/testsuite/tests/perf/compiler/all.T b/testsuite/tests/perf/compiler/all.T
index d9b0be509a..822ccb0026 100644
--- a/testsuite/tests/perf/compiler/all.T
+++ b/testsuite/tests/perf/compiler/all.T
@@ -67,7 +67,7 @@ test('T1969',
              # 2014-06-29 5949188 (x86/Linux)
              # 2015-07-11 6241108 (x86/Linux, 64bit machine) use +RTS -G1
              # 2016-04-06 9093608 (x86/Linux, 64bit machine)
-           (wordsize(64), 17285216, 15)]),
+           (wordsize(64), 19924328, 15)]),
              # 2014-09-10 10463640, 10  # post-AMP-update (somewhat stabelish)
                # looks like the peak is around ~10M, but we're
                # unlikely to GC exactly on the peak.
@@ -79,6 +79,7 @@ test('T1969',
              # 2015-07-11 11670120 (amd64/Linux)
              # 2015-10-28 15017528 (amd64/Linux) emit typeable at definition site
              # 2016-10-12 17285216 (amd64/Linux) it's not entirely clear why
+             # 2017-02-01 19924328 (amd64/Linux) Join points (#12988)
       compiler_stats_num_field('bytes allocated',
           [(platform('i386-unknown-mingw32'), 301784492, 5),
                                  #            215582916 (x86/Windows)
@@ -439,9 +440,10 @@ test('T5631',
 test('parsing001',
      [compiler_stats_num_field('bytes allocated',
           [(wordsize(32), 274000576, 10),
-           (wordsize(64), 581551384, 5)]),
+           (wordsize(64), 493730288, 5)]),
         # expected value: 587079016 (amd64/Linux)
         # 2016-09-01:     581551384 (amd64/Linux) Restore w/w limit (#11565)
+        # 2016-12-19:     493730288 (amd64/Linux) Join points (#12988)
        only_ways(['normal']),
       ],
      compile_fail, [''])
@@ -503,7 +505,7 @@ test('T5321Fun',
              # 2014-09-03: 299656164     (specialisation and inlining)
              # 10/12/2014: 206406188     #  Improvements in constraint solver
              # 2016-04-06: 279922360 x86/Linux
-            (wordsize(64), 525895608, 5)])
+            (wordsize(64), 498135752, 5)])
              # prev:       585521080
              # 29/08/2012: 713385808     #  (increase due to new codegen)
              # 15/05/2013: 628341952     #  (reason for decrease unknown)
@@ -526,6 +528,7 @@ test('T5321Fun',
                                          #    change, however. Namely I am
                                          #    quite skeptical of the downward
                                          #    "drift" reported above
+             # 31/01/2017: 498135752     #  Join points (#12988)
       ],
       compile,[''])
 
@@ -802,7 +805,7 @@ test('T9872d',
 test('T9961',
      [ only_ways(['normal']),
        compiler_stats_num_field('bytes allocated',
-          [(wordsize(64), 537297968, 5),
+          [(wordsize(64), 571246936, 5),
           # 2015-01-12    807117816   Initally created
           # 2015-spring   772510192   Got better
           # 2015-05-22    663978160   Fix for #10370 improves it more
@@ -811,6 +814,7 @@ test('T9961',
           # 2016-03-20    519436672   x64_64/Linux   Don't use build desugaring for large lists (#11707)
           # 2016-03-24    568526784   x64_64/Linux   Add eqInt* variants (#11688)
           # 2016-09-01    537297968   x64_64/Linux   Restore w/w limit (#11565)
+          # 2016-12-19    571246936   x64_64/Linux   Join points (#12988)
            (wordsize(32), 275264188, 5)
           # was           375647160
           # 2016-04-06    275264188   x86/Linux
@@ -934,8 +938,9 @@ test('T13035',
 test('T13056',
      [ only_ways(['optasm']),
        compiler_stats_num_field('bytes allocated',
-         [(wordsize(64), 520166912, 5),
+         [(wordsize(64), 546800240, 5),
          # 2017-01-06    520166912 initial
+         # 2017-01-31    546800240 Join points (#12988)
          ]),
      ],
      compile,
@@ -943,9 +948,10 @@ test('T13056',
 
 test('T12707',
      [ compiler_stats_num_field('bytes allocated',
-          [(wordsize(64), 1348865648, 5),
+          [(wordsize(64), 1280336112, 5),
           # initial:    1271577192
           # 2017-01-22: 1348865648  Allow top-level strings in Core
+          # 2017-01-31: 1280336112  Join points (#12988)
           ]),
      ],
      compile,
diff --git a/testsuite/tests/perf/haddock/all.T b/testsuite/tests/perf/haddock/all.T
index 8ec02cefcc..f037954263 100644
--- a/testsuite/tests/perf/haddock/all.T
+++ b/testsuite/tests/perf/haddock/all.T
@@ -5,7 +5,7 @@
 test('haddock.base',
      [unless(in_tree_compiler(), skip), req_haddock
      ,stats_num_field('bytes allocated',
-          [(wordsize(64), 32855223200, 5)
+          [(wordsize(64), 31115778088 , 5)
             # 2012-08-14: 5920822352 (amd64/Linux)
             # 2012-09-20: 5829972376 (amd64/Linux)
             # 2012-10-08: 5902601224 (amd64/Linux)
@@ -30,6 +30,7 @@ test('haddock.base',
             # 2015-12-17: 27812188000 (x86_64/Linux) - Move Data.Functor.* into base
             # 2016-02-25: 30987348040 (x86_64/Linux) - RuntimeRep
             # 2016-05-12: 32855223200 (x86_64/Linux) - Make Generic1 poly-kinded
+            # 2017-01-11: 31115778088 (x86_64/Linux) - Join points (#12988)
 
           ,(platform('i386-unknown-mingw32'), 4434804940, 5)
             # 2013-02-10:                     3358693084 (x86/Windows)
@@ -52,7 +53,7 @@ test('haddock.base',
 test('haddock.Cabal',
      [unless(in_tree_compiler(), skip), req_haddock
      ,stats_num_field('bytes allocated',
-          [(wordsize(64), 25478853176 , 5)
+          [(wordsize(64), 23272708864, 5)
             # 2012-08-14: 3255435248 (amd64/Linux)
             # 2012-08-29: 3324606664 (amd64/Linux, new codegen)
             # 2012-10-08: 3373401360 (amd64/Linux)
@@ -92,6 +93,7 @@ test('haddock.Cabal',
             # 2016-10-03: 21554874976 (amd64/Linux) - Cabal update
             # 2016-10-06: 23706190072 (amd64/Linux) - Cabal update
             # 2016-12-20: 25478853176 (amd64/Linux) - Cabal update
+            # 2017-01-14: 23272708864 (amd64/Linux) - Join points (#12988)
 
           ,(platform('i386-unknown-mingw32'), 3293415576, 5)
             # 2012-10-30:                     1733638168 (x86/Windows)
diff --git a/testsuite/tests/perf/join_points/Makefile b/testsuite/tests/perf/join_points/Makefile
new file mode 100644
index 0000000000..9101fbd40a
--- /dev/null
+++ b/testsuite/tests/perf/join_points/Makefile
@@ -0,0 +1,3 @@
+TOP=../../..
+include $(TOP)/mk/boilerplate.mk
+include $(TOP)/mk/test.mk
diff --git a/testsuite/tests/perf/join_points/all.T b/testsuite/tests/perf/join_points/all.T
new file mode 100644
index 0000000000..b6f6e40699
--- /dev/null
+++ b/testsuite/tests/perf/join_points/all.T
@@ -0,0 +1,28 @@
+# Only compile with optimisation
+def f( name, opts ):
+  opts.only_ways = ['optasm']
+
+setTestOpts(f)
+
+test('join001', normal, compile, [''])
+
+test('join002',
+  [stats_num_field('bytes allocated', [(wordsize(64), 2000290792, 5)])],
+  compile_and_run,
+  [''])
+test('join003',
+  [stats_num_field('bytes allocated', [(wordsize(64), 2000290792, 5)])],
+  compile_and_run,
+  [''])
+test('join004',
+  [stats_num_field('bytes allocated', [(wordsize(64), 48146720, 5)])],
+  compile_and_run,
+  [''])
+
+test('join005', normal, compile, [''])
+test('join006', normal, compile, [''])
+
+test('join007',
+  [stats_num_field('bytes allocated', [(wordsize(64), 50944, 5)])],
+  compile_and_run,
+  [''])
diff --git a/testsuite/tests/perf/join_points/join001.hs b/testsuite/tests/perf/join_points/join001.hs
new file mode 100644
index 0000000000..04dce36dc9
--- /dev/null
+++ b/testsuite/tests/perf/join_points/join001.hs
@@ -0,0 +1,16 @@
+{-# LANGUAGE BangPatterns #-}
+
+module Main where
+
+findDivBy :: Int -> [Int] -> Maybe Int
+findDivBy p ns
+    -- go should be a join point and should get worker/wrappered; the worker
+    -- must also be a join point (since it's mutually recursive with one).
+  = let go !p ns = case ns of n:ns' -> case n `mod` p of 0 -> Just n
+                                                         _ -> go p ns'
+                              []    -> Nothing
+    in case p of
+         0 -> error "div by zero"
+         _ -> go p ns
+
+main = print $ findDivBy 7 [1..10]
diff --git a/testsuite/tests/perf/join_points/join002.hs b/testsuite/tests/perf/join_points/join002.hs
new file mode 100644
index 0000000000..49aecdcc7f
--- /dev/null
+++ b/testsuite/tests/perf/join_points/join002.hs
@@ -0,0 +1,51 @@
+module Main where
+
+import Data.List
+
+-- These four functions should all wind up the same; they represent successive
+-- simplifications that should happen. (Actual details may vary, since find
+-- isn't quite defined this way, but the differences disappear by the end.)
+
+firstEvenIsPositive1 :: [Int] -> Bool
+firstEvenIsPositive1 = maybe False (> 0) . find even
+
+-- After inlining:
+
+firstEvenIsPositive2 :: [Int] -> Bool
+firstEvenIsPositive2 xs =
+  let go xs = case xs of x:xs' -> if even x then Just x else go xs'
+                         []    -> Nothing
+  in case go xs of Just n  -> n > 0
+                   Nothing -> False
+
+-- Note that go *could* be a join point if it were declared inside the scrutinee
+-- instead of outside. So it's now Float In's job to move the binding inward a
+-- smidge. *But* if it goes too far inward (as it would until recently), it will
+-- wrap only "go" instead of "go xs", which won't let us mark go as a join point
+-- since join points can't be partially invoked.
+--
+-- After Float In:
+
+firstEvenIsPositive3 :: [Int] -> Bool
+firstEvenIsPositive3 xs =
+  case let {-join-} go xs = case xs of x:xs' -> if even x then Just x
+                                                          else go xs'
+                                       []    -> Nothing
+       in go xs of
+    Just n  -> n > 0
+    Nothing -> False
+
+-- After the simplifier:
+
+firstEvenIsPositive4 :: [Int] -> Bool
+firstEvenIsPositive4 xs =
+  let {-join-} go xs = case xs of x:xs' -> if even x then x > 0 else go xs'
+                                  []    -> False
+  in go xs
+
+-- This only worked because go was a join point so that the case gets moved
+-- inside.
+
+{-# NOINLINE firstEvenIsPositive1 #-}
+
+main = print $ or $ [firstEvenIsPositive1 [1,3..n] | n <- [1..10000]]
diff --git a/testsuite/tests/perf/join_points/join002.stdout b/testsuite/tests/perf/join_points/join002.stdout
new file mode 100644
index 0000000000..bc59c12aa1
--- /dev/null
+++ b/testsuite/tests/perf/join_points/join002.stdout
@@ -0,0 +1 @@
+False
diff --git a/testsuite/tests/perf/join_points/join003.hs b/testsuite/tests/perf/join_points/join003.hs
new file mode 100644
index 0000000000..051c2d8bfe
--- /dev/null
+++ b/testsuite/tests/perf/join_points/join003.hs
@@ -0,0 +1,69 @@
+{-
+ - A variation on join002.hs that avoids Float Out issues. The join points in
+ - join002.hs may get floated to top level, which is necessary to allow in
+ - general, but which makes them into functions rather than join points, thus
+ - messing up the test.
+ -}
+
+module Main (
+  firstMultIsPositive1, firstMultIsPositive2, firstMultIsPositive3,
+  firstMultIsPositive4,
+
+  main
+) where
+
+import Data.List
+
+divides :: Int -> Int -> Bool
+p `divides` n = n `mod` p == 0
+
+infix 4 `divides`
+
+-- These four functions should all wind up the same; they represent successive
+-- simplifications that should happen. (Actual details may vary, since find
+-- isn't quite defined this way, but the differences disappear by the end.)
+
+firstMultIsPositive1 :: Int -> [Int] -> Bool
+firstMultIsPositive1 p = maybe False (> 0) . find (p `divides`)
+
+-- After inlining:
+
+firstMultIsPositive2 :: Int -> [Int] -> Bool
+firstMultIsPositive2 p xs =
+  let go xs = case xs of x:xs' -> if p `divides` x then Just x else go xs'
+                         []    -> Nothing
+  in case go xs of Just n  -> n > 0
+                   Nothing -> False
+
+-- Note that go *could* be a join point if it were declared inside the scrutinee
+-- instead of outside. So it's now Float In's job to move the binding inward a
+-- smidge. *But* if it goes too far inward (as it would until recently), it will
+-- wrap only "go" instead of "go xs", which won't let us mark go as a join point
+-- since join points can't be partially invoked.
+--
+-- After Float In:
+
+firstMultIsPositive3 :: Int -> [Int] -> Bool
+firstMultIsPositive3 p xs =
+  case let {-join-} go xs = case xs of x:xs' -> if p `divides` x then Just x
+                                                                 else go xs'
+                                       []    -> Nothing
+       in go xs of
+    Just n  -> n > 0
+    Nothing -> False
+
+-- After the simplifier:
+
+firstMultIsPositive4 :: Int -> [Int] -> Bool
+firstMultIsPositive4 p xs =
+  let {-join-} go xs = case xs of x:xs' -> if p `divides` x then x > 0
+                                                            else go xs'
+                                  []    -> False
+  in go xs
+
+-- This only worked because go was a join point so that the case gets moved
+-- inside.
+
+{-# NOINLINE firstMultIsPositive1 #-}
+
+main = print $ or $ [firstMultIsPositive1 2 [1,3..n] | n <- [1..10000]]
diff --git a/testsuite/tests/perf/join_points/join003.stdout b/testsuite/tests/perf/join_points/join003.stdout
new file mode 100644
index 0000000000..bc59c12aa1
--- /dev/null
+++ b/testsuite/tests/perf/join_points/join003.stdout
@@ -0,0 +1 @@
+False
diff --git a/testsuite/tests/perf/join_points/join004.hs b/testsuite/tests/perf/join_points/join004.hs
new file mode 100644
index 0000000000..1962cc266e
--- /dev/null
+++ b/testsuite/tests/perf/join_points/join004.hs
@@ -0,0 +1,30 @@
+{-
+ - A rather contrived example demonstrating the virtues of not floating join
+ - points outward.
+ -}
+
+module Main (main) where
+
+-- Calculate n `div` d `div` d by looping.
+
+{-# NOINLINE slowDivDiv #-}
+slowDivDiv :: Int -> Int -> Int
+slowDivDiv n d
+  = let {-# NOINLINE divPos #-}
+        divPos :: Int -> Int
+        divPos n0
+          = -- This function is a join point (all calls are tail calls), so it
+            -- never causes a closure allocation, so it doesn't help to float it
+            -- out. Thus -fno-join-points causes a ~25% jump in allocations.
+            let go n' i
+                  = case n' >= d of True  -> go (n' - d) (i + 1)
+                                    False -> i
+            in go n0 0
+    in case n >= 0 of True  -> divPos (divPos n)
+                      False -> divPos (-(divPos (-n)))
+                                 -- It's important that divPos be called twice
+                                 -- because otherwise it'd be a one-shot lambda
+                                 -- and so the join point would be floated
+                                 -- back in again.
+
+main = print $ sum [slowDivDiv n d | n <- [1..1000], d <- [1..1000]]
diff --git a/testsuite/tests/perf/join_points/join004.stdout b/testsuite/tests/perf/join_points/join004.stdout
new file mode 100644
index 0000000000..a3abc727e8
--- /dev/null
+++ b/testsuite/tests/perf/join_points/join004.stdout
@@ -0,0 +1 @@
+793515
diff --git a/testsuite/tests/perf/join_points/join005.hs b/testsuite/tests/perf/join_points/join005.hs
new file mode 100644
index 0000000000..7feec98c28
--- /dev/null
+++ b/testsuite/tests/perf/join_points/join005.hs
@@ -0,0 +1,23 @@
+{- Test of Worker/Wrapper operating on join points -}
+
+module Main where
+
+sumOfMultiplesOf :: Int -> [Int] -> Int
+sumOfMultiplesOf p ns
+  = {- This is a join point (and it will stay that way---it won't get floated to
+       top level because p occurs free). It should get worker/wrappered. -}
+    let go ns acc
+          = case ns of []    -> acc
+                       n:ns' -> case n `mod` p of 0 -> go ns' (acc + n)
+                                                  _ -> go ns'  acc
+    in go ns 0
+
+{-
+It's hard to test for this, but what should happen is that go gets W/W'd and the
+worker is a join point (else Core Lint will complain). Interestingly, go is
+*not* CPR'd, because then the worker couldn't be a join point, but once the
+simplifier runs, the worker ends up returning Int# anyway. See Note [Don't CPR
+join points] in WorkWrap.hs.
+-}
+
+main = print $ sumOfMultiplesOf 2 [1..10]
diff --git a/testsuite/tests/perf/join_points/join006.hs b/testsuite/tests/perf/join_points/join006.hs
new file mode 100644
index 0000000000..3c0b2ceecd
--- /dev/null
+++ b/testsuite/tests/perf/join_points/join006.hs
@@ -0,0 +1,22 @@
+module Main where
+
+{-# NOINLINE foo #-}
+foo :: Int -> Int
+foo x = x
+
+{-# RULES "foo/5" forall (f :: Int -> Int). foo (f 5) = foo (f 42) #-}
+-- highly suspect, of course!
+
+main = print $ foo (let {-# NOINLINE j #-}
+                        j :: Int -> Int
+                        j n = n + 1 in j 5)
+
+{-
+If we're not careful, this will get rewritten to
+
+  main = print $ let <join> j n = n + 1 in foo (j 42)
+
+which violates the join point invariant (can't invoke a join point from
+non-tail context). Solution is to refuse to float join points when matching
+RULES.
+-}
diff --git a/testsuite/tests/perf/join_points/join007.hs b/testsuite/tests/perf/join_points/join007.hs
new file mode 100644
index 0000000000..aa2f68c0bc
--- /dev/null
+++ b/testsuite/tests/perf/join_points/join007.hs
@@ -0,0 +1,42 @@
+-- Test of fusion in unfold/destroy style. Originally, unfold/destroy supported
+-- filter, but the constructors (here Done and Yield) couldn't be compiled away.
+-- Join points let us do this by pulling the case from sumS into the loop in
+-- filterS.
+
+{-# LANGUAGE GADTs #-}
+
+module Main (main) where
+
+data Stream a where Stream :: (s -> Step a s) -> s -> Stream a
+data Step a s = Done | Yield a s
+
+{-# INLINE sumS #-}
+sumS :: Stream Int -> Int
+sumS (Stream next s0) = go s0 0
+  where
+    go s acc = case next s of Done       -> acc
+                              Yield a s' -> go s' (acc + a)
+
+{-# INLINE filterS #-}
+filterS :: (a -> Bool) -> Stream a -> Stream a
+filterS p (Stream next s0) = Stream fnext s0
+  where
+    fnext s = seek s
+      where
+        -- should be a join point!
+        seek s = case next s of Done                   -> Done
+                                Yield a s' | p a       -> Yield a s'
+                                           | otherwise -> seek s'
+
+{-# INLINE enumFromToS #-}
+enumFromToS :: Int -> Int -> Stream Int
+enumFromToS lo hi = Stream next lo
+  where
+    next n | n > hi    = Done
+           | otherwise = Yield n (n+1)
+
+{-# NOINLINE test #-} -- for -ddump-simpl
+test :: Int -> Int -> Int
+test lo hi = sumS (filterS even (enumFromToS lo hi))
+
+main = print $ test 1 10000000
diff --git a/testsuite/tests/perf/join_points/join007.stdout b/testsuite/tests/perf/join_points/join007.stdout
new file mode 100644
index 0000000000..0a7ad1072c
--- /dev/null
+++ b/testsuite/tests/perf/join_points/join007.stdout
@@ -0,0 +1 @@
+25000005000000
diff --git a/testsuite/tests/perf/should_run/all.T b/testsuite/tests/perf/should_run/all.T
index 382c317a9b..592e63c274 100644
--- a/testsuite/tests/perf/should_run/all.T
+++ b/testsuite/tests/perf/should_run/all.T
@@ -436,10 +436,11 @@ test('T9203',
                       [ (wordsize(32), 84345136 , 5)
                       # was
                       # 2016-04-06     84345136 (i386/Debian) not sure
-                      , (wordsize(64), 95451192, 5) ]),
+                      , (wordsize(64), 84620888, 5) ]),
                       # was            95747304
                       # 2019-09-10     94547280 post-AMP cleanup
                       # 2015-10-28     95451192 emit Typeable at definition site
+                      # 2016-12-19     84620888 Join points
       only_ways(['normal'])],
      compile_and_run,
      ['-O2'])
@@ -447,9 +448,10 @@ test('T9203',
 test('T9339',
      [stats_num_field('bytes allocated',
                       [ (wordsize(32),    40046844, 5)
-                      , (wordsize(64),    80050760, 5) ]),
+                      , (wordsize(64),       50728, 5) ]),
                       # w/o fusing last: 320005080
                       # 2014-07-22:       80050760
+                      # 2016-08-17:          50728 Join points (#12988)
       only_ways(['normal'])],
      compile_and_run,
      ['-O2'])
diff --git a/testsuite/tests/roles/should_compile/Roles13.stderr b/testsuite/tests/roles/should_compile/Roles13.stderr
index 20206e28df..7e510d442e 100644
--- a/testsuite/tests/roles/should_compile/Roles13.stderr
+++ b/testsuite/tests/roles/should_compile/Roles13.stderr
@@ -1,13 +1,14 @@
 
 ==================== Tidy Core ====================
-Result size of Tidy Core = {terms: 63, types: 26, coercions: 5}
+Result size of Tidy Core
+  = {terms: 63, types: 26, coercions: 5, joins: 0/0}
 
--- RHS size: {terms: 2, types: 2, coercions: 0}
+-- RHS size: {terms: 2, types: 2, coercions: 0, joins: 0/0}
 convert1 :: Wrap Age -> Wrap Age
 [GblId, Arity=1, Caf=NoCafRefs]
 convert1 = \ (ds :: Wrap Age) -> ds
 
--- RHS size: {terms: 1, types: 0, coercions: 5}
+-- RHS size: {terms: 1, types: 0, coercions: 5, joins: 0/0}
 convert :: Wrap Age -> Int
 [GblId, Arity=1, Caf=NoCafRefs]
 convert =
@@ -15,42 +16,42 @@ convert =
   `cast` (<Wrap Age>_R -> Roles13.N:Wrap[0] Roles13.N:Age[0]
           :: ((Wrap Age -> Wrap Age) :: *) ~R# ((Wrap Age -> Int) :: *))
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $trModule1 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs]
 $trModule1 = "main"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $trModule2 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs]
 $trModule2 = GHC.Types.TrNameS $trModule1
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $trModule3 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs]
 $trModule3 = "Roles13"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $trModule4 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs]
 $trModule4 = GHC.Types.TrNameS $trModule3
 
--- RHS size: {terms: 3, types: 0, coercions: 0}
+-- RHS size: {terms: 3, types: 0, coercions: 0, joins: 0/0}
 Roles13.$trModule :: GHC.Types.Module
 [GblId, Caf=NoCafRefs]
 Roles13.$trModule = GHC.Types.Module $trModule2 $trModule4
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $tc'MkAge1 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs]
 $tc'MkAge1 = "'MkAge"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $tc'MkAge2 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs]
 $tc'MkAge2 = GHC.Types.TrNameS $tc'MkAge1
 
--- RHS size: {terms: 5, types: 0, coercions: 0}
+-- RHS size: {terms: 5, types: 0, coercions: 0, joins: 0/0}
 Roles13.$tc'MkAge :: GHC.Types.TyCon
 [GblId, Caf=NoCafRefs]
 Roles13.$tc'MkAge =
@@ -60,17 +61,17 @@ Roles13.$tc'MkAge =
     Roles13.$trModule
     $tc'MkAge2
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $tcAge1 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs]
 $tcAge1 = "Age"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $tcAge2 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs]
 $tcAge2 = GHC.Types.TrNameS $tcAge1
 
--- RHS size: {terms: 5, types: 0, coercions: 0}
+-- RHS size: {terms: 5, types: 0, coercions: 0, joins: 0/0}
 Roles13.$tcAge :: GHC.Types.TyCon
 [GblId, Caf=NoCafRefs]
 Roles13.$tcAge =
@@ -80,17 +81,17 @@ Roles13.$tcAge =
     Roles13.$trModule
     $tcAge2
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $tc'MkWrap1 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs]
 $tc'MkWrap1 = "'MkWrap"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $tc'MkWrap2 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs]
 $tc'MkWrap2 = GHC.Types.TrNameS $tc'MkWrap1
 
--- RHS size: {terms: 5, types: 0, coercions: 0}
+-- RHS size: {terms: 5, types: 0, coercions: 0, joins: 0/0}
 Roles13.$tc'MkWrap :: GHC.Types.TyCon
 [GblId, Caf=NoCafRefs]
 Roles13.$tc'MkWrap =
@@ -100,17 +101,17 @@ Roles13.$tc'MkWrap =
     Roles13.$trModule
     $tc'MkWrap2
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $tcWrap1 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs]
 $tcWrap1 = "Wrap"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $tcWrap2 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs]
 $tcWrap2 = GHC.Types.TrNameS $tcWrap1
 
--- RHS size: {terms: 5, types: 0, coercions: 0}
+-- RHS size: {terms: 5, types: 0, coercions: 0, joins: 0/0}
 Roles13.$tcWrap :: GHC.Types.TyCon
 [GblId, Caf=NoCafRefs]
 Roles13.$tcWrap =
diff --git a/testsuite/tests/simplCore/should_compile/Makefile b/testsuite/tests/simplCore/should_compile/Makefile
index 5a465d9818..ef3e74ad7f 100644
--- a/testsuite/tests/simplCore/should_compile/Makefile
+++ b/testsuite/tests/simplCore/should_compile/Makefile
@@ -47,6 +47,7 @@ T5658b:
 	$(RM) -f T5658b.o T5658b.hi
 	'$(TEST_HC)' $(TEST_HC_OPTS) -O -c T5658b.hs -ddump-simpl | grep -c indexIntArray
 # Trac 5658 meant that there were three calls to indexIntArray instead of two
+# (now four due to join-point discount causing W/W to stabilize unfolding)
 
 T5776:
 	$(RM) -f T5776.o T5776.hi
@@ -121,7 +122,7 @@ T13155:
 T13156:
 	$(RM) -f T13156.hi T13156.o
 	'$(TEST_HC)' $(TEST_HC_OPTS) -c T13156.hs -O -ddump-prep -dsuppress-uniques | grep "case"
-        # There should be a single 'case case'
+        # There should be a single 'case r @ GHC.Types.Any'
 
 .PHONY: T4138
 T4138:
diff --git a/testsuite/tests/simplCore/should_compile/T13156.hs b/testsuite/tests/simplCore/should_compile/T13156.hs
index cdc322af38..2ddfa2cefb 100644
--- a/testsuite/tests/simplCore/should_compile/T13156.hs
+++ b/testsuite/tests/simplCore/should_compile/T13156.hs
@@ -4,39 +4,42 @@ f g x = let r :: [a] -> [a]
             r = case g x of True -> reverse . reverse
                             False -> reverse
         in
-        r `seq` r `seq` True
+        r `seq` r `seq` r
 
 
 {- Expected -ddump-prep looks like this.
-   (Room for improvement on the case (case ..) line.)
+   (Case-of-type-lambda an oddity of Core Prep.)
 
--- RHS size: {terms: 9, types: 9, coercions: 0}
+-- RHS size: {terms: 9, types: 9, coercions: 0, joins: 0/0}
 T13156.f1 :: forall a. [a] -> [a]
 [GblId, Arity=1, Caf=NoCafRefs, Str=<S,1*U>, Unf=OtherCon []]
 T13156.f1 =
-  \ (@ a_aC4) (x_sNG [Occ=Once] :: [a]) ->
-    case GHC.List.reverse @ a x_sNG of sat_sNH { __DEFAULT ->
-    GHC.List.reverse1 @ a sat_sNH (GHC.Types.[] @ a)
+  \ (@ a) (x [Occ=Once] :: [a]) ->
+    case GHC.List.reverse @ a x of sat { __DEFAULT ->
+    GHC.List.reverse1 @ a sat (GHC.Types.[] @ a)
     }
 
--- RHS size: {terms: 13, types: 20, coercions: 0}
-T13156.f :: forall p. (p -> GHC.Types.Bool) -> p -> GHC.Types.Bool
+-- RHS size: {terms: 17, types: 28, coercions: 0, joins: 0/0}
+T13156.f
+  :: forall p.
+     (p -> GHC.Types.Bool) -> p -> [GHC.Types.Int] -> [GHC.Types.Int]
 [GblId,
  Arity=2,
  Caf=NoCafRefs,
  Str=<C(S),1*C1(U)><L,U>,
  Unf=OtherCon []]
 T13156.f =
-  \ (@ p_aBS)
-    (g_sNI [Occ=Once!] :: p -> GHC.Types.Bool)
-    (x_sNJ [Occ=Once] :: p) ->
-    case case g_sNI x_sNJ of {
-           GHC.Types.False -> GHC.List.reverse @ GHC.Types.Any;
-           GHC.Types.True -> T13156.f1 @ GHC.Types.Any
-         }
-    of
+  \ (@ p)
+    (g [Occ=Once!] :: p -> GHC.Types.Bool)
+    (x [Occ=Once] :: p) ->
+    case \ (@ a) ->
+           case g x of {
+             GHC.Types.False -> GHC.List.reverse @ a;
+             GHC.Types.True -> T13156.f1 @ a
+           }
+    of r [Dmd=<S,U>]
     { __DEFAULT ->
-    GHC.Types.True
+    case r @ GHC.Types.Any of { __DEFAULT -> r @ GHC.Types.Int }
     }
 
 -}
diff --git a/testsuite/tests/simplCore/should_compile/T13156.stdout b/testsuite/tests/simplCore/should_compile/T13156.stdout
index 51d10c851f..5aa8f6aa38 100644
--- a/testsuite/tests/simplCore/should_compile/T13156.stdout
+++ b/testsuite/tests/simplCore/should_compile/T13156.stdout
@@ -1,2 +1,4 @@
     case GHC.List.reverse @ a x of sat { __DEFAULT ->
-    case case g x of {
+    case \ (@ a1) ->
+           case g x of {
+    case r @ GHC.Types.Any of { __DEFAULT -> r @ a }
diff --git a/testsuite/tests/simplCore/should_compile/T3717.stderr b/testsuite/tests/simplCore/should_compile/T3717.stderr
index f9adeb28da..9bcc4f31ac 100644
--- a/testsuite/tests/simplCore/should_compile/T3717.stderr
+++ b/testsuite/tests/simplCore/should_compile/T3717.stderr
@@ -1,8 +1,9 @@
 
 ==================== Tidy Core ====================
-Result size of Tidy Core = {terms: 36, types: 15, coercions: 0}
+Result size of Tidy Core
+  = {terms: 36, types: 15, coercions: 0, joins: 0/0}
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T3717.$trModule4 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -10,7 +11,7 @@ T3717.$trModule4 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 20 0}]
 T3717.$trModule4 = "main"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T3717.$trModule3 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -19,7 +20,7 @@ T3717.$trModule3 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T3717.$trModule3 = GHC.Types.TrNameS T3717.$trModule4
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T3717.$trModule2 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -27,7 +28,7 @@ T3717.$trModule2 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 30 0}]
 T3717.$trModule2 = "T3717"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T3717.$trModule1 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -36,7 +37,7 @@ T3717.$trModule1 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T3717.$trModule1 = GHC.Types.TrNameS T3717.$trModule2
 
--- RHS size: {terms: 3, types: 0, coercions: 0}
+-- RHS size: {terms: 3, types: 0, coercions: 0, joins: 0/0}
 T3717.$trModule :: GHC.Types.Module
 [GblId,
  Caf=NoCafRefs,
@@ -47,7 +48,7 @@ T3717.$trModule =
   GHC.Types.Module T3717.$trModule3 T3717.$trModule1
 
 Rec {
--- RHS size: {terms: 10, types: 2, coercions: 0}
+-- RHS size: {terms: 10, types: 2, coercions: 0, joins: 0/0}
 T3717.$wfoo [InlPrag=[0], Occ=LoopBreaker]
   :: GHC.Prim.Int# -> GHC.Prim.Int#
 [GblId, Arity=1, Caf=NoCafRefs, Str=<S,1*U>]
@@ -59,7 +60,7 @@ T3717.$wfoo =
     }
 end Rec }
 
--- RHS size: {terms: 10, types: 4, coercions: 0}
+-- RHS size: {terms: 10, types: 4, coercions: 0, joins: 0/0}
 foo [InlPrag=INLINE[0]] :: Int -> Int
 [GblId,
  Arity=1,
diff --git a/testsuite/tests/simplCore/should_compile/T3772.stdout b/testsuite/tests/simplCore/should_compile/T3772.stdout
index 76936e336f..98a809d95f 100644
--- a/testsuite/tests/simplCore/should_compile/T3772.stdout
+++ b/testsuite/tests/simplCore/should_compile/T3772.stdout
@@ -1,9 +1,10 @@
 
 ==================== Tidy Core ====================
-Result size of Tidy Core = {terms: 40, types: 16, coercions: 0}
+Result size of Tidy Core
+  = {terms: 40, types: 16, coercions: 0, joins: 0/0}
 
 Rec {
--- RHS size: {terms: 10, types: 2, coercions: 0}
+-- RHS size: {terms: 10, types: 2, coercions: 0, joins: 0/0}
 $wxs :: GHC.Prim.Int# -> ()
 [GblId, Arity=1, Caf=NoCafRefs, Str=<S,1*U>]
 $wxs =
@@ -14,7 +15,7 @@ $wxs =
     }
 end Rec }
 
--- RHS size: {terms: 14, types: 5, coercions: 0}
+-- RHS size: {terms: 14, types: 5, coercions: 0, joins: 0/0}
 foo [InlPrag=NOINLINE] :: Int -> ()
 [GblId, Arity=1, Caf=NoCafRefs, Str=<S(S),1*U(U)>]
 foo =
@@ -26,7 +27,7 @@ foo =
     }
     }
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T3772.$trModule2 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -34,7 +35,7 @@ T3772.$trModule2 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 30 0}]
 T3772.$trModule2 = "T3772"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T3772.$trModule1 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -43,7 +44,7 @@ T3772.$trModule1 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T3772.$trModule1 = GHC.Types.TrNameS T3772.$trModule2
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T3772.$trModule4 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -51,7 +52,7 @@ T3772.$trModule4 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 20 0}]
 T3772.$trModule4 = "main"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T3772.$trModule3 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -60,7 +61,7 @@ T3772.$trModule3 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T3772.$trModule3 = GHC.Types.TrNameS T3772.$trModule4
 
--- RHS size: {terms: 3, types: 0, coercions: 0}
+-- RHS size: {terms: 3, types: 0, coercions: 0, joins: 0/0}
 T3772.$trModule :: GHC.Types.Module
 [GblId,
  Caf=NoCafRefs,
diff --git a/testsuite/tests/simplCore/should_compile/T4908.stderr b/testsuite/tests/simplCore/should_compile/T4908.stderr
index e9957bf9de..185b9b3529 100644
--- a/testsuite/tests/simplCore/should_compile/T4908.stderr
+++ b/testsuite/tests/simplCore/should_compile/T4908.stderr
@@ -1,8 +1,9 @@
 
 ==================== Tidy Core ====================
-Result size of Tidy Core = {terms: 68, types: 43, coercions: 0}
+Result size of Tidy Core
+  = {terms: 68, types: 43, coercions: 0, joins: 0/0}
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T4908.$trModule4 :: Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -10,7 +11,7 @@ T4908.$trModule4 :: Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 20 0}]
 T4908.$trModule4 = "main"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T4908.$trModule3 :: TrName
 [GblId,
  Caf=NoCafRefs,
@@ -19,7 +20,7 @@ T4908.$trModule3 :: TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T4908.$trModule3 = GHC.Types.TrNameS T4908.$trModule4
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T4908.$trModule2 :: Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -27,7 +28,7 @@ T4908.$trModule2 :: Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 30 0}]
 T4908.$trModule2 = "T4908"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T4908.$trModule1 :: TrName
 [GblId,
  Caf=NoCafRefs,
@@ -36,7 +37,7 @@ T4908.$trModule1 :: TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T4908.$trModule1 = GHC.Types.TrNameS T4908.$trModule2
 
--- RHS size: {terms: 3, types: 0, coercions: 0}
+-- RHS size: {terms: 3, types: 0, coercions: 0, joins: 0/0}
 T4908.$trModule :: Module
 [GblId,
  Caf=NoCafRefs,
@@ -47,7 +48,7 @@ T4908.$trModule =
   GHC.Types.Module T4908.$trModule3 T4908.$trModule1
 
 Rec {
--- RHS size: {terms: 19, types: 5, coercions: 0}
+-- RHS size: {terms: 19, types: 5, coercions: 0, joins: 0/0}
 T4908.f_$s$wf [Occ=LoopBreaker] :: Int -> Int# -> Int# -> Bool
 [GblId, Arity=3, Caf=NoCafRefs, Str=<L,A><L,1*U><S,1*U>]
 T4908.f_$s$wf =
@@ -62,7 +63,7 @@ T4908.f_$s$wf =
     }
 end Rec }
 
--- RHS size: {terms: 24, types: 13, coercions: 0}
+-- RHS size: {terms: 24, types: 13, coercions: 0, joins: 0/0}
 T4908.$wf [InlPrag=[0]] :: Int# -> (Int, Int) -> Bool
 [GblId,
  Arity=2,
@@ -85,7 +86,7 @@ T4908.$wf =
       0# -> GHC.Types.True
     }
 
--- RHS size: {terms: 8, types: 6, coercions: 0}
+-- RHS size: {terms: 8, types: 6, coercions: 0, joins: 0/0}
 f [InlPrag=INLINE[0]] :: Int -> (Int, Int) -> Bool
 [GblId,
  Arity=2,
diff --git a/testsuite/tests/simplCore/should_compile/T4930.stderr b/testsuite/tests/simplCore/should_compile/T4930.stderr
index 365584d3d0..9db97a5e1f 100644
--- a/testsuite/tests/simplCore/should_compile/T4930.stderr
+++ b/testsuite/tests/simplCore/should_compile/T4930.stderr
@@ -1,8 +1,9 @@
 
 ==================== Tidy Core ====================
-Result size of Tidy Core = {terms: 49, types: 19, coercions: 0}
+Result size of Tidy Core
+  = {terms: 44, types: 17, coercions: 0, joins: 0/0}
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T4930.$trModule4 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -10,7 +11,7 @@ T4930.$trModule4 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 20 0}]
 T4930.$trModule4 = "main"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T4930.$trModule3 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -19,7 +20,7 @@ T4930.$trModule3 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T4930.$trModule3 = GHC.Types.TrNameS T4930.$trModule4
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T4930.$trModule2 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -27,7 +28,7 @@ T4930.$trModule2 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 30 0}]
 T4930.$trModule2 = "T4930"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T4930.$trModule1 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -36,7 +37,7 @@ T4930.$trModule1 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T4930.$trModule1 = GHC.Types.TrNameS T4930.$trModule2
 
--- RHS size: {terms: 3, types: 0, coercions: 0}
+-- RHS size: {terms: 3, types: 0, coercions: 0, joins: 0/0}
 T4930.$trModule :: GHC.Types.Module
 [GblId,
  Caf=NoCafRefs,
@@ -47,24 +48,19 @@ T4930.$trModule =
   GHC.Types.Module T4930.$trModule3 T4930.$trModule1
 
 Rec {
--- RHS size: {terms: 23, types: 6, coercions: 0}
+-- RHS size: {terms: 18, types: 4, coercions: 0, joins: 0/0}
 T4930.$wfoo [InlPrag=[0], Occ=LoopBreaker]
   :: GHC.Prim.Int# -> GHC.Prim.Int#
 [GblId, Arity=1, Caf=NoCafRefs, Str=<S,U>]
 T4930.$wfoo =
   \ (ww :: GHC.Prim.Int#) ->
-    case case GHC.Prim.tagToEnum# @ Bool (GHC.Prim.<# ww 5#) of {
-           False -> GHC.Types.I# (GHC.Prim.+# ww 2#);
-           True ->
-             case T4930.$wfoo ww of ww1 { __DEFAULT -> GHC.Types.I# ww1 }
-         }
-    of
-    { GHC.Types.I# ipv ->
-    GHC.Prim.+# ww 5#
+    case GHC.Prim.tagToEnum# @ Bool (GHC.Prim.<# ww 5#) of {
+      False -> GHC.Prim.+# ww 5#;
+      True -> case T4930.$wfoo ww of { __DEFAULT -> GHC.Prim.+# ww 5# }
     }
 end Rec }
 
--- RHS size: {terms: 10, types: 4, coercions: 0}
+-- RHS size: {terms: 10, types: 4, coercions: 0, joins: 0/0}
 foo [InlPrag=INLINE[0]] :: Int -> Int
 [GblId,
  Arity=1,
diff --git a/testsuite/tests/simplCore/should_compile/T5658b.stdout b/testsuite/tests/simplCore/should_compile/T5658b.stdout
index 0cfbf08886..b8626c4cff 100644
--- a/testsuite/tests/simplCore/should_compile/T5658b.stdout
+++ b/testsuite/tests/simplCore/should_compile/T5658b.stdout
@@ -1 +1 @@
-2
+4
diff --git a/testsuite/tests/simplCore/should_compile/T7360.stderr b/testsuite/tests/simplCore/should_compile/T7360.stderr
index 2e387b27bc..b35c39931c 100644
--- a/testsuite/tests/simplCore/should_compile/T7360.stderr
+++ b/testsuite/tests/simplCore/should_compile/T7360.stderr
@@ -1,8 +1,9 @@
 
 ==================== Tidy Core ====================
-Result size of Tidy Core = {terms: 94, types: 48, coercions: 0}
+Result size of Tidy Core
+  = {terms: 94, types: 48, coercions: 0, joins: 0/0}
 
--- RHS size: {terms: 6, types: 3, coercions: 0}
+-- RHS size: {terms: 6, types: 3, coercions: 0, joins: 0/0}
 T7360.$WFoo3 [InlPrag=INLINE] :: Int -> Foo
 [GblId[DataConWrapper],
  Arity=1,
@@ -17,19 +18,19 @@ T7360.$WFoo3 =
   \ (dt [Occ=Once!] :: Int) ->
     case dt of { GHC.Types.I# dt [Occ=Once] -> T7360.Foo3 dt }
 
--- RHS size: {terms: 5, types: 2, coercions: 0}
+-- RHS size: {terms: 5, types: 2, coercions: 0, joins: 0/0}
 fun1 [InlPrag=NOINLINE] :: Foo -> ()
 [GblId, Arity=1, Caf=NoCafRefs, Str=<S,1*U>]
 fun1 = \ (x :: Foo) -> case x of { __DEFAULT -> GHC.Tuple.() }
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T7360.fun5 :: ()
 [GblId,
  Unf=Unf{Src=<vanilla>, TopLvl=True, Value=False, ConLike=False,
          WorkFree=False, Expandable=False, Guidance=IF_ARGS [] 20 0}]
 T7360.fun5 = fun1 T7360.Foo1
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T7360.fun4 :: Int
 [GblId,
  Caf=NoCafRefs,
@@ -38,7 +39,7 @@ T7360.fun4 :: Int
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T7360.fun4 = GHC.Types.I# 0#
 
--- RHS size: {terms: 16, types: 13, coercions: 0}
+-- RHS size: {terms: 16, types: 13, coercions: 0, joins: 0/0}
 fun2 :: forall a. [a] -> ((), Int)
 [GblId,
  Arity=1,
@@ -66,7 +67,7 @@ fun2 =
          }
      })
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T7360.$trModule4 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -74,7 +75,7 @@ T7360.$trModule4 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 20 0}]
 T7360.$trModule4 = "main"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T7360.$trModule3 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -83,7 +84,7 @@ T7360.$trModule3 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T7360.$trModule3 = GHC.Types.TrNameS T7360.$trModule4
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T7360.$trModule2 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -91,7 +92,7 @@ T7360.$trModule2 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 30 0}]
 T7360.$trModule2 = "T7360"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T7360.$trModule1 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -100,7 +101,7 @@ T7360.$trModule1 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T7360.$trModule1 = GHC.Types.TrNameS T7360.$trModule2
 
--- RHS size: {terms: 3, types: 0, coercions: 0}
+-- RHS size: {terms: 3, types: 0, coercions: 0, joins: 0/0}
 T7360.$trModule :: GHC.Types.Module
 [GblId,
  Caf=NoCafRefs,
@@ -110,7 +111,7 @@ T7360.$trModule :: GHC.Types.Module
 T7360.$trModule =
   GHC.Types.Module T7360.$trModule3 T7360.$trModule1
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T7360.$tc'Foo9 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -118,7 +119,7 @@ T7360.$tc'Foo9 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 30 0}]
 T7360.$tc'Foo9 = "'Foo3"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T7360.$tc'Foo8 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -127,7 +128,7 @@ T7360.$tc'Foo8 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T7360.$tc'Foo8 = GHC.Types.TrNameS T7360.$tc'Foo9
 
--- RHS size: {terms: 5, types: 0, coercions: 0}
+-- RHS size: {terms: 5, types: 0, coercions: 0, joins: 0/0}
 T7360.$tc'Foo3 :: GHC.Types.TyCon
 [GblId,
  Caf=NoCafRefs,
@@ -141,7 +142,7 @@ T7360.$tc'Foo3 =
     T7360.$trModule
     T7360.$tc'Foo8
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T7360.$tc'Foo7 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -149,7 +150,7 @@ T7360.$tc'Foo7 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 30 0}]
 T7360.$tc'Foo7 = "'Foo2"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T7360.$tc'Foo6 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -158,7 +159,7 @@ T7360.$tc'Foo6 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T7360.$tc'Foo6 = GHC.Types.TrNameS T7360.$tc'Foo7
 
--- RHS size: {terms: 5, types: 0, coercions: 0}
+-- RHS size: {terms: 5, types: 0, coercions: 0, joins: 0/0}
 T7360.$tc'Foo2 :: GHC.Types.TyCon
 [GblId,
  Caf=NoCafRefs,
@@ -172,7 +173,7 @@ T7360.$tc'Foo2 =
     T7360.$trModule
     T7360.$tc'Foo6
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T7360.$tc'Foo5 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -180,7 +181,7 @@ T7360.$tc'Foo5 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 30 0}]
 T7360.$tc'Foo5 = "'Foo1"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T7360.$tc'Foo4 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -189,7 +190,7 @@ T7360.$tc'Foo4 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T7360.$tc'Foo4 = GHC.Types.TrNameS T7360.$tc'Foo5
 
--- RHS size: {terms: 5, types: 0, coercions: 0}
+-- RHS size: {terms: 5, types: 0, coercions: 0, joins: 0/0}
 T7360.$tc'Foo1 :: GHC.Types.TyCon
 [GblId,
  Caf=NoCafRefs,
@@ -203,7 +204,7 @@ T7360.$tc'Foo1 =
     T7360.$trModule
     T7360.$tc'Foo4
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 T7360.$tcFoo2 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -211,7 +212,7 @@ T7360.$tcFoo2 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 20 0}]
 T7360.$tcFoo2 = "Foo"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 T7360.$tcFoo1 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -220,7 +221,7 @@ T7360.$tcFoo1 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 T7360.$tcFoo1 = GHC.Types.TrNameS T7360.$tcFoo2
 
--- RHS size: {terms: 5, types: 0, coercions: 0}
+-- RHS size: {terms: 5, types: 0, coercions: 0, joins: 0/0}
 T7360.$tcFoo :: GHC.Types.TyCon
 [GblId,
  Caf=NoCafRefs,
diff --git a/testsuite/tests/simplCore/should_compile/T9400.stderr b/testsuite/tests/simplCore/should_compile/T9400.stderr
index 92979b36b1..a8004dce8b 100644
--- a/testsuite/tests/simplCore/should_compile/T9400.stderr
+++ b/testsuite/tests/simplCore/should_compile/T9400.stderr
@@ -1,33 +1,34 @@
 
 ==================== Tidy Core ====================
-Result size of Tidy Core = {terms: 37, types: 22, coercions: 0}
+Result size of Tidy Core
+  = {terms: 37, types: 22, coercions: 0, joins: 0/0}
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $trModule1 :: Addr#
 [GblId, Caf=NoCafRefs]
 $trModule1 = "main"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $trModule2 :: TrName
 [GblId, Caf=NoCafRefs]
 $trModule2 = GHC.Types.TrNameS $trModule1
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 $trModule3 :: Addr#
 [GblId, Caf=NoCafRefs]
 $trModule3 = "T9400"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 $trModule4 :: TrName
 [GblId, Caf=NoCafRefs]
 $trModule4 = GHC.Types.TrNameS $trModule3
 
--- RHS size: {terms: 3, types: 0, coercions: 0}
+-- RHS size: {terms: 3, types: 0, coercions: 0, joins: 0/0}
 T9400.$trModule :: Module
 [GblId, Caf=NoCafRefs]
 T9400.$trModule = GHC.Types.Module $trModule2 $trModule4
 
--- RHS size: {terms: 22, types: 15, coercions: 0}
+-- RHS size: {terms: 22, types: 15, coercions: 0, joins: 0/0}
 main :: IO ()
 [GblId]
 main =
diff --git a/testsuite/tests/simplCore/should_compile/all.T b/testsuite/tests/simplCore/should_compile/all.T
index d63d0d1958..4cc11de737 100644
--- a/testsuite/tests/simplCore/should_compile/all.T
+++ b/testsuite/tests/simplCore/should_compile/all.T
@@ -24,7 +24,6 @@ test('simpl020', [], multimod_compile, ['simpl020', '-v0'])
 test('simpl-T1370', normal, compile, [''])
 test('T2520', normal, compile, [''])
 
-
 test('spec001', when(fast(), skip), compile, [''])
 test('spec002', normal, compile, [''])
 test('spec003', normal, compile, [''])
@@ -99,7 +98,7 @@ test('T4918', [], run_command, ['$MAKE -s --no-print-directory T4918'])
 # result of -ddump-simpl, which is never advertised to
 # be very stable
 test('T4945',
-     expect_broken(4945),
+     normal,
      run_command,
      ['$MAKE -s --no-print-directory T4945'])
 
diff --git a/testsuite/tests/simplCore/should_compile/par01.stderr b/testsuite/tests/simplCore/should_compile/par01.stderr
index 4ccb9d892b..bbcb9ef4fd 100644
--- a/testsuite/tests/simplCore/should_compile/par01.stderr
+++ b/testsuite/tests/simplCore/should_compile/par01.stderr
@@ -1,9 +1,10 @@
 
 ==================== CorePrep ====================
-Result size of CorePrep = {terms: 22, types: 10, coercions: 0}
+Result size of CorePrep
+  = {terms: 22, types: 10, coercions: 0, joins: 0/0}
 
 Rec {
--- RHS size: {terms: 7, types: 3, coercions: 0}
+-- RHS size: {terms: 7, types: 3, coercions: 0, joins: 0/0}
 Par01.depth [Occ=LoopBreaker] :: GHC.Types.Int -> GHC.Types.Int
 [GblId, Arity=1, Caf=NoCafRefs, Str=<L,U>, Unf=OtherCon []]
 Par01.depth =
@@ -13,27 +14,27 @@ Par01.depth =
     }
 end Rec }
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 Par01.$trModule4 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs, Unf=OtherCon []]
 Par01.$trModule4 = "main"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 Par01.$trModule3 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs, Str=m1, Unf=OtherCon []]
 Par01.$trModule3 = GHC.Types.TrNameS Par01.$trModule4
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 Par01.$trModule2 :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs, Unf=OtherCon []]
 Par01.$trModule2 = "Par01"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 Par01.$trModule1 :: GHC.Types.TrName
 [GblId, Caf=NoCafRefs, Str=m1, Unf=OtherCon []]
 Par01.$trModule1 = GHC.Types.TrNameS Par01.$trModule2
 
--- RHS size: {terms: 3, types: 0, coercions: 0}
+-- RHS size: {terms: 3, types: 0, coercions: 0, joins: 0/0}
 Par01.$trModule :: GHC.Types.Module
 [GblId, Caf=NoCafRefs, Str=m, Unf=OtherCon []]
 Par01.$trModule =
diff --git a/testsuite/tests/simplCore/should_compile/spec-inline.stderr b/testsuite/tests/simplCore/should_compile/spec-inline.stderr
index 0de46d181d..dda28c8926 100644
--- a/testsuite/tests/simplCore/should_compile/spec-inline.stderr
+++ b/testsuite/tests/simplCore/should_compile/spec-inline.stderr
@@ -1,8 +1,9 @@
 
 ==================== Tidy Core ====================
-Result size of Tidy Core = {terms: 178, types: 68, coercions: 0}
+Result size of Tidy Core
+  = {terms: 178, types: 68, coercions: 0, joins: 0/2}
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 Roman.$trModule4 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -10,7 +11,7 @@ Roman.$trModule4 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 20 0}]
 Roman.$trModule4 = "main"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 Roman.$trModule3 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -19,7 +20,7 @@ Roman.$trModule3 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 Roman.$trModule3 = GHC.Types.TrNameS Roman.$trModule4
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 Roman.$trModule2 :: GHC.Prim.Addr#
 [GblId,
  Caf=NoCafRefs,
@@ -27,7 +28,7 @@ Roman.$trModule2 :: GHC.Prim.Addr#
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 30 0}]
 Roman.$trModule2 = "Roman"#
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 Roman.$trModule1 :: GHC.Types.TrName
 [GblId,
  Caf=NoCafRefs,
@@ -36,7 +37,7 @@ Roman.$trModule1 :: GHC.Types.TrName
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 Roman.$trModule1 = GHC.Types.TrNameS Roman.$trModule2
 
--- RHS size: {terms: 3, types: 0, coercions: 0}
+-- RHS size: {terms: 3, types: 0, coercions: 0, joins: 0/0}
 Roman.$trModule :: GHC.Types.Module
 [GblId,
  Caf=NoCafRefs,
@@ -46,19 +47,19 @@ Roman.$trModule :: GHC.Types.Module
 Roman.$trModule =
   GHC.Types.Module Roman.$trModule3 Roman.$trModule1
 
--- RHS size: {terms: 1, types: 0, coercions: 0}
+-- RHS size: {terms: 1, types: 0, coercions: 0, joins: 0/0}
 lvl :: GHC.Prim.Addr#
 [GblId, Caf=NoCafRefs]
 lvl = "spec-inline.hs:(19,5)-(29,25)|function go"#
 
--- RHS size: {terms: 2, types: 2, coercions: 0}
+-- RHS size: {terms: 2, types: 2, coercions: 0, joins: 0/0}
 Roman.foo3 :: Int
 [GblId, Str=x]
 Roman.foo3 =
   Control.Exception.Base.patError @ 'GHC.Types.LiftedRep @ Int lvl
 
 Rec {
--- RHS size: {terms: 55, types: 9, coercions: 0}
+-- RHS size: {terms: 55, types: 9, coercions: 0, joins: 0/1}
 Roman.foo_$s$wgo [Occ=LoopBreaker]
   :: GHC.Prim.Int# -> GHC.Prim.Int# -> GHC.Prim.Int#
 [GblId, Arity=2, Caf=NoCafRefs, Str=<S,U><S,U>]
@@ -88,7 +89,7 @@ Roman.foo_$s$wgo =
     }
 end Rec }
 
--- RHS size: {terms: 74, types: 22, coercions: 0}
+-- RHS size: {terms: 74, types: 22, coercions: 0, joins: 0/1}
 Roman.$wgo [InlPrag=[0]] :: Maybe Int -> Maybe Int -> GHC.Prim.Int#
 [GblId,
  Arity=2,
@@ -132,7 +133,7 @@ Roman.$wgo =
         }
     }
 
--- RHS size: {terms: 9, types: 5, coercions: 0}
+-- RHS size: {terms: 9, types: 5, coercions: 0, joins: 0/0}
 Roman.foo_go [InlPrag=INLINE[0]] :: Maybe Int -> Maybe Int -> Int
 [GblId,
  Arity=2,
@@ -146,7 +147,7 @@ Roman.foo_go =
   \ (w :: Maybe Int) (w1 :: Maybe Int) ->
     case Roman.$wgo w w1 of ww { __DEFAULT -> GHC.Types.I# ww }
 
--- RHS size: {terms: 2, types: 0, coercions: 0}
+-- RHS size: {terms: 2, types: 0, coercions: 0, joins: 0/0}
 Roman.foo2 :: Int
 [GblId,
  Caf=NoCafRefs,
@@ -155,7 +156,7 @@ Roman.foo2 :: Int
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 Roman.foo2 = GHC.Types.I# 6#
 
--- RHS size: {terms: 2, types: 1, coercions: 0}
+-- RHS size: {terms: 2, types: 1, coercions: 0, joins: 0/0}
 Roman.foo1 :: Maybe Int
 [GblId,
  Caf=NoCafRefs,
@@ -164,7 +165,7 @@ Roman.foo1 :: Maybe Int
          WorkFree=True, Expandable=True, Guidance=IF_ARGS [] 10 20}]
 Roman.foo1 = GHC.Base.Just @ Int Roman.foo2
 
--- RHS size: {terms: 11, types: 4, coercions: 0}
+-- RHS size: {terms: 11, types: 4, coercions: 0, joins: 0/0}
 foo :: Int -> Int
 [GblId,
  Arity=1,