Don't eta-expand unsaturated primops

Previously, as described in Note [Primop wrappers], `hasNoBinding` would
return False in the case of `PrimOpId`s. This would result in eta
expansion of unsaturated primop applications during CorePrep. Not only
did this expansion result in unnecessary allocations, but it also meant
lead to rather nasty inconsistencies between the CAFfy-ness
determinations made by TidyPgm and CorePrep.

This fixes #16846.

1 files changed, 9 insertions, 14 deletions

diff --git a/compiler/basicTypes/Id.hs b/compiler/basicTypes/Id.hs
index e2dfe925b1..4dceb4bc03 100644
--- a/compiler/basicTypes/Id.hs
+++ b/compiler/basicTypes/Id.hs
@@ -524,9 +524,17 @@ hasNoBinding :: Id -> Bool
 -- Data constructor workers used to be things of this kind, but
 -- they aren't any more.  Instead, we inject a binding for
 -- them at the CorePrep stage.
+--
+-- 'PrimOpId's also used to be of this kind. See Note [Primop wrappers] in PrimOp.hs.
+-- for the history of this.
+--
+-- Note that CorePrep currently eta expands things no-binding things and this
+-- can cause quite subtle bugs. See Note [Eta expansion of hasNoBinding things
+-- in CorePrep] in CorePrep for details.
+--
 -- EXCEPT: unboxed tuples, which definitely have no binding
 hasNoBinding id = case Var.idDetails id of
-                        PrimOpId _       -> True        -- See Note [Primop wrappers]
+                        PrimOpId _       -> False   -- See Note [Primop wrappers] in PrimOp.hs
                         FCallId _        -> True
                         DataConWorkId dc -> isUnboxedTupleCon dc || isUnboxedSumCon dc
                         _                -> isCompulsoryUnfolding (idUnfolding id)
@@ -570,19 +578,6 @@ The easiest way to do this is for hasNoBinding to return True of all things
 that have compulsory unfolding.  Some Ids with a compulsory unfolding also
 have a binding, but it does not harm to say they don't here, and its a very
 simple way to fix #14561.
-
-Note [Primop wrappers]
-~~~~~~~~~~~~~~~~~~~~~~
-Currently hasNoBinding claims that PrimOpIds don't have a curried
-function definition.  But actually they do, in GHC.PrimopWrappers,
-which is auto-generated from prelude/primops.txt.pp.  So actually, hasNoBinding
-could return 'False' for PrimOpIds.
-
-But we'd need to add something in CoreToStg to swizzle any unsaturated
-applications of GHC.Prim.plusInt# to GHC.PrimopWrappers.plusInt#.
-
-Nota Bene: GHC.PrimopWrappers is needed *regardless*, because it's
-used by GHCi, which does not implement primops direct at all.
 -}
 
 isDeadBinder :: Id -> Bool