Canonicalize bignum literals

Before this patch Integer and Natural literals were desugared into "real"
Core in Core prep. Now we desugar them directly into their final ConApp
form in HsToCore. We only keep the double representation for BigNat#
(literals larger than a machine Word/Int) which are still desugared in
Core prep.

Using the final form directly allows case-of-known-constructor to fire
for bignum literals, fixing #20245.

Slight increase (+2.3) in T4801 which is a pathological case with
Integer literals.

Metric Increase:
    T4801
    T11545

1 files changed, 3 insertions, 5 deletions

diff --git a/compiler/GHC/StgToByteCode.hs b/compiler/GHC/StgToByteCode.hs
index 1ba0687a9b..e056dadc2b 100644
--- a/compiler/GHC/StgToByteCode.hs
+++ b/compiler/GHC/StgToByteCode.hs
@@ -1716,11 +1716,9 @@ pushLiteral padded lit =
           LitNumWord32  -> code Word32Rep
           LitNumInt64   -> code Int64Rep
           LitNumWord64  -> code Word64Rep
-          -- No LitInteger's or LitNatural's should be left by the time this is
-          -- called. CorePrep should have converted them all to a real core
-          -- representation.
-          LitNumInteger -> panic "pushAtom: LitInteger"
-          LitNumNatural -> panic "pushAtom: LitNatural"
+          -- No LitNumBigNat should be left by the time this is called. CorePrep
+          -- should have converted them all to a real core representation.
+          LitNumBigNat  -> panic "pushAtom: LitNumBigNat"
 
 -- | Push an atom for constructor (i.e., PACK instruction) onto the stack.
 -- This is slightly different to @pushAtom@ due to the fact that we allow