Use -funfolding-use-threshold=150 in fileio benchmark

The lines-unlines benchmark does not fuse well without this option.

docs/Build.md

@@ -13,9 +13,11 @@ the `SpecConstr` optimization in more cases.
 In some cases, you may need to use `-funfolding-use-threshold` to make sure
 that the combinators fuse. The default value of this option is `60`. Increasing
 this default value can be detrimental in general, therefore, increase only if
-you suspect an issue. As an example, a value of `75` is necessary and
-sufficient to fuse `S.chunksOf n (A.writeN n)`. Hopefully GHC will fix this so
-that it is not needed in future.
+you suspect an issue.  Hopefully GHC will fix this so that it is not needed in
+future.  Some known examples:
+
+* use a value of `75` to fuse `S.chunksOf n (A.writeN n)`.
+* use a value of `150` to fully fuse `S.splitSuffixOn`.
 
 At the very least `-O` compilation option is required. In some cases, the
 program may exhibit memory hog with default optimization options.  For example,

src/Streamly/Benchmark/FileIO/Stream.hs

@@ -59,6 +59,7 @@ import qualified Streamly.Internal as Internal
 import qualified Streamly.Streams.StreamD as D
 
 #ifdef INSPECTION
+import Foreign.Storable (Storable)
 import Streamly.Streams.StreamD.Type (Step(..), GroupState)
 import Test.Inspection
 #endif
@@ -233,6 +234,7 @@ inspect $ 'chunksOfD `hasNoType` ''AT.FlattenState
 inspect $ 'chunksOfD `hasNoType` ''D.ConcatMapUState
 #endif
 
+-- XXX splitSuffixOn requires -funfolding-use-threshold=150 for better fusion
 -- | Lines and unlines
 {-# INLINE linesUnlinesCopy #-}
 linesUnlinesCopy :: Handle -> Handle -> IO ()
@@ -245,7 +247,7 @@ linesUnlinesCopy inh outh =
       $ FH.read inh
 
 #ifdef INSPECTION
--- inspect $ hasNoTypeClasses 'linesUnlinesCopy
+inspect $ hasNoTypeClassesExcept 'linesUnlinesCopy [''Storable]
 -- inspect $ 'linesUnlinesCopy `hasNoType` ''Step
 -- inspect $ 'linesUnlinesCopy `hasNoType` ''AT.FlattenState
 -- inspect $ 'linesUnlinesCopy `hasNoType` ''D.ConcatMapUState

src/Streamly/Streams/StreamD.hs

@@ -1085,6 +1085,7 @@ splitBy predicate f (Stream step state) = Stream (step' f) (Just state)
 
     step' _ _ Nothing = return Stop
 
+-- XXX requires -funfolding-use-threshold=150 in lines-unlines benchmark
 {-# INLINE_NORMAL splitSuffixBy #-}
 splitSuffixBy :: Monad m
     => (a -> Bool) -> Fold m a b -> Stream m a -> Stream m b

streamly.cabal

@@ -598,6 +598,7 @@ benchmark array
 benchmark fileio
   import: bench-options
   type: exitcode-stdio-1.0
+  ghc-options: -funfolding-use-threshold=150
   hs-source-dirs: benchmark
   main-is: FileIO.hs
   if flag(benchmark)