On Fri, 5 Dec 2008, Ryan Ingram wrote:
You're testing the interpreted code, so it's not surprising that the naive version performs better; the interpretive overhead only applies to your bit of glue code.
I wanted to avoid the optimizer to do clever things itself.
Alternatively, at least compile the module with optimizations before running it in ghci:
ryani$ ghc -ddump-simpl -O2 -c foldlr.hs >foldlr.core (This gives you "functional assembly language" to look at for examining code generation)
ryani$ ghci foldlr.hs [...] Prelude FoldLR> :set +s Prelude FoldLR> test (1000000,'a') (0.39 secs, 70852332 bytes) Prelude FoldLR> testNaive (1000000,'a') (0.42 secs, 105383824 bytes)
There is still no clear advantage of foldl'r compared to foldl'rNaive, is it?