Just to see if we're talking about the same, here are the two versions of
the program I'm comparing
https://gist.github.com/michaelt/c8789970e8f096e56fff
chunkX.hs is the 'idiomatic' version, but I bring the definition of splitAt
into the module to make it
easier to compare to the other. chunkX just differs by pattern matching on
the Proxy constructors.
chunkX.hs takes half-again as long as chunkZ.hs,
real 0m5.332s -- with next
user 0m5.247s
sys 0m0.064s
real 0m3.631s -- without next
user 0m3.541s
sys 0m0.060s
Here are the two chunks of core, excluding what was the same
https://gist.github.com/michaelt/8f82c82a96cf77f162da
`next` is brought into the core for chunkX.hs , since it is 'inlinable',
and is specialized to IO, if I understand.
But it is a separate 'loop'. The loop for splitAt then refers to it, and
performs monadic extraction. It is
pretty much the same as the written program.
The core for the chunkZ.hs invades the written program much more.
I'm not sure how well I can articulate or appraise the differences.
--
You received this message because you are subscribed to the Google Groups
"Haskell Pipes" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to haskell-pipes+unsubscr...@googlegroups.com.
To post to this group, send email to haskell-pipes@googlegroups.com.
