naturalToWord vs 'clampNaturalToWord'
The short of it is, clampNaturalToWord :: Natural → Word clampNaturalToWord n = case n of NatS# w -> W# w _ -> maxBound results in much better space and time use than naturalToWord, which always takes the least significant Word of the Natural. Because naturalToWord takes the least significant Word, you always have to check for size before conversion. Even though you do check, GHC has to assume that, after conversion, the big Natural could have resulted in a small Word and will uselessly duplicate all branches for both cases. Is there anything wrong with clamping by default in this case? --Keith Sent from my phone with K-9 Mail.
We can easily add this function to ghc-bignum. I've opened https://gitlab.haskell.org/ghc/ghc/-/issues/18697 to track this. Sylvain On 16/09/2020 04:19, Keith wrote:
The short of it is,
clampNaturalToWord :: Natural → Word clampNaturalToWord n = case n of NatS# w -> W# w _ -> maxBound
results in much better space and time use than naturalToWord, which always takes the least significant Word of the Natural.
Because naturalToWord takes the least significant Word, you always have to check for size before conversion. Even though you do check, GHC has to assume that, after conversion, the big Natural could have resulted in a small Word and will uselessly duplicate all branches for both cases.
Is there anything wrong with clamping by default in this case?
--Keith
Sent from my phone with K-9 Mail.
_______________________________________________ Libraries mailing list Libraries@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries
Awsome: it’d be good to provide the explicitly word32 and word64 flavors
too for multi platform portability.
On Wed, Sep 16, 2020 at 4:59 AM Sylvain Henry
We can easily add this function to ghc-bignum.
I've opened https://gitlab.haskell.org/ghc/ghc/-/issues/18697 to
track this.
Sylvain
On 16/09/2020 04:19, Keith wrote:
The short of it is,
clampNaturalToWord :: Natural → Word
clampNaturalToWord n = case n of
NatS# w -> W# w
_ -> maxBound
results in much better space and time use than naturalToWord,
which always takes the least significant Word of the Natural.
Because naturalToWord takes the least significant Word, you always
have to check for size before conversion. Even though you do
check, GHC has to assume that, after conversion, the big Natural
could have resulted in a small Word and will uselessly duplicate
all branches for both cases.
Is there anything wrong with clamping by default in this case?
--Keith
Sent from my phone with K-9 Mail.
_______________________________________________
Libraries mailing list
Libraries@haskell.org
http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries
_______________________________________________
Libraries mailing list
Libraries@haskell.org
participants (3)
-
Carter Schonwald -
Keith -
Sylvain Henry