RE: calculating CRC
- From: William Adams <william_a_adams@xxxxxxx>
- To: <luajit@xxxxxxxxxxxxx>
- Date: Wed, 30 May 2012 16:18:11 +0000
Thanks for the pointers Mike. I'll take a dump and see what I get. :-)
I've debated using the #. I guess that's more universal as it will call a
length metamethod if it exists.
I've converted to the int array, and use the cast. I have yet to test on an
extremely large data set, but it just feels cleaner, meaner, and morally
correct.
-- William
===============================
- Shaping clay is easier than digging it out of the ground.
http://williamaadams.wordpress.com
http://www.thingiverse.com/WilliamAAdams
https://github.com/Wiladams
----------------------------------------
> Date: Wed, 30 May 2012 15:15:31 +0200
> From: mike-1205@xxxxxxxxxx
> To: luajit@xxxxxxxxxxxxx
> Subject: Re: calculating CRC
>
> William Adams wrote:
> > 1) Would I be better off doing a cast of 's' in the beginning,
> > and then using simple array access after that? Any great
> > advantage over just doing the getbytes? My assumption is it
> > would be faster, but I don't know of other tradeoffs.
>
> It should be slightly faster. Not all range checks can be
> eliminated for string.byte().
>
> > 2) Should I switch the Digest.consts table to being an array of
> > in32_t instead of a Lua table? Will it be more efficient in
> > some way, or some other benefit?
>
> It avoids bounds checks, type checks and plenty of double -> int
> conversions (in single-number mode, default on x86/x64). So yes,
> it would make a difference.
>
> BTW: string.len(s) is compatibility from Lua 5.0. Better use #s.
>
> BTW#2: Try -jdump on a sample snippet only testing this code. Even
> if you don't know much about assembler, you can still judge the
> length of the machine code inside the 'LOOP'. Shorter is better,
> less branches and less loads/stores are better. Beware: it's
> addictive to stare at compiler output. :-)
>
> --Mike
>
Other related posts:
