If you're still using
pip =: (256 #. (-&128 @: {.) , }.) @: (".;._2@:,&'.')
to encode your input data, giving eg
pip"1 ] 3{.ipst
1590633180 365671447 69450821
you should find an improvement in time by deferring the
_128 offset, eg
((_2147483648 * * )<.@+])256 #. (".;._2@:,&'.')"1 ] 3{.ipst
1590633180 365671447 69450821
This gives the same results as my suggested stuff with
shift, provided the same offset is used.
Mike
On 14/12/2011 1:30 PM, Alexander Epifanov wrote:
> I was trying to make shift, but i does not work correct, bacause first
> bit is sign, that is why 255.255.255.255 is -1, and 0.0.0.0 is 1 and
> it is not correct to compare them.
> 0.0.0.0 should be minint, 255.255.255.255 - maxint.
>
> Regards,
>
> On Wed, Dec 14, 2011 at 1:22 PM, Mike Day<[email protected]> wrote:
>> Your problem with deal is easily dealt (!) with:
>> The vocabulary entry for dyadic ?, "deal" is
>>
>> "x ? yis a list ofxitems randomly chosen/without repetition/fromi.y"
>>
>> So in particular, 5?5 is a random permutation of i.5 .
>>
>> The way to get what I think you want is
>>
>> ?10#5 NB. eg roll 10 5-sided dice
>> 4 2 0 0 2 0 3 1 4 1
>>
>> Though why you should want to generate ips when you
>> have so many already escapes me!
>>
>> As for "pip" etc, you _can_ use 32-bit integers in this fashion:
>>
>> NB. encode, ie shift the 4 elements left 3 2 1 0 bytes
>> 24 16 8 0 +/@:(33 b.) 148 99 57 121
>> _1805436551
>>
>> NB. you presumably need to be able to recover the input
>> NB. so you need this
>> mask =: 24 16 8 0 (33 b.) 255
>>
>> NB. to decode, ie unpack the 32-bit integer
>> _24 _16 _8 0 (33 b.) mask (17 b.) _1805436551
>> 148 99 57 121
>>
>> This is sort of equivalent to pip:
>> pip '148.99.57.121'
>> 2489530745
>> x:1805436551 + 2489530745
>> 4294967296
>> 2x^32
>> 4294967296
>>
>> If you're content to use such integers, you might find
>> some more improvement by avoiding the use of extended
>> integers.
>>
>> It's surely surprising that J significantly outperforms K .
>> Have you asked the K-gurus what they think, in the k4
>> forum?
>>
>> Mike
>>
>> On 14/12/2011 10:14 AM, Alexander Epifanov wrote:
>>> Hello again,
>>> Here my optimized version with ints, its 1.5 faster then float solution.
>>> http://dl.dropbox.com/u/34917039/ip.j
>>>
>>> The main thing: it is much faster than K!
>>>
>>> Could you let me know, is it possible to optimize time of "f" function?
>>> I tryed to represent IP like 4 ints: 192 168 1 1, but this solution
>>> works much slower.
>>>
>>> ips.txt: file with ip addresses
>>> ipr.txt: file with ip ranges: ip,ip
>>>
>>> And the other thing I cannot understand:
>>> 5?5
>>> 4 1 0 2 3
>>> 10?5
>>> |domain error
>>> | 10 ?5
>>> ???
>>>
>>> Thank you,
>>>
>>>
>> [snipped the rest]
>> ----------------------------------------------------------------------
>> For information about J forums see http://www.jsoftware.com/forums.htm
>
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
