Nice.
Except, I prefer shift to take the number of bits as a left argument.
Also, there's still the matter of doing it tacitly.
So, here's my rephrasing of your excellent work, for the explicit mix:
top=: <:2^32
SH =: top AND 34 b. NB. 32 bit shift
MP =: top AND + NB. 32 bit addition ("modplus")
emix =: verb define
for_j. 11 _2 8 _16 10 _4 8 _9 do.
'a b c d e f g h' =. y
i=. a XOR j SH b
y =. (b MP c), c, (d MP i), e, f, g, h, i
end.
)
And, here is a tacit equivalent:
rfold=: 1 :'u&.>/@,&.:(<"_1),:'
tmix=: _9 8 _4 10 _16 8 _2 11 umix rfold ]
imix=: 1 }. ], (0,[) XOR/@SH 2 {. ]
ymix=: MP/ .*&(8 8$1(8 58}),=i.8)
umix=: ymix@imix
Honestly, though, I prefer the explicit version. It's simpler, more
concise, and faster. But I think that that is more a direct
consequence of the (somewhat arbitrary) design of the algorithm than
anything else.
Thanks,
--
Raul
On Sat, Sep 5, 2015 at 12:33 AM, Michal Wallace
<[email protected]> wrote:
> It's just a hashing algorithm, mixing up the bits of data in a
> deterministic but irreversible way.
>
> If you notice that the "alphabet-distance" between the variable names on
> each line is always the same,
> my implementation might make more sense. Since the offsets are always the
> same, I'm just rotating the
> array and re-applying the same logic.
>
> Here's the c code that will tell us what the result should be for mix i.8:
>
> #include <stdio.h>
> int main() {
> int a,b,c,d,e,f,g,h;
> a=0;b=1;c=2;d=3;e=4;f=5;g=6;h=7;
> a^=b<<11; d+=a; b+=c;
> b^=c>>2; e+=b; c+=d;
> c^=d<<8; f+=c; d+=e;
> d^=e>>16; g+=d; e+=f;
> e^=f<<10; h+=e; f+=g;
> f^=g>>4; a+=f; g+=h;
> g^=h<<8; b+=g; h+=a;
> h^=a>>9; c+=h; a+=b;
> // a b c d e f g h
> printf("%d %d %d %d %d %d %d %d\n", a,b,c,d,e,f,g,h);
> }
>
> Answer:
>
> 68112070 67580715 540296643 2058 539239185 529307 67580712 539769534
>
> Here's the corrected code:
>
> top=: <:2^32
>
> SH =: top AND (34 b.)~ NB. 32 bit shift
>
> MP =: top AND + NB. 32 bit addition ("modplus")
>
> mix =: verb define
>
> for_i. 11 _2 8 _16 10 _4 8 _9 do.
>
> 'a b c d e f g h' =. y
>
> x =. a XOR b SH i
>
> y =. 1 |. x, (b MP c), c, (d MP x), e, f, g, h
>
> end.
>
> )
>
>
> assert (mix i.8) -: 68112070 67580715 540296643 2058 539239185 529307
> 67580712 539769534
>
>
>
>
>
>
>
> On Fri, Sep 4, 2015 at 10:34 PM, Dan Bron <[email protected]> wrote:
>
>> Michal Wallace wrote:
>> > I don't know whether or not this produces the correct results because I
>> > don't have any test data, but...
>>
>>
>> Yeah, that is troublesome. Unfortunately, it’s the same catch-22 I’m in.
>>
>> I’m transliterating the C code here:
>>
>> http://rosettacode.org/wiki/The_ISAAC_Cipher#C <
>> http://rosettacode.org/wiki/The_ISAAC_Cipher#C>
>>
>> as directly as possible into J, so that I can get a working program which
>> produces the expected outputs for the given inputs.
>>
>> Once I have a working program that I can test, interrogate, and reason
>> about, I’ll be in a much better position to refactor the code into
>> idiomatic, and, hopefully, elegant J.
>>
>> But the very reason I have to do it this cart-before-horse way is because
>> I don’t (yet) understand the algorithm on a conceptual level. So I’m
>> starting from the code.
>>
>> I guess what I was asking for in my previous email was for someone who
>> does or can easily grok the concepts underlying the code to express them in
>> J (which is a language I speak, so such code would teach me those concepts).
>>
>> Barring that, someone who is confident enough in his C to trust in his
>> translation of the macro would also suffice.
>>
>> (One big obstacle here, and I think more broadly to the lack of adoption
>> of ISAAC the author of that article laments is the majority of
>> easily-accessible artifacts dealing with it are code, rather than prose.)
>>
>> -Dan
>> ----------------------------------------------------------------------
>> For information about J forums see http://www.jsoftware.com/forums.htm
>>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
