Best array population & access optimization

[Dar Scott]

On Friday, June 20, 2003, at 07:37 PM, Mark Brownell wrote:

> unsigned long parray[]>From C & C++
> 0x243f6a88L, 0x85a308d3L, 0x13198a2eL, 0x03707344L, 0xa4093822L, 
> 0x299f31d0L,
> 0x082efa98L, 0xec4e6c89L, 0x452821e6L, 0x38d01377L, 0xbe5466cfL, 
> 0x34e90c6cL,
> 0xc0ac29b7L, 0xc97c50ddL, 0x3f84d5b5L, 0xb5470917L, 0x9216d5d9L, 
> 0x8979fb1bL,
>
> P_array From VB 5 version
> H243F6A88, H85A308D3, H13198A2E, H3707344, HA4093822, H299F31D0, 
> H82EFA98, HEC4E6C89,
> H452821E6, H38D01377, HBE5466CF, H34E90C6C, HC0AC29B7, HC97C50DD, 
> H3F84D5B5, HB5470917,
> H9216D5D9, H8979FB1B
>
> So signed numbers exist in the arrays of some versions in use.

I was thinking of the values rather than the variable types.

An signed variable that is greater than 32 bits can hold 32-bit 
unsigned numbers.

However, since Blowfish primitives are xor and addition and since most 
signed numbers are 2-complement it might not matter.

Even so, in Revolution once you bitAnd or bitXor, I would have expected 
unsigned values.  I have only seen signed values come up in some uses 
of baseCovert().

If you are interested in speeding up your Blowfish, I'd look at 
binaryEncode and binaryDecode for the 4-char to/from 32-bit number.  
That 32-bit number will really get converted to 8-byte floating, but 
there is no loss; I believe it should behave just as you expect.

I would also use those for processing P-box hex tables.

I always get byte order mixed up.

Dar Scott