Re: Skybuck's Universal Code 4 (memory efficient)

"Skybuck Flying" <> wrote in message news:...
>
> "Skybuck Flying" <> wrote in message news:...
>>
>> "Skybuck Flying" <> wrote in message
>> news:f3dsdk$p80$...
>>>
>>> "MooseFET" <> wrote in message
>>> news: oups.com...
>>>> On May 27, 8:20 am, "Skybuck Flying" <s...@hotmail.com> wrote:
>>>>> How would you encode a larger example with your system, for example
>>>>> the
>>>>> following data bits:
>>>>>
>>>>> 111010101010100110100101101110011
>>>> 33 Bits by my count.
>>>>
>>>>>
>>>>> Maybe that would clearify your idea a bit
>>>>
>>>> I stopped after 3 bits, so I'll pick up at 4 bits and again.
>>>>
>>>> The first 4 bits of the prefix are already stated as being 1111 so we
>>>> only need to deal with the next 6 bits.
>>
>> from previous post:
>>
>>> > 1100 - 0 plus the following bit as a pair
>>> > 1101 - 1 plus the following bit as a pair
>>> > 1110 - Three bits follow
>>> > 1111 - All other cases
>>
>>>> 00XXXX - The 4 bit value XXXX
>>>> 010XXX - Two more bits follow making a total of 5 bits
>>>> 0110XX - Four more bits follow making 6 bits
>>>> 01110X - Six bits follow making 7 bits
>>>> 011110 - 8 bits follow
>>>> 10NNNN - (9+NNNN) bits follow 9..24
>>>> 110NNN - Two more of N follow then (25+N) bits *
>>>> 1110NN - 4 more bits of N follow then (57+N) bits
>>>> 11110N - 6 more bits of N follow then (121+N) bits
>>>> 111111 - All other cases

Ok, I guess this table doesn't show the whole picture.

This table is supposed to be:

1100 0X
1101 1X
1110 XXX
1111 00XX XX
1111 010X XXXX
1111 0110 XXXX XX
1111 0111 0XXX XXXX
1111 0111 10XX XXXX XX *
1111 10NN NNNN NNNN NNN
1111 110N NNNN
1111 1110 NNNN NN
1111 1111 0NNN NNNN
1111 1111 11 - All other cases

* 10 bits overhead for only 8 bits.

Doesn't seem like your encoding is more efficient.

My encoding needs:

0x1 + 0x2 + 0x4 + 1x8 = 4 bits + 4 for markers = 8 bits overhead.

Total bits for my encoding: 16.
Total bits for your encoding: 18.

Your encoding uses 2 more bits !

Since I use a binary encoding system for the length, the longer the data
bits are the more efficient my encoding system will prove to be.

Binary is exponential after all

The "unary-like" marker bits for the binary length bits will prove to be a
very small overhead =D

Bye,
Bye,
Skybuck.