I would be interested to know what people think of this compressor.
Dwing's UDA v0.301
Well... it's only a slightly tweaked 0.300... I still prefer WinUDA 0.291 for it's GUI
Do not trust to UDA/WinUDA too much - it is just slightly modified PAQ6/8 compressor. I am familiar with sources of some versions of UDA. The older ones uses slightly modified PAQ6 with some "optimizations" like replacement of divisions to the shifts. For example:
a = b / 2;
a = b >> 1;
Will produce the same result, but the second variant is faster. However, with modern compilers, such trick is useless - a good compiler performs such optimizations automatically...
I do not know what is the last UDA, but looks like it is the simplified PAQ8 with some ASM insertions...
I don't understand why Dwing has omitted the amazing JPEG compression. This for me, is what makes PAQ8 much more interesting than previous versions.
I'm hoping that Dwing will very soon be releasing a more sophisticated version of WinUDA based on his modified PAQ8 (UDA) engine.
I know that UDA uses only a little bit changed PAQ algorithm. But it is similar to all these programs using various implementations of Deflate algorithm, isnt it?Originally Posted by encode
This means either that PAQs arent compiled with state-of-the-art compilers (because last two UDAs are about 6 times faster) or that the UDA engine has been tweaked for slightly worse compression and kinda faster speedOriginally Posted by encode
UDA 0.301 differs from 0.300 by improved IA32 and 24bitBMP+TIF filters (copied from readme) and unlike the older versions is available with source code included.
Probably, to get more speed...Originally Posted by LovePimple
Not quite right. Deflate is a standardized compression algorithm. PAQ is an experimental compression engine - i.e. the interest in compression. IMHO, UDA - rip... (but better rip compared to the KGB)Originally Posted by Black_Fox
ALL modern compilers (gcc, bcc32, ...) have such optimizations. Dwing just removed some models to improve speed in cost of compression. I think Matt Mahoney is the number one in data compression world, and he wrote efficient enough code. I do not think Dwing is more experienced in data compression than Matt...Originally Posted by Black_Fox
I agree.Originally Posted by encode