QuickLZ 1.10, The Fastest Just Got Faster
Posted by Sachin Garg on 27th November 2006 | Permanent Link
Since the time we first noticed it, QuickLZ has come a long way (like it now has its own domain name). To quote its author Lasse Reinhold, “Version 1.10 is out - kicks LZO’s b*** at decompression speed ;-)”
New in this version:
Now compatible with x86, x64, IA64, Alpha, SPARC, MIPS, POWER and the like. For x86/x64: Increased average speed of compression by 3 Mbyte/s and decompression by 19 Mbyte/s on the test machines. Programming interface: Some arguments are now void* instead of char*. Bugs: None (upgrade not critical).
If you are looking for more than decent compression at breathtaking speeds, look no further (if you are not convinced by results on QuickLZ homepage, check these).
December 1st, 2006 at 10:02 pm
Hello Sachin and Mark,
i am from Amsterdam and stay since 4 years in South Africa and i hope you can give me some directions.
India and S Africa do have a few things in common,
amongst others large rural parts,poor telecom infrstructure, low income and governments aiming to
improve it.Here they start fyi in 2008 with terrestrial digital radio / tv broacasting which will
be added to the analog signal until 2015.
Seperately either by wifi / wimax / landline the
telecom infrastr. will be improved.
i was wondering if the tv / radio signal will be done
with internet a VOIP phone feature can be added and
the question is if this is a realistic option on basis
of a municipal network with features like p2p via hdd and CPU,FON,Gridnetwork,I-Share , improving en/decoding
compaction, mesh, and the space for improvements,large
network, wht would you think would be the required
bitrates for streaming Standard definition via the
STB or mobile VOIP phone (roughly 3 / per tv - house)
apart from wifi/wimax there is DSL which will be upgraded, think they intend to shorten distance to the
houses.a broadband phone would allow some payment options.
As you know the BBC broadcasted radio and visual signal
a while ago, and Fraunhofer (has some nice decoders)
seems to have a range of transmitters, point is that
for the low income it must cost twice nothing and a basic access to internet would be great, innocense is
serious problem.i saw China starts with the AVS decoder, is cheaper as AVC.
i heard they decided for an own HD DVD stndrd and i herd a 24 yer old Indian student invented a new type
of dvd recording, do you have heard about it ?
Have a nice day,
Looking 4ward to hear,
Peter Ras
South Africa
December 2nd, 2006 at 7:40 pm
I can ofcourse share what I know and how things are in India. But myself being in an urban area, my view of situation in rural areas might not be all correct, and the fact that different rural areas are at different levels of development makes it more confusing :-)
There are places where infrastructure is more-or-less at par with developed counties’ and places where things are, to put it lightly, not too good.
I think terrestrial broadcasts in India are all analog, I haven’t heard of digital terrestrial broadcasts happening here anytime soon.
But we have satellite digital broadcasts which are distributed in local areas by cable-network-operators. Availability of these is limited to ‘developed’ areas which includes many rural areas too.
Only recently we are now having direct-to-home satellite broadcasts available, which ofcourse is digital and is technically ‘available’ everywhere but its affordability (both equipment and subscription costs) in rural areas is questionable.
Bitrates depend on codecs used, check this or this for bitrates with H.264
DSL is the primary broadband internet access means in Indian urban areas. Remote areas need to use radio towers or satellite access like VSAT.
In small towns where telephone exchanges cannot support DSL, we are having private operators who setup VSAT etc uplink and distribute hi-speed connections using wide-area Ethernet based networks.
I am not sure if AVS will be a good choice, atleast not outside China. You might want check out these posts on AVS at c10n
Yes, they are planning for a chinese only HD DVD standard and from what I have heard, only difference it has is it adds their AVS codec to the list of supported codecs :-)
That rainbow technology by the Indian student is a typical under-researched-over-hyped technology. And most of the hype is about how obviously flawed the concept is.
Another remarkable development is that China plans to provide free satellite digital broadcasts all over the country. They recently launched a satellite for this which unfortunately failed to deploy due to mechanical failure, but they will ofcourse try again.
December 24th, 2006 at 6:18 am
QuickLZ 1.10 has now been ported to C# by Shane Bryldt: http://www.codeproject.com/useritems/ManagedQLZ.asp
December 24th, 2006 at 8:40 am
This looks cool, congrats.
And Merry Christmas!!!
January 10th, 2007 at 11:37 pm
Now with streaming compression suitable for packets down to 200-300 bytes :-)
January 17th, 2007 at 9:50 am
I’m still waiting for a Linux port. LZOP’s just not doing it for me.
January 18th, 2007 at 2:51 pm
I’ll make the demo project gcc compatible around the 28′th and compile it for a few platforms.
January 29th, 2007 at 9:32 am
As promised - now the demo project is gcc compatible and pre-compiled for Linux i386.
March 16th, 2007 at 7:03 am
Demo project can now compress multiple files/directories and support file search patterns. Source available.
I think it’s the first multiple-file archiver (except some archivers in combination with tar, perhaps?) which is disk I/O bound! :)
April 19th, 2007 at 4:23 am
Version 1.30 beta is out, finally with variable compression ratio (popular request): http://www.quicklz.com/beta.html
Level 0 is both compressing faster and better than 1.20 and is using an interesting algorithm (explanation requires knowlege about LZ77/LZSS compression):
QuickLZ is using LZSS compression, but in level 0, when a string match is found, the hash value that found the match is outputted instead of the offset. At decompression the hash table is being reconstructed so that the input hash value can be translated to an offset by a lookup in the hash table.
It shares some ideas with the LZP-1 algorithm which, instead of outputting the offset, outputs the first byte of the match uncompressed. The advantage of both algorithm is faster and better compression on the cost of decompression speed.
Compiled binaries for Windows are abailable on the website.
March 20th, 2008 at 3:32 am
Been a while :)
Now with C# and Java ports, and the new interesting level 2.
March 20th, 2008 at 12:50 pm
Interesting, I am amazed by the speed difference in Java, C# and C versions.
IMHO for simple mathematics (like in compression algorithms) difference shouldn’t have been that significant.
March 30th, 2008 at 4:29 am
The compressor and decompressor are not using much math, just a single xor and a couple of adds and subs.
I think what’s holding it back is bounds checking and lack of pointers (I didn’t want to use unsafe code), and, for the Java version of compression, lack of unsigned integers (see the hack in fastreadN()).