Install: http://www.7-zip.org/
Source code: http://prdownloads.sourceforge.net/sevenzi...ar.bz2?download
I need pentium 4-optimized version (I`ll pack gigabytes of data, I need this 5-10% (IIRC) of speed gain.
Full Version: Anyone could compile latest 7-zip 4 me?
Noone?
http://www.cs.indiana.edu/~cshei/nb/7-Zip.zip
We had a rainy afternoon today in the midwest, so I unplugged my notebook (got a UPS, but still...) and got to work on this. Whew, 38 projects in the source code! 8 of them didn't compile (mostly crypto ones). Don't even try running this on anything without SSE2, it won't work. I ask just one thing: compare the speed of this with the regular install. I'm wondering if this is actually any faster (and I'm now too tired to figure it out myself).
We had a rainy afternoon today in the midwest, so I unplugged my notebook (got a UPS, but still...) and got to work on this. Whew, 38 projects in the source code! 8 of them didn't compile (mostly crypto ones). Don't even try running this on anything without SSE2, it won't work. I ask just one thing: compare the speed of this with the regular install. I'm wondering if this is actually any faster (and I'm now too tired to figure it out myself).
Test file info:
Size : 23.6 MB (24, 850, 910 bytes)
Contains : 292 Files, 37 Folders
My system info:
CPU: Pentium 4 Willamete 1.4 GHz
RAM: PC800 RDRAM
Create solid achieve enabled
7zip normal
1:41
7zip by chun-yu
1:22
7zip by chun yu (multi threading enabled)
1:30
Thanks Chun Yu, its faster than the non SSE2 opt 7zip
Size : 23.6 MB (24, 850, 910 bytes)
Contains : 292 Files, 37 Folders
My system info:
CPU: Pentium 4 Willamete 1.4 GHz
RAM: PC800 RDRAM
Create solid achieve enabled
7zip normal
1:41
7zip by chun-yu
1:22
7zip by chun yu (multi threading enabled)
1:30
Thanks Chun Yu, its faster than the non SSE2 opt 7zip
P4 2.4GHz
System Bus: 533 Mhz
Archive size: 77mb (FoxServ)
Max compression, solid, switches: mf=bt4b fb=192 d=20m hcf=on
Original: 6:57=417seconds
P4: 4:51=291seconds
P4 was 30,2% better than original. It _IS_ much!
Both archives are NOT bit identicall, both were tested and are OK.
EDIT:
For ZIP (same archive)
Max compression, fb=255 pass=4
P4: 1.52
ORIGINAL: 2.11
Archives are bit identicall.
System Bus: 533 Mhz
Archive size: 77mb (FoxServ)
Max compression, solid, switches: mf=bt4b fb=192 d=20m hcf=on
Original: 6:57=417seconds
P4: 4:51=291seconds
P4 was 30,2% better than original. It _IS_ much!
Both archives are NOT bit identicall, both were tested and are OK.
EDIT:
For ZIP (same archive)
Max compression, fb=255 pass=4
P4: 1.52
ORIGINAL: 2.11
Archives are bit identicall.
QUOTE(Cobra @ Aug 30 2003, 06:15 PM)
Both archives are NOT big identicall, both were tested and are OK.
Is one's size much different than the other?
Difference is in only few bytes...
The important question is: Is the decompressed archive bit-identical to the original tree?
QUOTE(Cobra @ Aug 30 2003, 01:15 PM)
Both archives are NOT bit identicall, both were tested and are OK.
It is bit identiccal when the "multi threading" is on
A few bytes different when the "multi threading" is off
But the "size on disk" are identical
Anyone here with P4 with Hyper-Threading option? Please test optimized version with HT option and w/o this option!
I tested same archive with normal compression, w/o additional switches.
Original: 2:33
P4: 1:48
P4-HT: 2.07
Please note, my CPU is not HT compatibile!
All three archives after decompression gives files bit-identical to source files.
Original: 2:33
P4: 1:48
P4-HT: 2.07
Please note, my CPU is not HT compatibile!
All three archives after decompression gives files bit-identical to source files.
Can anyone make forther testing? For some files P4 and normal speed are equal... 
QUOTE(Igor)
There are some undocumented switches for LZMA
lp - number of bits for Literal Pos context - from 0 to 4.
lc - number of bits in previous byte for Literal Context - from 0 to 8
pb - number of bits for Pos Context for other purposes - from 0 to 4.
Default values for them are
-m0lp=0 -m0lc=3 -m0pb=2
lp and pb are intended for periodical data when period is equal 2^value (where lp=value or pb=value).
lp - number of bits for Literal Pos context - from 0 to 4.
lc - number of bits in previous byte for Literal Context - from 0 to 8
pb - number of bits for Pos Context for other purposes - from 0 to 4.
Default values for them are
-m0lp=0 -m0lc=3 -m0pb=2
lp and pb are intended for periodical data when period is equal 2^value (where lp=value or pb=value).
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.