DOS ain't dead - 7-Zip 9.09 beta

Rugxulo

Usono,
13.12.2009, 10:41

7-Zip 9.09 beta (Announce)

Igor Pavlov released 7-Zip 9.09 beta on 12 December 2009.

Website: http://sourceforge.net/projects/sevenzip
Download: http://sourceforge.net/projects/sevenzip/files/7-Zip/9.09%20beta/7za909.zip/download

> What's new after 7-Zip 9.07 beta:
>
> * 7-Zip now can unpack Apple Partition Map (APM) disk images.
> * Speed optimizations in AES code for Intel's 32nm CPUs.
> * Speed optimizations in CRC calculation code for Intel's Atom CPUs.
> * Some bugs were fixed.
>
> What's new after 7-Zip 9.08 alpha:
>
> * VHD support was improved.
> * Some bugs were fixed.

DOS386

15.12.2009, 02:53

@ Rugxulo

7-Zip 9.09 beta

Post reply

> Igor Pavlov released 7-Zip 9.09 beta on 12 December 2009.

OK, it works ... but

> * Speed optimizations in AES code for Intel's 32nm CPUs.
> * Speed optimizations in CRC calculation code for Intel's Atom CPUs.

No speed up for old CPU's ...

> * Some bugs were fixed.

"All bugs were fixed." would be the end of 7-ZIP :clap:

---
This is a LOGITECH mouse driver, but some software expect here
the following string:*** This is Copyright 1983 Microsoft ***

Rugxulo

Usono,
15.12.2009, 08:34

@ DOS386

7-Zip 9.09 beta

Post reply

> > Igor Pavlov released 7-Zip 9.09 beta on 12 December 2009.
>
> OK, it works ... but
>
> > * Speed optimizations in AES code for Intel's 32nm CPUs.
> > * Speed optimizations in CRC calculation code for Intel's Atom CPUs.
>
> No speed up for old CPU's ...

9.08 alpha was faster than 9.07 beta on my 486, which (like Atom) is an in-order cpu. It was also faster on my P1.

> > * Some bugs were fixed.
>
> "All bugs were fixed." would be the end of 7-ZIP :clap:

I'll admit, kinda annoying that he says that without details. (It should go without saying otherwise....)

DOS386

18.12.2009, 01:37

@ Rugxulo

7-Zip 9.09 beta

Post reply

> 9.08 alpha was faster than 9.07 beta on my 486, which (like Atom) is an
> in-order cpu. It was also faster on my P1.

It was NOT faster on my P1 , see Igor's forum :-|

---
This is a LOGITECH mouse driver, but some software expect here
the following string:*** This is Copyright 1983 Microsoft ***

Rugxulo

Usono,
18.12.2009, 13:00

@ DOS386

7-Zip 9.09 beta

Post reply

> > 9.08 alpha was faster than 9.07 beta on my 486, which (like Atom) is an
> > in-order cpu. It was also faster on my P1.
>
> It was NOT faster on my P1 , see Igor's forum :-|

Using "7za b", 9.09 seems to be 3-5% faster than 4.57 on my P166 using latest stable HX (ignoring the partially buggy output). Anyways, it doesn't matter, 7-Zip is faster than almost anything, and the one or two apps that are barely faster aren't anywhere near as good compression ratio. So there. ;-)

DOS386

19.12.2009, 04:12

@ Rugxulo

7-Zip 9.09 buggy

Post reply

> Using "7za b", 9.09 seems to be 3-5% faster than 4.57 on my P166 using
> latest stable HX (ignoring the partially buggy output)

Heh ??? How buggy ???

> Anyways, it doesn't matter

Corrupt files don't matter :confused: ???

> 7-Zip is faster than almost anything, and the one or two
> apps that are barely faster aren't anywhere near as good compression
> ratio. So there.

But the bugs ???

---
This is a LOGITECH mouse driver, but some software expect here
the following string:*** This is Copyright 1983 Microsoft ***

Rugxulo Usono, 19.12.2009, 05:26 @ DOS386	7-Zip 9.09 buggy Post reply
	> > Using "7za b", 9.09 seems to be 3-5% faster than 4.57 on my P166 using > > latest stable HX (ignoring the partially buggy output) > > Heh ??? How buggy ??? > > > Anyways, it doesn't matter > > Corrupt files don't matter ??? I meant the numbers for one of the columns listed when benchmarking (Usage %) are obviously bogus (HX regression). Everything else is fine.

DOS386

19.12.2009, 14:11

@ Rugxulo

HX is buggy - NOT 7-ZIP

Post reply

> one of the columns listed when benchmarking (Usage %) are obviously bogus (HX regression)

Why don't you report a BUG ? Or did you by private Mail ?

[image]

Actually 2 columns are BOGUS ... and it IS a regression :-(

---
This is a LOGITECH mouse driver, but some software expect here
the following string:*** This is Copyright 1983 Microsoft ***

DOS386

25.12.2009, 15:18

@ Rugxulo

7-Zip 9.10 beta

Post reply

> Igor Pavlov released 7-Zip 9.09 beta on 12 December 2009

9.10 beta released 2009-12-22 :-)

http://sourceforge.net/projects/sevenzip/files/

What's new:

* Just 1 bug got fixed - related to encrypted ZIP archives
* Also updated LZMA SDK is now available

---
This is a LOGITECH mouse driver, but some software expect here
the following string:*** This is Copyright 1983 Microsoft ***

Rugxulo

Usono,
28.12.2009, 16:17

@ DOS386

7-Zip 9.10 beta

Post reply

> > Igor Pavlov released 7-Zip 9.09 beta on 12 December 2009
>
> 9.10 beta released 2009-12-22 :-)
>
> http://sourceforge.net/projects/sevenzip/files/
>
> What's new:
>
> * Just 1 bug got fixed - related to encrypted ZIP archives
> * Also updated LZMA SDK is now available

Seems the ANSI C decoder is very slightly changed:

1). "l" (list) can now display date and time (although still unpacks disregarding timestamp)
2). "cpuid" used (either _MSC_VER syntax or GCC), not exactly sure why, perhaps for AES instructions?? (dunno, didn't really look that close)

Rugxulo

Usono,
28.12.2009, 21:15

@ Rugxulo

7-Zip 9.10 beta

Post reply

> > 9.10 beta released 2009-12-22 :-)
>
> Seems the ANSI C decoder is very slightly changed:

Well, anyways, if anybody wants to test, here's my latest packaging of the ANSI C 7zdecode program (which is public domain, BTW):

http://sites.google.com/site/rugxulo/7zdec910.zip?attredirects=0&d=1

Archive: 7zdec910.zip Zip file size: 257043 bytes, number of entries: 9 -rw-a-- 2.0 fat 32534 b- stor 09-Dec-28 13:14 7zdec910.7z -rw-a-- 2.0 fat 6578 b- defN 09-Dec-28 13:51 7zdec910.txt -rwxa-- 2.0 fat 49668 b- defN 09-Dec-28 12:55 7zdecdj1.exe -rwxa-- 2.0 fat 47376 b- defN 09-Dec-28 12:54 7zdecdj2.exe -rwxa-- 2.0 fat 77828 b- defN 09-Dec-28 12:26 7zdecemx.exe -rwxa-- 2.0 fat 75264 b- defN 09-Dec-28 12:26 7zdecrsx.exe -rwxa-- 2.0 fat 30720 b- defN 09-Dec-28 12:25 7zdecw32.exe -rwxa-- 2.0 fat 39334 b- defN 09-Dec-28 12:25 7zdecwat.exe -rw-a-- 2.0 fat 3261 b- defN 09-Dec-28 13:38 license.d3x 9 files, 362563 bytes uncompressed, 256125 bytes compressed: 29.4%

DOS386

29.12.2009, 09:32

@ Rugxulo

7-Zip 9.10 beta

Post reply

> Seems the ANSI C decoder is very slightly changed:
> "l" (list) can now display date and time (although still unpacks disregarding
> "cpuid" used (either _MSC_VER syntax or GCC), not exactly sure why, perhaps
> Well, anyways, if anybody wants to test, here's my latest packaging of the
> ANSI C 7zdecode program (which is public domain, BTW):

COOL, I'll check it :-)

---
This is a LOGITECH mouse driver, but some software expect here
the following string:*** This is Copyright 1983 Microsoft ***

Rugxulo

Usono,
29.12.2009, 22:19

@ Rugxulo

7-Zip 9.10 beta

Post reply

>
> http://sites.google.com/site/rugxulo/7zdec910.zip?attredirects=0&d=1
>

I haven't updated this file online since yesterday. But I've still been playing around locally. Maybe Japheth (or anybody, really) can comment on this:

I'm starting to almost feel DJGPP v1 is silly to support. The only advantage is being usable when DPMI isn't / can't be available (XMS only, perhaps Win 2.1 / 386, don't have, never tried nor seen it), but I can't think of anybody needing that. Also, it's SFNs only and can only be compressed by UPX as "dos/mz" if you bind GO32.EXE to the start. Even then it's still too big compared to v2. Still good to test, though. And at least RSX will run it under DPMI (although Vista has no trouble anyways ... one of the few good surprises).

Borland C 5.5.1 for Win32 compiles with a few very minor changes in 7zcpuarc.c, but it needs TASM for the inline assembly. Not a huge deal, but the freeware BC55 didn't come with it, TC++ 2006 had it but is no longer online, and JWasm has some incompatibilities. Plus, it's bigger than the OpenWatcom/Win32 build, even when UPX'd. And it's not open source. (Yeah yeah, who cares?) But it also can work with WDOSX (as can OW/W32) although that disables LFNs (meh). The good news I've found is that you can write a 7zdecbor.def file with "STUB '7zdecdj2.exe'", so when you "-e7zdecbor" it will bind the DJGPPv2 .EXE as stub. Okay, that's a bigger .EXE, but if you really really only wanted "one" .EXE for DOS and all Windows (even Win64) then maybe that's the best idea. (For the record, it seems WLINK and even FASM's "format PE on ..." option don' t handle DOSX apps very well, e.g. DJGPPv2. Digital Mars allegedly doesn't have that limitation, but I've never tried it.)

Also tried Cygwin and MinGW, but the former doesn't work in HX "except with SHARE" (eh???), and the latter needs MSVCRT (or MSVCR71 renamed, in my case). Kinda pointless. I do seriously wonder why they link to that. It saves almost nothing vs. not using it. Oh well.

P.S. Cygwin 1.7 released, no more Win9x compatibility, hence probably even more difficult (if not impossible) to get working under HX. Sure, if you absolutely need UTF-8 console or GCC 4.3.4, but otherwise it seems annoying.

Rugxulo Usono, 31.12.2009, 06:59 @ Rugxulo	7-Zip 9.10 beta Post reply
	> > > > http://sites.google.com/site/rugxulo/7zdec910.zip?attredirects=0&d=1 > > > > I haven't updated this file online since yesterday. But I've still been > playing around locally. Maybe Japheth (or anybody, really) can comment on > this: Well, nobody cares , but I did slightly update the .ZIP today (only source and text updates, mostly minor, added/tweaked a few .BATs for other compiles).