IPB

Welcome Guest ( Log In | Register )

4 Pages V   1 2 3 > »   
Reply to this topicStart new topic
FLAC 1.3.1 has been released
ktf
post Nov 27 2014, 09:40
Post #1





Group: Members
Posts: 432
Joined: 22-March 09
From: The Netherlands
Member No.: 68263



After 18 months of development, FLAC 1.3.1 has been released. The trigger for this release was a heap overflow found by the Google Security Team.

Changes
  • Huge encoding speedup for i386 and x86_64 (up to 250% for 24-bit)
  • Slightly improved compression of compression levels 6, 7 and 8
  • Fixed CVE-2014-9028 (heap write overflow) and CVE-2014-8962
  • Output buffer has been added to Windows to mitigate fragmentation
  • Various build system improvements and bug fixes
For the complete changelog, see http://xiph.org/flac/changelog.html

The source is available here, official Windows binaries are supplied here.


--------------------
Music: sounds arranged such that they construct feelings.
Go to the top of the page
+Quote Post
Maurits
post Nov 27 2014, 10:08
Post #2





Group: Members
Posts: 413
Joined: 30-September 05
From: London, Europe
Member No.: 24805



Don't forget to mention the added decoding efficiency improvements for all bit depths. Particularly 24 bit on IA32.
Go to the top of the page
+Quote Post
audiophool
post Nov 27 2014, 11:26
Post #3





Group: Members
Posts: 77
Joined: 13-June 14
Member No.: 116389



The website still points to the 1.2.1b binary. (Navigate to http://xiph.org/flac/download.html, then click the link named "FLAC for Windows (command-line tools only)."

I do notice that the 1.3.1 binary has just been added to the sourceforge page, but it still defaults to 1.2.1 ("Looking for the latest version? Download flac-1.2.1b.exe (2.7 MB)."
Go to the top of the page
+Quote Post
eahm
post Nov 27 2014, 17:59
Post #4





Group: Members
Posts: 1169
Joined: 11-February 12
Member No.: 97076



ktf thanks!

audiophool, give them some time smile.gif
Go to the top of the page
+Quote Post
Maurits
post Nov 27 2014, 20:18
Post #5





Group: Members
Posts: 413
Joined: 30-September 05
From: London, Europe
Member No.: 24805



I'd be very interested to see an encoding speed comparison between 1.3.0 and 1.3.1 with one of the new Haswell based CPUs. They can make full use of the AVX2 improvements in the new version, many older CPUs only benefit from AVX1 or older SSE instruction sets.
Go to the top of the page
+Quote Post
IgorC
post Nov 28 2014, 03:12
Post #6





Group: Members
Posts: 1580
Joined: 3-January 05
From: ARG/RUS
Member No.: 18803





Thank You very much for speed and compression improvements to all involved people.


Until now FLACCL was faster on my setup (strong CPU and week GPU) comparing to FLAC 1.3.0.
But now 1.3.1 is faster. Sticking with it.



Go to the top of the page
+Quote Post
taminos
post Nov 28 2014, 09:31
Post #7





Group: Members
Posts: 14
Joined: 6-October 14
Member No.: 117479



is there plans for a patch/build to support old systems (xp sp3 barton 3000+) too?
Go to the top of the page
+Quote Post
ktf
post Nov 28 2014, 09:38
Post #8





Group: Members
Posts: 432
Joined: 22-March 09
From: The Netherlands
Member No.: 68263



QUOTE (taminos @ Nov 28 2014, 09:31) *
is there plans for a patch/build to support old systems (xp sp3 barton 3000+) too?

Why, doesn't it work on that system?


--------------------
Music: sounds arranged such that they construct feelings.
Go to the top of the page
+Quote Post
Nick.C
post Nov 28 2014, 09:44
Post #9


lossyWAV Developer


Group: Developer
Posts: 1813
Joined: 11-April 07
From: Wherever here is
Member No.: 42400



Excellent work - thanks to all of the developers involved! biggrin.gif


--------------------
lossyWAV -q X -a 4 --feedback 4| FLAC -8 ~= 320kbps
Go to the top of the page
+Quote Post
taminos
post Nov 28 2014, 10:13
Post #10





Group: Members
Posts: 14
Joined: 6-October 14
Member No.: 117479



@ktf: i believe the cpu is missing a feature and the build relies on that? when trying to encode in foobar or flacdrop a grey error console showed up.

btw: the official 1.3.0 also did not work but i found a mingw version that did the trick.

This post has been edited by taminos: Nov 28 2014, 10:23
Go to the top of the page
+Quote Post
ktf
post Nov 28 2014, 10:43
Post #11





Group: Members
Posts: 432
Joined: 22-March 09
From: The Netherlands
Member No.: 68263



QUOTE (taminos @ Nov 28 2014, 10:13) *
@ktf: i believe the cpu is missing a feature and the build relies on that?

AFAIK the build is build for i686, so at least MMX but no requirements for SSE. What error do you get? For 1.3.0, there have never been official binaries.


--------------------
Music: sounds arranged such that they construct feelings.
Go to the top of the page
+Quote Post
sundance
post Nov 28 2014, 10:54
Post #12





Group: Members
Posts: 179
Joined: 17-April 02
Member No.: 1804



I benchmarked the official 1.3.1 (32bit) compile from xiph.org with other flac compiles I collected during its "evolution":
CODE
Encoder                   Options                     Process time    FLAC file size
---------------------------------------------------------------------------------------
flac 1.3.0                -8                          77.548 sec      292.407.249 bytes
flac 1.3.1                -8                          56.831 sec      292.127.317 bytes
flac-git-7251201 (AiZ)    -l 12 -b 4096 -m -e -r 6    50.887 sec      292.407.763 bytes
flac 1.3.1                -l 12 -b 4096 -m -e -r 6    52.572 sec      292.407.763 bytes

WAV file size = 663.678.612 bytes
CPU used: Intel Core 2 Quad Q9500 @ 2.8 GHz

AiZ's post: http://www.hydrogenaud.io/forums/index.php...st&p=869644

Although GCC 4.9.0 was used for AiZ's compile, it is faster (~ 3% on my CPU) than the GCC compiles of 1.3.1.

@AiZ: Would you mind to provide "your" 1.3.1?

Go to the top of the page
+Quote Post
smok3
post Nov 28 2014, 12:47
Post #13


A/V Moderator


Group: Moderator
Posts: 1747
Joined: 30-April 02
From: Slovenia
Member No.: 1922



Can I do a static build (flac & metaflac) on wheezy? If so, how?


--------------------
PANIC: CPU 1: Cache Error (unrecoverable - dcache data) Eframe = 0x90000000208cf3b8
NOTICE - cpu 0 didn't dump TLB, may be hung
Go to the top of the page
+Quote Post
ktf
post Nov 28 2014, 13:15
Post #14





Group: Members
Posts: 432
Joined: 22-March 09
From: The Netherlands
Member No.: 68263



Use
CODE
make LDFLAGS='-static'
to make a static build


--------------------
Music: sounds arranged such that they construct feelings.
Go to the top of the page
+Quote Post
smok3
post Nov 28 2014, 14:03
Post #15


A/V Moderator


Group: Moderator
Posts: 1747
Joined: 30-April 02
From: Slovenia
Member No.: 1922



QUOTE (ktf @ Nov 28 2014, 13:15) *
Use
CODE
make LDFLAGS='-static'
to make a static build

thanks (that did the trick).

This post has been edited by smok3: Nov 28 2014, 14:14


--------------------
PANIC: CPU 1: Cache Error (unrecoverable - dcache data) Eframe = 0x90000000208cf3b8
NOTICE - cpu 0 didn't dump TLB, may be hung
Go to the top of the page
+Quote Post
ChronoSphere
post Nov 28 2014, 15:18
Post #16





Group: Members
Posts: 556
Joined: 11-March 07
Member No.: 41384



Interesting, my own VS2013 compile is about 10 seconds slower than the official (gcc-compiled?) on the same set of 1.51GB of wav files.
I actually expected the opposite considering gcc is not "native" to windows.

I followed the readme instructions, so nasm.exe was used. How can I control which instruction sets are used, or does it try to automatically use all possible optimizations? I'm on an AMD Phenom II CPU.
Go to the top of the page
+Quote Post
ktf
post Nov 28 2014, 16:01
Post #17





Group: Members
Posts: 432
Joined: 22-March 09
From: The Netherlands
Member No.: 68263



QUOTE (ChronoSphere @ Nov 28 2014, 15:18) *
I actually expected the opposite considering gcc is not "native" to windows.

FLAC doesn't use libraries for much of its work. The real work is done directly, so there isn't much difference between a native compiler and a non-native one.

QUOTE
I followed the readme instructions, so nasm.exe was used. How can I control which instruction sets are used, or does it try to automatically use all possible optimizations? I'm on an AMD Phenom II CPU.

This release uses function multiversioning, which is supported by GCC starting with release 4.9. IIRC, MSVC2013 supports it too, so it should indeed automatically check the CPU whether it supports the optimizations.

This post has been edited by ktf: Nov 28 2014, 16:03


--------------------
Music: sounds arranged such that they construct feelings.
Go to the top of the page
+Quote Post
taminos
post Nov 28 2014, 16:10
Post #18





Group: Members
Posts: 14
Joined: 6-October 14
Member No.: 117479



@ktf: maybe this is already helpful to you - cant test anymore (link broken), but i am pretty sure that it was even your 1.3.0 build that works on the barton cpu: http://www.hydrogenaud.io/forums/index.php...st&p=835599

regarding the specific error, i'll report back.
Go to the top of the page
+Quote Post
lvqcl
post Nov 28 2014, 16:53
Post #19





Group: Developer
Posts: 3464
Joined: 2-December 07
Member No.: 49183



There's a line in configure.ac:
CODE
        XIPH_ADD_CFLAGS([-msse2])

which means that SSE2 is required.

Here is a build that doesn't need SSE2: (see post #39)

This post has been edited by lvqcl: Nov 29 2014, 20:49
Go to the top of the page
+Quote Post
taminos
post Nov 28 2014, 18:32
Post #20





Group: Members
Posts: 14
Joined: 6-October 14
Member No.: 117479



thats it, is working now, although very slow compared to the 1.3.0 build, thank you. would be interresting to see a sse1 version, as the cpu is capable of that instruction.
Go to the top of the page
+Quote Post
lvqcl
post Nov 28 2014, 19:46
Post #21





Group: Developer
Posts: 3464
Joined: 2-December 07
Member No.: 49183



Really?

a) This build is able to use SSE/AVX-accelerated routines when possible.
b) Remember that -6, -7 and -8 presets were re-tuned. For example, -6 was very close to -5. Now it is noticeably slower but compresses better.
Or you can old presets by using the following settings:
CODE
old preset     equivalent settings
-6            -l 8 -b 4096 -m -r 6
-7            -l 8 -b 4096 -m -e -r 6
-8            -l 12 -b 4096 -m -e -r 6
Go to the top of the page
+Quote Post
ktf
post Nov 28 2014, 22:12
Post #22





Group: Members
Posts: 432
Joined: 22-March 09
From: The Netherlands
Member No.: 68263



QUOTE (audiophool @ Nov 27 2014, 11:26) *
I do notice that the 1.3.1 binary has just been added to the sourceforge page, but it still defaults to 1.2.1 ("Looking for the latest version? Download flac-1.2.1b.exe (2.7 MB)."

This has just been fixed


--------------------
Music: sounds arranged such that they construct feelings.
Go to the top of the page
+Quote Post
Heliologue
post Nov 28 2014, 22:21
Post #23





Group: Members
Posts: 108
Joined: 17-September 06
Member No.: 35303



It might also be worth pointing out that John33 has new Windows builds of 1.3.1 compiled with ICL 14 over at Rarewares
Go to the top of the page
+Quote Post
Makaki
post Nov 28 2014, 23:31
Post #24





Group: Members
Posts: 78
Joined: 20-May 13
From: Santo Domingo
Member No.: 108227



I've been using rarewares build for a while now, mostly because of the lack of official 1.3.0 builds for windows.

Now the question is, which one is faster / more efficient?
Go to the top of the page
+Quote Post
me7
post Nov 28 2014, 23:56
Post #25





Group: Members
Posts: 179
Joined: 23-August 06
Member No.: 34375



The Rarewares ICL x64 build is actually slower than the (official) GCC x64 build on my Haswell machine. Roughly 5% difference.
Go to the top of the page
+Quote Post

4 Pages V   1 2 3 > » 
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 20th December 2014 - 17:40