Help - Search - Members - Calendar
Full Version: Rev8 quick test
Hydrogenaudio Forums > Lossy Audio Compression > MP3 > MP3 - General
Enig
Dibrom, I wonder if you use fast compile or not.

I just quick test one Track and get a bigger file than the one encoded by previous Rev7 version.

Rev7: 6,299,480
Rev8: 6,548,379

The other reason may be that It was caused by quality improvement tweaks that we don't get detailed explain yet.

BTW, do you put the technics used in standard preset to extreme and insane and it's time to safely use those higher presets?
Enig
Hi Dibrom:

Thank you for your great work.

I have a question to ask you for a long time. When I encode an old-recorded mono Track, can I use "lame --alt-preset standard -b80" for better size/quality balance? Otherwise, almost all frames has 112k lower bitrate band in the result MP3 file. I think use "-b80" switch would shrink the result MP3 file and without loss of quality.

The default switch is "-b112" so I'm not so sure wether to change this value will cause a quality issue.
Dibrom
About the rev7 vs rev8 size increase you note, which compile of rev7 were you using?

If you were using my compile (MSVC) and not Mitiok's (ICL) that would make sense. ICL compiled binaries produce slightly larger filesizes than MSVC. So for a more direct comparison, make sure you use Mitiok's rev7 compile (http://mitiok.free.fr/lame_dm_rev7-bin_f.zip) to compare with my rev8.

If you were already using Mitiok's compile though then it might just be a slight difference in bitrate due to a changing around of some of the functions and it should be nothing to worry about because on other cases it should be slightly lower also.

About encoding mono sources, yes you can specify -b80 if you would like, it should be fine.
Enig
Tanks for you kindly reply.

I just compared your non-ICL compile with this one. I should be aware of it because the increased encoding speed. But for me the speed is not so important. Dibrom can you give us another non-ICL compile version of Rev8?
john33
Eniq

If speed doesn't matter to you, I can give you Rev.8 compiled with MinGW32 and then your files will be even smaller!! (At the same settings, of course!)

john33
Dibrom
MSVC Compile here:

http://static.hydrogenaudio.org/extra/lame_dm_rev8a-MSVC.zip

or ICL:

http://static.hydrogenaudio.org/extra/lame_dm_rev8a-ICL.zip

This is a slightly updated version so you should download at least one.
john33
OK, so now you can have a MinGW32 Rev.8a too!

john33
ff123
What, exactly, is causing the difference in bitrates?

Is there some sort of real-time clock in the code somewhere? A difference in dynamic memory allocation? Less accurate floating-point calculations?

ff123
john33
I can't give you any technical answer to that, but the MinGW32 versions seem to allocate more to 160kbps than 192kbps compared to the MSVC/Intel versions. There other slight differences in allocations, but the major difference seems to be there.

john33
Dibrom
QUOTE
Originally posted by ff123
What, exactly, is causing the difference in bitrates?

Is there some sort of real-time clock in the code somewhere?  A difference in dynamic memory allocation?  Less accurate floating-point calculations?


They have to do with changes the compiler makes to the code itself during aggressive optimizations. I'm not sure specifically what the differences in that regard are between GCC and MSVC, but as for ICL, it uses some extremely aggressive optimizations. It will reorder functions, change divisions to multiplications and stuff like that (I think), slightly alter other math related functions for speed at a slight cost in accuracy, make automatic use of mmx and sse (which will further alter some math functions), some other areas are modified as well such as conversions between floating point variables and integers (speed over accuracy), etc.

Most of the stuff in ICL you can disable progressively which will continuously result in a drop in speed until the output is much closer to MSVC. So far though, even with the very aggressive options used I haven't noticed a degredation in quality yet... only an increase in bitrate.
brett
hi dibrom. mac osx user here. can you point me to the source for an old-fashioned user compile of rev8? i'm presently using rev7. many thanks for a reply AND for your great work.

brett.
john33
Brett

Unless you are able to do a CVS sweep, then I suggest you wait until tomorrow morning when you should be able to get the latest source from:

ftp://cedric.vabo.cz/pub/linux/apps/lame/

This site provides daily updates from the CVS but the sweep this morning was too early to catch Dibrom's Rev.8, now Rev.8a.

john33
Dibrom
QUOTE
Originally posted by brett
hi dibrom. mac osx user here. can you point me to the source for an old-fashioned user compile of rev8? i'm presently using rev7. many thanks for a reply AND for your great work.


Here you go:

http://static.hydrogenaudio.org/extra/lame_dm_rev8a-src.zip
brett
thanks, dibrom. however, it wouldn't "./configure" on osx 10.1.1. the good news: the 12/15 version from john33's site is up and it compiles.

brett.
Dibrom
QUOTE
Originally posted by brett
thanks, dibrom. however, it wouldn't "./configure"  on osx 10.1.1. the good news: the 12/15 version from john33's site is up and it compiles.


Interesting.. the sources I posted were fresh from CVS to a .zip file.

Anyway, glad to hear you got it working somehow.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Invision Power Board © 2001-2009 Invision Power Services, Inc.