Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: rev10 (Read 4936 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

rev10

Reply #1
Hey there Dibrom, thanks for your preset, it makes LAME more easier for us users to use!

Just a few questions:

1.) Have these presets and modifcations made their way into the CVS for both DLL and CLI exe?

2.) Is there anymore left to do for the presets, as in quality or is it just now tuning the efficiency?

3.) How much more "revisions" can we expect from you?

Cheers and thanks for your effort in making LAME the best encoder available!

AgentMil
-=MusePack... Living Audio Compression=-

Honda - The Power of Dreams

rev10

Reply #2
Quote
Originally posted by AgentMil
Hey there Dibrom, thanks for your preset, it makes LAME more easier for us users to use!


You're welcome.

Just a few questions:

Quote
1.) Have these presets and modifcations made their way into the CVS for both DLL and CLI exe?


Only the command line interface.  I may work on the .dll some if I have time, but I don't know if this will happen before the next stable.  We'll see.

Quote
2.) Is there anymore left to do for the presets, as in quality or is it just now tuning the efficiency?


Well, there's always a possibility for more improvements.  It's mostly just a matter of time and know how, and finding areas where things can actually be improved realistically.  Right now I don't know of much left to do, but who knows when that might change

Quote
3.) How much more "revisions" can we expect from you?


Actually rev9a was going to be the last before kye pointed out a problem, which I was able to fix.  Unless that happens again (I kind of doubt it) then this should be the last before the stable.. at least the last that has anything to do with quality.

rev10

Reply #3
For anyone that wants them, all 3 versions of the unofficial DLLs updated to Rev10, will soon be available at:

http://www.inf.ufpr.br/~rja00/

as usual. If anyone wants them, the sources will be there too. I assume that DLL3 works OK since nobody has said otherwise.

Until such tiime as the DLLs have been officially updated, as per Dibrom's comments, and the necessary changes in CDex, etc have been made to accommodate them properly, I shall update the DLLs and sources at the above site on a daily/as changed basis.

john33

rev10

Reply #4
Hi!
tested the --abr mapping yesterday, and for some reason it doesnt work  no matter what you try todo you get a 128k mp3 ..
This isnt a serious problem of course, its just a quick fix that didnt work  ,  maybe the program (VobTomp3 aka BESweet) uses the dll in a weird way

Anyway, thanx for trying and have a Merry Christmas and a Happy New Year everyone!

/stefan

---- edit ----
I apologize for the excessive smiley use


rev10

Reply #6
hi!

firstly, thx for your great work, Dibrom! I appreciate it a lot!

a question:

Quote
Note: I recommend that everyone who wants to be sure about the quality get the ICL version, this is the version I test on and I have now noted at least one case where the MSVC version sounds ever so slightly worse (try --alt-preset fast standard with ICL vs MSVC on fatboy.wav). The difference is small, and this is the only case I have seen yet out of many, many files, but the difference is there so I recommend the ICL version to be sure.


could you explain this to me? as long as I remember, it is the ICL version which is hardly optimized, and it is less accurate in floating point arithmetics, changes divisions to multiplications and so on.... so how is it possible that its quality is higher?? (in fast modes actually) This is ridiculous.
I like MSVC, 'cause it produces smaller files, and, theoretically, it should produce higher quality mp3's, isn't it? I my case, speed isn't an issue.

but if i'm totally wrong, i'll be grateful if you explain it to me.

thx again!
olcios

rev10

Reply #7
Quote
Originally posted by john33
For anyone that wants them, all 3 versions of the unofficial DLLs updated to Rev10, will soon be available at:

http://www.inf.ufpr.br/~rja00/ 

as usual. If anyone wants them, the sources will be there too. I assume that DLL3 works OK since nobody has said otherwise.

Until such tiime as the DLLs have been officially updated, as per Dibrom's comments, and the necessary changes in CDex, etc have been made to accommodate them properly, I shall update the DLLs and sources at the above site on a daily/as changed basis.

john33


The DLLs are already there.

Regards;

Roberto.

rev10

Reply #8
Quote
as long as I remember, it is the ICL version which is hardly optimized, and it is less accurate in floating point arithmetics, changes divisions to multiplications and so on.... so how is it possible that its quality is higher?? (in fast modes actually) This is ridiculous.
I like MSVC, 'cause it produces smaller files, and, theoretically, it should produce higher quality mp3's, isn't it? I my case, speed isn't an issue.


Well, the issue you have to consider also is that psychoacoustics is by no means nearly perfect in Lame, so the (audible) effect of a bit more or less precise floating point calculations is not so clear. I believe there's not much quality difference between ICL and MSVC compiles, but since ICL tends to produce larger bitrate, the bits gotta go somewhere.
Of course it's theoretically possible that the bits go to the wrong place and even, that more important frequencies wont get as much bits as they should, because of less precise calculations. But this doesn't seem to be the case at least according to tests so far. ICL gives a bit higher bitrate and at least with fast setting and fatboy it increases audible quality.
Juha Laaksonheimo

rev10

Reply #9
Quote
Originally posted by olcios
could you explain this to me? as long as I remember, it is the ICL version which is hardly optimized


Well I'm not exactly certain what you mean by optimized.  As far as producing faster code, ICL is a much more advanced compiler than MSVC and certainly is more optimized in that regards.

Quote
, and it is less accurate in floating point arithmetics, changes divisions to multiplications and so on.... so how is it possible that its quality is higher?? (in fast modes actually) This is ridiculous.


Don't get too hung up on the details of it like that.  For one, multiplication instead of division, if done properly, shouldn't really result in a difference (certainly not *much* of a difference) at all.  It's just significantly faster for x86 processors to do because of the way they are designed.  Less accurate floating point math doesn't necessarily mean that' its horribly inaccurate either.  We are talking floats instead of doubles, but that's not *always* the case.  The compiler should only optimize in this regard during a vectorization, or when it can translate a function into assembly code which uses either mmx, sse, or some other optimizations like prefetches and similar things.

Besides that, less accurate doesn't mean worse quality at all.  There is no connection.  It simply could overshoot in some cases and undershoot in others.  Like JohnV said, LAME's psychoacoustics are far from perfect so it shouldn't come as much of a surprise than in this case the ICL version was better in quality.  The fact that it is usually a few higher kbps in bitrate also would suggest that if anything it is overshooting with quality.

Quote
I like MSVC, 'cause it produces smaller files, and, theoretically, it should produce higher quality mp3's, isn't it? I my case, speed isn't an issue.


It produces smaller files, but it certainly doesn't produce higher quality MP3's at all.  Don't relate less precise math operations in a few areas to quality, it doesn't work that way.  It could either be too generous with calculations, or less generous, but it doesn't have to be a particular one.  I'm siding with ICL because so far I have actually heard a difference in quality, where MSVC sounded worse.  Coupled with the fact that ICL produces slightly higher bitrates, I believe it's safer to say that it should usually sound slightly better if there is ever going to be a difference.

However, all of this is sort of moot now, because LAME 3.90 is out.. rev10 is obsolete

rev10

Reply #10
ok. thx Dibrom and JohnV

Olcios