Hello,

How do I get the closest equivalent to 3.90.3 aps in an official Lame release (e.g. 3.92)? I seem to remember something about -Z? But I can't find it via search.

Or, to put it another way, what were the changes made from 3.90 to 3.90.1, 3.90.2 and 3.90.3? Last time I asked this I was directed to version control, but I couldn't find the info there either!

Latest release is 3.97, not 3.92.
Simply use -V2 (or -V2 --vbr-new).

For changes, just read the changelog:
http://cvs.sourceforge.net/viewcvs.py/lame...y.html?rev=1.70
http://209.85.135.104/search?q=cache:dKqFt...mp;client=opera

I know the latest release is 3.97. I am asking about the old Hydrogen Audio branch (3.90.3) and near equivalents.

I know about the standard changelog. It does not contain any info on the changes in 3.90.1 to 3.90.3, which is what I asked about.

As guruboolez, V2. That is the "equivalent" to aps.

If you want 3.90.3 alt-preset standard use 3.90.3, "stealing" that command line expansion for 3.97 will do you no good. 3.90.3 aps had tunings specifically designed for the 3.90 psy-model. They wouldn't work on 3.97. Current -V2 has tunings for the 3.97 model, which wouldn't work elsewhere.

It's partly due to the work on the alt-presets that the normal presets in recent releases improved to better than the old 3.90.3 aps. They incorporate the old work, changed and updated in line with changes in the underlying encoder.

He's right; the "special" tunings were made to be included by default, starting from 3.94 and, along with the rest of LAME, have been updated/tuned since.

From the changelog:

I don't recall the precise detail and I'm not going to wade through the code to check ( ), but 3.90.2 was Dibrom's final tuned version of 3.90. 3.90.3 was the same as 3.90.2 except that the ExperimentalZ switch was included in some presets and disabled as a user switch. Dibrom sanctioned this change and I merely compiled it.

From memory, I believe that 3.92 was very close to 3.90.2.

Not a completely definitive answer, but the best I can manage.

3.97 --alt-preset standard ?

It's quite a while that I was concerned with such questions. IIRC I was astonished that 3.90.3 was released quite a lot after 3.90.2 - I beleive it was in December 2002 when there was already 3.91 or even 3.92 out. But as john33 said I remember that there were not essential changes and will have only reflected the -Z discussion that was out then.

Guess this special history has to do with Dibrom forking from Lame development. Must have been a bad time then.

From memory 3.90.2, 3.90.3, and 3.91 are all pretty much the same encoder apart from -Z switch default usage.

What I really don't understand: why is 3.90.3 not explixcitly available on ReallyRarewares? It's the most important historic version and IMO should be there. There was a discussion here before with the result that it was to be brought there, but nothing happened.

You should consider the whole history again. The most important version was 3.90, or 3.90.1 (kind of fork that appeared few time later to correct a bug, unofficial in regard of LAME development, official on HA.org) and maybe 3.90.2 (I can't even remember what changed), but certainly not 3.90.3 which appeared much later and which is defined by the addition of -Z -- switch that Dibrom was originally reluctant to use for efficiency reason [higher bitrate] and quality one [IIRC he said that it sometimes introduce audible issues].
In all cases, 3.90 (or the corrected 3.91 version) corresponds to the true big historic step of LAME. 3.90.3 is only an epiphenomenon in regard to LAME development.

What we see is that these historic details are going to be buried in the dark, and that they are not very essential. May be it's not for a good reason but what is in people's mind is 3.90.3.

What is missing is 3.90.3 in ReallyRarewares - or a clear stetement there like 'Heh, if you're out for old 3.90.3, forget about 3.90.3 and use 3.91. It's essentially the same thing with some minor improvements' or something like that.

Why would people want to grab 3.90.3? To me there could be 3 reasons:

#1: They think that encoder releases get better when aging. In this case, we should explain them that wine and encoders are quite different things.

#2: They want to do some comparisons against other releases. In this case they could use either 3.90/1 or 3.93.1.

#3: They want this release because it's recommended by some usenet ripping groups. In this case we must make it clear that we are not there to support ripping groups, and that they could go to a store or ask tracks directly to artists themselves.

Small correction: 3.90.3 is (or rather was) only significant for a few circumscribed group of people: a part of HA.org community and some 'enlighted' ripping groups (by "enlighted" I mean people refusing CBR full-stereo). Long time before you registered on this board, both 3.90.1 & 3.90.2 earned a great importance.
But all softwares using LAME (CDex, Winamp, MediaJubebox/MediaCenter...) didn't use any of this local evolution of LAME. These special encoders weren't recommended by LAME developer's either. In other words, their historical significance is completely null outside limited groups of people.


RRW is a site for old, outdated and deprecated software. It wouldn't make any sense to give some serious recommandation about encoders available for fun. People still wanting 3.90.3 for any other reason than curiosity have usually no clue about what they doing and about how to check the quality of 2 different encoders.

For anyone that really wants 3.90.3, it's here.

I will ask Roberto to add it to RRW.

Thanks a lot.

Yes, this topic was made by me and I'm still missing the 3.90.3 compile



guruboolez, I know what you mean. It's right that 3.90.3 is NOT a big step of Lame history, because of this small -Z change.
But I also think that RRW is a page where "good old wares" are available. And one of them simply is 3.90.3. Why are there other encoders like plugger available when almost no one cares about it anymore?
It's only a question about around 400 kilobytes and one encoder more in the "classic" lame list...



Why just let them use 3.90.3 therefore? Because then everybody would be satisfied and those who just want to use 3.90.3 then let them use it and those who just want do tests (like me with 3.90.3 and 3.97) with 3.90.3 then LET them do it. I used some old grab progs from RRW for my tests and that's all I needed. People who visit RRW mostly want to use "good old versions" again or make comparison tests etc.

My only advice: put lame 3.90.3 on RRW

Thanks john33. Finally someone answered my question!

Since asking I have spent several hours digging in the forum to try to find a more detailed answer. Based on what I've read:

3.90.2 is exactly the same as 3.91
3.90.3 added -Z for aps and ape

I didn't find anything regarding 3.90.1, but presumably it contains 1 or more of the fixes that ended up in 3.91.

Although the above seems to be the case from what was written, the info was extracted from various posts that were sometimes vague and contradictory, so if someone knows different from the above please say.

The reason I want to know is that I want to compare the old recommended version with the new one. But I don't actually want to use 3.90.3, because the -Qrcd issue seems like quite an awful state of affairs to me (being a C programmer), and that was not fixed till 3.92. See this thread:
http://www.hydrogenaudio.org/forums/index....ompiler++option

Yes I know 3.90.3 is the 'blessed' version, no comments that I should use that instead please. I will be using 3.92 (which is said to be almost the same) with -Z.

It's nothing to do with my original post, but I agree that 3.90.3 should be available on ReallyRarewares. It seems very odd that it isn't there.

Heh heh heh! Why didn't you say that in the first place? That's a fine and perfectly reasonable thing to do. We all got dismissive because it sounded, from your question and wording of equivalent, like you wanted to use 3.90.3 preset on 3.97 for actual encoding.

For the future, it you are making a weird request for testing purposes, say that in advance. We get many more people who have "personal theories" about various command line options then we do people with logical tests in mind. The replies will be much nicer. Sorry about the misunderstanding!

My recollection (and it has been several years) was that 3.90.2 came out very quickly after 3.90.1 to fix a VERY minor bug, something on the order of reporting itself to be the wrong version. I don't thing the encoding changed at all.

It's up to you but if you want to compare the old recommended version against the new one why not just take the recommended version as it was?

As for the -Qrcd thing if I understand this correctly
a) all the 3.90 tweeking and testing was done using this option
b) rounding errors are smaller when using this option.

It's a pity that language standards don't always make best use of math precision.
I once ran upon a similar case with JAVA standard being inferior to what the IEEE754 standard says and what the integrated math processors provide.

Java implements IEEE754 for SP and DP:

http://java.sun.com/docs/books/vmspec/2nd-....doc.html#33377

Extended support is optional, since most ISAs don't support it, and its on its way out anyway. As I understand it, if you're not getting IEEE-754 compliant fp math in java, you should file a bug report with the developer of your JVM.

Yes, it was about the extended precision which sometimes is very valuable for intermediate results (for instance with scalar products). Nice to hear it's on it's way with Java.

Java already allows Extended precision. By "on its way out" I mean that extended precision is being phased out in the not too distant future (basically, whenever Intel and MS agree to drop support for x87 entirely).

Hmmm... the future is not always brighter.