New faster LAME 3.90 compile

[MaTTeR]

Dibrom has managed to squeeze another 2-5% speed out of LAME 3.90 just by optimizing a compile. Here's the info-

"These *REQUIRE* a fairly recent processor to run correctly (they will crash otherwise). A Pentium Pro or later Intel processor or an Athlon or later AMD processor"

Download the new hacked
DLL (http://www.hydrogenaudio.org/extra/lame390-modified_dll-ICL-fast.zip)

Download the new "fast" compile here. (http://www.hydrogenaudio.org/extra/lame390-ICL-fast.zip)

These compiles are using the same gold 3.90 source code but just with optimizations for speed now. A big thx goes out to Dibrom for all his effort.

[ProfDrMorph]

hehe another one that reads multiple forums. I read about this before on www.hydrogenaudio.org.

Every processor with 686 architecture should work. I'm not sure but on the AMD K6 it MAY work, too. And please don't call this .dll a hacked .dll since it's just a recompiled .dll ;) Hacking is something different.

[MaTTeR]

Originally posted by ProfDrMorph

And please don't call this .dll a hacked .dll since it's just a recompiled .dll ;) Hacking is something different.

That's what it has been refered to on multiple forums such as r3mix and Hydrogen. I call it as I see it :p

[ProfDrMorph]

That's what it has been refered to on multiple forums such as r3mix and Hydrogen.
Who called it like that? I think they misunderstand what Dibrom did. I read the Hydrogen forum from time to time but I'm not posting there nor do I intensivly read it.

[DSPguru]

@DarkAvenger,
those dll's also support my floating-point interface.

[Pasqui]

This dll is called "hacked" because it allows you to use Dibrom's presets (even the ABR switch usefull for movies). Here is the readme included in the archive:

If there is anyone out there who is totally committed to using CDex with the Lame DLL and wants to use the new --alt-presets, I can offer a VERY QUICK AND DIRTY solution!! It is NOT pretty, but it works. It maps to the QUALITY options as follows:

1. Normal (q=5) = --alt-preset standard

2. Low (q=9) = --alt-preset fast standard

3. High (q=2) = --alt-preset extreme

4. Voice (for voice recordings) = --alt-preset insane

5. R3Mix preset = --r3mix

6. Very High Quality (q=0) = remains as originally UNLESS VBR Method VBR-ABR is selected. If VBR-ABR is selected, then --alt-preset nnn (ie. --alt preset ABR) is performed for ABR bitrates between 80kbps and 320kbps. With ABR selected, all other switches are non-functional.

For items 1, 2, 3, 4, and 5 , all other switches are non-functional. Even if you can change the settings, they are ignored.

If you use Peter Pawlowski's Cool_lame filter, this DLL will also work with that BUT you only get items 1 thru 4 because of the preset limitations within the filter options.

If anyone is interested, you can have the binary and/or the amended source. This is compiled with MinGW32, so it's not the quickest, but it works.

These can be downloaded from:

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

DON'T FORGET THIS IS AN UNOFFICIAL RELEASE. HOWEVER, IF POSSIBLE AND I AM ABLE, I WILL PRODUCE THIS AS LONG AS IT IS REQUIRED, FOR EACH NEW RELEASE OF DIBROM'S Rev's AND ONCE THEY BECOME PART OF THE CVS. THE CHANGES TO THE SOURCE CODE ARE REALLY QUITE TRIVIAL.

[DSPguru]

why are bringing all hydrogen forum messages and pasting them here ?

[MaTTeR]

Originally posted by DSPguru
what's the idea of bringing all hydrogen forum messages and pasting them here ?

Well I thought people might be interested in newer "optimized" versions here. Not everyone scours r3mix and Hyrdrogen forums for the latest encoding info. I'll stop and everyone can fend for themselves :p

[DSPguru]

@matter
your news was relevant indeed, that's ok :).
but i don't want it to start a "mirroring" phenomenon.

[Psyche]

@DSPGuru
If I recall correctly, your modifications to lame_enc.dll where included in the official 3.90 release (and in the 'hacked' :D dll), but it can't be used in BeSweet because your prog doesn't support the q switch (it maps alt-presets to q).
If I'm not wrong (again) you use the BladeEnc interface... why? If you used the lame interface you could add (almost) every switch of CLI lame.
Anyway, I wanted to tell you that if you used the LQP presets (which mainly change q) maybe the modified dll would work, but now after having a look at lame sources I'm not so sure.
OK, good work after all. :)

[DSPguru]

my modifications to v3.90 was adding a floating-point interfacae to lame_enc.dll, i still haven't released my solution to using presets on the dll.
the "hacked" dlls are offering their solution to the issue, but ATM, i prefer to wait for a better solution.

here's a COOL test, to know if a program uses my floating-point interface :
1. encode a mp3 using vbr and save the result.
2. encode same input again using vbr, but with gain of -50db.

now, the first mp3 would hold music, the second would be SILENT.
but..
if my floating-point interface was used : the first mp3 and second mp3 would have the SAME bitrate at any point of time.

Cheers,
Dg.