I downloaded Mitiok's build of LAME 3.96 when it was released and the file size of the EXE is 197KB. I just downloaded the Rarewares compile and it is 186KB. Should I be concerned that there is a problem with one of these builds due to the file size difference?

Generally, there is no cause to be concerned. I think Mitiok and John33 used a different compiler or different compiler setting, hence the difference.

There might be a problem with any of the files, but this is not simply due to the exe sizes.

I think (though I might be wrong) that these two lame.exes should give bit-identical mp3's (or at least the decompressed wav's should be bit-identical).

Maybe someone with more knowledge will come to our aid

... I don't think so... just two different compiles. The one at Rarewares is made by John33 using ICL4.5, the one at Mitiok is made (I presume) by Dimitry Kutsanov using MSC. There is another floating around compiled by Gabriel Bouvigne and probably many others home brewed versions.

I think you can pick the one you love the best. They could also generate slightly different files due to different math libraries.

At one point I used to think that Mitiok compiles could be considered as the "unoficial reference" ones, but I'm not sure about that anymore. Maybe a lame developer could shed some light on that...

Sergio

Edit: oops.. I'm a slow typer... kalmark got first!

Does anyone know why John33 compiled lame.exe and lame_enc.dll with ICL 4.5 but lameACM.acm with ICL 7.1?

Because the ICL4.5 compiler does not produce a viable ACM compile. This is nothing to worry about as the 4.5 and 7.1 compiled encoders produce bit identical results, it's just that the 4.5 compiler generates slightly faster code.

I suppose you meant to say it generates the code slightly faster.

Else, how could one of two bit identical codes be faster?
greets Tz

"the code is faster" = "the program build with compiler X needs less time for the same job then the same program build with compiler Y". The result is identical, i.e the mp3s encoded with both encoders are (should be) bit-identical.

I should probably have said "generates slightly faster executables".

I have compiled 3.96 on Mac OS X yesterday. A "make test" shows me 7694 diffs from the testfile. Is this not a little bit to much? On Linux i got 400 diffs. But 7694 on Mac OS X....

BadHorsie

Having done some limited encoding with both builds using --preset standard, I found the output to be bit for bit identical, but John33's build to be about 9% faster on a P4.

Here, on AthlonXP, LAME from Mitiok's web-site is a bit faster, output is bit-identical.

EDIT: Oops, it's not bit-identical, but the differences seem to be in mp3 header only...