I downloaded the source code from muspeack.net and compiled mppenc myself...
I've noticed that my compile encodes around 17x, whereas the precompiled versions on musepack.net only run about 15.5x
I was wondering what was different here, I didn't change anything in the Makefile, so why would they be different?
Is there any certain reason i should use the precompiled binary?
and also, I can't get mppdec to compile from source, it fails on wave_out.c every time
I was wondering what was different here, I didn't change anything in the Makefile, so why would they be different?
[a href="index.php?act=findpost&pid=271873"][{POST_SNAPBACK}][/a]
You probably have a different gcc version...
Is there any certain reason i should use the precompiled binary?
[a href="index.php?act=findpost&pid=271873"][{POST_SNAPBACK}][/a]
Probably not.. since gcc is a quite reliable tool, so your mppenc should encode safely
thanks...i figured i was just being too paranoid
We kept the -ffast-math flag in the makefile and this is the reason your compile is faster. It is very likely to fail on teh_sample (http://www.hydrogenaudio.org/forums/index.php?showtopic=30758), though, and will cause other problems with other samples. Disable it. It will be removed from the makefile too.
Probably not.. since gcc is a quite reliable tool, so your mppenc should encode safely
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=271889")
GCC is reliable indeed, but the MPPenc code doesn't seem to be.
[a href="http://www.hydrogenaudio.org/forums/index.php?showtopic=30758&]http://www.hydrogenaudio.org/forums/index....howtopic=30758&[/url]
-ffast-math is there, as it makes things run faster.
But parts of it, that appear to cause teh_sample artifacts, explicitely disabled in Makefile.
Besides musepack.net mppenc was built with -O1, as -O2 and up caused another problems.
But parts of -O2, that seem to be responsible, are also disabled in Makefile.
So in sum if you build it yourself, it's (hopefully) still safe enough.
Anyways you can always check by doing some encodes with musepack.net build and your compile and comparing the results.
I see...i will admit, i've heard a few artifacts in the songs i've encoded using --fast-math (although it could just be the fact i upgraded my audio hardware, it's much easier to abx between different bitrates now)
that was the exact flag that worried me in the Makefile
I think i'll stick with the official builds
I'll have to figure out the difference between the dynamicly and staticlly linked builds (i know its common *nix knowledge, but i've forgotten over the years)
a "static" compile includes - within the executable - the linked libraries that a "dynamic" compile depends on.
later