Which binaries ?

[fredthemasterofpuppets]

Well, I read the Q&A, and there are 3 binaries :
Nic's binaries: http://nic.dnsalias.com/
Koepi's binaries: http://www.roeder.goe.net/~koepi/
uManiacs binaries: http://xvid.hopto.org/
Are they really different ? Wich one should I use ?
thanks

[ivan_alias]

If you are new to XviD (which I assume you are) then it wont make much difference really.

Run a few encodes with the default settings and see how you get on. After that ask the question. (when you have had time to read up on the differences)

There is no 'best build' as XviD is still in the alpha development stage!

Just pick one and give it a go!

[fredthemasterofpuppets]

Thanks for your quick answer, for now I used Nic's Binaries, and I did encode one movie, the result is great, I should test more !
Thank you again !

[AndyP]

Hi

As far as I know,

uManiacs are probably the most up-to-date as they are build everytime the CVS changes, but are not optimised when compiled, do not include Nic's directshow filter and use the default motion estimation (PMVFast)

Koepis are also pretty up-to-date, although sometimes slightly more experimental these days than they used to be (which is great :) ) and are compiled with intel compilers (optimised), usually include Nic DS filter and usually use EZPS motion estimation (arguably better but not yet complete). These are what I use at the moment.

Nics are as Koepis but have not been updated recently (use a different lumi-masking, and older curve compression code) but I think still sometimes look better.

Hope this helps

Andy

[SirDavidGuy]

Another alternative is to compile your own.

You get all the goodies, like B-Frames and EPZS^2, without having to wait.

Also, ICL6 is avaliable for 30 days free, which means you can compile it as highly optimized as you want...

[Sigmatador]

and when the 30 days are finished, you can change your computer's date or request a new download for get a new license file ^^

[Koepi]

I must have missed something during my computer science lessons.

I can't optimize my binaries as I want, I have clear limitiations by the compiler.

But maybe I'm just too unfamiliar with the intel compilers and you can do magic that I can't. Would be nice if you spread your knowledge, so I can give even _more_ optimized binaries.

Regards,
Koepi

[AndyP]

Koepi, Sorry if I offended you. I was just trying to help as you and others have helped me in the past.

Will go back to lurking now.

Andy

[Koepi]

AndyP,

that wasn't addressed to you, your explanation was very fine.

I was refering to

"SirDavidGuy
Another alternative is to compile your own.

You get all the goodies, like B-Frames and EPZS^2, without having to wait.

Also, ICL6 is avaliable for 30 days free, which means you can compile it as highly optimized as you want..."

:)
Best regards,
Koepi

[Marc FD]

never used Intel Compilers.
do you need to make changes in the code ??
i heard intel compilers sometimes optimise too much, making the prog crash...

@koepi
i know you use the intel compilers for your XviD build, but i think the gain is very low, right ? 5% or something like that ??

[Koepi]

15-25% is more appropiate.
You have to use a set of compiler flags to gain that much though, I'll PM then to you if you like.

Regards,
Koepi

[Marc FD]

i think it's due to XviD ME who is plain C ....

25% !! that make me want to play with intel compilers.
maybe it could be used to compile faster releases version of my avisynth filters ;)

PS : yes some hint would be _very_ appreciated

[SirDavidGuy]

Originally posted by Koepi
I must have missed something during my computer science lessons.

I can't optimize my binaries as I want, I have clear limitiations by the compiler.

But maybe I'm just too unfamiliar with the intel compilers and you can do magic that I can't. Would be nice if you spread your knowledge, so I can give even _more_ optimized binaries.

/me groans...

You know what I mean ;)

Although, last time I checked on a 2 min clip, my builds were a few seconds faster than yours... ;)

[tiki4]

The box said "Requires Windows 98 or Better". I don't understand why it wont work on Linux!
Call me Dave.

:D :D :D

[SirDavidGuy]

Originally posted by tiki4


:D :D :D

The funny thing is, I'm now actually back on Windows.

All my term papers, have to be submitted in the latest version of Micro$oft Office, my programs have to be compiled into PE format, etc.

[Koepi]

I'm still missing the setings that allow your binaries to be faster.

Seems to be just posing and boosting without any fundamental background? That would be annoying. So please, let us know, finaqlly, you had a week to tweak...! (well, 2 days, but that doesn't rhyme ;) )

[SirDavidGuy]

Originally posted by Koepi
I'm still missing the setings that allow your binaries to be faster.

Seems to be just posing and boosting without any fundamental background? That would be annoying. So please, let us know, finaqlly, you had a week to tweak...! (well, 2 days, but that doesn't rhyme ;) )

I've been coding a sector patcher, be patient!

But the settings came from a discussion on the XviD forum, on the best compilation settings, which (I believe, I can't access the forum, obviously) you participated in.

[Koepi]

Yupp, and I use those settings that came up there. And they are really fast (though they disallow usage on pII/k6/k6-2/...).

So I don't think there's much more that you can do. And the speed boost is already _really_ amazing. Sorry for the rude post above, my honest apologies, I was drunk and upset about some other threads already(NEVER post when you're drunk. I stuck a notice against my monitor ;) ).

But please, if you have even better compiler settings, post them!

Best regards,
Koepi

[SirDavidGuy]

Sorry for the rude post above, my honest apologies, I was drunk and upset about some other threads already

No offense taken.

(I'll put a similiar warning above my monitor ;-) )

[lovelace]

Originally posted by Koepi
Yupp, and I use those settings that came up there. And they are really fast (though they disallow usage on pII/k6/k6-2/...).

I'm using uManics binaries, but I'd love to try those experimentals by Koepi. Bad for me, I'm working on K6-2 :( . Long time ago, his binaries gave best quality on my encodings. But now, they are to fast for my machine...
Maybe, when XviD development goes on, and Koepi has nothing else to do (but to get drunken :cool: ), he could place an additional binary-version, not that strong optimized (e.g. for pII/k6/k6-2/... ).
I'll lay down to his feet for now and then :rolleyes:, since I'm too stupid, lazy or whatever, to compile the code myself.

Hoping,
lovelace

[athos]

As the Xvid forums cannot be accessed atm, would someone please post those ICC settings that had been agreed on in that forum as being the best? If it is not appropriate to post them here for some reason, would someone please be so kind as to send them to me in a PM?

[Koepi]

do you mean ICL by ICC?

Btw., there were no "best" settings. there were just real nice settings.

[cult]

or maybe u mean acc?

[athos]

I mean Intel C++ Compiler, i guess it abbreviates to ICL somehow :)

I tried theses settings (have not yet compared performance, sorry):

in Configuration Properties:
General -> Whole Program Optimization = Yes
Intel(r) Specific -> Compiler Name = Intel C++ Compiler (icl.exe)
C/C++ -> Optimization -> Optimization = Custom
C/C++ -> Optimization -> Global Optimizations = Yes (/Ob2)
C/C++ -> Optimization -> Inline Function Expansion = Any Suitable (/Ob2)
C/C++ -> Optimization -> Enable Intrinsic Functions = Yes (/Oi)
C/C++ -> Optimization -> Favor Size or Speed = Favor Fast Code (/Ot)
C/C++ -> Optimization -> Optimize for Processor = Pentium Pro and Above (/G6)
C/C++ -> Optimization -> Optimizae for Windows Applications Yes (/GA)
Command Line -> Additional Options -> "/O3"

The command line looks like this (for xvidcore):/nologo /W3 /Og /Ob2 /Oi /Ot /G6 /GA /D "NDEBUG,ARCH_X86,WIN32,_LIB" /GF /Gm
/EHsc /ML /GS /Gy /YX"StdAfx.h" /Fp".\Release/core.pch" /Fo".\Release/"
/Fd".\Release/" /Gd /TPPresumably, "/O3" is added after that.

I use nasm-0.98.34 for the assembly stuff, works great after i renamed it from nasmw.exe to nasm.exe. Also, I am using ICL 6.0 with Visual Studio .NET (ie VC++ 7.0)

About O3:Enables -O2 option with more aggressive optimization, for example, prefetching, scalar replacement, and loop transformations. Optimizes for maximum speed, but does not guarantee higher performance unless loop and memory access transformation take place.(from the ICL help)

I realize that there might not be any "best" settings, but I'm looking for the "best yet agreed upon" ("real nice":)) or something like that from people who are more experienced than me in this area.