Help - Search - Members - Calendar
Full Version: Any ogg encoders support SMP?
Hydrogenaudio Forums > Lossy Audio Compression > Ogg Vorbis > Ogg Vorbis - General
riiv
I saw something about poggenc on the xiph mailing list but got bored of searching. heh. Do oggdropXP or oggenc support SMP?

Also a feature request. Since ogg files are the most self describing compressed file format (love it), all I'd suggest is a option to replicate the options used on a file encode eg.
"oggenc --advanced-encode-option lowpass_frequency=18 -c"CMD=-q6 --advanced-encode-option lowpass_frequency=18"
john33
QUOTE(riiv @ Dec 19 2002 - 06:32 AM)
I saw something about poggenc on the xiph mailing list but got bored of searching. heh.  Do oggdropXP or oggenc support SMP?

To answer your first question, no! rolleyes.gif
QUOTE(riiv @ Dec 19 2002 - 06:32 AM)
Also a feature request.  Since ogg files are the most self describing compressed file format (love it), all I'd suggest is a option to replicate the options used on a file encode eg.
"oggenc --advanced-encode-option lowpass_frequency=18 -c"CMD=-q6 --advanced-encode-option lowpass_frequency=18"

With oggdropXPd, that information is already available in the log file, if you are using it. However, it could easily be added as an extra tag if there is a demand for it. I can't speek for the official oggenc, but I could similarly modify oggenc2.1.

Anyone else have a view on this?
Peter Harris
QUOTE(riiv @ Dec 19 2002 - 01:32 AM)
I saw something about poggenc on the xiph mailing list but got bored of searching. heh.  Do oggdropXP or oggenc support SMP?

I seem to recall someone doing an auto-threading compile with the new Intel compiler. The gain in speed was unspectacular.

You can always just do what I do on my SMP box: Run two oggencs at the same time. I was doing this even before I noticed that EAC can automatically run as many oggencs as you like.
QuantumKnot
The Intel compilers support an option called OpenMP I believe which uses the hyperthreading feature of new Pentium 4s. But CNET recently did a benchmark of Hyperthreading and found it either improved the performance negligibly or actually degraded performance. So I guess it isn't really worth it until Intel improves it smile.gif
ogg
QUOTE
But CNET recently did a benchmark of Hyperthreading and found it either improved the performance negligibly or actually degraded performance
Guess they didn't do any multitasking benchmarks, or tests with truly multithreaded software, to show the benefit. HT is better than no HT, you just have to do the right tests to show it. (SMP/HT amatures like CNET, and Tomshardware wouldn't know this.)
QuantumKnot
QUOTE(ogg @ Dec 20 2002 - 10:29 AM)
QUOTE
But CNET recently did a benchmark of Hyperthreading and found it either improved the performance negligibly or actually degraded performance
Guess they didn't do any multitasking benchmarks, or tests with truly multithreaded software, to show the benefit. HT is better than no HT, you just have to do the right tests to show it. (SMP/HT amatures like CNET, and Tomshardware wouldn't know this.)

Well, they did the benchmarks on Photoshop 7 which one would expect to be one of the most truly multithreaded programs there is out there. The only performance improvements that were noticed (and they were small) were in some select Photoshop filters.
ogg
PhotoShop is not multithreaded, only some of the plugins are. This is why I said "truly" multithreaded. A lot of programs claim to be multithreaded, but only parts are.
Ardax
Even in PhotoShop, using a multithreaded plugin probably isn't going to give you an appreciable speedup.

Read Ace's HT Article to get some useful information about HT.

That said, I think that a highly tuned version of oggenc could probably get a decent speedup via HT. There may be other optimizations that would yield speedups for a wider range of CPUs though.
QuantumKnot
Well, there is only one way to find out. B)

Compile the oggenc sources with the Intel compiler with the -openmp and -parallel switches and test it on an HT-enabled Pentium 4. smile.gif
JohnMK
John33 compiled oggenc 1.0 with ICL7 with all the pertinent switches -- no automatic multithreading could be squeezed out of it. So I think it's clear that oggenc will have to be rewritten to support SMP. I really don't know if it's feasible however. I don't know anyone who has the time or even the inclination to do that.
JohnMK
QUOTE(QuantumKnot @ Dec 19 2002 - 06:31 PM)
Well, there is only one way to find out.  B)


That comment is fallacious. It's not quite as simple as you've reduced it down to.
QuantumKnot
QUOTE(JohnMK @ Dec 28 2002 - 04:22 AM)
QUOTE(QuantumKnot @ Dec 19 2002 - 06:31 PM)
Well, there is only one way to find out.  B)


That comment is fallacious. It's not quite as simple as you've reduced it down to.

Maybe you didn't notice the smiley I had there. rolleyes.gif

Of course it's not that simple.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Invision Power Board © 2001-2008 Invision Power Services, Inc.