Hydrogenaudio Forums > Any integer lossless encoder to fit "iRockbox"?

Tang

Feb 12 2005, 13:44

Hello,
As some may know the Rockbox iRiverport for iHP1xx is on his way thanks to LinusM releasing his precious "bootloader"...

Actually the team is concentrating his work on the AUDIO part... The great priority of course is decoding (especially MP3 even if a lot of audio codecs are in the TODOlist!)

Anyway the team have made some searchs for encoding part and they complaining they don't find much things that seems suitable for the iRiverport since they need INTEGER encoders that won't overdrive the iHP processor...

So i've asked on the HA "rockbox iRiverport" thread for LOSSY integer encoder with Open-source license of course (check HERE)...

As there is a lossless section i make this thread to ask the same thing but for an interger lossless open source encoder!

I will be very kind of you to put all information you know about such an encoder!

Thanks,

Tanguy

rjamorim

Feb 12 2005, 16:01

All lossless encoders use integer math only. They can't risk using floats and introduce rounding errors to the audio streams.

I believe WavPack would be an excellent choice, because it's a very fast encoder (not using any x86-only ASM tricks like Monkey's) and the lossy mode would be perfect for people wanting to encode smaller files (because of space constraints). No other high quality lossy codec is freely available in an integer-only implementation, AFAIK.

Tang

Feb 12 2005, 16:14

QUOTE(rjamorim @ Feb 12 2005, 02:01 PM)

All lossless encoders use integer math only. They can't risk using floats and introduce rounding errors to the audio streams.

I believe WavPack would be an excellent choice, because it's a very fast encoder (not using any x86-only ASM tricks like Monkey's) and the lossy mode would be perfect for people wanting to encode smaller files (because of space constraints). No other high quality lossy codec is freely available in an integer-only implementation, AFAIK.

Thanks Roberto

I'm gonna transmit this and your answer in the other thread to Rockbox...

HotshotGG

Feb 12 2005, 18:00

QUOTE(rjamorim @ Feb 12 2005, 02:01 PM)

No other high quality lossy codec is freely available in an integer-only implementation, AFAIK.

What does that make Tremor then? I thought LGPL was free too? are you just talking about it terms lossless/lossy hybrid?

rjamorim

Feb 12 2005, 18:02

QUOTE(HotshotGG @ Feb 12 2005, 10:00 PM)

What does that make Tremor then? I thought LGPL was free too? are you just talking about it terms lossless/lossy hybrid?

We are talking about encoders, sire

HotshotGG

Feb 12 2005, 18:04

QUOTE

We are talking about encoders, sire

Ahh.. Ok master I apologize read it too fast

Tang

Feb 12 2005, 18:09

QUOTE(HotshotGG @ Feb 12 2005, 04:04 PM)

QUOTE

We are talking about encoders, sire

Ahh.. Ok master I apologize read it too fast

Eh eh you're not the only one...

emtee

Feb 12 2005, 20:24

It'd also be nice to point the rockbox team to the Which is the best lossless codec thread, just in case they're unsure of wavpack's excellent capabilities. Maybe they can do something cool with the hybrid mode.

jcoalson

Feb 13 2005, 03:38

(minor correction, lossless encoders can use floating point math. e.g. libFLAC uses floating point in the LPC analysis stage.)

the latest release of libFLAC has a compiler define (FLAC__INTEGER_ONLY_LIBRARY) that builds the whole library (decoder, encoder, etc.) with integer only.

Josh

rjamorim

Feb 13 2005, 05:58

Oops. I stand corrected.

Tang

Feb 13 2005, 10:43

QUOTE(jcoalson @ Feb 13 2005, 01:38 AM)

(minor correction, lossless encoders can use floating point math. e.g. libFLAC uses floating point in the LPC analysis stage.)

the latest release of libFLAC has a compiler define (FLAC__INTEGER_ONLY_LIBRARY) that builds the whole library (decoder, encoder, etc.) with integer only.

Josh

Hi,
Thanks for these correction... in fact seems that for now the Rbx team get info directly here in the HA threads... It's the best way... So there is no need for me to report them any new information..

I guess they will subscribe to HA board when they will encounter big issue and would contact any precise member or ask something to community...

BTW I'm very thankfull to every HA people...

Best regards,
Tanguy

PS: I'll check the Rbx HA thread to add any new infomation if it appears necessary... in fac I'm gonna add something NOW (latest progress) so stay tuned...

Twinsen

Apr 12 2005, 07:35

Hi!
Actually there is a really fast free (non-commercial) mp3 encoder. I tested it on Intel XScale PXA255 @400MHz, 100MHz 32bit SDRAM. This processor is quite common in PDA's, and has no floating point unit! It took 40sec to compress 60sec audio, so it works real time.
Input: 44100Hz 16bit stereo
Output: 128kbps 44.1KHz mp3
The encoder is the sample mp3 encoder for the Intel Performance Primitives (IPP) library. It can be freely downloaded from intel's website!

Tang

Apr 12 2005, 09:11

QUOTE(Twinsen @ Apr 12 2005, 05:35 AM)

Hi!
Actually there is a really fast free (non-commercial) mp3 encoder. I tested it on Intel XScale PXA255 @400MHz, 100MHz 32bit SDRAM. This processor is quite common in PDA's, and has no floating point unit! It took 40sec to compress 60sec audio, so it works real time.
Input: 44100Hz 16bit stereo
Output: 128kbps 44.1KHz mp3
The encoder is the sample mp3 encoder for the Intel Performance Primitives (IPP) library. It can be freely downloaded from intel's website!

Hi Twinsen,
Thanks for the notice, i'm gonna relay this to Rockbox immediatly.. Thanks a lot...

Best regards,
Tanguy

rjamorim

Apr 12 2005, 09:22

QUOTE(Twinsen @ Apr 12 2005, 10:35 AM)

The encoder is the sample mp3 encoder for the Intel Performance Primitives (IPP) library. It can be freely downloaded from intel's website!

I'm afraid that encoder is incompatible with the GPL, therefore it would conflict with the rest of the RockBox code.

Also, I would suspect it's not optimized at all for M68K.

Tang

Apr 12 2005, 09:30

QUOTE(rjamorim @ Apr 12 2005, 07:22 AM)

QUOTE(Twinsen @ Apr 12 2005, 10:35 AM)

The encoder is the sample mp3 encoder for the Intel Performance Primitives (IPP) library. It can be freely downloaded from intel's website!

I'm afraid that encoder is incompatible with the GPL, therefore it would conflict with the rest of the RockBox code.

Also, I would suspect it's not optimized at all for M68K.

MMM indeed if it isn't GPL compatible it's a fatal issue..

The non optimisation should not be a real matter since there is no mp3 integer encoder optimized for M68k...

Cheers Roberto

rjamorim

Apr 12 2005, 09:36

QUOTE(Tang @ Apr 12 2005, 12:30 PM)

MMM indeed if it isn't GPL compatible it's a fatal issue..

Intel is delightfully vague on that subject

http://www.intel.com/software/products/noncom/faq.htm

QUOTE

Q. If I use the non-commercial product to build my product, can I open source it?
A. The FAQ for GNU GPL has information related to this question. Please refer to the GNU website "Frequently Asked Questions about the GNU GPL" at: http://www.gnu.org/licenses/gpl-faq.html For further clarification, please contact the Free Software Foundation at: [...]

Tang

Apr 12 2005, 09:40

Eh eh very sympathetic as a perfect "langue-de-bois" exemple!

handy_man_dan

Jun 21 2005, 22:29

QUOTE(Twinsen @ Apr 12 2005, 07:35 AM)

Hi!
Actually there is a really fast free (non-commercial) mp3 encoder. I tested it on Intel XScale PXA255 @400MHz, 100MHz 32bit SDRAM. This processor is quite common in PDA's, and has no floating point unit! It took 40sec to compress 60sec audio, so it works real time.
Input: 44100Hz 16bit stereo
Output: 128kbps 44.1KHz mp3
The encoder is the sample mp3 encoder for the Intel Performance Primitives (IPP) library. It can be freely downloaded from intel's website!

Twinsen,
Thanks for this performance data. I had recently built the IPP encoder sample and tested at 44.1Khz, 16 bit stereo in and 192kbps out. Performance on a very similar PDA was about 4x SLOWER than real time so I put the IPP encoder sample on the sidelines. I am using eVC4.1 with all optimization turned on. What compiler did you use and any special compiler switches?? Thanks