Sluggish when filling buffer

[PHK Brasil]

♪
Hi.

When my iPod 5G 60gb fills up the buffer, it takes a long time to do it on Rockbox, comparing to Apple's firmware.

Maybe 1 or 2 minutes, aproximately.

Is like it was "fighting" to fill the buffer. Maybe is cause CPU is busy with any others works, like Crossfeed.

But without Crossfeed, EQs, and others, it keeps taking more time than Apple's firmware. Is this normal???

This allways happened in all builds I used.

Anybody knows how much time to fill up the 32MB buffer and the 64MB patched one?


Thanks.
♪

[Llorean]

How do you know Apple is using the full buffer?

[PHK Brasil]

Hi Llorean!

How do you know Apple is using the full buffer?

Answering: Just putting a long playlist to play. When Apple's firmware buffer or rebuffer, it takes a short time to do it.

On Rockbox, the HD spins and spins and spins...

And about my question? Any answer?

Thanks again.
♪

[Llorean]

No, what you've shown me is that your evidence is that Apple spends less time buffering, not that they fill the same amount of the buffer. Their firmware is several megabytes, and evidence suggests that all of it is in RAM at once, which means they have much less RAM available for audio buffer, and so could reasonably take less time to fill.

That combined with the fact that Rockbox's processor is busier in general because as we've said many times playback needs optimizations easily accountes for Rockbox taking more time. It's very likely filling more space and a bit slower of a speed.

[PHK Brasil]

♪
Hi Llorean! Thank you for the ultra fast reply!

I will make a test, putting a long playlist in Apple's firmware and seeing when the HD spins to rebuffer.

Then I'll go to Rockbox firmware doing the same playlist and comparing.

Thanks a lot again. I'll be back.

♪

[PHK Brasil]

♪

No, what you've shown me is that your evidence is that Apple spends less time buffering, not

that they fill the same amount of the buffer. Their firmware is several megabytes, and

evidence suggests that all of it is in RAM at once, which means they have much less RAM

available for audio buffer, and so could reasonably take less time to fill.

iPod 5G 30gb = 32MB total RAM
iPod 5G 60gb = 64MB total RAM (Hmmm...at least 32MB for audio buffer)

Both uses same firmware. So, iPod 5G 60gb has at least 32MB extra for audio buffer + the same ammount of MBs of 30gb's model audio buffer.
Is much more ammount of audio buffer than Rockbox (only 32MB).

Time to fill the audio buffer:

-Apple OS takes around 7 seconds while PLAYING

-Rockbox takes 12s to fill 32MB audio buffer (PAUSED, no EQs or Crossfeed)

-Rockbox takes 50s to fill 32MB audio buffer (PLAYING, no EQs or Crossfeed)

-Rockbox takes 75s to fill 32MB audio buffer (PLAYING, Crossfeed ON)

Just a quick comparison, but it depends of bitrates, audio format, etc, while playing and filling buffer.

Thanks.
♪

[soap]

But LLorean's basic question was "how often does Apple firmware rebuffer?"  I don't think he ever disputed that Apple FW appears to spend less time buffering.  
This can be tested quite simply by creating a playlist of high-bitrate music that is playable on both AppleFW and Rockbox, starting the playlist in one firmware, measuring how long the hard drive spins, then measuring how long until the drive spins again.  Repeat with firmware #2.  Preferably done with a playlist longer than the buffer of your player, for we know Rockbox will refill the buffer, even if you create a playlist of only one song.

[nls]

IIUC the portalplayer processors have dma that can be used for reading from the disk and should be a lot faster than what rockbox does now, and that is probably used in the OF. This combined with the fact that the cpu is generally rather busy in rockbox on portalplayer based DAP's makes buffering slow.

[Llorean]

As Soap said, I never denied that Rockbox likely buffers more slowly. What I'd like to establish is *how much* more slowly.

[Sentertainment]

I'm quite sure rockbox has a larger buffer too....it seems to be buffering 2-4 tracks ahead on my iPod.

[PHK Brasil]

I'm quite sure rockbox has a larger buffer too....it seems to be buffering 2-4 tracks ahead on my iPod.

Sentertainment, not in the case o iPod video 60gb. They have 64MB of total RAM, being 40MB for audio buffer.
Rockbox standard has only 30MB.

But LLorean's basic question was "how often does Apple firmware rebuffer?"  I don't think he ever disputed that Apple FW appears to spend less time buffering.  
This can be tested quite simply by creating a playlist of high-bitrate music that is playable on both AppleFW and Rockbox, starting the playlist in one firmware, measuring how long the hard drive spins, then measuring how long until the drive spins again.

Yes, exactly what I did.

As Soap said, I never denied that Rockbox likely buffers more slowly. What I'd like to establish is *how much* more slowly.

I showed that Apple's firmware spends less time buffering, filling more amount of buffer than Rockbox. Approximated 40MB against 30MB of Rockbox.

Sorry, if I'm doing a confusion here.

The complete results, tested with 320kbps CBR, iPod vídeo 5G 60gb

1)Apple's firmware fills 40MB buffer, and rebuffer after 17 minutes playing music

Time filling buffer:      -8s PLAYING


2)Rockbox’s firmware standard fills 30MB buffer, and rebuffer after 12:40 minutes

Time filling buffer:      -15s PAUSED
 Â                             -50s PLAYING
 Â                             -90s PLAYING / CROSSFEED ON


3)Rockbox’s firmware (64MB patch) fills 60MB buffer, and rebuffer after 25:20 minutes.

Time filling buffer:      -15sPAUSED
 Â                             -50s PLAYING
 Â                             -80s PLAYING / CROSSFEED ON

Basically, they are slowly when CPU is busy. PAUSED, they are much faster.

But, supposing iPods video 5G 30GB & 60gb have the same firmware... So, the iPod 30gb in Apple's OS has only about 10MB for audio buffer??

Maybe, anybody who has a 30gb one could test and tell us, please.

Bye, and sorry’bout my english.
♪