All Projects

http://web.mit.edu/mgg6/www/Game_Scoop_Presents__Predicting_Nintendos_2013_E3.mp3

The ID3v2 tag in this file contains a fair amount of random metadata which seems to confuse Rockbox’s parser such that album art is not displayed. Extracting the album art using foobar2000, deleting the existing tag and then reattaching it fixes display of the embedded JPEG so the jpeg itself is fine.

Tested with today’s build (7c7fa36) on device (e200v1) and in sim.

My old linux machine is apparently too out of date to easily get latex packages needed to compile the manual, so I’ve blind written this without testing. If someone could double check that they compile cleanly, I can commit them.

This patch changes some of the links on the side bar of the front page to reflect the changes to the wiki over the past few years.

1a) Remove the FAQ link since it is out of date and not too relevant to most targets. Furthermore, individual port wiki pages still link the relevant parts.

1b) Add a link to the port status wiki page in its place.

2a) Remove the docs index link. Also not kept up to date, and less useful now that we have dedicated development and port status ‘portal’ pages that link to the same information.

2b) Add a link to the development guide in its place.

Thoughts? My goal was basically to stop funneling new users towards info that is based around the old Archos, iRiver and (PP) iPods. IMO that was rather confusing to new users who may not be aware of this project’s history. Instead I’d like to funnel them towards the port status page (so they can see if their target meets their needs) and the manual.

http://184.168.139.11/non_working.m4a

mp4.c errors out with “MP4 metadata error. errno 22, samples 9507840, frequency 44100. filesize 0”.

Plays in other software without error.

If I insert a microsd card after the bootloader splash, but before the main menu comes up, rockbox will usually freeze with no message. Repeatable on my player with a 16GB microSD card (works fine once booted).

Tested r29548.

http://churfirstenblick.ch/radio/radiozu.mp4

This file fails with the following message in the sim:

Read metadata for /radiozu.mp4
MP4 metadata error. errno 0, samples 2428928, frequency 44100, filesize 0
parsing /radiozu.mp4 failed (format: AAC)

This is printed in apps/metadata/mp4.c because the read_mp4_container returns a filesize of 0 bytes. The file seems to play ok in foobar2000.

http://samples.mplayerhq.hu/A-codecs/lossless/luckynight.rmvb

Trying to play this file, or even just building the database with it on the disk will result in a divide by zero error, at least on the Fuze V1.

Since we don’t actually support real audio lossless, my guess is that the parser is not careful enough about rejecting lossless files in a real media container.

Mostly for historical reasons, rockbox requires the user to hold down a key to enter charge mode when plugging in USB on a number of targets. Unfortunately, this has several problems on some targets, including:

1) The key is often quite different on different devices, even those with fairly similar layouts, so its often not obvious what to press.

For example, the gigabeat S and sansas all have a center select buttons, and a menu/home type button. On the S, the menu button enters charge mode. On the Sansas, select does it.

2) On some devices, its very difficult to hold the USB charge mode while inserting USB for mechanical reasons.

For example, on the Nano 2G or Sansa Clip, its extremely easy to hit another button while trying to force a USB plug in because the devices are both tiny and very sensitive.

To address the latter, the current code defines a USBPOWER_BTN_IGNORE, which is ignored if its also pressed. However for nearly all targets, the button is defined to something uesless, presumably because its purpose is not commented in the code.

This patch throws all that out. If you click any button while inserting USB, you will get charge mode. Asking on IRC, people seemed in favor of this approach. However, it is possible there are targets out there where this could make it difficult to enter USB mode due to awkward button placement. If so, this patch should be updated to work around that.

I'd appreciate feedback.

Work in progress iQMF using the ARMv5E 32×16 multiply instructions. So far this converts libatrac to using 16 bit iQMF coefficients, and then prescales everything going into the iQMF by 8 bits, then the output another 7 bits. Unfortunately, the sim audio is somewhat garbled because the prescaling isn’t performed correctly (scaling in the fixmul15 function works, but not prescaling the output of the imdct for some reason). I’m obviously missing something about how data is used in the codec, and either not scaling all the samples, or double scaling some of them.

I haven’t tested the asm code, but its fairly simple so it shouldn’t be too hard to fix once the sim is working. The ASM code is also not properly scheduled. Once its working it can be arranged to avoid pipeline stalls.

test_codec currently only works when the test file is smaller then the audio buffer. This makes it fail on low memory targets. This patch corrects that.

Volume tests of RB, the OF, and different amp settings are here:

http://anythingbutipod.com/forum/showpost.php?p=471215&postcount=383

RMAA tests driving a load for the OF, RB and various settings are here:

http://rmaa.elektrokrishna.com/index.php?dir=Comparisons/

Specifically, these two entries for the current SVN build and the patched one:

http://rmaa.elektrokrishna.com/Comparisons/Clip%2B%20-%20Rockbox%20Volume%200x16%20-%20Official%20build%2016%20Ohm.htm

http://rmaa.elektrokrishna.com/Comparisons/Clip%2B%20-%20Rockbox%20Volume%200x1B%20-%20Timar%20build%2016%20Ohm.htm

The AS chips use a two step volume control with separate levels for the DAC and headphone amp. This patch increases the DAC volume 7dB. So basically, +6dB in the current build (which does not distort) becomes 0dB in the patch (which also does not distort). The new +6dB corresponds to actually over driving the amp, and does distort (but probably not enough to notice in real music verses a full scale test tone).

Interestingly the OF can drive as loud as our new +6dB with very little distortion, which might suggest we don’t have one of the voltages set right somewhere. If we ever get a JTAGed as3525v2 player it would be great to dump the AS3543 registers.

libmad is full of code that doesn’t take into account pipeline interlocks on ARM9 and above. This fixes a bit of it, and gives a ~200kHz improvement.

Additionally, I noticed some of the ASM defines (PROD_ODDBACK_A, etc) are never used. Looking at the SVN logs, they probably never should have been committed. Anyone have an idea about them?

Currently on PP, AMS, and probably others the “Running Time” screen doesn’t actually get reset when rockbox is charged. The only way it’ll actually get reset that I can find is if the user inserts a charger while the running screen is open or the nvram is cleared.

This patch changes default_event_handler_ex to reset the running time on USB_DISCONNECT or CHARGER_DISCONNECT events.

Presently if the power button is held, but another button is accidentally pressed before the power down begins, the device does not shut down.

This is moderately annoying to me on the e200, but extremely annoying on the Clip, since on the Clip is really tiny and the volume buttons take up the entire side opposite to the power button.

This patch changes the shutdown to ignore other button presses. So if you hold volume and power, you still get to shutdown.

I’m not sure if this is safe on various targets.

Disables compiling various crossfade specific code in pcmbuf.c on low memory (AMS Sansa mainly) targets since this code is unused on them. Saves a bit under 2KB according to bloat-o-meter.

JdGordon suggested and I tend to agree that this should be made dependent on HAVE_CROSSFADE in the config-XXX.h files instead of just checking for MEMORYSIZE > 2.

This patch adds a malloc implementation to the codec library (malloc, calloc, realloc, free) based on Daniel’s dbestfit. Tremor is adapted to use it.

This patch decreases the worst case vorbis memory usage by at least 200KB by properly freeing temporary buffers. No other changes are made, and the old broken malloc implementation is left alone (speex still uses it for the moment).

While we have argued that malloc is not a good thing to have, removing it from tremor is probably not realistic, and the previous solution (tomal’s attempts to use temporary mallocs) is unsatisfactory. Furthermore, once libfaad is fixed, this change will enable us to drastically shrink the codec buffer.

The current day’s IRC logs are downloaded rather then rendered as webpages in google chrome. I’m guessing (without knowing anything at all about webpages) that this is a problem with headers MIME type or whatever.

http://www.rockbox.org/irc/log-today

This patch reorganizes variables by removing a seldom used array while putting a more frequently used one in IRAM.

This gives a ~2% speedup on PP5024, but I would be interested in any Coldfire results, where the difference will hopefully be larger.

test_codec doesn’t provide some of the threading functions in the codec interface because they’re not provided by the plugin interface. This cases test_codec to crash on SPC files, and any future multithreaded codecs.

This patch adds the needed functions to the plugin API, resolving the issue.

My measurements of the COP and boost power consumptions for the PP5024 suggest that load balancing rockbox as much as possible across both of the cores is optimal for battery life. To test this, I’ve put the synthesis filter from MAD on the coprocessor. This reduces CPU load sufficiently that the CPU doesn’t need to boost at 30MHz, and probably very little at 24MHz.

Unfortunately, theres still some glitching, either because of some cache coherency problem, or because I don’t properly synchronize the two threads.

I’ve removed the Vorbis, WMA and AAC IMDCT code and replaced it with a single IMDCT called from the codeclib which is based on the IMDCT used by Vorbis. Because Vorbis and WMA already used this code, there should be no functional changes for these codecs. AAC gains a surprisingly small increase in speed (178%→202% realtime or a savings of about 5.3MHz on the 192k test track from the CodecPerformanceComparison page with Sansa e200). Unless I’m missing something, the enormously complex FFT used in FAAD may not have been so slow afterall, though the the Vorbis IMDCT is faster still.

My 1 test album sounds fine on PP, but I’d appreciate any testing on Coldfire. Benchmarks would also be nice.

I've ported some of the ASM code from the WMA codec to libfaad with help from ammicon. This gives me a minor speedup on PP (saves ~1MHz). I'm curious to see how much it saves on Coldfire (and if it sounds alright).

I’d like to commit my initial work in converting rockbox to run as an application relatively soon because its fairly difficult to keep in sync.

In order to facilitate review and cleanup, I’ve split the patch into two portions. The first is just ifdef nonsense needed to allow “application” builds to use different settings then sim builds. This is likely to be non-controversial, though I wouldn’t mind any suggestions what the define for “rockbox is being compiled as an application on another OS” should be. Right now I use “SDL” but thats not quite right.

The second patch is the actual code needed to make this work: changes to the configure script, some code in the target tree, keymaps, etc. I’ve left most of the target tree stuff out of this patch since its not much good without a working port to some other device. Otherwise its just a copy of the uisimulator/sdl code in the target tree, and that seems kind of a mess. Instead I just compile the uisimulator folder and use it for SDL functions.

For now I’ve created a “generic” entry in the target tree. I’m not sure if this is the right name, but its basically just code that is specific to the application project, but does not assume any specific CPU. However, if we ever got an SDL port to an ARM target, this wouldn’t quite work either since there’d need to be an arm/sdl folder in addition to any generic code in generic/sdl.

For the actual target specific stuff, I’ve used the gigabeat F keymap, screen size, etc so that WPSes still work. It seems as sensible as any other target but I’m open to suggestions.

Finally for this to work, you will need to duplicate one of the bmps in the uisimulator/sdl folder and name it “UI-sdl.bmp”. I’m not really sure what the app bitmap should look like, or if one is even needed, so I haven’t done anything about this yet.

Thanks for any input on how I should proceed.

The Tremor IMDCT is much, much faster then the WMA (and almost certainly the FAAD) one. This patch copies the Tremor IMDCT over to libwma with minor changes to remove it’s connection to Vorbis.

The resulting decoder sounds correct on the Sim and ARM, while being 4MHz faster on PP5024.

I have not tested it on Coldfire, but expect some speedup and hopefully no reduction in quality.

Based on battery discharge data provided by Tomasz Wasilczyk using  FS#8070  and posted to the Sansa battery benchmark wiki page.

I can commit this whenever, but if other people want to run battery benches too, I could average them and perhaps get a better overall curve.

User "shotofadds" requested that someone upload this for him, evidently due to some problem using filespray. His description:

"Once applied, configure should list a new target (33: Cowon D2) and its bootloader build should generate a rockbox.iaudio file suitable for uploading with tcctool. This consists of a simple test program with basic LCD, button and I2C drivers. Happy hacking!"

Once he has registered and provided his full name, we can see about turning this FS entry to him, and then committing his patch.

(edit: fixed 'Cowon' typo in title)

At preglow’s request, here are the two patches that have been floating around for seeking in WMA.

The first works, but is slow and inaccurate. Theres a good chance you’ll end up in the next song if you seek too far.

The second doesn’t work at all for some reason, but should be much faster when fixed. It works like the first, but does a binary search for the correct ASF packet and attempts to restart playback at the new one. I spent ~15 minutes debugging it and ran out of time. If someone else wants to take a look at it, I’m sure its not that hard to fix. Actually seeking in general with WMA is quite simple, I just don’t have time to debug it.

Otherwise I’ll finish it as soon as my workload eases up.

Theres been some talk about using a theme based on Cabbie as the default theme in rockbox.

So far, variations of it exist for the Gigabeat, Ipod Video, Sansa e200 and H10.

The Gigabeat version can be seen here:

http://www.duke.edu/~mgg6/rockbox/gigabeat.png

For now, I’m using the gigabeat version as the reference.

Todo:

Port to the other targets, and clean up the preexisting ports, remove non-gpled graphics.

Come up with a version of the rockbox logo to put in the top of the WPS screen to replace the custom logo in the previous cabbie versions.

Switch everything over to appropriate, included fonts. So far I’m using Nimbus.