Even though the DMA buffer itself does not move the ISR copies from a movable
buffer into the static commit buffer. To ensure this copying yields consistent
data it must not be interrupted by this ISR..
Also bump the commit buffer size to 2k, this should reduce the overhead
considerably because many clips are smaller than that (especially on
swcodec/speex).
Change-Id: I4e1ab83074f31fc91b51a58baa1df55ce659ac73
The voice engine can now request more voice data during decoding, it does
not require the entire clip to be available before start of decoding anymore.
Therefore the commit buffer does not need to hold an entire voice clip anymore,
and can be made greatly smaller.
Change-Id: I3eca9026448e725b9b8d0dae1efca0ad185371da
This unifies the talk.c for all possible voice payload. .talk clips are placed
onto the same unified clip cache, along with normal clips. This allows for more
effecient memory usage.
The cache handling makes a slight difference between normal clips and .talk
ones: .talk clips can be cached multiple and are always freed first.The extra
logic to avoid loading multiple copies of .talks is not necessary because the
will be freed first anyway.
Change-Id: I88d056a0a613b129f5875f50fdb757b58bac0a42
This unifies the talk.c for all targets. The only separation is left is
TALK_PROGRESSIVE_LOAD: When this is defined the talk buffer will not be
initially prefilled. This is useful for super slow storage or when the buffer
is not large enough to prefill it with useful clips (the prefill code could
be made smarter too).
The buffer size can be adjusted. By default lowmem uses 100k while
other targets load the entire file. The bigger the more clips can be cached
but with diminishing returns.
Change-Id: Ife38fb101c53093117e6638c40c65f7d177a31d4
Previously the clip cache of TALK_PARTIAL_LOAD reserved space N clips, each slot
was as big as the maximum sized clip which was necessary to replace clips
in-memory in MRU-style.
The cache management now uses buflib to allocate and free each clip, using the
clip's real size. This allows the clip cache to be much more compact, because
no space is wasted for the max. sized clip. This makes use of buflib's ability
to easily manage differently-sized memory chunks by moving them to make free
space.
As an example: for english.voice TALK_PARTIAL_LOAD allocated 288k in advance.
for just 64 clips. With this patch ~70 clips can be stored in a 100k buffer.
This, the memory usage is cut by 2/3 and almost optimal (there's still the
buflib per-alloc cookie overhead).
As a result the TALK_PARTIAL_LOAD buffer is restricted to 100k which still
allows for more clips than previously, on average.
Change-Id: I257654071e9a95770cd6db2c2765f020befce412
This is necessary because when voice is active audio is disabled. But only
audio was able to shrink it's buffer to let other memory allocs succeed.
talk needs to be able to do this too when it owns the audio buffer exclusively.
Change-Id: Idea8ab90da7169f977c0c766cccb42c4fe6d6e81
When the policy is not set, it'll by default not give the clip buffer away.
Callers of core_alloc_maximum() suffer from this. However, the thumbnail
buffer can be easily freed when needed because nothing needs to be
reloaded from disk when it is reallocated (thumbnail clips are loaded on
demand, when in the file browser). Do this to give core_alloc_maximum() callers
a better chance to succeed with the default talk buffer policy.
Change-Id: I8c0da29c520612ca903f6c930bd7c74ae97eca3b
On hwcodec talk.c has the entire audio buffer (not just parts of it), therefore
it must give up everything and cannot count on core_alloc_maximum() to return
the remaining space. This is equivalent to it was handled before 22e802e.
You could probaby do smarter and shrink for example the .talk clip buffer
but is it really worth it?
Change-Id: Idc3431c59fb41b05338559c615093358c5d8ed9b
This fixes the radioart crash that was the result of buffering.c working
on a freed buffer at the same time as buflib (radioart uses buffering.c for the
images). With this change the buffer is owned by buflib exclusively so this
cannot happen.
As a result, audio_get_buffer() doesn't exist anymore. Callers should call
core_alloc_maximum() directly. This buffer needs to be protected as usual
against movement if necessary (previously it was not protected at all which
cased the radioart crash), To get most of it they can adjust the willingness of
the talk engine to give its buffer away (at the expense of disabling voice
interface) with the new talk_buffer_set_policy() function.
Change-Id: I52123012208d04967876a304451d634e2bef3a33
When allocating the voice buffer, it's supposed to start at the beginning
of the audio buffer, not at the end of the voice buffer. ;-D
Might clear up a thing or two.
Change-Id: I94796ff21090bcc56813cdc569957a1a9178abcd
Buffers are not allocated and thread is not created until the first
call where voice is required.
Adds a different callback (sync_callback) to buflib so that other
sorts of synchonization are possible, such as briefly locking-out the
PCM callback for a buffer move. It's sort of a messy addition but it
is needed so voice decoding won't have to be stopped when its buffer
is moved.
Change-Id: I4d4d8c35eed5dd15fb7ee7df9323af3d036e92b3
Use generic void * and size_t and make mp3_play_data and its callback
agree on types. Use mp3_play_callback_t instead of prototyping
right in the function call (so it's not so messy to look at). Change
doesn't appear to require plugin API version increment.
Change-Id: Idcab2740ee316a2beb6e0a87b8f4934d9d6b3dd8
* Fix .talk clips on hwcodec. Voice does have the entire audio buffer available there.
* Get rid of the separate TALK_PROGRESSIVE_LOAD in favour of the more advanced
TALK_PARTIAL_LOAD i.e. use the latter on the Ondios as well. This gets rid of quite
some ifdefing, and has the advantage that the voice file can be larger than the buffer
(at a slight binsize cost).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@30916 a1c6a512-1295-4272-9138-f99709370657
Since r30308 the talk buffer was set to NULL if e.g. a plugin called
audio_get_buffer() to steal the talk buffer. Since there's no audio_release_buffer() kind of function
the talk buffer was never set back again.
When trying to talk try to get the audio buffer with audio_get_buffer() as well,
which works until the audio buffer gets properly reinitialized.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@30840 a1c6a512-1295-4272-9138-f99709370657
The buflib memory allocator is handle based and can free and
compact, move or resize memory on demand. This allows to effeciently
allocate memory dynamically without an MMU, by avoiding fragmentation
through memory compaction.
This patch adds the buflib library to the core, along with
convinience wrappers to omit the context parameter. Compaction is
not yet enabled, but will be in a later patch. Therefore, this acts as a
replacement for buffer_alloc/buffer_get_buffer() with the benifit of a debug
menu.
See buflib.h for some API documentation.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@30380 a1c6a512-1295-4272-9138-f99709370657
Do it the hwcodec way which doesn't need a buffer_alloc(). The buffer for the
.talk files is now allocated together with the voicefile buffer.
Should also fix a panic when the .talk file buffer was allocated late at runtime.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@30335 a1c6a512-1295-4272-9138-f99709370657
The buffer_offset paramter of audio_init_recording() is removed as it
was unused in both implementations.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@30310 a1c6a512-1295-4272-9138-f99709370657
Namely, introduce buffer_get_buffer() and buffer_release_buffer().
buffer_get_buffer() aquires all available and grabs a lock, attempting to
call buffer_alloc() or buffer_get_buffer() while this lock is locked will cause
a panicf() (doesn't actually happen, but is for debugging purpose).
buffer_release_buffer() unlocks that lock and can additionally increment the
audiobuf buffer to make an allocation. Pass 0 to only unlock if buffer was
used temporarily only.
buffer_available() is a replacement function to query audiobuflen, i.e. what's
left in the buffer.
Buffer init is moved up in the init chain and handles ipodvideo64mb internally.
Further changes happened to mp3data.c and talk.c as to not call the above API
functions, but get the buffer from callers. The caller is the audio system
which has the buffer lock while mp3data.c and talk mess with the buffer.
mpeg.c now implements some buffer related functions of playback.h, especially
audio_get_buffer(), allowing to reduce #ifdef hell a tiny bit.
audiobuf and audiobufend are local to buffer.c now.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@30308 a1c6a512-1295-4272-9138-f99709370657
On these targets the full voice file can't be loaded so only load 64
clips at a time (the size of the queue)
Voice now works on clipv1
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@27962 a1c6a512-1295-4272-9138-f99709370657
now playback still works if voicing is enabled on the clipv1
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@26047 a1c6a512-1295-4272-9138-f99709370657
This is to a) to cleanup firmware/common and firmware/include a bit, but also b) for Rockbox as an application which should use the host system's c library and headers, separating makes it easy to exclude our files from the build.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@25850 a1c6a512-1295-4272-9138-f99709370657
