Make the host API and IPC to set and get component runtime data and values
generic with no bespoke commands at the IPC level.
Create a generic structure that descibes the type and size of data and
whether that data is appended to the IPC message or is available to
be DMAed from host.
Update the components to use the new structure. Some components like
SRC, tone and tone still require some work to export thier ABI.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
This patch adds to ipc.h enabling for two new components to read and
write test vectors via file I/O. They are specific to test bench usage
on development computer only and are not real SOF components. This update
avoids need to edit this file for test bench build.
Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>
This patch fixes issues in setting up the tone generator for correct
sample rate and channels number. Component IPC was missing latest needed
changes. The buffers handling issue is fixed. Obsolete commented code is
removed.
Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>
These conversion filters are no more used and will be replaced later with
new. The deletion is because the stop-band spec was changed from 0.51xFs
to 0.50xFs (no intentional aliasing of transition band) so the file
names of new default high-quality filters will be with different filename.
The filter spec is appended to file name to differentiate different
bandwidth filters for same rates fraction.
Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>
This patch fixes bugs in sink and source buffers handling and adds
support for short coefficient type to save RAM in platforms with less
space such as BYT. The patch introduces also a RAM saving minimum
conversions set between 16, 44.1, and 48 kHz. A new possible rate 18.9 kHz
was added to supported rates indication bits. SRC filters quality is also
adjusted for lower default performance. The higher performance will need
>16bit coefficients usage. The larger conversion set and 32 bit
filter coefficients can be easily restored with the SRC tools in rimage.
Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>
Current definitions are missing 12 and 64 kHz, have
typos (11250, 22500), list a weird 40kHz and don't
have extension capabilities for non-standard rates
Before we create ABI problems with the Linux kernel, let's add all
the sampling frequencies needed before we submit stuff upstream
(the same list needs to be reflected in the kernel include/uapi)
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Improve support for I2S and DSP modes alongside using topology data to
define the mode configuration. Add trace errors to reject any unsupported
configurations.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Some components have multiple sources/sink and the correct host side
component must be used when installing params. Make sure the correct
component is used here by tracking the previous component and passing it
to the params installation.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Messages starting at offset 1 remove any ambiguity over whether the
message has been constructed correctly.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Complain loudly in trace if we have components overruning or underruning
buffers. Components should check buffer free/avail before use.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Perform a check in the copy() function to make sure there is enough
data to copy to/from host buffers.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Add an XRUN handler into the pipeline to report bag overrun and underrun
from component buffers to the host. An XRUN on a component will now cause
a IPC XRUN message to each PCM interface source/sink to the component.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Make sure the preloader can detect when all buffers are loaded prior
to trigger and add a limiter to make sure we dont pre-load forever.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Remove the unused pointer to host parameters in params(). This was only
useful to the host component. Provide some component frame configuration
variables to give more flaxability than host params.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Use a generic structure containing common DAI settings for configuration.
This structure also contains a tail of DAI specific data that can also
be used by DAI drivers.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Improve IPC error reporting and values to host by tracing all failures
and making sure errors are returned when detected.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Allow pipeline host and DAI components to be timestamped in a platform
specific manner so that host DMA & DAI DMA position alongside any local
DSP timestamp can be returned to the host drivers.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Components will now each have a 64bit position counter that will indicate
the processing position in bytes for each component. This can be used
by time stamping, trace and debug.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
If values using in SOF_GLB_TYPE are > 0x08 then the highest bit is set. If this
bit is set then when the data is shifted back (for example in iGS in intel-ipc)
it will shift arithmetically and leave the sign bit which will cause comparisons
to fail. Ex:
#define FOO 0x08
#define BAR 0x08U
#define LSHIFT(x) (x << 28)
#define RSHIFT(x) (x >> 28)
int main() {
printf("%x, %x\n", RSHIFT(LSHIFT(FOO)), RSHIFT(LSHIFT(BAR)));
// fffffff8, 8
}
Singed-off-by: Curtis Malainey <cujomalainey@google.com>
Host DMA needs it's own timeout as its not bound to any synchronous data
transfer timings.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Move all the static scheduling params out of the platform headers and into
the static pipeline source as they are no longer used by dynamic pipeline
code.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
The completion condition may complete very close to the timeout. This will
result in an error being returned alongside completion. make sure we dont
return an error in this case.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
This is a large tightly coupled patch that adds support for pipeline
definition and scheduling configuration via host IPC.
The intention is that hosts can create and configure pipelines dynamically
at runtime based on use case. This involves defining a pipeline topologies
and scheduling configuration. It also includes configuration of
individual components within a pipeline.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Minor debugging issue, when verbose trace is turned on it will not compile
because these variables are incorrect.
Signed-off-by: Curtis Malainey <cujomalainey@google.com>
We should copy from [inbox_base + offset(0 here)] src to
msg->rx_data dst here, otherwise, copying to 0 dst will
crash the firmware here.
Reported-by: Luo Xionghu <xionghu.luo@intel.com>
Signed-off-by: Keyon Jie <yang.jie@linux.intel.com>
This change fixes the warnings and errors returned by checkpatch.pl in
source code files eq_fir.c, eq_iir.c, fir.h, and tone.c.
Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>
SRC sets the sink and downstream pipeline PCM rate, period_count and
period_bytes in src_params() according to received out_rate via IPC.
Fail in delay lines allocation results to -ENOMEM error.
Function call comp_set_sink_params() is renamed to
comp_buffer_sink_params().
Added void to functions and corresponding headers without parameters.
Plus white space and long lines fixes for all SRC source files.
Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>
Liam Girdwood