Resource usage¶
The mic array unit requires several kinds of hardware resources, including ports, clock blocks, chanends, hardware threads, compute time (MIPS) and memory.
This page attempts to capture the requirements for each hardware type with relevant configurations.
Warning
The usage information below applies when the default usage model is used. Resource usage in an application which uses custom mic array sub-components will depend crucially on the specifics of the customization.
Discrete Resources¶
Resource |
Count |
|---|---|
port |
3 |
clock block |
1 (SDR) 2 (DDR) |
chanend |
4 |
thread |
1 or 2 |
Ports¶
In all configurations, the mic array unit requires 3 of the xcore.ai device’s hardware ports. Two of these ports (for the master audio clock and PDM clock) must be 1-bit ports. The third (PDM capture port) can be 1-, 4- or 8-bit, depending on the microphone count and SDR/DDR configuration.
Clock Blocks¶
In applications which use an SDR microphone configuration, the mic array unit requires 1 of the xcore.ai device’s 5 clock blocks. This clock block is used both to generate the PDM clock from the master audio clock and as the PDM capture clock.
In applications which use a DDR microphone configuration, the mic array unit requires 2 of the xcore.ai device’s 5 clock blocks. One clock is used to generate the PDM clock from the master audio clock, and the other is used as the PDM capture clock (which must operate at different rates in a DDR configuration).
Chanends¶
Chanends are a hardware resource which allow threads (possibly running on different tiles) to communicate over channels. The mic array unit requires 4 chanends. Two are used for communication between the PDM rx service and the decimation thread. Two more are needed for transferring completed frames from the mic array unit to other application components.
Threads¶
The PDM rx service can run either as a stand-alone thread or as an interrupt within the decimator thread. Accordingly, the mic array requires one thread when the PDM rx service runs in interrupt mode, and two threads when it runs as a stand-alone thread.
Running PDM rx as a stand-alone thread modestly reduces the mic array unit’s MIPS consumption by eliminating the context switch overhead of an interrupt. The cost of that is one hardware thread.
Note
When configured as an interrupt, PDM rx ISR is typically configured on the decimation thread, but this is not a strict requirement. The PDM rx interrupt can be configured for any thread on the same tile as the decimation thread. They must be on the same tile because shared memory is used between the two contexts.
Compute¶
The compute requirement of the mic array unit depends strongly on the actual configuration being used. The compute requirement is expressed in millions of instructions per second (MIPS) and is approximately linearly related to many of the configuration parameters.
Each tile of an xcore.ai device has 8 hardware threads and a 5 stage pipeline. The exact calculation of how many MIPS are available to a thread is complicated, and is, in general, affected by both the number of threads being used, as well as the work being done by each thread.
As a rule of thumb, however, the core scheduler will offer each thread a minimum
of CORE_CLOCK_MHZ/8 millions of instruction issue slots per second (~MIPS),
and no more than CORE_CLOCK_MHZ/5 millions of issue slots per second, where
CORE_CLOCK_MHZ is the core CPU clock rate (specified as SystemFrequency in the XN file).
With a core clock rate of 600 MHz, that means that each core should expect at least 75 MIPS.
Table Estimated MIPS (per configuration) shows the mic array MIPS by profiling an application that includes the
mic array. The application used to generate the MIPS numbers runs the default mic
array API (so the decimator running in a single hardware thread) with all defines set to their default values as listed
in Configuration defines (mic_array_conf_default.h) except for MIC_ARRAY_CONFIG_MIC_COUNT
and MIC_ARRAY_CONFIG_USE_PDM_ISR. These two (along with the output sampling rate) are varied to build the
different configurations that are profiled.
mic count |
PDM RX |
output samp freq |
MIPS |
|---|---|---|---|
1 |
ISR |
16000 |
14.146 |
1 |
ISR |
32000 |
17.234 |
1 |
ISR |
48000 |
21.305 |
1 |
THREAD |
16000 |
12.930 |
1 |
THREAD |
32000 |
15.954 |
1 |
THREAD |
48000 |
19.961 |
2 |
ISR |
16000 |
29.310 |
2 |
ISR |
32000 |
34.621 |
2 |
ISR |
48000 |
41.934 |
2 |
THREAD |
16000 |
27.006 |
2 |
THREAD |
32000 |
32.285 |
2 |
THREAD |
48000 |
39.533 |
Note
The MIPS numbers scale approximately linearly with the number of microphones. Although the table lists values only for the 1- and 2-mic configurations, these results can be extrapolated to estimate MIPS for configurations with a higher microphone count. If a given configuration cannot be accommodated within a single hardware thread, the decimator can be split across multiple threads to distribute the compute load. This approach is not supported by the default mic array API. An example of a custom multi-threaded decimator implementation can be found in app_par_decimator.
Memory¶
The memory cost of the mic array unit has three parts: code, stack and data. Code is the memory needed to store compiled instructions in RAM. Stack is the memory required to store intermediate results during function calls, and data is the memory used to store persistent objects, variables and constants.
Table Memory usage (in bytes) reports the memory usage of two minimal applications that include the mic array:
one using the default mic array API and another using a custom
configuration created by instantiating a MicArray object.
Both applications are built for 1 or 2 microphones with a 16 kHz output sample rate, with the other compile-time parameters set to their default values as defined in Configuration defines (mic_array_conf_default.h).
Memory for higher microphone counts can be extrapolated from the 1- and 2-mic numbers.
Across different sampling rates, memory usage with the default API remains unchanged since the default API compiles all the decimation filters included in the library. For custom usage, the data memory across different sampling rates varies depending on the decimation filters included.
Config |
Available on tile |
Used |
Stack |
Code |
Data |
|---|---|---|---|---|---|
1mic_custom |
524288 |
12572 |
572 |
8070 |
3930 |
1mic_default |
524288 |
16580 |
636 |
9058 |
6886 |
2mic_custom |
524288 |
13964 |
580 |
8278 |
5106 |
2mic_default |
524288 |
18084 |
636 |
9378 |
8070 |
Note
The default API requires more memory than the custom configuration. The additional code memory comes from the wrapper and abstraction code included in the default API. The increased data memory results from the inclusion of filter coefficients for all filters provided by the library, whereas the custom build includes only the coefficients required by the specific MicArray instance.
Note
A minimal empty application (no mic array) occupies ~4.6 KiB (Stack: 356 B, Code: 3754 B, Data: 542 B). Subtract this baseline from the reported totals to isolate mic array overhead.