samples: i2s_codec: Refine i2s_codec sample #96976
Conversation
ZhaoxiangJin
commented
Oct 3, 2025
ZhaoxiangJin
commented
- Configure BLOCK count by using kconfig option 'EXTRA_BLOCKS' instad of fixed value. This allows user to customize the RAM usage based on their platform.
- Keep the PCM data in flash to avoid large RAM usage.
- Refine the sample to get the SAMPLE_BIT_WIDTH and BYTES_PER_SAMPLE from Kconfig options.
|
Hello @TomasBarakNXP, @VitekST, please help review. Thanks. |
| 0xc3, 0xc8, 0xc8, 0xc8, 0xc8, 0x38, 0xcf, 0x38, 0xcf, 0x1c, 0xd7, 0x1c, 0xd7, 0x37, 0xe0, | ||
| 0x37, 0xe0, 0x45, 0xea, 0x45, 0xea, 0xf8, 0xf4, 0xf8, 0xf4}; | ||
| const unsigned int __16kHz16bit_stereo_sine_pcm_len = sizeof(__16kHz16bit_stereo_sine_pcm); | ||
| static const unsigned int __16kHz16bit_stereo_sine_pcm_len = sizeof(__16kHz16bit_stereo_sine_pcm); |
| int "Extra block size" | ||
| default 32 | ||
| help | ||
| Count of DMIC channels to capture and process. |
There was a problem hiding this comment.
Is this really the count of DMIC channels?
ZhaoxiangJin
force-pushed
the
refine-i2s-codec-example
branch
from
3f5e03a to
a587e39
Compare
ZhaoxiangJin
force-pushed
the
refine-i2s-codec-example
branch
from
a587e39 to
dd50879
Compare
|
Update: |
| 0xc3, 0xc8, 0xc8, 0xc8, 0xc8, 0x38, 0xcf, 0x38, 0xcf, 0x1c, 0xd7, 0x1c, 0xd7, 0x37, 0xe0, | ||
| 0x37, 0xe0, 0x45, 0xea, 0x45, 0xea, 0xf8, 0xf4, 0xf8, 0xf4}; | ||
| const unsigned int __16kHz16bit_stereo_sine_pcm_len = sizeof(__16kHz16bit_stereo_sine_pcm); | ||
| static const unsigned int __16kHz16bit_stereo_sine_pcm_len = |
There was a problem hiding this comment.
const is probably not necessary here. See the reported issue: when cast to (void *) in main.c, const is lost.
There was a problem hiding this comment.
Hello @TomasBarakNXP, currently, both __16kHz16bit_stereo_sine_pcm and __16kHz16bit_stereo_sine_pcm_len are modified with const. I'm not sure what the "reported issue" you're referring to is. However, I understand that it's worth keeping the const declarations in both places.
Why is it worth keeping the const declaration for __16kHz16bit_stereo_sine_pcm?
- Using const for storing the PCM table in flash can significantly save RAM (also stated this motivation in the comments at the top of sine.h).
__16kHz16bit_stereo_sine_pcmis used here:
mem_block = (void *)&__16kHz16bit_stereo_sine_pcm;
ret = i2s_buf_write(i2s_dev_codec, mem_block, block_size);
The prototype of i2s_buf_write is: int i2s_buf_write(const struct device *dev, void *buf, size_t size), and the implementation only memcpy data from buf to the driver-allocated TX slab and then transmits it via DMA, buf is not modified (see drivers/i2s/i2s_common.c: memcpy(mem_block, (void *)buf, size)).
The code mem_block = (void)&__16kHz16bit_stereo_sine_pcm; simply casts the parameter to "lose const" to accommodate the void parameter of i2s_buf_write. Since the function only reads it, this is safe; the only drawback is that it's not perfectly const-correct. In other words, the lack of const API parameters is due to historical reasons/interface design issues, the actual source buffer is read-only.
Why is it worth keeping the const for __16kHz16bit_stereo_sine_pcm_len?
- To prevent accidental modification, it explicitly states that "this is read-only data." Any attempt to write to it will result in a compilation failure or an error when read-only segment protection is enabled, preventing potential risks such as length corruption and out-of-bounds access.
- Storing it in a read-only segment conserves RAM. Although the savings are not much.
| them and the codec will be expected to consume them. | ||
|
|
||
| config EXTRA_BLOCKS | ||
| int "Extra block size" |
There was a problem hiding this comment.
| int "Extra block size" | |
| int "Number of extra blocks" |
- Configure BLOCK count by using kconfig option 'EXTRA_BLOCKS' instad of fixed value. This allows user to customize the RAM usage based on their platform. - Keep the PCM data in flash to avoid large RAM usage. - Refine the sample to get the SAMPLE_BIT_WIDTH and BYTES_PER_SAMPLE from Kconfig options. Signed-off-by: Zhaoxiang Jin <Zhaoxiang.Jin_1@nxp.com>
ZhaoxiangJin
force-pushed
the
refine-i2s-codec-example
branch
from
dd50879 to
b50fec9
Compare
|
