AM62P-Q1: encode error

z lin

Part Number: AM62P-Q1

hi experts ,

SDK version: 11.01

we use V4l2 interface to encode, sometimes, when call

xioctl(handle->fd, VIDIOC_QBUF, &buf)) function, report "contiguous chunk is too small 212992/2119680" error. when this error occours, need to restart system, then the encode function may work normally.

Meanwhile, we found that the remaining CMA is sufficient.

16 days ago

0 z lin 16 days ago

Prodigy 80 points

our encode image size is 1920*736

0 Suren Porwar 15 days ago in reply to z lin

TI__Mastermind 31680 points

Hi,

Can you help us reproduce this issue on our EVM?

Best Regards,

Suren

0 z lin 14 days ago in reply to Suren Porwar

Prodigy 80 points

Hi,

Is the memory allocated by the

gbm_bo_create() function physically contiguous?

Best Regards,

linz

0 Suren Porwar 5 days ago in reply to z lin

TI__Mastermind 31680 points

Please share your pipeline or parameters you are setting for the encoder?

How can we reproduce the issue?

Best Regards,

Suren

0 z lin 2 days ago in reply to Suren Porwar

Prodigy 80 points

Hi,

our pipeline is camera->GPU->encode(H.264).

Previous we use

V4L2_MEMORY_DMABUF memory to encode(the buffer use gbm_bo_create() functiion allocated), power on, power off system sometime will reproduce this issue,

The probability is 1%. you can check why use V4L2_MEMORY_DMABUF memory cause this issue

now we use

V4L2_MEMORY_MMAP memory, don't reproduce this issue.

Processors

Processors forum

AM62P-Q1: encode error