Hello all,
I am trying to run an Android naughat 7.1 in imx7 sabresd board. I used the pre-built image given by the nxp and tested the image with the following device.
The performance of Android is very poor with the pre-built image
I also tried to add support for the display TFT3P5581-E with the resolution 480x854, I find the same performance.User 57%, System 41%, IOW 0%, IRQ 0%
User 340 + Nice 15 + Sys 253 + Idle 5 + IOW 0 + IRQ 0 + SIRQ 0 = 613
PID USER PR NI CPU% S #THR VSS RSS PCY Name
User 46%, System 43%, IOW 0%, IRQ 0%
User 298 + Nice 0 + Sys 275 + Idle 62 + IOW 0 + IRQ 0 + SIRQ 0 = 635
PID USER PR NI CPU% S #THR VSS RSS PCY Name
User 41%, System 44%, IOW 0%, IRQ 0%
User 262 + Nice 0 + Sys 285 + Idle 88 + IOW 0 + IRQ 0 + SIRQ 0 = 635
PID USER PR NI CPU% S #THR VSS RSS PCY Name
User 54%, System 42%, IOW 0%, IRQ 0%
User 358 + Nice 0 + Sys 280 + Idle 19 + IOW 0 + IRQ 0 + SIRQ 2 = 659
PID USER PR NI CPU% S #THR VSS RSS PCY Name
But one thing i observed is that, when i tried to run the image in a lower resolution screen (i.e)LCD8000-3T, i could see some better animation when compared to higher resolution displays. But the performance is no way similar to iMX6.
I understand the iMX7 does not have GPU, but is this the expected performance of the device ?
Hi,
Unfortunately, it is the expected performance specifically by the fact that you mentioned. The absent of GPU is a big constrain for Android; the i.MX7 support is limited, intended to be used on specific applications.
Hope this will be useful for you.
Best regards!
/Carlos
-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------
Hi carlos, thanks for the reply. I went through the code and found that the OS uses pixelflinger to render the Android graphics using the OpenGLes1.0 . I also see that there are some neon code to do some blend works.
Is the pixelflinger is fully optimised ?. Does it utilise the neon core efficiently ?
I am trying to offload the rendering load to the neon core from the CPU core.
Have anyone tried to optimise the pixelflinger like this ?