stm32mp1 spi transaction slow down main application

Ara.1 · ‎2021-09-29

Hi All

I have disabled the second core to validate one of old product, which runs on single core,

and same application cpu usage i have seen is 90-95 %

and same application runs and does the same spi transaction, i am seeing the cpu usage is 57%

all constraint are same except the core

following is the top of stm32mp1

top - 04:18:53 up 37 min, 1 user, load average: 1.70, 1.65, 1.51

Tasks: 107 total, 4 running, 103 sleeping, 0 stopped, 0 zombie

%Cpu(s): 4.5 us, 93.4 sy, 0.0 ni, 1.5 id, 0.3 wa, 0.0 hi, 0.3 si, 0.0 st

9 root 20 0 0 0 0 S 2.3 0.0 0:05.56 ksoftirqd/0

78 root 20 0 0 0 0 I 1.6 0.0 0:03.88 kworker/0:2-pm

768 root 20 0 4556 1544 1204 R 1.0 0.2 0:01.27 top

612 root 20 0 113608 6668 4068 R 57.6 0.7 21:52.52 pxl

78 root 20 0 0 0 0 I 16.8 0.0 1:44.42 kworker/0:2-pm

63 root 20 0 0 0 0 S 8.2 0.0 2:43.77 spi1

62 root -51 0 0 0 0 S 7.4 0.0 2:30.84 irq/30-44005000

9137 root 20 0 0 0 0 I 2.4 0.0 0:01.96 kworker/0:0-pm

10 root 20 0 0 0 0 R 0.6 0.0 0:10.91 rcu_preempt

768 root 20 0 4556 1544 1204 R 0.6 0.2 0:18.68 top

9 root 20 0 0 0 0 R 0.3 0.0 0:10.92 ksoftirqd/0

what am observing , is that the SPI irq is running with very high priority -51, which makes our APP(pxl) deviates from cpu usage from 90 to 57,

is there any solution, to get pxl use cpu 90-95%

Olivier GALLIEN · ‎2021-09-29

Hi @Ara.1 ,

Sorry it's not clear.

Do you mean that you application pxl is not working properly ?

CPU usage is not meaningful as such.

What was the old core performance ? ( mips, frequency )

Olivier

Olivier GALLIEN
In order to give better visibility on the answered topics, please click on 'Accept as Solution' on the reply which solved your issue or answered your question.

Ara.1 · ‎2021-09-29

application $APP is working for the number of while (1){} ,

number of execution rate per sec are reduced when spi in action,

Ara.1 · ‎2021-09-29

I have checked the drivers/spi/spi-stm32.c driver, while transfer spi

[ 78.789827] transfer of 4 bytes (4 data frames)

[ 78.794371] dma disabled

[ 78.796905] timeout set to 10 msecs

how to enable dma for SPI ?

Ara.1 · ‎2021-09-29

if we increase the nice the normal thread

63 root 25 5 0 0 0 S 8.2 0.0 1:19.19 spi1

spi normal thread...

this brings back $APP normal.

root@mp1som:~# top

top - 02:54:47 up 17 min, 1 user, load average: 2.01, 1.99, 1.44

Tasks: 110 total, 3 running, 107 sleeping, 0 stopped, 0 zombie

%Cpu(s): 4.6 us, 93.8 sy, 0.0 ni, 1.0 id, 0.3 wa, 0.0 hi, 0.3 si, 0.0 st

MiB Mem : 935.4 total, 810.0 free, 47.7 used, 77.6 buff/cache

MiB Swap: 0.0 total, 0.0 free, 0.0 used. 852.2 avail Mem

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND

540 root 20 0 113636 6744 4176 R 64.7 0.7 10:56.47 pxl

62 root -51 0 0 0 0 S 12.1 0.0 1:57.49 irq/30-44005000

63 root 25 5 0 0 0 S 8.2 0.0 1:19.19 spi1

23 root 20 0 0 0 0 I 6.9 0.0 0:29.57 kworker/0:1-pm

3289 root 20 0 4556 1604 1264 R 1.0 0.2 0:03.61 top

Olivier GALLIEN · ‎2021-09-30

Hi @Ara.1 ,

Ok you application doing while (1) it take all remaining CPU and is a good indicator of CPU taken by other threads.

Anyway I don't get you last status :

"if we increase the nice the normal thread"

What do you mean ? Can you rephrase ?

Does this mean point is close ?

Thx

Olivier

Olivier GALLIEN
In order to give better visibility on the answered topics, please click on 'Accept as Solution' on the reply which solved your issue or answered your question.