AM3352 cpu usage for spidev_test

Hi,

Sorry for the late response. I had to test this on my side.

Biser, is correct in his statement, that the while loop is loading the cpu.  In my understanding it is fine to see this behavior, when you test spi like this (with a  loop in spidev_test.c)

I'll try to explain. I work with a 500KHz speed of the spi transfer. 

1. Here are the tests (I wasn't optimizing my programs, so cpu loads are far worst than what you report, but behaviour should be similar on your side) I ran on my BBB, running kernel 4.1:

1.1) loop the spi transfer directly in spidev_test.c, as bellow:

if (input_tx) {

             size = strlen(input_tx+1);

             tx = malloc(size);

             rx = malloc(size);

             size = unescape((char *)tx, input_tx, size);

             while (1) {

                    transfer(fd, tx, rx, size);

                    free(rx);

                    free(tx);

              }

             } else {

                    while (1)

                          transfer(fd, default_tx, default_rx, sizeof(default_tx));

}

1.2) execute  a shell script that calls the spidev_test.c, compiled as spi:

#!/bin/sh                                                                                                                            

while true                                                                                                                          

do                                                                                                                                  

./spi                                                                                                                              

done

As a result CPU load is ~70%, because of the while (1) loop. Result was similar (even worse) when I comment the transfer() call in 1.1).

On the other hand: 

2.1) I use the spidev_test.c, without a while loop in it: 

if (input_tx) {

             size = strlen(input_tx+1);

             tx = malloc(size);

             rx = malloc(size);

             size = unescape((char *)tx, input_tx, size);

                    transfer(fd, tx, rx, size);

                    free(rx);

                    free(tx);

             } else {

                          transfer(fd, default_tx, default_rx, sizeof(default_tx));

}

2.2). And simulate the spi transfer with a loop in the shell script, like: 

#!/bin/sh
while true
do
./spi
done 

CPU load is 14%. Here is the explanation: 

In user space a C program is executed at several hundred KHz speed. So, when you call the spidev_test.c WITH a while loop in it the loop increments with a speed approximately equal to the spi tranfer executed by the transfer() call, so CPU waits for the transfer (since they are equally fast) which loads the processor. 

On the other hand a shell script is executed far more slower (several KHz speed), so when the shell script steps in the transfer() it takes longer time to go to the next call in the spidev_test.c & the spi transfer is executed by then => cpu does not wait. 

The behavior is similar in your case: 
When working with 400KHz spi transfer speed, the spidev_test.c is executed equally fast as the spi transfer => cpu waits & is loaded. 

When working with 24MHz spi transfer speed, the spidev_test.c takes longer to step through the calls, allowing the transfer() to execute while the pc steps over the program. 

Hope this helps. 

Best Regards, 
Yordan