about performance

2020-02-27T14:20:30.801000Z

Hi Andreaz

do you mean it will be faster after inlining

The test code is for a external library, so it is compiled by gnu of vxworks.

about performance

2020-02-27T14:20:29.465000Z

Watch out for this code, a decent c compiler might inline that to the following, however the CodeSys compiler will not!

for (index = 0;index < cycles;index++)
i = 200 * 100;

After inlining

i = 200 * 100 * cycles

about performance

2020-02-27T14:20:28.937000Z

I did a test which show the performance of my cpu board by 1 million times multiplication.
First I did it directly in vxworks by c language. It spent 2500us on finishing the multiplication.
Then I make a external library by c language. It spent about 11666us on the test.
I don't know why the performance is so poor when I do it by codesys.
How can I improve the performance?

the code just like below:

void CDECL CDECL_EXT func1__main(func1_main_struct *p)
{
UINT32 startTick, endTick, multipIntTick;
INT32 index, cycles;
INT32 i;

cycles = p->pInstance->varIn;
CAL_LogAdd(STD_LOGGER, COMPONENT_ID, LOG_INFO, ERR_OK, 0, "---cycles=%d---\n", cycles);

startTick = sysTimeBaseLGet();         //get the ticks
for (index = 0;index < cycles;index++)
    i = 200 * 100;
endTick = sysTimeBaseLGet();     
if (endTick < startTick)
    multipIntTick = (endTick + 0xFFFFFFFF - startTick) / 50;    //get us
else 
    multipIntTick = (endTick - startTick) / 50;

startTick = sysTimeBaseLGet();
CAL_LogAdd(STD_LOGGER, COMPONENT_ID, LOG_INFO, ERR_OK, 0, "---Int=%d---\n", multipIntTick);
p->pInstance->varOut = multipIntTick;

}

Recent posts to about performance

about performance

about performance

about performance