[Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[eric]

hello:
        I use xenomai-2.6.4 with kernel ipipe-gch-ipipe-3.0-imx6q.tar.bz2   with sample test  , build  the flow module with  timer 250us cycle,if I make a user space application with memory copy ,the kernel latency will very big ,why??? thanks
latency :
pmax: 0.508624 ms
pmax: 0.533823 ms
pmax: 0.553197 ms
pmax: 0.603770 ms
pmax: 0.613982 ms
pmax: 0.638025 ms
pmax: 0.662129 ms
pmax: 0.699245 ms
pmax: 0.780533 ms
pmax: 0.800770 ms
pmax: 0.802066 ms
pmax: 0.813921 ms
pmax: 0.816330 ms
pmax: 0.820429 ms
pmax: 0.820483 ms
pmax: 0.821783 ms
pmax: 0.827649 ms
pmax: 0.833197 ms
pmax: 0.837606 ms
pmax: 0.843921 ms
pmax: 0.874033 ms
pmax: 0.892414 ms
pmax: 0.892788 ms




usr application:
#include <stdio.h>
int main()
{
char a[1024*1024]={9};
char b[1024*1024];
    while(1){
    memcpy(b,a,1024*800);
    }
return 0;
}


kernel module   code :


#define TASK_PRIO 99 // 99 is Highest RT priority
#define TASK_MODE 0 // No flags
#define TASK_STKSZ 0 // default Stack size


#define TASK_PERIOD 250000ll // in nano seconds = 500 ms
long long pmax,start,stop;
static rtdm_task_t td;    //RTDM Real time task pointer
char logo[40*1024]={0};
char mem_spvm[40*1024];
long long runcount=0;
void periodic (void *arg) {
long max,pmax;


for (;;) {
rtdm_task_wait_period(); //deschedule until next period


stop =rtdm_clock_read();
memset(mem_spvm,0,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));
memset(mem_spvm,127,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));


//stop = rtdm_clock_read();
if(runcount<1000)
{
pmax=0;
}
if(pmax<(stop-start)){
  pmax=stop-start;
  printk("pmax: %ld.%06ld ms\n",(long)pmax/1000000,(long)pmax%1000000);
}
start=stop;
runcount++;
}
}




int init_module(void) {
   return rtdm_task_init (&td, "periodic", &periodic, NULL, TASK_PRIO,TASK_PERIOD);
}


void cleanup_module(void) {
rtdm_task_destroy(&td);
}


MODULE_LICENSE("GPL");   

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[Gilles Chanteperdrix]

On Tue, Dec 08, 2015 at 06:43:02PM +0800, eric wrote:
> hello:
> I use xenomai-2.6.4 with kernel
>         ipipe-gch-ipipe-3.0-imx6q.tar.bz2 with sample test , build
>         the flow module with timer 250us cycle,if I make a user
>         space application with memory copy ,the kernel latency
>         will very big ,why??? thanks

Do you have the same problem if you use a recent mainline kernel?
Such as 3.14 or 3.18 ?

-- 
					    Gilles.
https://click-hack.org

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[eric]

It is much better with 3.18.20 ,it is  the test result with xenomai-3.0.1   500us cycle ,still a little big ,thanks!
-sh-2.05b# [  154.812032] pmax: 0.597303 ms
[  245.216863] pmax: 0.600940 ms
[  249.992395] pmax: 0.603288 ms
[  277.280006] pmax: 0.610470 ms










At 2015-12-08 19:23:20, "Gilles Chanteperdrix" <gilles.chanteperdrix@xenomai.org> wrote:
>On Tue, Dec 08, 2015 at 06:43:02PM +0800, eric wrote:
>> hello:
>> I use xenomai-2.6.4 with kernel
>>         ipipe-gch-ipipe-3.0-imx6q.tar.bz2 with sample test , build
>>         the flow module with timer 250us cycle,if I make a user
>>         space application with memory copy ,the kernel latency
>>         will very big ,why??? thanks
>
>Do you have the same problem if you use a recent mainline kernel?
>Such as 3.14 or 3.18 ?
>
>-- 
>					    Gilles.
>https://click-hack.org

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[Gilles Chanteperdrix]

On Mon, Dec 14, 2015 at 10:13:57AM +0800, eric wrote:
> It is much better with 3.18.20 ,it is  the test result with xenomai-3.0.1   500us cycle ,still a little big ,thanks!
> -sh-2.05b# [  154.812032] pmax: 0.597303 ms
> [  245.216863] pmax: 0.600940 ms
> [  249.992395] pmax: 0.603288 ms
> [  277.280006] pmax: 0.610470 ms

I am not sure about what you measure, but if you observe a 500us
latency, you have a problem.

What latency does Xenomai latency test give?

-- 
					    Gilles.
https://click-hack.org

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[eric]

kernel space:




#define TASK_PRIO 99 // 99 is Highest RT priority
#define TASK_MODE 0 // No flags
#define TASK_STKSZ 0 // default Stack size
#define TASK_PERIOD 500000ll // in nano seconds = 500 ms
long long pmax,start,stop;
static rtdm_task_t td;    //RTDM Real time task pointer
char logo[40*1024];
char mem_spvm[40*1024];
long long runcount=0;




void periodic (void *arg) {
long max,pmax;
for (;;) {
rtdm_task_wait_period(NULL); //deschedule until next period
stop =rtdm_clock_read();
memset(mem_spvm,0,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));
memset(mem_spvm,127,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));
memset(mem_spvm,0,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));
memset(mem_spvm,127,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));
memset(mem_spvm,0,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));
memset(mem_spvm,127,sizeof(logo));
if(runcount<1000)
{
pmax=0;
}
if(pmax<(stop-start)){
  pmax=stop-start;
  printk("pmax: %ld.%06ld ms\n",(long)pmax/1000000,(long)pmax%1000000);
}
start=stop;
runcount++;
}
}
int init_module(void) {
   return rtdm_task_init (&td, "periodic", &periodic, NULL, TASK_PRIO,TASK_PERIOD);
}


void cleanup_module(void) {
rtdm_task_destroy(&td);
}


user space:


#include <stdio.h>
int main()
{
char a[1024*1024]={0xff};
char b[1024*1024];
int i=0;
while(1){


for(i=1000000;i>0;i--)
system("sync");
memcpy(b,a,1024*800);
}
return 0;
}







At 2015-12-14 14:42:53, "Gilles Chanteperdrix" <gilles.chanteperdrix@xenomai.org> wrote:
>On Mon, Dec 14, 2015 at 10:13:57AM +0800, eric wrote:
>> It is much better with 3.18.20 ,it is  the test result with xenomai-3.0.1   500us cycle ,still a little big ,thanks!
>> -sh-2.05b# [  154.812032] pmax: 0.597303 ms
>> [  245.216863] pmax: 0.600940 ms
>> [  249.992395] pmax: 0.603288 ms
>> [  277.280006] pmax: 0.610470 ms
>
>I am not sure about what you measure, but if you observe a 500us
>latency, you have a problem.
>
>What latency does Xenomai latency test give?
>
>-- 
>					    Gilles.
>https://click-hack.org

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[Gilles Chanteperdrix]

On Mon, Dec 14, 2015 at 03:57:01PM +0800, eric wrote:
> kernel space:

This does not answer my question. here is my question again, so that
you can read it again:

> >What latency does Xenomai latency test give?

-- 
					    Gilles.
https://click-hack.org

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[eric]

I use latency test ,it show this




-sh-2.05b# ./latency 
   0"000.000| WARNING: [main] cannot open RTDM device /dev/rtdm/memdev-private: No such file or directory
   0"000.000| WARNING: [main] cannot map private umm area: No such file or directory
   0"000.000| BUG in init_bind(): [main] (CONFIG_DEVTMPFS_MOUNT not enabled?)
-sh-2.05b# ./latency  -t 1
   0"000.000| WARNING: [main] cannot open RTDM device /dev/rtdm/memdev-private: No such file or directory
   0"000.000| WARNING: [main] cannot map private umm area: No such file or directory
   0"000.000| BUG in init_bind(): [main] (CONFIG_DEVTMPFS_MOUNT not enabled?)








At 2015-12-14 16:03:17, "Gilles Chanteperdrix" <gilles.chanteperdrix@xenomai.org> wrote:
>On Mon, Dec 14, 2015 at 03:57:01PM +0800, eric wrote:
>> kernel space:
>
>This does not answer my question. here is my question again, so that
>you can read it again:
>
>> >What latency does Xenomai latency test give?
>
>-- 
>					    Gilles.
>https://click-hack.org

[Xenomai] Compileerror rt_imx_uart.c on kernel 3.10.53 ARM

[Wolfgang Netbal]

Dear All,

when I activated the serial realtime driver 
(CONFIG_XENO_DRIVERS_IMX_UART) in my kernel 3.10.53
I get the following compile errors.

drivers/xenomai/serial/rt_imx_uart.c: In function 'rt_imx_uart_probe':

drivers/xenomai/serial/rt_imx_uart.c:2032:20: error: 'struct imxuart_platform_data' has no member named 'init'

   if (pdata && pdata->init) {

                     ^

drivers/xenomai/serial/rt_imx_uart.c:2033:14: error: 'struct imxuart_platform_data' has no member named 'init'

    err = pdata->init(pdev);

               ^

drivers/xenomai/serial/rt_imx_uart.c:2095:20: error: 'struct imxuart_platform_data' has no member named 'exit'

   if (pdata && pdata->exit)

                     ^

drivers/xenomai/serial/rt_imx_uart.c:2096:8: error: 'struct imxuart_platform_data' has no member named 'exit'

    pdata->exit(pdev);

         ^

drivers/xenomai/serial/rt_imx_uart.c: In function 'rt_imx_uart_remove':

drivers/xenomai/serial/rt_imx_uart.c:2126:20: error: 'struct imxuart_platform_data' has no member named 'exit'

   if (pdata && pdata->exit)

                     ^

drivers/xenomai/serial/rt_imx_uart.c:2127:8: error: 'struct imxuart_platform_data' has no member named 'exit'

    pdata->exit(pdev);



I found out that in kernel 3.10.53 the function pointer to exit and init 
function in struct imxuart_platform_data is missing.

I tried to find a newer version of rt_imx_uart.c in xenomai 3.0 but the 
file is the same.

Do you have any idea how to fix this ?

Kind regards
Wolfgang

Re: [Xenomai] Compileerror rt_imx_uart.c on kernel 3.10.53 ARM

[Gilles Chanteperdrix]

On Mon, Dec 14, 2015 at 10:04:17AM +0100, Wolfgang Netbal wrote:
> Dear All,
> 
> when I activated the serial realtime driver 
> (CONFIG_XENO_DRIVERS_IMX_UART) in my kernel 3.10.53
> I get the following compile errors.
> 
> drivers/xenomai/serial/rt_imx_uart.c: In function 'rt_imx_uart_probe':
> 
> drivers/xenomai/serial/rt_imx_uart.c:2032:20: error: 'struct imxuart_platform_data' has no member named 'init'
> 
>    if (pdata && pdata->init) {
> 
>                      ^
> 
> drivers/xenomai/serial/rt_imx_uart.c:2033:14: error: 'struct imxuart_platform_data' has no member named 'init'
> 
>     err = pdata->init(pdev);
> 
>                ^
> 
> drivers/xenomai/serial/rt_imx_uart.c:2095:20: error: 'struct imxuart_platform_data' has no member named 'exit'
> 
>    if (pdata && pdata->exit)
> 
>                      ^
> 
> drivers/xenomai/serial/rt_imx_uart.c:2096:8: error: 'struct imxuart_platform_data' has no member named 'exit'
> 
>     pdata->exit(pdev);
> 
>          ^
> 
> drivers/xenomai/serial/rt_imx_uart.c: In function 'rt_imx_uart_remove':
> 
> drivers/xenomai/serial/rt_imx_uart.c:2126:20: error: 'struct imxuart_platform_data' has no member named 'exit'
> 
>    if (pdata && pdata->exit)
> 
>                      ^
> 
> drivers/xenomai/serial/rt_imx_uart.c:2127:8: error: 'struct imxuart_platform_data' has no member named 'exit'
> 
>     pdata->exit(pdev);
> 
> 
> 
> I found out that in kernel 3.10.53 the function pointer to exit and init 
> function in struct imxuart_platform_data is missing.
> 
> I tried to find a newer version of rt_imx_uart.c in xenomai 3.0 but the 
> file is the same.
> 
> Do you have any idea how to fix this ?

Maybe the file is adapted to newer kernel versions? Such as Linux
3.14 or Linux 3.18 ?

Please try, if you want to answer a random mail, to remove any
reference-to or in-reply-to header, so that your mail does not
appear in a middle of an unrelated thread. Look at the mess you have
done here:
https://xenomai.org/pipermail/xenomai/2015-December/thread.html

-- 
					    Gilles.
https://click-hack.org

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[eric]

   Above i use the wrong version about latency test , now it works and the latency test 
 
RTD|     22.513|     39.832|     60.649|       0|     0|     11.269|     63.939
RTD|     27.225|     40.068|     61.134|       0|     0|     11.269|     63.939
RTD|     27.270|     40.531|     60.103|       0|     0|     11.269|     63.939
RTD|     14.876|     39.540|     61.451|       0|     0|     11.269|     63.939
RTD|     15.890|     39.760|     68.512|       0|     0|     11.269|     68.512
RTD|     10.951|     29.915|     61.269|       0|     0|     10.951|     68.512
RTD|     10.981|     12.596|     15.118|       0|     0|     10.951|     68.512
RTD|     10.814|     12.648|     16.284|       0|     0|     10.814|     68.512
RTD|     10.920|     12.636|     15.057|       0|     0|     10.814|     68.512
RTT|  00:09:38  (periodic user-mode task, 1000 us period, priority 99)
RTH|----lat min|----lat avg|----lat max|-overrun|---msw|---lat best|--lat worst
RTD|     10.784|     12.672|     15.465|       0|     0|     10.784|     68.512
RTD|     11.071|     12.602|     15.374|       0|     0|     10.784|     68.512
RTD|     10.904|     12.632|     14.859|       0|     0|     10.784|     68.512
RTD|     10.934|     12.639|     15.056|       0|     0|     10.784|     68.512
RTD|     11.010|     12.575|     15.146|       0|     0|     10.784|     68.512







At 2015-12-14 16:30:40, "eric" <ericvic@163.com> wrote:

I use latency test ,it show this




-sh-2.05b# ./latency 
   0"000.000| WARNING: [main] cannot open RTDM device /dev/rtdm/memdev-private: No such file or directory
   0"000.000| WARNING: [main] cannot map private umm area: No such file or directory
   0"000.000| BUG in init_bind(): [main] (CONFIG_DEVTMPFS_MOUNT not enabled?)
-sh-2.05b# ./latency  -t 1
   0"000.000| WARNING: [main] cannot open RTDM device /dev/rtdm/memdev-private: No such file or directory
   0"000.000| WARNING: [main] cannot map private umm area: No such file or directory
   0"000.000| BUG in init_bind(): [main] (CONFIG_DEVTMPFS_MOUNT not enabled?)








At 2015-12-14 16:03:17, "Gilles Chanteperdrix" <gilles.chanteperdrix@xenomai.org> wrote:
>On Mon, Dec 14, 2015 at 03:57:01PM +0800, eric wrote:
>> kernel space:
>
>This does not answer my question. here is my question again, so that
>you can read it again:
>
>> >What latency does Xenomai latency test give?
>
>-- 
>					    Gilles.
>https://click-hack.org





 

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[Gilles Chanteperdrix]

On Mon, Dec 14, 2015 at 05:05:48PM +0800, eric wrote:
>    Above i use the wrong version about latency test , now it works and the latency test 
>  
> RTD|     22.513|     39.832|     60.649|       0|     0|     11.269|     63.939
> RTD|     27.225|     40.068|     61.134|       0|     0|     11.269|     63.939
> RTD|     27.270|     40.531|     60.103|       0|     0|     11.269|     63.939
> RTD|     14.876|     39.540|     61.451|       0|     0|     11.269|     63.939
> RTD|     15.890|     39.760|     68.512|       0|     0|     11.269|     68.512
> RTD|     10.951|     29.915|     61.269|       0|     0|     10.951|     68.512
> RTD|     10.981|     12.596|     15.118|       0|     0|     10.951|     68.512
> RTD|     10.814|     12.648|     16.284|       0|     0|     10.814|     68.512
> RTD|     10.920|     12.636|     15.057|       0|     0|     10.814|     68.512
> RTT|  00:09:38  (periodic user-mode task, 1000 us period, priority 99)
> RTH|----lat min|----lat avg|----lat max|-overrun|---msw|---lat best|--lat worst
> RTD|     10.784|     12.672|     15.465|       0|     0|     10.784|     68.512
> RTD|     11.071|     12.602|     15.374|       0|     0|     10.784|     68.512
> RTD|     10.904|     12.632|     14.859|       0|     0|     10.784|     68.512
> RTD|     10.934|     12.639|     15.056|       0|     0|     10.784|     68.512
> RTD|     11.010|     12.575|     15.146|       0|     0|     10.784|     68.512
> 

So the latency is below 100us, much less than 500us. It seems a bit
high though, do you have any debug option enabled? Do you run this
with a real-time load?

-- 
					    Gilles.
https://click-hack.org

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[eric]

>
>So the latency is below 100us, much less than 500us. It seems a bit
>high though, do you have any debug option enabled? Do you run this
>with a real-time load?
>
>-- 
>					    Gilles.
>https://click-hack.org

        I can not run real-time load ,it always stop  the flowing step.  So  i use some other load program .And I do not  enabled  debug option.
root@imx6qdlsolo:/opt/x3/bin# ./xeno-test
Started child 29228: /bin/sh /usr/xenomai/bin/xeno-test-run-wrapper ./xeno-test
++ echo 0
++ testdir=/opt/imx6/smp3/xenomai-3.0.1/x3/bin
++ /opt/imx6/smp3/xenomai-3.0.1/x3/bin/smokey --run
arith OK
bufp OK
iddp OK
leaks OK

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[Gilles Chanteperdrix]

On Tue, Dec 15, 2015 at 08:54:56AM +0800, eric wrote:
> >
> >So the latency is below 100us, much less than 500us. It seems a bit
> >high though, do you have any debug option enabled? Do you run this
> >with a real-time load?
> >
> >-- 
> >					    Gilles.
> >https://click-hack.org
> 
>         I can not run real-time load ,it always stop  the flowing step.  So  i use some other load program .And I do not  enabled  debug option.
> root@imx6qdlsolo:/opt/x3/bin# ./xeno-test
> Started child 29228: /bin/sh /usr/xenomai/bin/xeno-test-run-wrapper ./xeno-test
> ++ echo 0
> ++ testdir=/opt/imx6/smp3/xenomai-3.0.1/x3/bin
> ++ /opt/imx6/smp3/xenomai-3.0.1/x3/bin/smokey --run
> arith OK
> bufp OK
> iddp OK
> leaks OK

I was rather thinking about running the user-space load which
creates high latencies with your own latency test.

But if you are interested in running xeno-test, you should check its
syntax:
- either by running it with the -h or --help argument
- or by checking its manual page, either on your disk or on xenomai
website:
https://xenomai.org/documentation/xenomai-3/html/man1/xeno-test/
- or by reading the page where we explain how we do the benchmarks
we publish:
https://xenomai.org/2014/07/benchmarking-with-xeno-test/

Regards.

-- 
					    Gilles.
https://click-hack.org

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[eric]

ok，thanks  very much！
At 2015-12-15 14:17:46, "Gilles Chanteperdrix" <gilles.chanteperdrix@xenomai.org> wrote:
>On Tue, Dec 15, 2015 at 08:54:56AM +0800, eric wrote:
>> >
>> >So the latency is below 100us, much less than 500us. It seems a bit
>> >high though, do you have any debug option enabled? Do you run this
>> >with a real-time load?
>> >
>> >-- 
>> >					    Gilles.
>> >https://click-hack.org
>> 
>>         I can not run real-time load ,it always stop  the flowing step.  So  i use some other load program .And I do not  enabled  debug option.
>> root@imx6qdlsolo:/opt/x3/bin# ./xeno-test
>> Started child 29228: /bin/sh /usr/xenomai/bin/xeno-test-run-wrapper ./xeno-test
>> ++ echo 0
>> ++ testdir=/opt/imx6/smp3/xenomai-3.0.1/x3/bin
>> ++ /opt/imx6/smp3/xenomai-3.0.1/x3/bin/smokey --run
>> arith OK
>> bufp OK
>> iddp OK
>> leaks OK
>
>I was rather thinking about running the user-space load which
>creates high latencies with your own latency test.
>
>But if you are interested in running xeno-test, you should check its
>syntax:
>- either by running it with the -h or --help argument
>- or by checking its manual page, either on your disk or on xenomai
>website:
>https://xenomai.org/documentation/xenomai-3/html/man1/xeno-test/
>- or by reading the page where we explain how we do the benchmarks
>we publish:
>https://xenomai.org/2014/07/benchmarking-with-xeno-test/
>
>Regards.
>
>-- 
>					    Gilles.
>https://click-hack.org

Re: [Xenomai] ipipe-gch-ipipe-3.0-imx6q latency

[eric]

            And with x86 dual core the same test the result quite good just 13us latency , but just imx6q for a long time the result not very well.






At 2015-12-14 15:57:01, "eric" <ericvic@163.com> wrote:

kernel space:




#define TASK_PRIO 99 // 99 is Highest RT priority
#define TASK_MODE 0 // No flags
#define TASK_STKSZ 0 // default Stack size
#define TASK_PERIOD 500000ll // in nano seconds = 500 ms
long long pmax,start,stop;
static rtdm_task_t td;    //RTDM Real time task pointer
char logo[40*1024];
char mem_spvm[40*1024];
long long runcount=0;




void periodic (void *arg) {
long max,pmax;
for (;;) {
rtdm_task_wait_period(NULL); //deschedule until next period
stop =rtdm_clock_read();
memset(mem_spvm,0,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));
memset(mem_spvm,127,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));
memset(mem_spvm,0,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));
memset(mem_spvm,127,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));
memset(mem_spvm,0,sizeof(logo));
memcpy(mem_spvm,logo,sizeof(logo));
memset(mem_spvm,127,sizeof(logo));
if(runcount<1000)
{
pmax=0;
}
if(pmax<(stop-start)){
  pmax=stop-start;
  printk("pmax: %ld.%06ld ms\n",(long)pmax/1000000,(long)pmax%1000000);
}
start=stop;
runcount++;
}
}
int init_module(void) {
   return rtdm_task_init (&td, "periodic", &periodic, NULL, TASK_PRIO,TASK_PERIOD);
}


void cleanup_module(void) {
rtdm_task_destroy(&td);
}


user space:


#include <stdio.h>
int main()
{
char a[1024*1024]={0xff};
char b[1024*1024];
int i=0;
while(1){


for(i=1000000;i>0;i--)
system("sync");
memcpy(b,a,1024*800);
}
return 0;
}







At 2015-12-14 14:42:53, "Gilles Chanteperdrix" <gilles.chanteperdrix@xenomai.org> wrote:
>On Mon, Dec 14, 2015 at 10:13:57AM +0800, eric wrote:
>> It is much better with 3.18.20 ,it is  the test result with xenomai-3.0.1   500us cycle ,still a little big ,thanks!
>> -sh-2.05b# [  154.812032] pmax: 0.597303 ms
>> [  245.216863] pmax: 0.600940 ms
>> [  249.992395] pmax: 0.603288 ms
>> [  277.280006] pmax: 0.610470 ms
>
>I am not sure about what you measure, but if you observe a 500us
>latency, you have a problem.
>
>What latency does Xenomai latency test give?
>
>-- 
>					    Gilles.
>https://click-hack.org