xilinx_zynq_zedboard cpu_idle_report implementation

[Gedare Bloom]

Hi,
_Thread_Idle has been removed.

I don't think there is a direct handle available for the idle threads
anymore. You should see the IDLE thread in the cpuuse report. If you
need to find the idle threads specifically you might be able to loop
through the Object tables (like the cpuuse report does) and compare
the name with IDLE. Directly accessing _Thread_Idle (and the Object
tables) like this is a violation of the RTEMS userspace API.

If a compelling reason for a direct handle to the idle thread is
given, it may be possible for RTEMS to store such a handle in the
per-cpu tables, which was done previously but was removed.

-Gedare

On Wed, Jan 8, 2014 at 9:53 AM, Giovanni Macciocu <G.Macciocu at sron.nl> wrote:
> Hi All,
>
> details:
>
> Rtems head: 7 Jan. 2014
> BSP: xilinx_zynq_zedboard
> Goal: cpu_idle_report_implementation
>
> I'm adjusting a piece of code we've used succesfully with rtems 4.9 and rtems 4.10 in combination with the Sparc BSP (Leon processor).
> This code is used to measure the idle time of the processor.
> I'm adjusting this code in order to be used with the recent rtems head of 7 Jan. 2014 and the BSP of the xilinx_zynq_zedboard
>
> I've added the following include
> #include <rtems/score/objectimpl.h>
> in order to access the 'Object_Information' object
>
> The only problem I'm still having is that I cannot find an equal implementation of '_Thread_Idle'
> in rtems 4.9 and 4.10 (sparc bsp) this is implemented at line 435 of sparc-rtems4.10/include/rtems/score/thread.h (SCORE_EXTERN Thread_Control *Thread_Idle;)
> however in 4.11 with the xilinx_zynq_zedboard bsp I cannot find a similar implementation.
>
> file: rtems/cpukit/libmisc/cpuuse/cpuidlereport.c
>
> // CPU Usage Reporter
>
> #ifdef HAVE_CONFIG_H
> #include "config.h"
> #endif
>
> #include <rtems.h>
>
> #include <assert.h>
> #include <string.h>
> #include <stdlib.h>
> #include <stdio.h>
> #include <ctype.h>
> #include <inttypes.h>
>
> #include <rtems/cpuuse.h>
> #include <rtems/bspIo.h>
>
> // added this in v4.11 in order to acces 'Objects_Information'
> #include <rtems/score/objectimpl.h>
>
> #if defined(RTEMS_ENABLE_NANOSECOND_CPU_USAGE_STATISTICS)
> #include <rtems/score/timespec.h>
> #endif
>
> #ifdef RTEMS_ENABLE_NANOSECOND_CPU_USAGE_STATISTICS
> extern struct timespec    CPU_usage_Uptime_at_last_reset;
> #endif
>
> uint32_t rtems_cpu_idle_report( void )
> {
>   uint32_t ival, fval, promille = 1000;
>
>   /*
>    *  When not using nanosecond CPU usage resolution, we have to count
>    *  the number of "ticks" we gave credit for to give the user a rough
>    *  guideline as to what each number means proportionally.
>    */
> #ifdef RTEMS_ENABLE_NANOSECOND_CPU_USAGE_STATISTICS
>
>   struct timespec uptime, total, rane/
>
>   _TOD_Get_uptime( &uptime );
>   _Timespec_Subtract( &CPU_usage_Uptime_at_last_reset, &uptime, &total );
>
>   ran = _Thread_Idle->cpu_time_used;
>   _Timespec_Divide( &ran, &total, &ival, &fval );
>
> #else
>
>   uint32_t             i;
>   uint32_t             api_index;
>   Thread_Control      *the_thread;
>   Objects_Information *information;
>   uint32_t             total_units = 0;
>
>   for ( api_index = 1 ; api_index <= OBJECTS_APIS_LAST ; api_index++ ) {
>
>     if ( !_Objects_Information_table[ api_index ] )
>       continue;
>     information = _Objects_Information_table[ api_index ][ 1 ];
>     if ( information ) {
>       for ( i=1 ; i <= information->maximum ; i++ ) {
>         the_thread = (Thread_Control *)information->local_table[ i ];
>
>         if ( the_thread )
>           total_units += the_thread->cpu_time_used;
>       }
>     }
>   }
>
>   ival = (total_units) ?  _Thread_Idle->cpu_time_used * 10000 / total_units : 0;
>   fval = ival % 100;
>   ival /= 100;
>
> #endif
>
>   promille = ival * 10 + fval / 100;
>   return promille;
> }
>
> Any ideas on what I'm missing here?
>
> Thanks,
>
> Giovanni Macciocu, Software Design Engineer
> SRON Netherlands Institute for Space Research
>
>
>
> _______________________________________________
> rtems-devel mailing list
> rtems-devel at rtems.org
> http://www.rtems.org/mailman/listinfo/rtems-devel