Frontend user code

From MidasWiki
Revision as of 18:22, 21 August 2014 by Suz (talk | contribs)
Jump to navigation Jump to search

Links


Introduction

This section describes the features of the user-written part of a Frontend, referred to here as frontend.c. To make a custom frontend, users will usually modify one of the templates provided in the MIDAS package for their particular hardware and other requirements. A partial list of the templates provided is shown below:

Hardware Filename Directory (MIDAS package) Purpose Language
VME fevmemodules.c ../midas/examples/Triumf/c/ Access to VME modules C
VME fevme.cxx ../midas/examples/Triumf/c++/ Access to VME modules C++
CAMAC frontend.c ../midas/examples/experiment/ Access to CAMAC modules C
RS485 bus mscb_fe.c ../midas/examples/slowcont/ Slow controls C
EPICS frontend.c ../midas/examples/epics/ Slow controls C


The features of a typical frontend program are best explained by reference to examples of the user code provided in the Midas Package.

Frontend code

The following sections refer to the templates for user frontend code found in the MIDAS packages (see Introduction). Most of the examples are taken from
http://ladd00.triumf.ca/~daqweb/doc/midas/examples/Triumf/c/fevmemodules.c

For a slow control frontend see also Slow Control System.

Include files

The following example (from template frontend file fevmemodules.c - see above) shows the standard include files needed for VME access with VMIC. The user may add any other include files as needed. In this case, header files for several device drivers have also been added. The optional file experim.h is often included. This is a special include file for ease of communication between the C code and the ODB. The example fevmemodules.c includes
http://ladd00.triumf.ca/~daqweb/doc/midas/examples/Triumf/c/experim.h.
The file experim.h (except when using one of the examples) is generated by the user (see experim.h). The example below shows a typical list of include files for a frontend:

#include <stdio.h>     // C standard headers
#include <stdlib.h>
#include "midas.h"     // MIDAS include file
#include "mvmestd.h"   // VME header file
#include "vmicvme.h"   // VMIC header file
#include "vmeio.h"     // optional hardware header files
#include "v1190B.h"
#include "v792.h"
#include "vf48.h"
#include "v1729.h"
#include "experim.h"   // user-created with odbedit make command


Global declarations

The following example (from template frontend file fevmemodules.c - see Introduction) shows the global declaration. The declarations are system wide. Some may be changed to suit the user, but none should not be removed.

frontend_name
This value can be modified to reflect the purpose of the code
frontend_call_loop
If set to TRUE, the function frontend_loop() runs after every equipment loop. If FALSE, frontend_loop() does not run. The user can add suitable code to this routine if desired (e.g. to check for a condition).
display_period
The time interval (defined in milliseconds) between the refresh of a frontend status display. The value of zero disables the display. If the frontend is started in the background with the display enabled, the stdout should be redirected to the null device to prevent the process from hanging.
max_event_size
specifies the maximum size (in bytes) of the expected event.
event_buffer_size
specifies the maximum size (in bytes) of the buffer to be allocated by the system.

See below for an example of global declarations from a frontend.

/* The frontend name (client name) as seen by other MIDAS clients   */
char *frontend_name = "fevmemodules";
/* The frontend file name, don't change it */
char *frontend_file_name = __FILE__;
                                                                    //
/* frontend_loop is called periodically if this variable is TRUE    */
BOOL frontend_call_loop = FALSE;
                                                                    //
/* a frontend status page is displayed with this frequency in ms */
INT display_period = 000;
                                                                    //
/* maximum event size produced by this frontend */
INT max_event_size = 200000;
                                                                    //
/* maximum event size for fragmented events (EQ_FRAGMENTED) */
INT max_event_size_frag = 5 * 1024 * 1024;
                                                                    //
/* buffer size to hold events */
INT event_buffer_size = 10 * 100000;


Global User Declarations

After the global declarations, the user may add his or her own declarations. The following example (from template frontend file fevmemodules.c - see Introduction) defines various VME hardware parameters. An example of hardware declarations in a frontend are shown below:

/* Hardware */
MVME_INTERFACE *myvme;
                     //
/* VME base addresses */
DWORD VMEIO_BASE = 0x780000;
DWORD VTDC0_BASE = 0xF10000;
.......
/* Globals */
#define  N_ADC 100
#define  N_TDC 100
#define  N_PTS 5000


System Function Prototypes

These prototypes declare the pre-defined system functions which should be present.

INT frontend_init();
INT frontend_exit();
INT begin_of_run(INT run_number, char *error);
INT end_of_run(INT run_number, char *error);
INT pause_run(INT run_number, char *error);
INT resume_run(INT run_number, char *error);
INT frontend_loop();


Readout Function Prototypes

Following the previous group is a second group of prototypes, which define the readout functions. These depend on the defined equipments, and run when the respective equipment is triggered. In this example, two equipments will be defined, so there are two prototypes. The user functions will be described in detail in the following sections.

INT read_trigger_event(char *pevent, INT off);
INT read_scaler_event(char *pevent, INT off);

If using an interrupt, callback function prototypes are also included

extern void interrupt_routine(void);
void register_cnaf_callback(int debug);


Bank Definition and Equipment List

See Bank Definitions and Equipment List for detailed information.

However, this may be easier to follow if first the sequence of operations, and examples of the the different types of events (polled/interrupt or periodic etc.) and the functions they require in the frontend are introduced.

Sequence of Operations in the frontend

The following table shows the sequence of operations of the Frontend System functions. These subroutines must be present in frontend.c, but the contents are coded by the user. These functions are called by mfe.c at the appropriate time. The System Transition functions are associated with a particular Run Transition as shown below:

System Function System Transition Function Associated Transition Action
frontend_init() Runs once after system initialization, before Equipment registration.
begin_of_run() TR_START Runs after system statistics reset at each begin-of-run request.
pause_run() TR_PAUSE Runs at each pause-run request.
resume_run() TR_RESUME Runs at each resume-run request.
end_of_run() TR_STOP Runs at each end-of-run request.
frontend_exit() Runs once before any Slow Control Equipment exit


Each defined Equipment has the option to force itself to run at individual transition times (see Equipment ReadOn Flag), so that its equipment function will be called on a certain transition (or combination of transitions).

The system transition functions all run prior to the equipment functions. This gives the system the chance to take basic action on the transition request (e.g. enable/disable interrupt) before the equipment runs. All the transition routines run with a Transition Sequence number of 500 (the default). This allows users to add additional functions in the frontend that will run before or after any of the transitions (such as a prestart() or a poststop() function). See Run Transition Priority for more information.

Function frontend_init

No parameters.

This function runs once only at the application startup. Users may perform hardware checking, loading/setting of global variables, hot-link settings to the ODB etc. in frontend_init(), e.g.


INT frontend_init()
{
  set_equipment_status(equipment[idx].name, "Initializing...", "yellow");
  ....
  // Open VME interface
  status = mvme_open(&myvme, 0);
                                  //
  // Set am to A24 non-privileged Data
  mvme_set_am(myvme, MVME_AM_A24_ND);
                                  //
  // Set dmode to D32
  mvme_set_dmode(myvme, MVME_DMODE_D32)
  ....
                                  //
  set_equipment_status(equipment[idx].name, "OK", "green");
  return SUCCESS;
}


Reporting Equipment Status

If running with run control utility mhttpd, a frontend can send an update to the main status page, to report on its progress, using the function set_equipment_status (see above example). This is useful when hardware can take a long time to respond.


Function begin_of_run

Parameters:

  • INT run number provides the number of the current run being started
  • char * error can be used for returning a message to the system. This message string will be logged into the midas.log file (see Message System.

This function is called every time a run start transition occurs, i.e. at begin-of-run. It allows the updating of user parameters, and the loading/setup/clearing of hardware. At the exit of this function, the acquisition should be armed and ready to test the interrupt (if used), e.g.


INT begin_of_run (INT runnumber, char * error)
{
  ......
  // Set am to A24 non-privileged Data
  mvme_set_am(myvme, MVME_AM_A24_ND);
  // Set dmode to D32
  mvme_set_dmode(myvme, MVME_DMODE_D32);
                                     //
  //-------- ADCs -------------------
  v792_Setup(myvme, VADC0_BASE, 2);
  v792_ThresholdWrite(myvme, VADC0_BASE,
      (WORD *)&(ts.v792.threshold1));
  v792_DataClear(myvme, VADC0_BASE);
  csr = v792_CSR1Read(myvme, VADC0_BASE);
  printf("Data Ready ADC0: 0x%x\n", csr);
  ........
  // Disable interrupt
  mvme_write_value(myvme, VLAM_BASE+4, 0x0);
  // Reset Latch
  mvme_write_value(myvme, VLAM_BASE, 0x1);
  // Clear pending interrupts
  mvme_write_value(myvme, VLAM_BASE+8, 0x0);
  // Enable interrupt
  inRun = 1;
  mvme_write_value(myvme, VLAM_BASE+4, inRun);
  return SUCCESS;
}


Functions pause/resume_run

Parameters:

  • INT run number provides the number of the current run being paused/resumed.
  • char * error can be used for returning a message to the system. This message string will be logged into the midas.log file (see Message System.

These two functions are called upon "Pause" and "Resume" command respectively. Any code relevant to the upcoming run state can be included,e.g.


INT pause_run (INT run_number, char * error)
{
  disable_trigger();
  // Disable interrupt
  inRun = 0;
  mvme_write_value(myvme, VLAM_BASE+4, inRun);
  return SUCCESS;
}
                           //
INT resume_run (INT run_number, char * error)
{
  enable_trigger();
  inRun = 1;
  mvme_write_value(myvme, VLAM_BASE+4, inRun);
  return SUCCESS;
}

Function end_run

Parameters:

  • INT run number provides the number of the current run being ended.
  • char * error can be used for returning a message to the system. This message string will be logged into the midas.log file (see Message System.

This function is called at every "stop run" transition. It provides the opportunity to disable the hardware, e.g.

INT end_of_run(INT run_number, char *error)
{
  // Stop DAQ for seting up the parameters
  vf48_AcqStop(myvme, VF48_BASE);
                      //
  done = 0;
  stop_req = 0;
  inRun = 0;
  // Disable interrupt
  mvme_write_value(myvme, VLAM_BASE+4, inRun);
  trig_level = 0;
  // Close run gate
  vmeio_AsyncWrite(myvme, VMEIO_BASE, 0x0);
  return SUCCESS;
}


Function frontend_exit

Parameters: none

The function runs When the frontend program is shut down. Can be used to release any locked resources like memory, communications ports etc. e.g.

 function frontend_exit()
 {
    mvme_close(gVme);
    return;
 }

Function frontend_loop

Parameters: none

If the flag frontend_call_loop is set TRUE, this routine is called when the frontend is idle or once between every event. In the following example, it is being used to check for a timeout:

...
BOOL frontend_call_loop = TRUE;
...
                                //
INT frontend_loop()
{
  char str[128];
                  //
  if (stop_req && done==0) {
     db_set_value(hDB,0,"/logger/channels/0/Settings/Event limit", &evlimit, sizeof(evlimit), 1, TID_DWORD);
     if (cm_transition(TR_STOP, 0, str, sizeof(str), ASYNC, FALSE) != CM_SUCCESS) {
        cm_msg(MERROR, "VF48 Timeout", "cannot stop run: %s", str);
     }
     inRun = 0;
     // Disable interrupt
     mvme_write_value(myvme, VLAM_BASE+4, inRun);
     done = 1;
     cm_msg(MERROR, "VF48 Timeout","VF48 Stop requested");
  }
  return SUCCESS;
}



Event Types and Triggers

The frontend supports several different types of event trigger. The event trigger type is specified by the Equipment Flag in the Equipment Declaration. Common event types are "polled events" where the Equipment Flag is EQ_POLLED, "interrupts events" where the Flag is EQ_INTERRUPT, and "periodic events" where the Flag is EQ_PERIODIC. The name of the associated readout routine is specified in the Equipment Declaration for each event type.

Polled and interrupt events (see Event Types) require several extra functions to handle the hardware that periodic events do not require. These are described below.


Function poll_event

Parameters:

  • INT source
  • INT count
  • BOOL test

If the Equipment Type is EQ_POLLED, the poll_event() routine will be called as often as possible over the corresponding poll time (e.g. 500ms) given by each polling equipment.

In this case, the Equipment declaration would have this form:

   EQUIPMENT equipment[] = {
     { "Trigger",            // equipment name
       {
         ...
         EQ_POLLED,          // equipment type
         ...
         500,                // poll for 500ms 
         ...
         "", "", "",},
         read_trigger_event,    // readout routine 
      ...
        

The user must provide suitable code in the routine poll_event(), e.g.

INT poll_event(INT source, INT count, BOOL test)
{
   /* Polling routine for events. Returns TRUE if event  is available. If test equals TRUE, don't return. 
      The test flag is used to time the polling 
    */
    int i;
    int lam = 0;
                                //
    for (i = 0; i < count; i++, lam++) {
      lam = vmeio_CsrRead(myvme, VMEIO_BASE);
      if (lam)
        if (!test)
          return lam;
    }
    return 0;
 }

An event readout routine must also be provided by the user.


Function interrupt_configure

  • INT cmd
  • INT source
  • PTYPE adr

If the Equipment Type is EQ_INTERRUPT, an interrupt configuration routine called interrupt_configure() must be provided by the user. The interrupt configuration routine has the following declaration:

/*-- Interrupt configuration --------------------------*/
INT interrupt_configure(INT cmd, INT source, PTYPE adr)
{
 int vec = 0;
 switch (cmd) 
 {
   case CMD_INTERRUPT_ENABLE:
     if (inRun) mvme_write_value(myvme, VLAM_BASE+4, 0x1);
     break;
                                     //
   case CMD_INTERRUPT_DISABLE:
     if (inRun) mvme_write_value(myvme, VLAM_BASE+4, 0x0);
     break;
                                     //
   case CMD_INTERRUPT_ATTACH:
     mvme_set_dmode(myvme, MVME_DMODE_D32);
     mvme_interrupt_attach(myvme, INT_LEVEL, INT_VECTOR, 
               (void *)adr, &myinfo);
     mvme_write_value(myvme, VLAM_BASE+0x10, INT_VECTOR);
     vec = mvme_read_value(myvme, VLAM_BASE+0x10);
     printf("Interrupt Attached to 0x%x for vector:0x%x\n",
                    adr, vec&0xFF);
     break;
                                     //
   case CMD_INTERRUPT_DETACH:
     printf("Interrupt Detach\n");
     break;
  }
  return SUCCESS;
}

Under the four commands listed above, the user must implement the hardware operation needed to perform the requested action. In the Midas drivers directory examples can be found of such an interrupt code for CAMAC. See source code such as hyt1331.c,ces8210.c

An event readout routine must also be provided by the user in the frontend.

Event Readout routine

In the case of POLLED or INTERRUPT events, the event readout routine is called an Interrupt readout routine.

An event readout routine (called when an event occurs) is usually of the form

INT function_name ( char *pevent ... )
{
  INT event_size;
  ........  // read data from hardware
  ........  // pack into banks depending on format
  ........
  return (event_size);
}

where the first argument of the readout function (pevent) provides the pointer to the newly constructed event, and points to the first valid location for storing the data.

NOTE
  • The return value is the event size, and must be the number of bytes collected in this function.
  • The event serial number will be incremented by one for every call to the readout routine, as long as the returned size is non-zero.
  • If the returned value is set to zero, the event will be dismissed and the serial number to that event will be decremented by one.


General readout function

If the Equipment type is EQ_INTERRUPT or EQ_POLLED (i.e. Polled or Interrupt event) see also Polled or Interrupt readout routine.

In this case (e.g. a periodic event) the Equipment declaration may have this form:

   EQUIPMENT equipment[] = {
     { "Scaler",           // equipment name
        {    
           ...
           EQ_PERIODIC     // equipment type
           0,              // interrupt source (ignored) 
           "MIDAS",        // event format
           ...
           10000,          // period (read every 10s)
           ...
           "", "", "",},
      read_scaler_event,   // readout routine 
      ...

A readout event will often need to send out data. This is done using one of the supported event structures usually "MIDAS" format data banks. The bank format ("MIDAS" in the example above) is declared in the Equipment List Parameters Format field.

An example of a scaler readout routine read_scaler_event() where the data is read out into MIDAS data banks is shown below.

INT read_scaler_event(char *pevent, INT off)
{
   DWORD *pdata, a;
                              //
   /* init bank structure; */
   bk_init(pevent);
                              //
   /* create SCLR bank */
   bk_create(pevent, "SCLR", TID_DWORD, &pdata);
                              //
   /* read scaler bank (CAMAC) */
   for (a = 0; a < N_SCLR; a++)
      cam24i(CRATE, SLOT_SCLR, a, 0, pdata++);
                              //    
   /* close SCLR bank */
   bk_close(pevent, pdata);
                              //
   /* return event size in bytes */
   return bk_size(pevent);
}

Some other examples of event readout routines in this document are

Many other examples of readout routines can be found in the frontend templates in the MIDAS package.

Polled or Interrupt readout routine

In the case of a Polled or Interrupt event, the content of the memory location pointed to by pevent (see Event Readout routine) prior to its use in the readout function, contains the interrupt source bitwise register. This feature can be exploited in order to identify which hardware module has triggered the readout when multiple interrupts have been assigned to the same readout function.

The examples below show a VME interrupt source for a given equipment. Depending whether USE_INT is defined, the Equipment will either use a Polled or an Interrupt mechanism. The Equipment declaration is of the form:

EQUIPMENT equipment[] = {
              //
   {"Trigger",  /* equipment name */
      ...
#ifdef USE_INT
      EQ_INTERRUPT, /* equipment type */
#else
      EQ_POLLED,    /* equipment type */
#endif
  /* interrupt source: crate 0, all stations */
      LAM_SOURCE(0, 0x0),
      ....
      "", "", "",
    },
    read_trigger_event, /* readout routine */
    NULL, NULL,
    trigger_bank_list,
}
 

Note that the LAM_SOURCE macro simply codes the parameters into a bitwise register.

The readout routine would contains code such as

INT read_trigger_event(char *pevent, INT off)
{
#if defined VADC0_CODE
  DWORD  *pdata;
#endif
                       //
#if defined VADC0_CODE
  /* read ADC0 data */
  v792_EvtCntRead(myvme, VADC0_BASE, &evtcnt);
  ........
  /* Read Event */
  v792_EventRead(myvme, VADC0_BASE, pdata, &nentry);
  ........
  v792_DataClear(myvme, VADC0_BASE);
#endif
                      //
  ........
  return (size);
}

The data will be packed into banks as described for the general readout function above. The examples fevmemodules.c (VME) and frontend.c (CAMAC) (see Introduction contain a complete example of read_trigger_event().


Manual Trigger

Another type of frontend event trigger supported is the "manual trigger", where the Equipment Flag is EQ_MANUAL_TRIGGER. This flag causes an extra button to be present on the mhttpd web interface to enable the user to trigger the event.

In some cases, the same readout code may be used for two types of event: a manual trigger and (say) a poll event. It is possible to determine whether the readout of an event was triggered by a manual trigger or a regular trigger by adding a call to the event header Macro DATA_SIZE in the readout routine:

 flag = DATA_SIZE(pevent);

If the result is

 *  flag = 0 normal call
 *  flag = 1 manual trigger

It is also possible for a backend MIDAS client (such as an analyzer or custom data archiver) to trigger a manual trigger event. The client controls when an event is sent by means of a function that requests an event by triggering the event sending mechanism with a RPC call.

With a frontend Equipment declaration of a manually triggered event of the form:

   { "Histo",             /* equipment name */
      2, 0,                 /* event ID, trigger mask */
      "SYSTEM",             /* event buffer */
      EQ_MANUAL_TRIG,     /* equipment type */
      0,    
      .......
   }
   

the code fragment to manually trigger this event is:

 int main(unsigned int argc,char **argv)
 {
     .......
     bm_request_event(hBufEvent, 2, TRIGGER_ALL, GET_ALL, &request_id, process_event_TD);
     .......
 }

When it is time to save the data during the run, the function below is called:

BOOL trigger_histo_event(void)
{
  HNDLE hconn;
  BOOL event_triggered;
                           //
  event_triggered = FALSE;
  ...................
  if (run_state == STATE_RUNNING) {   // Check the frontend client exists
       if( cm_exist(ClientName,TRUE))  {    
           status = cm_connect_client (ClientName, &hconn);
           if(status != RPC_SUCCESS)
               cm_msg(MERROR,"trigger_histo_event","Cannot connect to frontend \"%s\" (%d)",
                      ClientName,status);
           else  {     // successfully connected to frontend client
               status = rpc_client_call(hconn, RPC_MANUAL_TRIG, 2); // trigger a histo event
               if (status != CM_SUCCESS)
                   cm_msg(MERROR,"trigger_histo_event","Error triggering event from frontend (%d)",status);
               else {  // successfully triggered event
                   event_triggered=TRUE;
                   status =cm_disconnect_client(hconn, FALSE);
                   if (status != CM_SUCCESS)
                       cm_msg(MERROR,"trigger_histo_event","Error disconnecting client (%d)",status);
               }
           }
       }
       else
           cm_msg(MERROR,"trigger_histo_event","Frontend client %s not running (%d)",
                  ClientName,status);
   } 
   return(event_triggered);
}


Deferred Transistion

This option permits the user to postpone any transition issued by any requester until some condition is satisfied. For example:

  • It may not be advisable to pause or stop a run until some hardware has turned off a particular valve.
  • The start of the acquisition system should be postponed until the beam rate has been stable for a given period of time.
  • While active, a particular acquisition system should not be interrupted until the "cycle" is completed.

In these examples, any application having access to the state of the hardware can register to be a "transition Deferred" client. The MIDAS system will then catch any transition request and postpone the trigger of such a transition until the condition is satisfied.

The Deferred transition requires 3 steps for setup

  1. Register for the deferred transition
  2. Provide a callback function to serve the deferred transition
  3. Implement the condition code

The following example demonstrates this process:

 BOOL transition_PS_requested=FALSE; // global
                                             //
 INT frontend_init()
 {
   // register for deferred transition
                                             //
   cm_register_deferred_transition(TR_STOP, wait_end_cycle);
   cm_register_deferred_transition(TR_PAUSE, wait_end_cycle);
   ...  
 }
/*
*/ 
 //-- Deferred transition callback
 BOOL wait_end_cycle(int transition, BOOL first)
 {
   if (first) {
     transition_PS_requested = TRUE;
     return FALSE;
   }
                       //
   if (end_of_mcs_cycle){
     transition_PS_requested = FALSE;
     end_of_mcs_cycle = FALSE;
     return TRUE;
   }
   else
     return FALSE;
 }
/*
*/
 INT read_mcs_event(char *pevent, INT offset)
 {  
     ...     // read out data at end of cycle
     ...
     end_of_mcs_cycle = TRUE; // end of cycle 
                              //
     if (!transition_PS_requested)
        start_cycle(); // start a new cycle 
     return bk_size(pevent);
  }


In the example above,

  • The frontend code is registered for PAUSE and STOP using cm_register_deferred_transition(). The second argument wait_end_cycle is the declaration of the callback function.
  • The callback function wait_end_cycle will be called as soon as the transition is requested with the Boolean flag first set to TRUE.
  • By setting transition_PS_requested TRUE , the user will be provided with the acknowledgment of the transition request.
  • By returning FALSE from the callback, the transition is prevented from occurring.
  • As soon as the user condition is satisfied (end_of_mcs_cycle = TRUE), the return code in the callback will be set to TRUE and the requested transition will be issued.

While the transition is Deferred, the odb key /runinfo/Requested transition will contain the transition code, and the d mhttpd webserver main status page will indicate that a deferred transition is in progress.

Once in deferred state, an odbedit override command can be issued to force the transition to happen,

>odbedit
odb> stop now    (or "start now")

or /runinfo/Requested transition can be set to 0. The transition will then take place on the next stop or start command.