pfm_get_event_encoding man page

pfm_get_event_encoding — get raw event encoding


#include <perfmon/pfmlib.h>

int pfm_get_event_encoding(const char *str,int dfl_plm, char **fstr, int *idx, uint64_t *code, int *count);


This function is used to retrieve the raw event encoding corresponding to the event string in str. The string may contain unit masks and modifiers. The default privilege level mask is passed in dfl_plm. It may be used depending on the event.

This function is deprecated. It is superseded by pfm_get_os_event_encoding() where the OS is set to PFM_OS_NONE. Encoding is retrieve through the pfm_pmu_encode_arg_t structure.

The following examples illustrates the transition:

   int i, count = 0;
   uint64_t *codes;

   ret = pfm_get_event_encoding("RETIRED_INSTRUCTIONS", PFM_PLM3, NULL, &codes, &count);
   if (ret != PFM_SUCCESS)
      err(1", cannot get encoding %s", pfm_strerror(ret));

   for(i=0; i < count; i++)
      printf("count[%d]=0x%"PRIx64"\n", i, codes[i]);

is equivalent to:

   pfm_pmu_encode_arg_t arg;
   int i;

   memset(&arg, 0, sizeof(arg));
   arg.size = sizeof(arg);

   ret = pfm_get_os_event_encoding("RETIRED_INSTRUCTIONS", PFM_PLM3, PFM_OS_NONE, &arg);
   if (ret != PFM_SUCCESS)
      err(1", cannot get encoding %s", pfm_strerror(ret));

   for(i=0; i < arg.count; i++)
      printf("count[%d]=0x%"PRIx64"\n", i,[i]);


The encoding may take several 64-bit integers. The function can use the array passed in code if the number
of entries passed in count is big enough. However, if both *codes is NULL and count
is 0, the function allocates the memory necessary to store the encoding. It is up to the caller to
eventually free the memory. The number of 64-bit entries in codes is reflected in *count upon
return regardless of whether the codes was allocated or used as is. If the number of 64-bit integers is
greater than one, then the order in which each component is returned is PMU-model specific. Refer to the PMU
specific man page.

The raw encoding means the encoding as mandated by the underlying PMU model. It may not be directly suitable
to pass to a kernel API. You may want to use API-specific library calls to ensure the correct encoding is passed.

If fstr is not NULL, it will point to the fully qualified event string upon successful return. The string
contains the event name, any umask set, and the value of all the modifiers. It reflects what the encoding will
actually measure. The function allocates the memory to store the string. The caller must eventually free the

Here is a example of how this function could be used:
#include <inttypes.h>
#include <err.h>
#include <perfmon/pfmlib.h>
int main(int argc, char **argv)
   uint64_t *codes 0;
   int count = 0;
   int ret;

   ret = pfm_initialize();
   if (ret != PFMLIB_SUCCESS)
      err(1", cannot initialize library %s", pfm_strerror(ret));

   ret = pfm_get_event_encoding("RETIRED_INSTRUCTIONS", PFM_PLM3, NULL, &codes, &count);
   if (ret != PFM_SUCCESS)
      err(1", cannot get encoding %s", pfm_strerror(ret));

   for(i=0; i < count; i++)
      printf("count[%d]=0x%"PRIx64"\n", i, codes[i]);

   return 0;


The function returns in *codes the encoding of the event and in *count the number of 64-bit integers to support that encoding. Upon success, PFM_SUCCESS is returned otherwise a specific error code is returned.



The code argument is too small for the encoding.


The code or count argument is NULL.


Not enough memory.


Event not found.


Invalid event attribute (unit mask or modifier)


Invalid modifier value.


attribute already set, cannot be changed.


Missing unit mask.


Unit masks or features cannot be combined into a single event.


Stephane Eranian <>

See Also



September, 2009 Linux Programmer's Manual