]> git.kernelconcepts.de Git - karo-tx-linux.git/commit
powerpc/mm: Add trace point for tracking hash pte fault
authorAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Tue, 14 Apr 2015 07:35:57 +0000 (13:05 +0530)
committerMichael Ellerman <mpe@ellerman.id.au>
Wed, 10 Jun 2015 04:06:29 +0000 (14:06 +1000)
commitcfcb3d80a28380ba027331eb548ba309c4b66559
treedfdfc88aaa2a101d5136a7d88eeea9325a342677
parent809fac67c8530704a214e2c476349692e2fb9299
powerpc/mm: Add trace point for tracking hash pte fault

This enables us to understand how many hash fault we are taking
when running benchmarks.

For ex:
-bash-4.2# ./perf stat -e  powerpc:hash_fault -e page-faults /tmp/ebizzy.ppc64 -S 30  -P -n 1000
...

 Performance counter stats for '/tmp/ebizzy.ppc64 -S 30 -P -n 1000':

       1,10,04,075      powerpc:hash_fault
       1,10,03,429      page-faults

      30.865978991 seconds time elapsed

NOTE:
The impact of the tracepoint was not noticeable when running test. It was
within the run-time variance of the test. For ex:

without-patch:
--------------

 Performance counter stats for './a.out 3000 300':

       643      page-faults               #    0.089 M/sec
  7.236562      task-clock (msec)         #    0.928 CPUs utilized
 2,179,213      stalled-cycles-frontend   #    0.00% frontend cycles idle
17,174,367      stalled-cycles-backend    #    0.00% backend  cycles idle
 0      context-switches          #    0.000 K/sec

       0.007794658 seconds time elapsed

And with-patch:
---------------

 Performance counter stats for './a.out 3000 300':

       643      page-faults               #    0.089 M/sec
  7.233746      task-clock (msec)         #    0.921 CPUs utilized
 0      context-switches          #    0.000 K/sec

       0.007854876 seconds time elapsed

 Performance counter stats for './a.out 3000 300':

       643      page-faults               #    0.087 M/sec
       649      powerpc:hash_fault        #    0.087 M/sec
  7.430376      task-clock (msec)         #    0.938 CPUs utilized
 2,347,174      stalled-cycles-frontend   #    0.00% frontend cycles idle
17,524,282      stalled-cycles-backend    #    0.00% backend  cycles idle
 0      context-switches          #    0.000 K/sec

       0.007920284 seconds time elapsed

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
arch/powerpc/include/asm/trace.h
arch/powerpc/mm/hash_utils_64.c