On Tue, Sep 4, 2012 at 7:42 AM, Arkaprava Basu wrote:
> Thanks Arun for the reply. Do you know whether the patches that you mentioned
> merged with kernel mainline or not? If not, can I trouble you to point me to
> which patches you are talking about?
Not merged yet.
http://marc.info/?l=linux-
On Sat, Sep 1, 2012 at 5:07 PM, Arkaprava Basu wrote:
> Hello,
>
> I am new to perfmon. I need to use PEBS to collect a trace of processor
> state (and ultimately generate address trace) for DTLB miss events. I was
> wondering if somebody can point to some example on how I can use libpfm to
>
On 6/8/12 9:46 AM, William Cohen wrote:
> In the meantime it would be good to get libpfm upstream fixed so swig doesn't
> complain about this error. I think that the attached patch eliminates this
> problem. This is based on the approach mentioned in a similar swig bug
> report:
>
> http://sour
@@
-#!/usr/bin/env python
#
# Copyright (c) 2008 Google, Inc.
# Contributed by Arun Sharma
diff --git a/python/src/session.py b/python/src/session.py
index df22ea6..f598aa4 100644
--- a/python/src/session.py
+++ b/python/src/session.py
@@ -1,4 +1,3 @@
-#!/usr/bin/env python
#
# Copyright (c) 2008
On Tue, Apr 24, 2012 at 6:51 AM, William Cohen wrote:
> Hi Arun,
>
> I don't have much experience with python and I am not that familiar with how
> these scripts are suppose to be used. Is there some description how these
> are suppose to be used?
If everything works fine, you should be able t
On Wed, Mar 28, 2012 at 1:56 PM, William Cohen wrote:
> libpfm-python.x86_64: E: non-executable-script
> /usr/lib64/python2.6/site-packages/perfmon/pmu.py 0644L /usr/bin/env
> libpfm-python.x86_64: E: non-executable-script
> /usr/lib64/python2.6/site-packages/perfmon/session.py 0644L /usr/bin/en
On Sat, May 28, 2011 at 4:56 AM, heechul Yun wrote:
>
> Can anyone tell me if this, exiting from WFI on every page fault, is
> correct behavior? If so, why?
Without knowing the full context, I'd look into the possibility that
the kernel is trying to shoot down TLB entries on the other processors
---
lib/pfmlib_amd64.c |3 ---
lib/pfmlib_common.c |8 +---
lib/pfmlib_intel_x86.c |2 --
lib/pfmlib_perf_event.c |3 +--
lib/pfmlib_perf_event_pmu.c | 11 ---
perf_examples/self_count.c |4 +++-
perf_examples/self_pipe.c |5 ++---
Useful for getting stack traces for hardware events not handled
by PERF_TYPE_HARDWARE.
Signed-off-by: Arun Sharma
Cc: Ingo Molnar
Cc: Frederic Weisbecker
Cc: Mike Galbraith
Cc: Paul Mackerras
Cc: Peter Zijlstra
Cc: Stephane Eranian
Cc: Thomas Gleixner
Cc: Tom Zanussi
Cc: linux-ker
2011/3/20 陳韋任 :
> And I am wondering is it possible that libpfm4 exposes LBR to users
> even perf_events doesn't?
No - the kernel needs to support this functionality by exposing it via
a system call. libpfm4 is a user level convenience library that
converts an event name into a structure expecte
2011/3/17 陳韋任 :
> The "better match" just means better than usual sampling mechanisms,
> right? Arun mentioned that "PEBS is known to be precise, but not
> accurate". So we still cannot get the exact IP which causes the event.
> Am I right?
Appendix B.2.3.3 Precise Execution Events in this doc:
On Thu, Mar 17, 2011 at 8:43 AM, Arun Sharma wrote:
>
> Stephane has disabled the uncore event support in libpfm4 (since the
> patch is still not upstream yet). But by uncommenting 2-3 lines of
> code, you should be able to get uncore events working.
>
This is the patch I
[ + Lin Ming, linux-perf-users ]
On Thu, Mar 17, 2011 at 5:35 AM, Easley wrote:
> EventNum and UmaskValue is control bits, but I can't find which address of
> msr to wirte.
Try Ming Lin's kernel patch here:
http://thread.gmane.org/gmane.linux.kernel/1071064
and my patch to link the perf tool
On Wed, Mar 16, 2011 at 7:07 PM, 陳韋任 wrote:
> You mean PEBS makes the skid disappered? Does `perf` or libpfm4 support this?
> The example I showed was ran on a x86 machine, it seems `perf` does not use
> PEBS
> feature.
>
It depends on the suffix. Try perf record -e branch-misses:p
PEBS is kn
libpfm4 is a library that takes a CPU vendor documentation
compatible string and fills out perf_event_attr. Typical
use case: user wants to look at events other than the
ones supported by perf using symbolic names.
Signed-off-by: Arun Sharma
Cc: Ingo Molnar
Cc: Frederic Weisbecker
Cc: Mike
This is necessary because swig recognizes the T *output, but
not struct T *output as an output argument.
---
python/src/perfmon_int.i |7 ++-
1 files changed, 6 insertions(+), 1 deletions(-)
diff --git a/python/src/perfmon_int.i b/python/src/perfmon_int.i
index db6f4c5..b34c6fe 100644
---
On Wed, Mar 2, 2011 at 7:20 PM, Lin Ming wrote:
>> +#ifdef LIBPFM4
>> + printf("\n");
>> + pfm_for_all_pmus(i) {
>> + int ret;
>> + pfm_pmu_info_t pinfo;
>> + int count;
>> + int k;
>> +
>
> Need to initialize pinfo before passing
On Thu, Mar 03, 2011 at 09:32:55AM +0800, Lin Ming wrote:
>
> I got a build error with latest libpfm4 git tree(bff5dff),
>
> CC util/parse-events.o
> cc1: warnings being treated as errors
> util/parse-events.c: In function ‘print_events’:
> util/parse-events.c:1087: error: implicit declaratio
This is an incremental patch on top of my earlier -v2 patch.
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index a7e5765..81c154e 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -1045,6 +1045,7 @@ static void print_umasks(pfm_event
of libpfm4.
Signed-off-by: Arun Sharma
Cc: Ingo Molnar
Cc: Frederic Weisbecker
Cc: Mike Galbraith
Cc: Paul Mackerras
Cc: Peter Zijlstra
Cc: Stephane Eranian
Cc: Thomas Gleixner
Cc: Tom Zanussi
---
tools/perf/Makefile| 13 +
tools/perf/feature-tests.mak | 11
On Sat, Feb 26, 2011 at 11:50 AM, Sergey Blagodurov
wrote:
> 1) pfmon can monitor more events than there are actual physical
> counters on the PMU with the event sets mechanism.
perf can do event multiplexing too.
>
> 2) pfmon can print intermediate counts while in system-wide mode.
>
You can
This is for supporting uncore events.
I tested this diff with Lin Ming's uncore patch against the kernel
http://thread.gmane.org/gmane.linux.kernel/1071064
diff --git a/include/perfmon/perf_event.h b/include/perfmon/perf_event.h
index defc88e..de5691e 100644
--- a/include/perfmon/perf_event.h
+++
I had to hack around one thing to make perf+libpfm4 work with uncore events:
--- a/lib/pfmlib_perf_event.c
+++ b/lib/pfmlib_perf_event.c
@@ -153,7 +153,7 @@ pfmlib_perf_event_encode(void *this, const char *str, int
dfl_plm, void *data)
attr->exclude_user = !(plm & PFM_PLM3);
at
BTW, pfm_for_each_event() still refers to the obsolete/unimplemented
pfm_get_event_first(). We should either fix up the macro or kill it
completely.
-Arun
--
The ultimate all-in-one performance toolkit: Intel(R) Parall
On Tue, Feb 15, 2011 at 11:48:17PM +0100, Stephane Eranian wrote:
>
> Why do you need to remove #if 0 code?
I don't need to. Feel free to ignore that part of the patch :)
-Arun
--
The ultimate all-in-one performance to
* Remove unimplemented functions from headers (so swig doesn't generate code
for them)
* Rename the output argument so swig can handle it properly.
* Pass the new pfm_os_t arg properly
---
include/perfmon/pfmlib.h |3 +--
lib/pfmlib_common.c | 44 --
On Mon, Feb 14, 2011 at 3:01 PM, Can Hankendi wrote:
> Hi again,
>
> What I'm looking for is creating performance traces. Let's say I've a
> program running for 5 seconds, and I wanna collect performance data for
> every 10 ms. In the end, I'll have a trace with 500 stats (1 stat for each
> 10ms).
On Fri, Feb 11, 2011 at 12:54 PM, Can Hankendi wrote:
> I receive an error when I try to use -A. ""Error: unknown switch"
> How can I update the tool to use -A ability.
Your perf binary is not new enough. Please rebuild from source.
git log tools/perf/builtin-stat.c
[...]
commit f5b4a9c3ab53d544
[ bcc: perfmon2-devel, +linux-perf-users ]
On Fri, Feb 11, 2011 at 11:16 AM, Can Hankendi wrote:
> Hi,
>
> I'm using 2.6.38 kernel and trying to run performance tests for individual
> cores on my system. But it seems like current kernel doesn't include the
> patches for perf events. I tried to co
libpfm4 is a library that takes a CPU vendor documentation
compatible string and fills out perf_event_attr. Typical
use case: user wants to look at events other than the
ones supported by perf using symbolic names.
Signed-off-by: Arun Sharma
Cc: Ingo Molnar
Cc: Frederic Weisbecker
Cc: Mike
Hi All,
I spent some time packaging libpfm4 for rpm based systems. %patch0
should be necessary only if your python is installed somewhere other
than /usr.
-Arun
libpfm4.spec
Description: Binary data
python-install-path.patch
Description: Binary data
--
---
python/src/session.py |2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/python/src/session.py b/python/src/session.py
index f90e541..df22ea6 100644
--- a/python/src/session.py
+++ b/python/src/session.py
@@ -59,7 +59,7 @@ class SystemWideSession(Session):
self.cp
On Mon, May 5, 2008 at 2:42 PM, Vivek Thakkar <[EMAIL PROTECTED]> wrote:
> I am also reading the documents mentioned by Anoop. It looks like IBS
> supports time based sampling and not event based sampling ala PEBS. I am not
> sure how much lossy the information would be. Anyways, a related questi
Patch attached.
-Arun
python-multi-cpu.patch
Description: Binary data
-
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.n
This might be of interest to users of python bindings:
I had an interface to the linux scheduler APIs in glibc as a part of another
module. I've moved it into it's own namespace (linux.sched) now:
http://code.google.com/p/pycpuid/source/browse/trunk/linux/sched/sched_int.i
Example code:
http://
On Tue, Mar 25, 2008 at 1:55 PM, stephane eranian <[EMAIL PROTECTED]>
wrote:
> Arun,
>
> The Python bindings patch has now being applied to CVS.
> Please verify that I got everything correctly as I encountered
> some conflicts while merging your latest patches.
>
Hi Stephane,
Looks like this li
On Tue, Mar 25, 2008 at 1:55 PM, stephane eranian <[EMAIL PROTECTED]>
wrote:
> Arun,
>
> The Python bindings patch has now being applied to CVS.
> Please verify that I got everything correctly as I encountered
> some conflicts while merging your latest patches.
Looks good to me. Thanks.
At som
vent() for common event types,
which return strings.
-Arun
Fix some issues with examples.
* We only support self profiling. Not other threads
* The user needs to specify events to monitor
Signed-off-by: Arun Sharma <[EMAIL PROTECTED]>
--- a/python/self.py Tue Mar 25 07:33:50 2008 +
+++
On Mon, Mar 24, 2008 at 8:21 PM, Arun Sharma <[EMAIL PROTECTED]> wrote:
>
> Stephane, your changes look good to me. But I'm almost done
> refactoring the examples roughly as follows:
Updated patch attached. I think I captured most of your comments in
this patch - but I
On Mon, Mar 24, 2008 at 5:04 PM, stephane eranian
<[EMAIL PROTECTED]> wrote:
> Hello Arun,
>
> Thanks for the patch.
>
> I made made some changes to tthe examples. They both had problems
> reading the 2 events. See the attached updated versions.
>
Stephane, your changes look good to me. But I'm
hon related patches I've posted to the
list so far.
-Arun
PS: The strings os_err_t, pfm_err_t don't matter. We could rename them
to something else if you like.
Python bindings for libpfm.
Signed-off-by: Arun Sharma <[EMAIL PROTECTED]>
--- a/include/perfmon/perfmon.h Fri Mar 21
On Fri, Mar 21, 2008 at 2:58 PM, stephane eranian <[EMAIL PROTECTED]>
wrote:
>
> > http://www.swig.org/Doc1.3/Python.html#Python_nn44
> >
> This one looks easy to do for the syscall calls, only 12 of them.
That's 12 syscalls + 29 library calls + future API changes.
> But we'd have to write one
On Fri, Mar 21, 2008 at 2:45 PM, Arun Sharma <[EMAIL PROTECTED]> wrote:
> On Fri, Mar 21, 2008 at 2:19 PM, stephane eranian <[EMAIL PROTECTED]>
> wrote:
>
> > Arun,
> >
> > Not all functions are syscalls here. only the first 12. The others are
> > ju
on. The other method is
%exception:
http://www.swig.org/Doc1.3/Python.html#Python_nn44
But we'd have to write one exception block per function.
Also, my earlier patch breaks self.py and sys.py. This incremental patch
should fix it.
-Arun
Some functions can return non-zero values on succe
Hi Stephane,
The attached patch converts libpfm errors into python exceptions.
Please review.
-Arun
On Thu, Mar 20, 2008 at 3:54 PM, Arun Sharma <[EMAIL PROTECTED]> wrote:
> Hi Stephane,
>
> Here's an initial version of my python bindings patch.
>
> The patch depends
or perfmon/libpfm.
Signed-off-by: Arun Sharma <[EMAIL PROTECTED]>
--- libpfm-old/python/Makefile 1970-01-01 00:00:00.0 +
+++ libpfm-py.hg/python/Makefile 2008-03-20 22:27:10.796894000 +
@@ -0,0 +1,9 @@
+all:
+ ./setup.py build
+
+install:
+ ./setup.py install
+
+clean:
46 matches
Mail list logo
