- 12 Jul, 2013 40 commits
-
-
Jiri Olsa authored
Adding 'struct perf_hpp_fmt' into hpp callbacks, so commands can access their private data. It'll be handy for diff command in future to be able to access file related data for each column. Signed-off-by:
Jiri Olsa <jolsa@redhat.com> Reviewed-by:
Namhyung Kim <namhyung@kernel.org> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/n/tip-7vy2m18574b1bicoljn8e9lw@git.kernel.orgSigned-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com>
-
Greg Price authored
For example, in an application with an expensive function implemented with deeply nested recursive calls, the default call-graph presentation is dominated by the different callchains within that function. By ignoring these callees, we can collect the callchains leading into the function and compactly identify what to blame for expensive calls. For example, in this report the callers of garbage_collect() are scattered across the tree: $ perf report -d ruby 2>- | grep -m10 ^[^#]*[a-z] 22.03% ruby [.] gc_mark --- gc_mark |--59.40%-- mark_keyvalue | st_foreach | gc_mark_children | |--99.75%-- rb_gc_mark | | rb_vm_mark | | gc_mark_children | | gc_marks | | |--99.00%-- garbage_collect If we ignore the callees of garbage_collect(), its callers are coalesced: $ perf report --ignore-callees garbage_collect -d ruby 2>- | grep -m10 ^[^#]*[a-z] 72.92% ruby [.] garbage_collect --- garbage_collect vm_xmalloc |--47.08%-- ruby_xmalloc | st_insert2 | rb_hash_aset | |--98.45%-- features_index_add | | rb_provide_feature | | rb_require_safe | | vm_call_method Signed-off-by:Greg Price <price@mit.edu> Tested-by:
-
Andi Kleen authored
List the kernel supplied pmu event aliases in perf list It's better when the users can actually see them. Signed-off-by:
Andi Kleen <ak@linux.intel.com> Link: http://lkml.kernel.org/r/1366480949-32292-2-git-send-email-andi@firstfloor.orgSigned-off-by:
-
Andi Kleen authored
When an event fails to parse and it's not in a new style format, try to parse it again as a cpu event. This allows to use sysfs exported events directly without //, so you can use perf record -e mem-loads ... instead of perf record -e cpu/mem-loads/ Signed-off-by:
-
Adrian Hunter authored
As evident from 'machine__process_fork_event()' and 'machine__process_exit_event()' the 'pid' member of struct thread is actually the tid. Rename 'pid' to 'tid' in struct thread accordingly. Signed-off-by:
Adrian Hunter <adrian.hunter@intel.com> Acked-by:
David Ahern <dsahern@gmail.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Namhyung Kim <namhyung@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/1372944040-32690-13-git-send-email-adrian.hunter@intel.comSigned-off-by:
-
Adrian Hunter authored
The 'size' variable includes the header so must be at least 'sizeof(struct perf_event_header)'. Error out immediately if that is not the case. Also don't byte-swap the header until it is actually "fetched" from the mmap region. Signed-off-by:
-
Adrian Hunter authored
The same lines of code are used in three places. Make it a new function '__perf_evlist__munmap()'. Signed-off-by:
-
Adrian Hunter authored
The name parameter is constant, declare it so. Signed-off-by:
-
Adrian Hunter authored
By default, perf inject should "repipe" all events including 'finished_round'. Signed-off-by:
-
Adrian Hunter authored
The 'inject' command expects to get a reference to 'struct perf_inject' from its 'tool' member. For that to work, 'tool' needs to be a parameter of all tool callbacks. Make it so. Signed-off-by:
-
Adrian Hunter authored
The 'machine' parameter is unused in 'perf_event__repipe_synth()' and some callers pass NULL anyway. So remove it. Signed-off-by:
-
Namhyung Kim authored
The 'rules' means that every second line of the tree view has a shaded background, which makes it easier to see which cell belongs to which row in the tree view. It can be useful for a tree view that has a lot of rows. Reviewed-by:
Pekka Enberg <penberg@kernel.org> Signed-off-by:
-
Namhyung Kim authored
If callchain is displayed, add "row-activated" signal handler for handling double-click or pressing ENTER key action. Reviewed-by:
-
Namhyung Kim authored
Sometimes it's annoying to see when some symbols have very wierd long names. So it might be a good idea to make column size changable. Reviewed-by:
-
Namhyung Kim authored
Display callchain percent value in the overhead column. Signed-off-by:
-
Namhyung Kim authored
Display callchain information in the symbol column. It's only enabled when recorded with -g and has symbol sort key. Reviewed-by:
-
Namhyung Kim authored
The GtkTreeStore can save items in a tree-like way. This is a preparation for supporting callgraphs in the hist browser. Reviewed-by:
-
Namhyung Kim authored
For some reason it consumed quite amount of compile time when declared as local variable, and it disappeared when moved out of the function. Moving other variables/tables didn't help. On my system this single-file-change build time reduced from 11s to 3s. Signed-off-by:
-
Namhyung Kim authored
They're internals of ftrace ring-buffer and not used in perf code directly. As it now resides on libtraceevent/kbuffer.h, just get rid of them. Signed-off-by:
-
Namhyung Kim authored
It's useless to call the read_trace_init() function at this time as we don't need a returned pevent and it makes me confusing. :) Signed-off-by:
-
Namhyung Kim authored
It's confusing to have same name for two difference functions which does something opposite way. Since what they do in this file is read *AND* writing some of tracing metadata files, rename them to record_*() looks better to me. Signed-off-by:
-
Namhyung Kim authored
It's the only user of the variable, so move it. Signed-off-by:
-
Namhyung Kim authored
They're not used anywhere and same information is kept in a pevent already. So let's get rid of them. Signed-off-by:
-
Namhyung Kim authored
The header_page file describes the format of the ring buffer page which is used by ftrace (not perf). And size of "commit" field (I guess it's older name was 'size') represents the real size of long type used for kernel. So update the pevent's long size. Signed-off-by:
-
Namhyung Kim authored
It seems perf does not parse header_event file so we can skip it as we do for header_page file. Signed-off-by:
-
Namhyung Kim authored
They're not used anywhere, just make them local variables. Signed-off-by:
-
Namhyung Kim authored
Save size of long type of system to struct pevent. Since original static variable was not used anywhere, just get rid of it. Signed-off-by:
-
Namhyung Kim authored
We now have page_size field in struct pevent, save the actual size of the system. Signed-off-by:
-
Namhyung Kim authored
kbuffer code is for parsing ftrace ring-buffer binary data and used for trace-cmd. Move the code here in order to be used more widely. Signed-off-by:
Steven Rostedt <rostedt@goodmis.org> Acked-by:
-
Namhyung Kim authored
The page size of traced system can be different than current system's because the recorded data file might be analyzed in a different machine. In this case we should use original page size of traced system when accessing the data file, so this information needs to be saved. Signed-off-by:
-
Namhyung Kim authored
Sometimes it'd be useful if existing trace_seq can be reused. But currently it's impossible since there's no API to reset the trace_seq. Let's add trace_seq_reset() for this case. Signed-off-by:
-
Namhyung Kim authored
If pevent_register_event_handler() received a string literal as @sys_name or @event_name parameter, it emitted a warning about const qualifier removal. Since they're not modified in the function we can make it have const qualifier. Signed-off-by:
-
Namhyung Kim authored
It's came from trace-cmd's kernelshark which is not a part of libtraceevent. Signed-off-by:
-
Namhyung Kim authored
The html_install, img_install, install_plugin and install_python are unused in the Makefile. Get rid of them. Signed-off-by:
Steven Rostedt <rostedt@goodmig.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1370323231-14022-2-git-send-email-namhyung@kernel.orgSigned-off-by:
-
David Ahern authored
List heads are currently allocated way down the function chain in __add_event and add_tracepoint and then freed when the scanner code calls parse_events_update_lists. Be more explicit with where memory is allocated and who should free it. With this patch the list_head is allocated in the scanner code and freed when the scanner code calls parse_events_update_lists. Signed-off-by:
-
David Ahern authored
No need to malloc the memory for it. Signed-off-by:
-
David Ahern authored
Function should only be freeing the entries in the list in case of failure, as those were allocated there, not the list_head itself. Signed-off-by:
-
David Ahern authored
Fixes valgrind complaint: ==1870== Syscall param write(buf) points to uninitialised byte(s) ==1870== at 0x4E3F5B0: __write_nocancel (in /lib64/libpthread-2.14.90.so) ==1870== by 0x449D7C: perf_evlist__start_workload (evlist.c:846) ==1870== by 0x427BC1: cmd_record (builtin-record.c:561) ==1870== by 0x419D72: run_builtin (perf.c:319) ==1870== by 0x4195F2: main (perf.c:376) ==1870== Address 0x7feffcdd7 is on thread 1's stack Signed-off-by:
-
Runzhen Wang authored
Power7 supports over 530 different perf events but only a small subset of these can be specified by name, for the remaining events, we must specify them by their raw code: perf stat -e r2003c <application> This patch makes all the POWER7 events available in sysfs. So we can instead specify these as: perf stat -e 'cpu/PM_CMPLU_STALL_DFU/' <application> where PM_CMPLU_STALL_DFU is the r2003c in previous example. Before this patch is applied, the size of power7-pmu.o is: $ size arch/powerpc/perf/power7-pmu.o text data bss dec hex filename 3073 2720 0 5793 16a1 arch/powerpc/perf/power7-pmu.o and after the patch is applied, it is: $ size arch/powerpc/perf/power7-pmu.o text data bss dec hex filename 15950 31112 0 47062 b7d6 arch/powerpc/perf/power7-pmu.o For the run time overhead, I use two scripts, one is "event_name.sh", which contains 50 event names, it looks like: # ./perf record -e 'cpu/PM_CMPLU_STALL_DFU/' -e ..... /bin/sleep 1 the other one is named "event_code.sh" which use corresponding events raw code instead of events names, it looks like: # ./perf record -e r2003c -e ...... /bin/sleep 1 below is the result. Using events name: [root@localhost perf]# time ./event_name.sh [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.002 MB perf.data (~102 samples) ] real 0m1.192s user 0m0.028s sys 0m0.106s Using events raw code: [root@localhost perf]# time ./event_code.sh [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.003 MB perf.data (~112 samples) ] real 0m1.198s user 0m0.028s sys 0m0.105s Signed-off-by:Runzhen Wang <runzhen@linux.vnet.ibm.com> Acked-by:
Michael Ellerman <michael@ellerman.id.au> Cc: icycoder@gmail.com Cc: linuxppc-dev@lists.ozlabs.org Cc: Michael Ellerman <michael@ellerman.id.au> Cc: Paul Mackerras <paulus@samba.org> Cc: Runzhen Wang <runzhew@clemson.edu> Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com> Cc: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com> Link: http://lkml.kernel.org/r/1372407297-6996-3-git-send-email-runzhen@linux.vnet.ibm.comSigned-off-by:
-
Jiri Olsa authored
There's no point of having out_delete label with perf_session__delete call within __cmd_report function, because it's called at the end of the cmd_report function. The speed up due to commenting out the perf_session__delete at the end does not seem relevant anymore. Measured speedup for ~1GB data file with 222466 FORKS events is around 0.5%. $ perf report -i perf.data.delete -P perf_session__delete -s parent + 99.51% [other] + 0.49% perf_session__delete Signed-off-by:
-
