Skip to contentSkip to conversation

perf bench: Add "all" pseudo subsystem and "all" pseudo suite

PATCH Superseded Hitoshi Mitake 2 participants 6 messages Dec 12, 2009 – Dec 14, 2009 1 message not in the archive perf events

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

PATCH 1/1 perf bench: Add "all" pseudo subsystem and "all" pseudo suite

This patch adds new "all" pseudo subsystem and "all" pseudo suite. These are for testing all subsystem and its all suite, or all suite of one subsystem.

(This patch also contains a few trivial comment fixes for bench/* and output style fixes. I judged that there are no necessity to make them into individual patch.)

Example of use:

| % ./perf bench sched all                      # Test all suite of sched subsystem
| # Running sched/messaging benchmark...
| # 20 sender and receiver processes per group
| # 10 groups == 400 processes run
|
|      Total time: 0.414 [sec]
|
| # Running sched/pipe benchmark...
| # Extecuted 1000000 pipe operations between two tasks
|
|      Total time: 10.999 [sec]
|
|       10.999317 usecs/op
|           90914 ops/sec
|
| % ./perf bench all                            # Test all suite of all subsystem
| # Running sched/messaging benchmark...
| # 20 sender and receiver processes per group
| # 10 groups == 400 processes run
|
|      Total time: 0.420 [sec]
|
| # Running sched/pipe benchmark...
| # Extecuted 1000000 pipe operations between two tasks
|
|      Total time: 11.741 [sec]
|
|       11.741346 usecs/op
|           85169 ops/sec
|
| # Running mem/memcpy benchmark...
| # Copying 1MB Bytes from 0x7ff33e920010 to 0x7ff3401ae010 ...
|
|      808.407437 MB/Sec

Below is description of another issue.

As a matter of fact, I fixed output of "./perf bench all" a little. Because the raw output is this,

| # Running sched/messaging benchmark...
| # 20 sender and receiver processes per group
| # 10 groups == 400 processes run
|
|      Total time: 0.420 [sec]
|
| # Running sched/pipe benchmark...
| # Extecuted 1000000 pipe operations between two tasks
|
|      Total time: 11.741 [sec]
|
|       11.741346 usecs/op
|           85169 ops/sec
|
| # Running mem/memcpy benchmark...
| # Copying 1MB Bytes from 0x7ff33e920010 to 0x7ff3401ae010 ...
|
|      808.407437 MB/Sec
|
|
| # Running mem/memcpy benchmark...             # Second mem !?
| # Copying 1MB Bytes from 0x7ff33e920010 to 0x7ff3401ae010 ...
|
|      841.042893 MB/Sec

mem subsystem was executed twice!
Something is going wrong, and I tested with the modification of all_subsystem() function,

original:
static void all_subsystem(void)

{
	int i;
	for (i = 0; subsystems[i].suites; i++)
		all_suite(&subsystems[i]);
}

modified,
static void all_subsystem(void)

{
	int i;
	for (i = 0; subsystems[i].suites; i++) {
	        printf("start of loop:%d\n", i);
		all_suite(&subsystems[i]);
	        printf("end of loop:%d\n", i);
	}
}

The result is,

| start of loop:0
| # Running sched/messaging benchmark...
| # 20 sender and receiver processes per group
| # 10 groups == 400 processes run
|
|      Total time: 0.459 [sec]
|
| # Running sched/pipe benchmark...
| # Extecuted 1000000 pipe operations between two tasks
|
|      Total time: 12.264 [sec]
|
|   12.264828 usecs/op
|           81533 ops/sec
|
| end of loop:0
| start of loop:1
| # Running mem/memcpy benchmark...
| # Copying 1MB Bytes from 0x7f2c5e526010 to 0x7f2c5fdb4010 ...
|
|      846.023689 MB/Sec
|
| end of loop:1               # !!!
|                             # !!!
| end of loop:0               # !!! <- WHAT IS THIS!!?
| start of loop:1
| # Running mem/memcpy benchmark...
| # Copying 1MB Bytes from 0x7f2c5e526010 to 0x7f2c5fdb4010 ...
|
|      839.630563 MB/Sec
|
| end of loop:1

Something terrible happend :-( Is this bug of gcc?
I disassembled builtin-bench.o, but highly optimized assembly was hard to read... If you have some ideas to solve this problem,
or know issues similar to it, could you tell me?

My gcc's description:

| % gcc --version
| gcc (Debian 4.3.2-1.1) 4.3.2
| Copyright (C) 2008 Free Software Foundation, Inc.
| This is free software; see the source for copying conditions.  There is NO
| warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
| % gcc -v
| Using built-in specs.
| Target: x86_64-linux-gnu
| Configured with: ../src/configure -v --with-pkgversion='Debian 4.3.2-1.1' --with-bugurl=file:///usr/share/doc/gcc-4.3/README.Bugs --enable-languages=c,c++,fortran,objc,obj-c++ --prefix=/usr --enable-shared --with-system-zlib --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --enable-nls --with-gxx-include-dir=/usr/include/c++/4.3 --program-suffix=-4.3 --enable-clocale=gnu --enable-libstdcxx-debug --enable-objc-gc --enable-mpfr --enable-cld --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu
| Thread model: posix
| gcc version 4.3.2 (Debian 4.3.2-1.1)

But I think that this patch has no problem and is commitable. I'll work on this with gdb again...

Cc Peter Zijlstra <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Frederic Weisbecker <[email protected]>
---
 tools/perf/bench/sched-messaging.c |    2 +-
 tools/perf/bench/sched-pipe.c      |    2 +-
 tools/perf/builtin-bench.c         |   57 +++++++++++++++++++++++++++++++++--
 3 files changed, 55 insertions(+), 6 deletions(-)
tools/perf/bench/sched-messaging.c +1 -1
index 605a2a9..e249eb2 100644
@@ -1,6 +1,6 @@
 /*
  *
- * builtin-bench-messaging.c
+ * sched-messaging.c
  *
  * messaging: Benchmark for scheduler and IPC mechanisms
  *
tools/perf/bench/sched-pipe.c +1 -1
index 238185f..260dca2 100644
@@ -1,6 +1,6 @@
 /*
  *
- * builtin-bench-pipe.c
+ * sched-pipe.c
  *
  * pipe: Benchmark for pipe()
  *
tools/perf/builtin-bench.c +53 -4

116 lines not on this page · raw message

Signature · 1 line
1.6.5.2
#2 lore raw
Quoted 6 lines · Hitoshi Mitake
| # Running mem/memcpy benchmark...             # Second mem !?
| # Copying 1MB Bytes from 0x7ff33e920010 to 0x7ff3401ae010 ...
|
|      841.042893 MB/Sec

mem subsystem was executed twice!

i saw something similar some time ago but thought it got fixed. Maybe some fork() related trouble?

Ingo

#3 lore raw

PATCH v2 1/1 perf bench: Add "all" pseudo subsystem and "all" pseudo suite

Sorry, you are right, Ingo.
This was my careless miss. sched-pipe.c returned twice. I wronged gcc by a false charge, sorry :(

This is version 2.

This patch adds new "all" pseudo subsystem and "all" pseudo suite. These are for testing all subsystem and its all suite, or all suite of one subsystem.

(This patch also contains a few trivial comment fixes for bench/* and output style fixes. I judged that there are no necessity to make them into individual patch.)

Example of use:

| % ./perf bench sched all                      # Test all suite of sched subsystem
| # Running sched/messaging benchmark...
| # 20 sender and receiver processes per group
| # 10 groups == 400 processes run
|
|      Total time: 0.414 [sec]
|
| # Running sched/pipe benchmark...
| # Extecuted 1000000 pipe operations between two tasks
|
|      Total time: 10.999 [sec]
|
|       10.999317 usecs/op
|           90914 ops/sec
|
| % ./perf bench all                            # Test all suite of all subsystem
| # Running sched/messaging benchmark...
| # 20 sender and receiver processes per group
| # 10 groups == 400 processes run
|
|      Total time: 0.420 [sec]
|
| # Running sched/pipe benchmark...
| # Extecuted 1000000 pipe operations between two tasks
|
|      Total time: 11.741 [sec]
|
|       11.741346 usecs/op
|           85169 ops/sec
|
| # Running mem/memcpy benchmark...
| # Copying 1MB Bytes from 0x7ff33e920010 to 0x7ff3401ae010 ...
|
|      808.407437 MB/Sec
Cc Peter Zijlstra <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Frederic Weisbecker <[email protected]>
---
 tools/perf/bench/sched-messaging.c |    2 +-
 tools/perf/bench/sched-pipe.c      |    5 ++-
 tools/perf/builtin-bench.c         |   57 +++++++++++++++++++++++++++++++++--
 3 files changed, 57 insertions(+), 7 deletions(-)
tools/perf/bench/sched-messaging.c +1 -1
index 605a2a9..e249eb2 100644
@@ -1,6 +1,6 @@
 /*
  *
- * builtin-bench-messaging.c
+ * sched-messaging.c
  *
  * messaging: Benchmark for scheduler and IPC mechanisms
  *
tools/perf/bench/sched-pipe.c +3 -2
index 238185f..2354f38 100644
@@ -1,6 +1,6 @@
 /*
  *
- * builtin-bench-pipe.c
+ * sched-pipe.c
  *
  * pipe: Benchmark for pipe()
  *
@@ -87,7 +87,8 @@ int bench_sched_pipe(int argc, const char **argv,
 	if (pid) {
 		retpid = waitpid(pid, &wait_stat, 0);
 		assert((retpid == pid) && WIFEXITED(wait_stat));
-		return 0;
+	} else {
+		exit(0);
 	}
 
 	switch (bench_format) {
tools/perf/builtin-bench.c +53 -4

116 lines not on this page · raw message

Signature · 1 line
1.6.5.2
#4 lore raw
Quoted 3 lines · Hitoshi Mitake

Sorry, you are right, Ingo.
This was my careless miss. sched-pipe.c returned twice. I wronged gcc by a false charge, sorry :(

Dont worry about it too much - GCC is guilty enough times :)

Quoted 1 line

This is version 2.

Great patch, applied, thanks!

Ingo

#5 lore raw
Quoted 8 lines · Ingo Molnar
Quoted 3 lines · Hitoshi Mitake

Sorry, you are right, Ingo.
This was my careless miss. sched-pipe.c returned twice. I wronged gcc by a false charge, sorry :(

Dont worry about it too much - GCC is guilty enough times :)

Oh +)

Quoted 4 lines
Quoted 2 lines

This is version 2.

Great patch, applied, thanks!

I received email from tip-bot, thanks!
Hitoshi

#6 lore raw
In reply to Hitoshi Mitake
Commit-ID:  2044279d1e07c90edd64324f09c7adf1acfc42e5
Gitweb:     http://git.kernel.org/tip/2044279d1e07c90edd64324f09c7adf1acfc42e5
Author:     Hitoshi Mitake <[email protected]>
AuthorDate: Sun, 13 Dec 2009 17:01:59 +0900
Committer:  Ingo Molnar <[email protected]>
CommitDate: Mon, 14 Dec 2009 08:51:19 +0100

perf bench: Add "all" pseudo subsystem and "all" pseudo suite

This patch adds a new "all" pseudo subsystem and an "all" pseudo suite. These are for testing all subsystem and its all suite, or all suite of one subsystem.

(This patch also contains a few trivial comment fixes for bench/* and output style fixes. I judged that there are no necessity to make them into individual patch.)

Example of use:

| % ./perf bench sched all                      # Test all suites of sched subsystem
| # Running sched/messaging benchmark...
| # 20 sender and receiver processes per group
| # 10 groups == 400 processes run
|
|      Total time: 0.414 [sec]
|
| # Running sched/pipe benchmark...
| # Extecuted 1000000 pipe operations between two tasks
|
|      Total time: 10.999 [sec]
|
|       10.999317 usecs/op
|           90914 ops/sec
|
| % ./perf bench all                            # Test all suites of all subsystems
| # Running sched/messaging benchmark...
| # 20 sender and receiver processes per group
| # 10 groups == 400 processes run
|
|      Total time: 0.420 [sec]
|
| # Running sched/pipe benchmark...
| # Extecuted 1000000 pipe operations between two tasks
|
|      Total time: 11.741 [sec]
|
|       11.741346 usecs/op
|           85169 ops/sec
|
| # Running mem/memcpy benchmark...
| # Copying 1MB Bytes from 0x7ff33e920010 to 0x7ff3401ae010 ...
|
|      808.407437 MB/Sec
Cc Peter Zijlstra <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Frederic Weisbecker <[email protected]>

LKML-Reference: <[email protected]>

---
 tools/perf/bench/sched-messaging.c |    2 +-
 tools/perf/bench/sched-pipe.c      |    5 ++-
 tools/perf/builtin-bench.c         |   57 +++++++++++++++++++++++++++++++++--
 3 files changed, 57 insertions(+), 7 deletions(-)
tools/perf/bench/sched-messaging.c +1 -1
index 605a2a9..e249eb2 100644
@@ -1,6 +1,6 @@
 /*
  *
- * builtin-bench-messaging.c
+ * sched-messaging.c
  *
  * messaging: Benchmark for scheduler and IPC mechanisms
  *
tools/perf/bench/sched-pipe.c +3 -2
index 238185f..2354f38 100644
@@ -1,6 +1,6 @@
 /*
  *
- * builtin-bench-pipe.c
+ * sched-pipe.c
  *
  * pipe: Benchmark for pipe()
  *
@@ -87,7 +87,8 @@ int bench_sched_pipe(int argc, const char **argv,
 	if (pid) {
 		retpid = waitpid(pid, &wait_stat, 0);
 		assert((retpid == pid) && WIFEXITED(wait_stat));
-		return 0;
+	} else {
+		exit(0);
 	}
 
 	switch (bench_format) {
tools/perf/builtin-bench.c +53 -4

116 lines not on this page · raw message