qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH v2 1/3] scripts/performance: Add topN_perf.py script


From: Aleksandar Markovic
Subject: Re: [PATCH v2 1/3] scripts/performance: Add topN_perf.py script
Date: Sat, 20 Jun 2020 11:59:40 +0200



петак, 19. јун 2020., Ahmed Karaman <ahmedkhaledkaraman@gmail.com> је написао/ла:
Python script that prints the top N most executed functions in QEMU
using perf.

Example Usage:

Don't use capitalization when not appropriate. This is better:

Example of usage:
 
topN_perf.py -n 20 -- /path/to/qemu program -program -flags

If '-n' is not specified, the default is 25.


The command line above is hardly example of usage. Furthermore, it is unclear of what you meant with "program -program". Other things are unclear too (what about qemu options?) The command line, as is now, is somewhere between a syntax description and an example of usage, but is neither a syntax description nor example of usage.

From the script, it looks that there is "-h" too. Not mentioned in commit message ar all, and it shouldbe.

I would suggest that you rework this section inthis way, for the sake of clarity:

Syntax: (

topN_perf.py [-h] [-n <number of displayed top functions >  -- <qemu executable> [<qemu executable options>] <target executable> [<target execurable options>]

-h - .........explain here
-n - .........explain here

Example of usage:

topN_perf.py -n 20 -- qemu-arm coulomb_double-arm

Example of output:


..... and here you continue with:



 No.  Percentage  Name                       Caller
----  ----------  -------------------------  -------------------------
   1      16.25%  float64_mul                qemu-x86_64
   2      12.01%  float64_sub                qemu-x86_64
   3      11.99%  float64_add                qemu-x86_64
   4       5.69%  helper_mulsd               qemu-x86_64
   5       4.68%  helper_addsd               qemu-x86_64
   6       4.43%  helper_lookup_tb_ptr       qemu-x86_64
   7       4.28%  helper_subsd               qemu-x86_64
   8       2.71%  f64_compare                qemu-x86_64
   9       2.71%  helper_ucomisd             qemu-x86_64
  10       1.04%  helper_pand_xmm            qemu-x86_64
  11       0.71%  float64_div                qemu-x86_64
  12       0.63%  helper_pxor_xmm            qemu-x86_64
  13       0.50%  0x00007f7b7004ef95         [JIT] tid 491
  14       0.50%  0x00007f7b70044e83         [JIT] tid 491
  15       0.36%  helper_por_xmm             qemu-x86_64
  16       0.32%  helper_cc_compute_all      qemu-x86_64
  17       0.30%  0x00007f7b700433f0         [JIT] tid 491
  18       0.30%  float64_compare_quiet      qemu-x86_64
  19       0.27%  soft_f64_addsub            qemu-x86_64
  20       0.26%  round_to_int               qemu-x86_64

Signed-off-by: Ahmed Karaman <ahmedkhaledkaraman@gmail.com>
---
 scripts/performance/topN_perf.py | 115 +++++++++++++++++++++++++++++++
 1 file changed, 115 insertions(+)
 create mode 100755 scripts/performance/topN_perf.py

diff --git a/scripts/performance/topN_perf.py b/scripts/performance/topN_perf.py
new file mode 100755
index 0000000000..53fa503d8a
--- /dev/null
+++ b/scripts/performance/topN_perf.py
@@ -0,0 +1,115 @@
+#!/usr/bin/env python3
+
+#  Print the top N most executed functions in QEMU using perf.
+#  Example Usage:
+#  topN_perf.py -n 20 -- /path/to/qemu program -program -flags
+#

The similat comments from commit message section apply here too.
 
+#   If '-n' is not specified, the default is 25.
+#
+#  This file is a part of the project "TCG Continuous Benchmarking".
+#
+#  Copyright (C) 2020  Ahmed Karaman <ahmedkhaledkaraman@gmail.com>
+#  Copyright (C) 2020  Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
+#
+#  This program is free software: you can redistribute it and/or modify
+#  it under the terms of the GNU General Public License as published by
+#  the Free Software Foundation, either version 2 of the License, or
+#  (at your option) any later version.
+#
+#  This program is distributed in the hope that it will be useful,
+#  but WITHOUT ANY WARRANTY; without even the implied warranty of
+#  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+#  GNU General Public License for more details.
+#
+#  You should have received a copy of the GNU General Public License
+#  along with this program. If not, see <https://www.gnu.org/licenses/>.
+
+import argparse
+import os
+import subprocess
+import sys
+
+
+# Parse the command line arguments
+parser = argparse.ArgumentParser(usage='topN_perf.py [-h] [-n TOP_FUNCTIONS] --'
+                                ' /path/to/qemu program -[flags PROGRAM_FLAGS]')
+

Same.

+parser.add_argument('-n', dest='top', type=int, default=25,
+                    help='Specify the number of top functions to print.')
+
+parser.add_argument('command', type=str, nargs='+', help=argparse.SUPPRESS)
+
+args = parser.parse_args()
+
+# Extract the needed variables from the args
+command = args.command
+top = args.top
+
+# Insure that perf is installed
+check_perf = subprocess.run(["which", "perf"], stdout=subprocess.DEVNULL)
+if check_perf.returncode:
+    sys.exit("Please install perf before running the script!")
+

OK. This is good.
 
+# Insure user has previllage to run perf
+check_previlage = subprocess.run(["perf", "stat", "ls", "/"],
+                              stdout=subprocess.DEVNULL, stderr=subprocess.PIPE)
+if check_previlage.returncode:
+    sys.exit(check_previlage.stderr.decode("utf-8") +
+             "\nOr alternatively, you can run the script with sudo privileges!")

I would avoid mixing stderr message and your message practically in one sentence. Better:

error: <text from stderr>
You must run the script with sudo pivilages, or, alternatively, set kernel Xxx option to....
 
+
+# Run perf record
+perf_record = subprocess.run((["perf", "record"] + command),
+                             stdout=subprocess.DEVNULL, stderr=subprocess.PIPE)
+if perf_record.returncode:
+    os.unlink('perf.data')
+    sys.exit(perf_record.stderr.decode("utf-8"))
+
+# Save perf report output to tmp.perf.data
+with open("tmp.perf.data", "w") as output:
+    perf_report = subprocess.run(
+        ["perf", "report", "--stdio"], stdout=output, stderr=subprocess.PIPE)
+    if perf_report.returncode:
+        os.unlink('perf.data')
+        output.close()
+        os.unlink('tmp.perf.data')
+        sys.exit(perf_report.stderr.decode("utf-8"))
+

I am really confused by using both perf.data and tmp.perf.data names. Why? They are in entirely different format. "tmp.perf.data should be perf.top-function.list, let's say, if I am not mistaken about its meaning and usage.

+# Read the reported data to functions[]
+functions = []
+with open("tmp.perf.data", "r") as data:
+    # Only read lines that are not comments (comments start with #)
+    # Only read lines that are not empty
+    functions = [line for line in data.readlines() if line and line[0]
+                 != '#' and line[0] != "\n"]
+
+# Limit the number of top functions to "top"
+number_of_top_functions = top if len(functions) > top else len(functions)
+
+# Store the data of the top functions in top_functions[]
+top_functions = functions[:number_of_top_functions]
+
+# Print information headers

# Print table header
 
+print('{:>4}  {:>10}  {:<30}  {}\n{}  {}  {}  {}'.format('No.',
+                                                         'Percentage',
+                                                         'Name',
+                                                         'Caller',
+                                                         '-' * 4,
+                                                         '-' * 10,
+                                                         '-' * 30,
+                                                         '-' * 25))
+
+
+# Print top N functions
+for (index, function) in enumerate(top_functions, start=1):
+    function_data = function.split()
+    function_percentage = function_data[0]
+    function_name = function_data[-1]
+    function_caller = ' '.join(function_data[2:-2])
+    print('{:>4}  {:>10}  {:<30}  {}'.format(index,
+                                             function_percentage,
+                                             function_name,
+                                             function_caller))
+
+# Remove intermediate files
+os.unlink('perf.data')
+os.unlink('tmp.perf.data')
-- 

Thanks,
Aleksandar.
 
2.17.1


reply via email to

[Prev in Thread] Current Thread [Next in Thread]