Add Script for Dissecting QEMU Execution
v3 cover Add Script for Dissecting QEMU Execution
Hi,
This series adds the dissect.py script which breaks down the execution of QEMU into three main phases:
code generation, JIT execution, and helpers execution.
It prints the number of instructions executed by QEMU in each of these three phases, plus the total number of executed instructions.
To learn more about how the script works and for further usage instructions, please check the "Dissecting QEMU Into Three Main Parts" report posted as part of the "TCG Continuous Benchmarking" GSoC project.
Report link:
https://lists.nongnu.org/archive/html/qemu-devel/2020-06/msg09441.html
Best regards,
Ahmed Karaman
v2->v3:
- Fix a misalignment in a comment line.
- Use tempfile.TemporaryDirectory() for handling intermediate files.
Ahmed Karaman (1):
scripts/performance: Add dissect.py script
scripts/performance/dissect.py | 166 +++++++++++++++++++++++++++++++++ 1 file changed, 166 insertions(+) create mode 100755 scripts/performance/dissect.py
Signature · 1 line
2.17.1
PATCH v3 1/1 scripts/performance: Add dissect.py script
Python script that dissects QEMU execution into three main phases:
code generation, JIT execution and helpers execution.
Syntax:dissect.py [-h] -- <qemu executable> [<qemu executable options>] \
<target executable> [<target executable options>]
[-h] - Print the script arguments help message.
Example of usage:dissect.py -- qemu-arm coulomb_double-arm
Example output:
Total Instructions: 4,702,865,362
Code Generation: 115,819,309 2.463% JIT Execution: 1,081,980,528 23.007% Helpers: 3,505,065,525 74.530%
--- scripts/performance/dissect.py | 166 +++++++++++++++++++++++++++++++++ 1 file changed, 166 insertions(+) create mode 100755 scripts/performance/dissect.py
scripts/performance/dissect.py new file +166 -0
@@ -0,0 +1,166 @@ +#!/usr/bin/env python3 + +# Print the percentage of instructions spent in each phase of QEMU +# execution. +# +# Syntax: +# dissect.py [-h] -- <qemu executable> [<qemu executable options>] \ +# <target executable> [<target executable options>] +# +# [-h] - Print the script arguments help message. +# +# Example of usage: +# dissect.py -- qemu-arm coulomb_double-arm +# +# This file is a part of the project "TCG Continuous Benchmarking". +# +# Copyright (C) 2020 Ahmed Karaman <[email protected]> +# Copyright (C) 2020 Aleksandar Markovic <[email protected]> +# +# 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 +import tempfile + + +def get_JIT_line(callgrind_data): + """ + Search for the first instance of the JIT call in + the callgrind_annotate output when ran using --tree=caller + This is equivalent to the self number of instructions of JIT. + + Parameters: + callgrind_data (list): callgrind_annotate output + + Returns: + (int): Line number + """ + line = -1 + for i in range(len(callgrind_data)): + if callgrind_data[i].strip('\n') and \ + callgrind_data[i].split()[-1] == "[???]": + line = i + break + if line == -1: + sys.exit("Couldn't locate the JIT call ... Exiting.") + return line + + +def main(): + # Parse the command line arguments + parser = argparse.ArgumentParser( + usage='dissect.py [-h] -- ' + '<qemu executable> [<qemu executable options>] ' + '<target executable> [<target executable options>]') + + parser.add_argument('command', type=str, nargs='+', help=argparse.SUPPRESS) + + args = parser.parse_args() + + # Extract the needed variables from the args + command = args.command + + # Insure that valgrind is installed + check_valgrind = subprocess.run( + ["which", "valgrind"], stdout=subprocess.DEVNULL) + if check_valgrind.returncode: + sys.exit("Please install valgrind before running the script.") + + # Save all intermediate files in a temporary directory + with tempfile.TemporaryDirectory() as tmpdirname: + # callgrind output file path + data_path = os.path.join(tmpdirname, "callgrind.data") + # callgrind_annotate output file path + annotate_out_path = os.path.join(tmpdirname, "callgrind_annotate.out") + + # Run callgrind + callgrind = subprocess.run((["valgrind", + "--tool=callgrind", + "--callgrind-out-file=" + data_path] + + command), + stdout=subprocess.DEVNULL, + stderr=subprocess.PIPE) + if callgrind.returncode: + sys.exit(callgrind.stderr.decode("utf-8")) + + # Save callgrind_annotate output + with open(annotate_out_path, "w") as output: + callgrind_annotate = subprocess.run( + ["callgrind_annotate", data_path, "--tree=caller"], + stdout=output, + stderr=subprocess.PIPE) + if callgrind_annotate.returncode: + sys.exit(callgrind_annotate.stderr.decode("utf-8")) + + # Read the callgrind_annotate output to callgrind_data[] + callgrind_data = [] + with open(annotate_out_path, 'r') as data: + callgrind_data = data.readlines() + + # Line number with the total number of instructions + total_instructions_line_number = 20 + # Get the total number of instructions + total_instructions_line_data = \ + callgrind_data[total_instructions_line_number] + total_instructions = total_instructions_line_data.split()[0] + total_instructions = int(total_instructions.replace(',', '')) + + # Line number with the JIT self number of instructions + JIT_self_instructions_line_number = get_JIT_line(callgrind_data) + # Get the JIT self number of instructions + JIT_self_instructions_line_data = \ + callgrind_data[JIT_self_instructions_line_number] + JIT_self_instructions = JIT_self_instructions_line_data.split()[0] + JIT_self_instructions = int(JIT_self_instructions.replace(',', '')) + + # Line number with the JIT self + inclusive number of instructions + # It's the line above the first JIT call when running with --tree=caller + JIT_total_instructions_line_number = JIT_self_instructions_line_number-1 + # Get the JIT self + inclusive number of instructions + JIT_total_instructions_line_data = \ + callgrind_data[JIT_total_instructions_line_number] + JIT_total_instructions = JIT_total_instructions_line_data.split()[0] + JIT_total_instructions = int(JIT_total_instructions.replace(',', '')) + + # Calculate number of instructions in helpers and code generation + helpers_instructions = JIT_total_instructions-JIT_self_instructions + code_generation_instructions = total_instructions-JIT_total_instructions + + # Print results (Insert commas in large numbers) + # Print total number of instructions + print('{:<20}{:>20}\n'. + format("Total Instructions:", + format(total_instructions, ','))) + # Print code generation instructions and percentage + print('{:<20}{:>20}\t{:>6.3f}%'. + format("Code Generation:", + format(code_generation_instructions, ","), + (code_generation_instructions / total_instructions) * 100)) + # Print JIT instructions and percentage + print('{:<20}{:>20}\t{:>6.3f}%'. + format("JIT Execution:", + format(JIT_self_instructions, ","), + (JIT_self_instructions / total_instructions) * 100)) + # Print helpers instructions and percentage + print('{:<20}{:>20}\t{:>6.3f}%'. + format("Helpers:", + format(helpers_instructions, ","), + (helpers_instructions/total_instructions)*100)) + + +if __name__ == "__main__": + main()
Signature · 1 line
2.17.1
Re: [PATCH v3 1/1] scripts/performance: Add dissect.py script
Quoted 19 lines · Ahmed Karaman
Python script that dissects QEMU execution into three main phases:
code generation, JIT execution and helpers execution.Syntax:
dissect.py[-h] -- <qemu executable> [<qemu executable options>] \
<target executable> [<target executable options>][-h] - Print the script arguments help message.
Example of usage:
dissect.py-- qemu-arm coulomb_double-armExample output:
Total Instructions: 4,702,865,362Code Generation: 115,819,309 2.463% JIT Execution: 1,081,980,528 23.007% Helpers: 3,505,065,525 74.530%
Quoted 6 lines
Signed-off-by Ahmed Karaman <[email protected]>Reviewed-by Aleksandar Markovic <[email protected]>--- scripts/performance/dissect.py | 166 +++++++++++++++++++++++++++++++++ 1 file changed, 166 insertions(+) create mode 100755 scripts/performance/dissect.py
Thanks for the update Ahmed!
Philippe Mathieu-Daudé
AppliedRe: [PATCH v3 1/1] scripts/performance: Add dissect.py script
In reply to Ahmed KaramanQuoted 19 lines · Ahmed Karaman
Python script that dissects QEMU execution into three main phases:
code generation, JIT execution and helpers execution.Syntax:
dissect.py[-h] -- <qemu executable> [<qemu executable options>] \
<target executable> [<target executable options>][-h] - Print the script arguments help message.
Example of usage:
dissect.py-- qemu-arm coulomb_double-armExample output:
Total Instructions: 4,702,865,362Code Generation: 115,819,309 2.463% JIT Execution: 1,081,980,528 23.007% Helpers: 3,505,065,525 74.530%
Quoted 6 lines
Signed-off-by Ahmed Karaman <[email protected]>Reviewed-by Aleksandar Markovic <[email protected]>--- scripts/performance/dissect.py | 166 +++++++++++++++++++++++++++++++++ 1 file changed, 166 insertions(+) create mode 100755 scripts/performance/dissect.py
Thanks, applied to my python-next tree:
https://gitlab.com/philmd/qemu/commits/python-next