Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
From:   Ivan Babrou <ivan@cloudflare.com>
Date:   Sat, 30 Nov 2019 13:20:08 -0800
Message-ID: <CABWYdi32BXz4W0VyYR2hV5A58rcegCaF4Jk__QmyiTJsr7=b3A@mail.gmail.com>
Subject: Misaligned output of perf stat --topdown
To:     linux-kernel <linux-kernel@vger.kernel.org>
Cc:     kernel-team <kernel-team@cloudflare.com>,
        Arnaldo Carvalho de Melo <acme@kernel.org>,
        Jiri Olsa <jolsa@redhat.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hey,

I'm observing misaligned output of perf stat --topdown:

$ sudo perf stat --topdown --cpu 1 -a find /state > /dev/null

 Performance counter stats for 'system wide':

                                    retiring      bad speculation
 frontend bound        backend bound
S0-D0-C1           1                30.4%                        13.4%
                       38.0%                18.2%

$ sudo perf stat --topdown --cpu 1 -a find /state > /dev/null

 Performance counter stats for 'system wide':

                                    retiring      bad speculation
 frontend bound        backend bound
S0-D0-C1           1                27.5%                 6.5%
               24.1%                        41.9%

$ sudo perf stat --topdown --cpu 1 -a find /state > /dev/null

 Performance counter stats for 'system wide':

                                    retiring      bad speculation
 frontend bound        backend bound
S0-D0-C1           1                29.6%                        12.3%
                       23.7%                        34.4%

It may be hard to notice in narrow plaintext email, so here's a gist:

* https://gist.github.com/bobrik/ea5ddd8eb1629c350c898093f39ac7ee

I think there are two issues:

* Sometimes output is misaligned, which may or may not depend on the
length of the printed value (<10.0% is misaligned, >=10.0% is ok)
* Values are never really aligned to headers (except for maybe the first column)

This is on 5.4.0.