When preparing for the Amazon paperback version, images need to be monochrome, otherwise, the quality will not be great.

I'm running win10 wsl 1.0 (windows subsystem for linux). I hope it run on only commandline because I don't have X windows installed (too big!) I've also installed MiKTeX using ubuntu 20.04 guide and finished up with shared set up. This is the portion of the error when building the book:

./chapters/13-Glossary/13-0 Glossary.md'
./chapters/14-Appendix/Appendix-A.md
./chapters/14-Appendix/Appendix-B.md
[WARNING] Note with key '12' defined at line 29 column 1 but not used.
[WARNING] Note with key '4' defined at line 30 column 1 but not used.

pandoc-fignos: Bad reference: @fig:Averages1.

pandoc-fignos: Bad reference: @fig:Averages2.

pandoc-fignos: Bad reference: @fig:BB_default.

pandoc-fignos: Bad reference: @fig:BB_better.

pandoc-fignos: Bad reference: @fig:FuncGroup_default.

pandoc-fignos: Bad reference: @fig:FuncGroup_better.

pandoc-fignos: Bad reference: @fig:MT_Scaling.

pandoc-fignos: Bad reference: @fig:MT_cycles.
pdflatex: error while loading shared libraries: libQt5Core.so.5: cannot open shared object file: No such file or directory

$

what is libQt5Core.so.5 ? i've also installed pdflatex but no change in the error :(
any idea what i'm missing?

Currently, we have in Chapter 7. CPU Features For Performance Analysis:

7-1 Top-Down Microarchitecture Analysis (TMA)
7-2 Last Branch Record (LBR)
7-3 Precise Event Based Sampling (PEBS)
7-4 Intel Processor Traces (PT)

Not all of this stuff is relevant to the broad audience, but some people would still like to know this.
So maybe the appendix would be a better place for some of the sections.
For example, Intel PT is a good candidate.

Because __rdtsc is not serializable CPU is free to reorder the execution of this instruction which could lead to wrong latency measurements results.

Instead the serializable version __rdtscp should be used.

I was really surprised not seeing the Ulrich Drepper's book "What every programmer should know about memory" in the References chapter. It was written in 2007 but the information explained there is still actual.

I believe it worth to add it to the References list for next reading.

Following is a link to online book:
https://people.freebsd.org/~lstewart/articles/cpumemory.pdf

Looks like on page 96 there is a typo. The text there references to dependency introduced inside the loop. The loop starts at line 5, really. But the dependency is on the line 6.

The error log:

Traceback (most recent call last):
  File "D:\workspace\perf-book-main\export_book.py", line 74, in <module>
    lines = f.readlines()
UnicodeDecodeError: 'gbk' codec can't decode byte 0xa2 in position 3820: illegal multibyte sequence

the reason is that file encoding is not utf-8. I add the param of encoding="utf-8", it build suceed。

texFile = 'book.tex'
editTexFile = 'book_edit.tex'

with open(texFile, 'r', encoding="utf-8") as f:
    lines = f.readlines()

with open(editTexFile, 'w', encoding="utf-8") as g:
    chapterRefs = []
    prev = ""

Section 3.7 starts with "in contrast to the MIMD approach described in the previous section"

But the previous section is about virtual memory. Abbreviation MIMD is also never explained (although it can be inferred from the context).

There is formating problem on page 239. Book version release Q1.2024

Thank you for the greate book

On page 37 on the bottom the the text is cut:

In the digital version of the book the text is printed correctly.

Sorry to trobule you, but where is the book.tex

Nice book!

I have pandoc 2.16.2 and pandoc-crossref (built with pandoc 2.16) installed, however, I got the error LaTeX Error: Environment pandoccrossrefsubfigures undefined.

I ignored all the error messages and finally got a pdf book with 176 pages. I don't know if it is a complete book.

Here is the detailed message

Underfull \hbox (badness 10000) in paragraph at lines 1442--1442
[][][]\T1/lmr/m/n/9 (+20) Book ``Work-load Mod-el-ing for Com-puter Sys-tems Pe
r-for-mance Eval-u-a-tion'' -
[23 <./img//1/CompDist2.png>]

! LaTeX Error: Environment pandoccrossrefsubfigures undefined.

See the LaTeX manual or LaTeX Companion for explanation.
Type  H <return>  for immediate help.
 ...

l.1470 \begin{pandoccrossrefsubfigures}

Env. info

macOS Big Sur

$ pandoc --version
pandoc 2.16.2
Compiled with pandoc-types 1.22.1, texmath 0.12.3.3, skylighting 0.12.1,
citeproc 0.6, ipynb 0.1.0.2
User data directory: /Users/gavin/.local/share/pandoc
Copyright (C) 2006-2021 John MacFarlane. Web:  https://pandoc.org
This is free software; see the source for copying conditions. There is no
warranty, not even for merchantability or fitness for a particular purpose.

$ pandoc-crossref --version
pandoc-crossref v0.3.12.1 git commit 0a47293d7d47aa763e82204e13deabc8f1a041c4 (HEAD) built with Pandoc v2.16.2, pandoc-types v1.22.1 and GHC 8.10.7

$ pdflatex --version
pdfTeX 3.14159265-2.6-1.40.19 (TeX Live 2018)
kpathsea version 6.3.0
Copyright 2018 Han The Thanh (pdfTeX) et al.
There is NO warranty.  Redistribution of this software is
covered by the terms of both the pdfTeX copyright and
the Lesser GNU General Public License.
For more information about these matters, see the file
named COPYING and the pdfTeX source.
Primary author of pdfTeX: Han The Thanh (pdfTeX) et al.
Compiled with libpng 1.6.34; using libpng 1.6.34
Compiled with zlib 1.2.11; using zlib 1.2.11
Compiled with xpdf version 4.00

$ bibtex --version
BibTeX 0.99d (TeX Live 2018)
kpathsea version 6.3.0
Copyright 2018 Oren Patashnik.
There is NO warranty.  Redistribution of this software is
covered by the terms of both the BibTeX copyright and
the Lesser GNU General Public License.
For more information about these matters, see the file
named COPYING and the BibTeX source.
Primary author of BibTeX: Oren Patashnik.

Thanks for this wonderful book. It's a pleasure to read it and there's a lot to learn from it.

Section 6.1.1 paragraph "Step1: Identify the Bottleneck" mentions:

As of Linux kernel 4.8, perf has an option --topdown used in perf stat command that prints TMA Level 1 metrics.

and that the benchmark was run with:

We ran the experiments on the machine equipped with Intel Core i5-8259U CPU (Skylake based) and 16GB of DRAM (DDR4 2400 MT/s), running 64-bit Ubuntu 20.04 (kernel version 5.13.0-27).

Is this correct or could further clarification be helpful?

I cannot run the example because I receive the error System does not support topdown. I have the following configuration:

• Kernel version: 5.15.0-92-generic
• Alder Lake, i7-12850HX

There's also a stackoverflow post mentioning that topdown is only available for Alderlake on Linux 6.5

In the text snippet above it would be beneficial to have an example of what is "minimum alignment requirement". When I read that sentence for the first time I was sure that "minimum alignment requirements" is 64 bytes. But later, after I read that sentence 5 times I understood that here you mean values like 4, 8 or16.

Magic Trace is an amazing tool using Intel PT to render execution traces, see https://github.com/janestreet/magic-trace?tab=readme-ov-file#getting-started

The book mentions that Intel PT has poor support, but this tool definitely closes a huge gap in terms of debugging and profiling, and should be mentioned.

If that sounds good I could have a look at writing a section on it.

Hello,

Firstly, I greatly appreciate your effort. Thank you so much for reading this.

I do have a comment, which is more of a question. Would it be possible to provide additional release formats, such as A5 PDF (for printing two pages on a single A4), or better yet, to simplify or automate the build process to support more recent Pandoc versions, without relying directly on copied binaries, if possible?

Hi, I was confused by this formula on page 31:
"Time per instruction on pipelined machine = Time per instruction on non-pipelined machine / Number of pipe stage"

My understanding is that the time per instruction is the same on pipelined machine or non-pipelined machine, since pipeline only improves the throughput but not the latency. If the "Time per instruction on pipelined machine" is decreased compared to non-pipeline machine, the only reason is that the frequency increases. Is there any misunderstanding here?

First of all, thanks for making this great book!
During keep forwarding with your book, I found some typo(?) which may lead to misunderstanding about __builtin__prefetch work (such as "what is that variable a? where it comes from?")

This line should be like this

..
__builtin_prefetch(arr + j, 0, 1); // well before the load
..

Thanks. Dear @dendibakh

It is hard to notice the table title in the book.

For example, on page 12 the font of the Table 1 title is the same as font the rest of text in the book and it took me a while to find the Table 1 referenced elsewhere :)