Hacker News Reader

Comment by zahllos | Hacker News Reader

Better option: Computer Organization and Design by Patterson and Hennessy, ISBN 978-0124077263. Comes in MIPS (original version now in its 5th edition), ARM and RISC-V focused flavors, covers "instructions unique to architecture X" as part of its discussion.

View on HN

I could not agree more with this recommendation. For folks who don't yet have this book but are interested, there is a set of online companion materials[0] that are freely available and contain a good bit of useful information.

[0] https://www.elsevier.com/books-and-journals/book-companion/9...

I had no idea they did RISC-V and ARM versions! Very cool.

This was the book for the "weed-out" class I took my first semester in college. Nearly failed that class, but it opened my eyes to so much more of computing, beyond just "coding".

> Nearly failed that class

I did fail it in university. Took it again and got an A-, and now love working with assembly and computer architecture.

The struggle make your knowledge crystallize even better, on top of overcoming the failure. I almost wish anyone to fail one important class in university.

I deeply regret the courses I barely passed. And the courses for which I had prior knowledge to pass comfortably but not engage on the nuances.

That was me with a high performance scientific computing class.

Does anyone have an idea how bad is the ARM version? The reviews online say they did not fully adapt it from MIPS, and that the instruction set is partially made up instead of using the real one, which means listings won't work on real hardware.

>which means listings won't work on real hardware.

They will. LEG is a subset of ARM. What won't work is ARM software on LEG hardware.

I would recommend getting the current edition, which has replaced ARM/LEG with the industry standard RISC-V.

Does RISC-V version include GPU?

I am more inclined for ARM version since that's applicable to both all current Apple hardware and school Chromebooks.

The ISA doesn't really matter (at least, not between MIPS, ARMv8, and RISC-V) when you're first diving into computer organisation. That's why one book can be sold with three different ISAs. The overarching concepts are mostly the same: you have digital logic, IEEE-754, basic assembly (add, sub, and, not, xor, load, store, shift), pipelining and hazards, caches.

Only afterwards do the differences between the ISAs come to play, when you have complex vector and ISA-specific instructions, superscalar pipelines, atomic instructions, etc.

It matters for engagement, e.g. if you can actually run examples painlessly.

The open standard ISA RISC-V has the advantage there, with open source hardware implementations (like C906 and C910) and good tooling and simulators.

RISC-V is rapidly building the strongest ecosystem.

Overall I enjoyed this book. However their opening paragraph to floating point numbers is offensive.

"programming languages support numbers with fractions, which are called reals in mathematics. Here are some examples of reals:

3.14159265... (pi)

2.71828... (e)"

>fractions are called reals in mathematics

>immediately gives pi and e as an example

Isn’t this technically true though? Fractions are just rational reals. The examples listed are still reals, just irrational.

It's bad pedagogy. But it is also technically incorrect. You can have fractions with complex numbers.

Good point! I’m not a mathematician so the pedagogy here doesn’t really irk me. But, I do also believe that teachers should be held to a higher standard, especially when writing out educational text. Thanks for the clarification!

Also, I bet the authors weren’t intentionally trying to lead readers astray with that statement. Most books have a preface with some point of contact if you find errors. I’m sure if you wrote in the authors would be happy to fix this up :)

i think i can see the nitpick here: pi and e aren't representable as rational numbers, and aren't representable exactly in IEEE754

I wouldn't say it's a nitpick. Rationals are fractions of integers, not just any fractions at all. Rationals are _examples_ of reals, but it's bad pedagogy to say "fractions, which are called reals" because it sounds like the authors are giving a definition of reals. While they technically did not say this was a definition, it is sloppy.

Fractions = rational numbers, which are an infinitely small subset of the real numbers. The rest of the reals are irrational numbers, of which pi and e are examples.

Fraction doesn't have to be rational.

I agree it's bad phrasing, but isn't the only real error here using "with fractions" rather than "with fractional digits"? Perhaps this is an idiomatic difference between different English dialects, such as American vs British? I'm not sure. If you read fractions as meaning fractional digits, the inaccuracy goes away completely, I think.

I would also use "real number" instead of "real", but that may be more of a stylistic choice rather than idiomatic.

This book is the reason I was able to pass my computer architecture exams in grad school. Just lugged the 800 page monstrosity around an entire semester and read it every chance I could. A really wonderful book ... I should get a recent copy and re-read it again.

God this book fucking rules! Used this in school and was the only textbook I found myself just working through outside of class for my own personal enrichment and entertainment. Can't recommend it enough

Agreed. Appendix B of this book in and of itself is an incredible summary of how a computer works generally.

Waiting for the GPU covering version. (I do have an older version of the book though).

If I recall correctly, there is indeed a chapter about GPUs in the fifth edition. Maybe it's an appendix, but I'm pretty certain! :)

The description of the ARM edition says Nvidia GPU is used throughout the book for comparison

I managed to get my hands on a really old version of this book (maybe 20 yrs old - no ARM I don't think). Do you think that is still good? Or should I just get the new one?

The description on the official site says the target is grad students. Is it fair to say that is not an introduction to the subject?

It depends what you mean or want by way of introduction. Are you interested at all in how your CPU does what it does and why, or were you interested in the "learn x86, ARM and RISC-V" tag-line that was originally on the book title. Do you imagine you will read this book and be writing all your software in assembler straight after? That won't happen, most likely - if you want that and insist you do, I'd suggest the flatassembler forums.

The COD book I recommended was recommended to me by an EE undergrad who took it on one of their courses, and I think it is suitable as an introduction to how computers actually work, with plenty of real-world examples and mentions of architecture features that indicate specific tradeoffs. Of course, whether that is suitable is really going to depend on the person reading it.

The Packt book looks like it might be attempting to do the same thing - except I am slightly worried it likely digresses a bit into virtualization, self driving cars and quantum computers (based on the Packt TOC and github repo). I'm going to say this is scope creep - I don't know how deep it goes into virtualization, but personally I think understanding how x86 does virtualization is much less useful than understanding the wider problem of how to partition hardware. If it doesn't say LPAR somewhere in there or discuss mainframes, I'd be disappointed. Self-driving cars do not, in my opinion, belong in a book on how modern computers work, and I'm not sure the author's experiments with IBM's quantum computer cloud thingabob are really what I want out of such a book either.

That said, I've not read the Packt book and I can't say anything about the earlier content. I have read the COD book - but I wish I'd come across it earlier than I did. Still, I got something out of it.

Some graduate textbooks are highly specialized and for niche fields - my formal education is in mathematics and I can definitely say this about some graduate maths books - but this is not one of them. I think it is widely applicable to anyone who a) wants an introduction to how computers do their stuff or b) already broadly knows, but has some gaps and wants to fill them in.

Thanks for the recommendation. I bought the book.

And FWIW the book you recommended is for undergrads and is a much more accessible read, according to the authors. I saw "Computer Architecture: A Quantitative Approach" and got confused.

>Comes in MIPS (original version now in its 5th edition), ARM and RISC-V focused flavors

They're not flavors, they're editions. They have changed their ISA over time.

The current edition uses RISC-V.

Reply on news.ycombinator.com