Re: Identifying CPU and current OS
- To: debian-user@lists.debian.org
- Subject: Re: Identifying CPU and current OS
- From: Greg <curtyshoo@gmail.com>
- Date: Wed, 1 Oct 2025 13:32:17 -0000 (UTC)
- Message-id: <[🔎] slrn10dqbb1.1sj.curtyshoo@einstein.home.arpa>
- References: <276ffa8a-b448-369a-9fd6-61a7d878a5ce@access.net> <98mnql-2q3.ln1@anthive.com> <0ec7e8e5-a51a-b05c-1655-d9c318a2b1f9@access.net> <f63ea2a4-9d1c-11f0-8000-00163eeb5320@msgid.mathom.us> <slrn10dlc3t.1df.curtyshoo@einstein.home.arpa> <f1823e30-9d4c-11f0-8000-00163eeb5320@msgid.mathom.us> <slrn10dnua7.1bu.curtyshoo@einstein.home.arpa> <df5921bc-9e03-11f0-8000-00163eeb5320@msgid.mathom.us> <slrn10do04o.1bu.curtyshoo@einstein.home.arpa> <f7a3b102-9e07-11f0-8000-00163eeb5320@msgid.mathom.us>
On 2025-09-30, Michael Stone <mstone@debian.org> wrote:
> On Tue, Sep 30, 2025 at 04:08:56PM -0000, Greg wrote:
>> In computing, a word is any processor design's natural unit of data. A word is
>> a fixed-sized datum handled as a unit by the instruction set or the hardware of
>> the processor.
>
> It has that definition (among others), one which mapped cleanly to
> historic processors, but which is much messier when dealing with modern
> CPUs. E.g., amd64 mostly has 64bit instruction operands and types, but
> there's also an 80 bit floating point type and with AVX-512 extensions
> has 512 bit registers and can transfer data to/from memory 512 bits at a
> time in a single instruction. So is the word size 512? 80? 64?
> Conversely, if someone confidently says "its word size is 64", what does
> that tell you/what can you do with that information? If it's a term
> divorced from practical applications then it's just a historical
> curiosity. If someone has to add a bunch of asterisks and explain away
> functionality that doesn't cleanly fit into the model, how relevant is
> the model?
>
I think the key word here is *natural* (the machine's general-purpose
integer registers, the ones used for addresses and integer arithmetic,
not the specialized, vector register widths).
Reply to: