[TUHS] Kernel Sizes

[TUHS] Kernel Sizes

[Rudi Blom]

16K kernels?

Well, I came late to the UNIX world. I only remember getting a SCO
UNIX 3.2V4.2 kernel onto a single 3-1/2" diskette and still have all I
wanted including (with scripts on a second diskette of course :-) ).
That was 25 years ago.

Mind you, from there I moved to Digital UNIX /vmunix 9,613.712, Tru64
17.930.976 to HP-UX 11iv2  66.161.464 and HP-UX 11iv3 127.929.032

Of course, I also got lots more functionality I'm supposed to want and need :-)

Cheers,
rudi

[TUHS] Kernel Sizes

[Warner Losh]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 604 bytes --]

On Jan 21, 2018 9:40 PM, "Grant Taylor via TUHS" <tuhs at minnie.tuhs.org>
wrote:

Weren’t these (or their source code) recently released by the IP owner?

Virtually Fun has an article about installing and running v8, including an
image.

https://virtuallyfun.com/2017/03/30/research-unix-v8/

I expect you can find more there and the TUHS pages that he references.


I've already looked. There is the source but no kernels...

Warner
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20180121/fdbca25e/attachment.html>

[TUHS] Kernel Sizes

[Grant Taylor]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 739 bytes --]

Weren’t these (or their source code) recently released by the IP owner?

Virtually Fun has an article about installing and running v8, including an image.

https://virtuallyfun.com/2017/03/30/research-unix-v8/

I expect you can find more there and the TUHS pages that he references.



-- 
Grant. . . .
unix || die
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20180121/97e07b5f/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2338 bytes
Desc: not available
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20180121/97e07b5f/attachment.bin>

[TUHS] Kernel Sizes

[Warner Losh]

On Sat, Jan 20, 2018 at 5:19 PM, Warren Toomey <wkt at tuhs.org> wrote:

> Unfortunately, we don't have a kernel from 1st Edition or 4th Edition.
>

Do we have v8, v9 or v10 kernels? I looked at the tarballs in the archives
and couldn't find unix, bsd, or kernel in any of them. Is there some name
that's used that I've managed to miss out on?

Warner
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20180121/6a831ce5/attachment.html>

[TUHS] Kernel Sizes

[Steve Nickolas]

On Mon, 22 Jan 2018, Greg 'groggy' Lehey wrote:

> On Saturday, 20 January 2018 at 21:24:27 -0700, Warner Losh wrote:
>> On Sat, Jan 20, 2018 at 7:13 PM, Steve Johnson <scj at yaccman.com> wrote:
>>
>>> While you're at it, I heard once that the latest GCC *manual* (>500 pp at
>>> last look) was larger than "the whole Unix distribution".  Is there any
>>> truth to that?
>>>
>>
>> compressed (gz), the entire v7 tape 3.5MB. The compressed (gz) source is
>> 250k. Uncompressed tape is 11.5MB while the source is 1.1MB.
>>
>> gcc 7.2.0 compressed (gz) is 105MB. gcc/doc directory is 13.5MB:
>>
>>> tar tvf gcc-7.2.0.tar.xz gcc-7.2.0/gcc/doc | awk '{a += $5;} END {print
>> a;}'
>> 13,470,317
>>
>> So yea, gcc 7.2 manual is larger than the v7 distribution tapes.
>
> It would be interesting to compare llvm.  Here are the most recent
> FreeBSD packages (effectively installation tarballs).  gcc pales by
> comparison:
>
> -rw-r--r--  1 root  wheel  90,805,148 23 Sep 15:53 /var/cache/pkg/gcc6-6.4.0_1-13072ceeab.txz
> -rw-r--r--  1 root  wheel  90,819,108 30 Sep 15:18 /var/cache/pkg/gcc6-6.4.0_2-d83317c1d0.txz
> -rw-r--r--  1 root  wheel  305,101,072 16 Nov 18:30 /var/cache/pkg/llvm40-4.0.1_4-06c71eb2eb.txz
> -rw-r--r--  1 root  wheel  341,208,492 19 Dec 17:05 /var/cache/pkg/llvm50-5.0.0_6-b3fff834c7.txz

Yikes, and I thought *GNU* sofware was a pig. o.O

-uso.

[TUHS] Kernel Sizes

[Paul Ruizendaal]

> A self-imposed limit of 16K held in v1, and was quite fully utilized.
> When the iernel was rewritten in C, the limit (perhaps larger by then)
> influenced the C compiler. More than one optimization was stimulated
> by the need to keep the kernel in bounds.
> 
> Doug


The LSX kernel was kept within a self-imposed limit of 16KB as well.

I've often thought that LSX was V5 'regressed' to the concepts of
V1, which was facing similar hardware constraints. Is that a
reasonable view?

For example, LSX has a maximum of three processes that are swapped
in and out in a stack-like fashion. Only one process is ever in core.

Is that how V1 was organized?

I realize that LSX was API compatible with V5/V6 and I don't mean
'regressed' in that sense.

Paul

[TUHS] Kernel Sizes

[Greg 'groggy' Lehey]

On Saturday, 20 January 2018 at 21:24:27 -0700, Warner Losh wrote:
> On Sat, Jan 20, 2018 at 7:13 PM, Steve Johnson <scj at yaccman.com> wrote:
>
>> While you're at it, I heard once that the latest GCC *manual* (>500 pp at
>> last look) was larger than "the whole Unix distribution".  Is there any
>> truth to that?
>>
>
> compressed (gz), the entire v7 tape 3.5MB. The compressed (gz) source is
> 250k. Uncompressed tape is 11.5MB while the source is 1.1MB.
>
> gcc 7.2.0 compressed (gz) is 105MB. gcc/doc directory is 13.5MB:
>
>> tar tvf gcc-7.2.0.tar.xz gcc-7.2.0/gcc/doc | awk '{a += $5;} END {print
> a;}'
> 13,470,317
>
> So yea, gcc 7.2 manual is larger than the v7 distribution tapes.

It would be interesting to compare llvm.  Here are the most recent
FreeBSD packages (effectively installation tarballs).  gcc pales by
comparison:

-rw-r--r--  1 root  wheel  90,805,148 23 Sep 15:53 /var/cache/pkg/gcc6-6.4.0_1-13072ceeab.txz
-rw-r--r--  1 root  wheel  90,819,108 30 Sep 15:18 /var/cache/pkg/gcc6-6.4.0_2-d83317c1d0.txz
-rw-r--r--  1 root  wheel  305,101,072 16 Nov 18:30 /var/cache/pkg/llvm40-4.0.1_4-06c71eb2eb.txz
-rw-r--r--  1 root  wheel  341,208,492 19 Dec 17:05 /var/cache/pkg/llvm50-5.0.0_6-b3fff834c7.txz

Greg
--
Sent from my desktop computer.
Finger grog at lemis.com for PGP public key.
See complete headers for address and phone numbers.
This message is digitally signed.  If your Microsoft mail program
reports problems, please read http://lemis.com/broken-MUA
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 163 bytes
Desc: not available
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20180122/fdc292bc/attachment.sig>

[TUHS] Kernel Sizes

[Doug McIlroy]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 714 bytes --]

> I seem to remember that at some point early on, we spent some of our
> capital budget on buying another 16K bytes for the PDP-11.  As I
> recall, the deal was that the OS got 8K and users got 8K.  I also
> recall that that purchase was 20% of our capital budget for the
> year...

> Doug, did I remember this correctly?

> Steve

Things did happen that way. I don't specifically remember the
numbers, but those you state have the ring of truth.

Memory came in (expensive) 4K increments. When the keepers of
Unix in the Charlotte wire center wanted to add 4K, their
management consented only on the condition that the wizards
in Research would confirm that roposed new functionality
really required it.

doug

[TUHS] Kernel Sizes

[Doug McIlroy]

>> A self-imposed limit of 16K held in v1

> 16k words or 16k bytes?

Bytes.

doug

[TUHS] Kernel Sizes

[Warner Losh]

On Sat, Jan 20, 2018 at 7:13 PM, Steve Johnson <scj at yaccman.com> wrote:

> While you're at it, I heard once that the latest GCC *manual* (>500 pp at
> last look) was larger than "the whole Unix distribution".  Is there any
> truth to that?
>

compressed (gz), the entire v7 tape 3.5MB. The compressed (gz) source is
250k. Uncompressed tape is 11.5MB while the source is 1.1MB.

gcc 7.2.0 compressed (gz) is 105MB. gcc/doc directory is 13.5MB:

% tar tvf gcc-7.2.0.tar.xz gcc-7.2.0/gcc/doc | awk '{a += $5;} END {print
a;}'
13,470,317

So yea, gcc 7.2 manual is larger than the v7 distribution tapes.

Warner
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20180120/131a79d8/attachment.html>

[TUHS] Kernel Sizes

[Steve Johnson]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1123 bytes --]

While you're at it, I heard once that the latest GCC *manual* (>500 pp
at last look) was larger than "the whole Unix distribution".  Is
there any truth to that?

Steve

----- Original Message -----
From: "Dan Stromberg" <drsalists@gmail.com>
To:"Warner Losh" <imp at bsdimp.com>
Cc:"The Eunuchs Hysterical Society" <tuhs at tuhs.org>
Sent:Sat, 20 Jan 2018 18:08:43 -0800
Subject:Re: [TUHS] Kernel Sizes

 On Sat, Jan 20, 2018 at 10:11 AM, Warner Losh <imp at bsdimp.com> wrote:
 > For a presentation I'm doing this summer on FreeBSD, I thought it
would be
 > cool to get the kernel sizes for various old flavors of Unix. I see
numbers

 You might find http://stromberg.dnsalias.org/~strombrg/working-set/
interesting.

 It intentionally fills virtual memory, and measures how much must
must
 be malloc'd and filled with gibberish, to cause thrashing.

 Subtracting that from the amount of physmem in the machine, gives a
 sort of measure of OS overhead.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20180120/21f6159a/attachment.html>

[TUHS] Kernel Sizes

[Dan Stromberg]

On Sat, Jan 20, 2018 at 10:11 AM, Warner Losh <imp at bsdimp.com> wrote:
> For a presentation I'm doing this summer on FreeBSD, I thought it would be
> cool to get the kernel sizes for various old flavors of Unix. I see numbers

You might find http://stromberg.dnsalias.org/~strombrg/working-set/ interesting.

It intentionally fills virtual memory, and measures how much must must
be malloc'd and filled with gibberish, to cause thrashing.

Subtracting that from the amount of physmem in the machine, gives a
sort of measure of OS overhead.

[TUHS] Kernel Sizes

[Andy Kosela]

On Saturday, January 20, 2018, Warren Toomey <wkt at tuhs.org> wrote:

> On Sat, Jan 20, 2018 at 11:11:24AM -0700, Warner Losh wrote:
>
>>   For a presentation I'm doing this summer on FreeBSD, I thought it would
>>   be cool to get the kernel sizes for various old flavors of Unix. I see
>>   numbers for v5, v6 and v7 in the tuhs tree view, and it appears these
>>   versions are complete enough for me to extract the kernels themselves.
>>   However, I see nothing prior to that.
>>
>
> Github has this project: https://github.com/DoctorWkt/unix-jun72
> with a June 1972 Unix kernel. Doing a "make", the generated build/unix is:
>
> -rwxrwxrwx 1 wkt wkt 36432 Jan 21 10:10 build/unix
>
> but I don't have a PDP-11 "size" command to give details.
>
> http://www.tuhs.org/Archive/Distributions/Research/Dennis_v3/ has a Unix
> system written in C, timestamped August 31, 1973 (just before Fourth
> Edition).
> Inside nsys.tar.gz you will find:
>
> -rw-r--r-- 0/0           26820 1973-09-24 03:41 u
>
> which is the kernel image.
>
> Unfortunately, we don't have a kernel from 1st Edition or 4th Edition.
>
>
>
Comparing size of kernels is cool and fun, but IMHO comparing system calls
is a more valuable metric as to measure the kernel bloat.

It would be interesting to compare number of implemented system calls in
various UNIX operating systems along with those kernel sizes, e.g., V7 had
around 50 system calls, current FreeBSD and Linux have more than 500...

--Andy
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20180120/38a6db7a/attachment-0001.html>

[TUHS] Kernel Sizes

[Warren Toomey]

On Sat, Jan 20, 2018 at 11:11:24AM -0700, Warner Losh wrote:
>   For a presentation I'm doing this summer on FreeBSD, I thought it would
>   be cool to get the kernel sizes for various old flavors of Unix. I see
>   numbers for v5, v6 and v7 in the tuhs tree view, and it appears these
>   versions are complete enough for me to extract the kernels themselves.
>   However, I see nothing prior to that.

Github has this project: https://github.com/DoctorWkt/unix-jun72
with a June 1972 Unix kernel. Doing a "make", the generated build/unix is:

-rwxrwxrwx 1 wkt wkt 36432 Jan 21 10:10 build/unix

but I don't have a PDP-11 "size" command to give details.

http://www.tuhs.org/Archive/Distributions/Research/Dennis_v3/ has a Unix
system written in C, timestamped August 31, 1973 (just before Fourth Edition).
Inside nsys.tar.gz you will find:

-rw-r--r-- 0/0           26820 1973-09-24 03:41 u

which is the kernel image.

Unfortunately, we don't have a kernel from 1st Edition or 4th Edition.

Cheers, Warren

[TUHS] Kernel Sizes

[Michael Kjörling]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 939 bytes --]

On 20 Jan 2018 11:11 -0700, from imp at bsdimp.com (Warner Losh):
> For a presentation I'm doing this summer on FreeBSD, I thought it would be
> cool to get the kernel sizes for various old flavors of Unix. I see numbers
> for v5, v6 and v7 in the tuhs tree view, and it appears these versions are
> complete enough for me to extract the kernels themselves. However, I see
> nothing prior to that. The archives seem to be somewhat incomplete, but I'm
> wondering if people have sizes for earlier versions.

https://github.com/dspinellis/unix-history-repo/tree/Research-V1-Snapshot-Development
looks like it might be useful.

It's possible that Warren might want to add some of that to the TUHS
archives as well.

-- 
Michael Kjörling • https://michael.kjorling.se • michael at kjorling.se
  “The most dangerous thought that you can have as a creative person
              is to think you know what you’re doing.” (Bret Victor)

[TUHS] Kernel Sizes

[Random832]

On Sat, Jan 20, 2018, at 13:11, Warner Losh wrote:
> For a presentation I'm doing this summer on FreeBSD, I thought it would be
> cool to get the kernel sizes for various old flavors of Unix. I see numbers
> for v5, v6 and v7 in the tuhs tree view, and it appears these versions are
> complete enough for me to extract the kernels themselves. However, I see
> nothing prior to that.

There is a possibly V2 era kernel (two, actually) on the s1-bits tape image, but it's been too long since I looked at this and I don't recall exactly how to extract them. I posted about this at the time, but there wasn't any interest - I'd determined that s1 was in fact an "init tape" as described in V1 boot.7 and V3 bproc.8. Those manpages do mention the size that was allocated on disk/tape for each kernel in those eras: 6K and 7-8K words [so twice that number of bytes], respectively.

To determine how much headroom was in this allocation, I just now went through and checked the s1-bits file for empty 512-byte blocks. It consists of 25 blocks of data (12800 bytes), followed by 4 blocks of zeros. I think that region of the tape was the bootloaders followed by the "cold boot" kernel. Then there are 22 blocks (11264 bytes) of data [then 10 blocks zeros], which was IIRC the other kernel (the "warm boot" kernel, which did not contain code for reinitializing the filesystem). The rest of the tape (not kernel images) is 490 blocks (245 KB) of data, followed by 27 blocks of 0xFF.

[TUHS] Kernel Sizes

[Donald ODona]

At 20 Jan 2018 18:12:45 +0000 (+00:00) from Warner Losh <imp at bsdimp.com>:
> For a presentation I'm doing this summer on FreeBSD, I thought it would be

bsd4.3 289792
V7 55876
SysVr3 702832

[TUHS] Kernel Sizes

[Warner Losh]

For a presentation I'm doing this summer on FreeBSD, I thought it would be
cool to get the kernel sizes for various old flavors of Unix. I see numbers
for v5, v6 and v7 in the tuhs tree view, and it appears these versions are
complete enough for me to extract the kernels themselves. However, I see
nothing prior to that. The archives seem to be somewhat incomplete, but I'm
wondering if people have sizes for earlier versions. Later versions are
more problematic because they move to new hardware, instruction sets, etc.
For this graph to be meaningful, it would need to be pdp-11 only, though
I'm of the opinion more data is better than less. I'll also be extracting
the different V7 commercial kernels: V7M, Ultrix and Venix and the BSD 2.x
releases, but those appear to be intact enough in the archives to extract
data on my own. I've heard rumors there's a SysV for the pdp-11, but I've
not been able to locate images for that. I don't need the actual images,
just sizes with some reference for the source of the data. It's just for
one slide in the talk, so I don't want to burn a ton of time on it...

The larger picture is that I've written what's effectively modprobe for
FreeBSD and will be talking about it in detail. How it's like modprobe, how
it's different, how all the pieces fit together, history of similar
efforts, etc. Part of the driving force here is the bloated FreeBSD GENERIC
kernel.

Of course, I'll share the final report I'm planning on sizes with the group.

Thanks for any data you can provide.

Warner
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20180120/fea76a1a/attachment.html>