[9fans] largest memory plan 9 system to date?

[9fans] largest memory plan 9 system to date?

[ron minnich]

[-- Attachment #1: Type: text/plain, Size: 782 bytes --]

I am wondering, what is the largest memory you have seen on a Plan 9
system?

In my case, it's been 32 GiB on an amd64. Is there something larger out
there?

One of the (many) reasons for NIX 2M page sizes was the limited address
space of the "virtual mmu" in port/. To deal with heap larger than 1GiB,
Nemo had this very clever idea, of using GiB PTEs when your heap hit the 1
GiB mark. This meant we needed only 512 PTEs for the first GiB (2M pages),
and then 511 PTEs to reach 512 GiB. But we never had that much memory, of
course.

ron

------------------------------------------
9fans: 9fans
Permalink: https://9fans.topicbox.com/groups/9fans/Ta60ba03129fc4e5b-M37f879d02295454986b9f2cc
Delivery options: https://9fans.topicbox.com/groups/9fans/subscription

[-- Attachment #2: Type: text/html, Size: 1345 bytes --]

Re: [9fans] largest memory plan 9 system to date?

[Noam Preil]

[-- Attachment #1: Type: text/plain, Size: 340 bytes --]

I have a 64g thread ripper.

Probably. I'd only just set it up a day before the fire, still needs cleaned :/

------------------------------------------
9fans: 9fans
Permalink: https://9fans.topicbox.com/groups/9fans/Ta60ba03129fc4e5b-M10f83496ae4f4f99578bd5d7
Delivery options: https://9fans.topicbox.com/groups/9fans/subscription

[-- Attachment #2: Type: text/html, Size: 1042 bytes --]

Re: [9fans] largest memory plan 9 system to date?

[ori]

From the hacked thinkpad X201 I'm typing this on:

        % cat /dev/swap
        68491845632 memory
        4096 pagesize
        524288 kernel
        1343866/15661303 user
        0/320000 swap
        23811/15661303 reclaim
        54772160/55446976/2147073280 kernel malloc
        72373216/72834080/1932365952 kernel draw
        3296/65568/16777216 kernel secret

It's excessive, but that's what it came with; it's occasionally
useful for VMs in VMX. The machine was purchased before the
current shortages.

Quoth ron minnich <rminnich@gmail.com>:
> I am wondering, what is the largest memory you have seen on a Plan 9
> system?
> 
> In my case, it's been 32 GiB on an amd64. Is there something larger out
> there?
> 
> One of the (many) reasons for NIX 2M page sizes was the limited address
> space of the "virtual mmu" in port/. To deal with heap larger than 1GiB,
> Nemo had this very clever idea, of using GiB PTEs when your heap hit the 1
> GiB mark. This meant we needed only 512 PTEs for the first GiB (2M pages),
> and then 511 PTEs to reach 512 GiB. But we never had that much memory, of
> course.
> 
> ron

------------------------------------------
9fans: 9fans
Permalink: https://9fans.topicbox.com/groups/9fans/Ta60ba03129fc4e5b-Mfc5dedda79d328590ede9fc1
Delivery options: https://9fans.topicbox.com/groups/9fans/subscription

Re: [9fans] largest memory plan 9 system to date?

[Ron Minnich]

followup question: what is the largest process you've ever seen run? I
am guessing a vm but maybe not?

On Tue, Mar 3, 2026 at 2:58 PM <ori@eigenstate.org> wrote:
>
> From the hacked thinkpad X201 I'm typing this on:
>
>         % cat /dev/swap
>         68491845632 memory
>         4096 pagesize
>         524288 kernel
>         1343866/15661303 user
>         0/320000 swap
>         23811/15661303 reclaim
>         54772160/55446976/2147073280 kernel malloc
>         72373216/72834080/1932365952 kernel draw
>         3296/65568/16777216 kernel secret
>
> It's excessive, but that's what it came with; it's occasionally
> useful for VMs in VMX. The machine was purchased before the
> current shortages.
>
> Quoth ron minnich <rminnich@gmail.com>:
> > I am wondering, what is the largest memory you have seen on a Plan 9
> > system?
> >
> > In my case, it's been 32 GiB on an amd64. Is there something larger out
> > there?
> >
> > One of the (many) reasons for NIX 2M page sizes was the limited address
> > space of the "virtual mmu" in port/. To deal with heap larger than 1GiB,
> > Nemo had this very clever idea, of using GiB PTEs when your heap hit the 1
> > GiB mark. This meant we needed only 512 PTEs for the first GiB (2M pages),
> > and then 511 PTEs to reach 512 GiB. But we never had that much memory, of
> > course.
> >
> > ron

------------------------------------------
9fans: 9fans
Permalink: https://9fans.topicbox.com/groups/9fans/Ta60ba03129fc4e5b-Mca8d35e2047d9974796ef72f
Delivery options: https://9fans.topicbox.com/groups/9fans/subscription

Re: [9fans] largest memory plan 9 system to date?

[Noam Preil]

[-- Attachment #1: Type: text/plain, Size: 272 bytes --]

I've run venti with 32g of cache before :)

------------------------------------------
9fans: 9fans
Permalink: https://9fans.topicbox.com/groups/9fans/Ta60ba03129fc4e5b-Me3eac509bc6f20ff48ab47fd
Delivery options: https://9fans.topicbox.com/groups/9fans/subscription

[-- Attachment #2: Type: text/html, Size: 904 bytes --]

Re: [9fans] largest memory plan 9 system to date?

[ori]

Quoth Ron Minnich <rminnich@p9f.org>:
> followup question: what is the largest process you've ever seen run? I
> am guessing a vm but maybe not?
> 

Probably a vm, but hard to say.

The fs can get pretty big (both cwfs and gefs), simply because they
default to 1/4 of installed memory for cache; git can get pretty big
when indexing because of large blobs and long delta chains.

As a side note: malloc used to limit pools to 2 gigs; today we
allow more than that, but single allocations still have the 2 gig
limit, because we haven't audited code for array index overflows.


------------------------------------------
9fans: 9fans
Permalink: https://9fans.topicbox.com/groups/9fans/Ta60ba03129fc4e5b-Mf51896c4f03e54107ed36b82
Delivery options: https://9fans.topicbox.com/groups/9fans/subscription