[9fans] Cute plan9/inferno client?

[9fans] Cute plan9/inferno client?

[Lawrence E. Bakst]

http://www.semiaccurate.com/2011/01/26/compulab-announces-tegra-2-powered-trim-slice/

Re: [9fans] Cute plan9/inferno client?

[erik quanstrom]

On Thu Jan 27 12:37:34 EST 2011, ml@iridescent.org wrote:
> http://www.semiaccurate.com/2011/01/26/compulab-announces-tegra-2-powered-trim-slice/
>

nvidia.  those are the closed-source everything guys, right?

- erik

Re: [9fans] Cute plan9/inferno client?

[Lawrence E. Bakst]

At 12:38 PM -0500 1/27/11, erik quanstrom wrote:
>On Thu Jan 27 12:37:34 EST 2011, ml@iridescent.org wrote:
>> http://www.semiaccurate.com/2011/01/26/compulab-announces-tegra-2-powered-trim-slice/
>>
>
>nvidia.  those are the closed-source everything guys, right?
>
>- erik

I deleted a line from the post that said the big issue will be the quality of the hardware doc, if there even is any. Still, if it ends up running linux there might be some hope.

The Tegra 3 would make a nice small ARM based server/terminal with a quad core Cortex A9.

I wouldn't rule out getting doc on everything but the 3D unit. Even in that case, some of Nvidia's 3D GPUs instruction sets have been reverse engineered.

In case you haven't noticed all of the major 3D chips guys are closed source.

I once had a high level discussion with some 802.11 chip companies about open source drivers. Their biggest fear seems to be in revealing all of the various bugs in their chips. Which I thought was interesting.

leb
--
leb@iridescent.org

Re: [9fans] Cute plan9/inferno client?

[Lawrence E. Bakst]

Forgot this Tegra road map link:
http://www.brightsideofnews.com/news/2011/1/23/nvidia-thinks-world-domination-tegra-2-3d-in-january2c-tegra-3-by-fall.aspx
--
leb@iridescent.org

Re: [9fans] Cute plan9/inferno client?

[erik quanstrom]

> I deleted a line from the post that said the big issue will be the
> quality of the hardware doc, if there even is any.  Still, if it ends
> up running linux there might be some hope.

the problem is that the linux driver guy is generally getting it straight
from the source, and not from the docs.  so (from experience)
existence of a linux driver does not imply that there are docs,
or if there are, a new driver could be written from them.

> I wouldn't rule out getting doc on everything but the 3D unit.  Even
> in that case, some of Nvidia's 3D GPUs instruction sets have been
> reverse engineered.

if you have time for that kind of nonsense, go ahead.  there
are full specs on the intel video hardware, and nobody has
done a think with that.

> In case you haven't noticed all of the major 3D chips guys are closed
> source.

ati?

- erik

Re: [9fans] Cute plan9/inferno client?

[hiro]

> if you have time for that kind of nonsense, go ahead.  there
> are full specs on the intel video hardware, and nobody has
> done a think with that.

Intel is dead.

Re: [9fans] Cute plan9/inferno client?

[erik quanstrom]

On Thu Jan 27 21:46:53 EST 2011, 23hiro@googlemail.com wrote:
> > if you have time for that kind of nonsense, go ahead.  there
> > are full specs on the intel video hardware, and nobody has
> > done a think with that.
>
> Intel is dead.

your wishes don't make reality.  they are >50% of the graphics chipset market.


- erik

Re: [9fans] Cute plan9/inferno client?

[Skip Tavakkolian]

probably not, but sure smells like it.

On Thu, Jan 27, 2011 at 6:45 PM, hiro <23hiro@googlemail.com> wrote:
>> if you have time for that kind of nonsense, go ahead.  there
>> are full specs on the intel video hardware, and nobody has
>> done a think with that.
>
> Intel is dead.
>
>

Re: [9fans] Cute plan9/inferno client?

[Lawrence E. Bakst]

A few comments.

1. It was a 1 line post to the group. I'm not pushing anything, have no vested interest, I just wanted the list to be aware of it. You don't like Nvida or closed source, that's fine with me. The Tegra is probably one of the best ARM chips out there and certainly the most integrated given the graphics and video hardware it has.

I'll point that companies like TI, that used to hate open source are now converts. Many of the Gumstix Overo modules that seem to be so beloved here, have PowerVR SGX chips on them that are 100% closed source.

2. All previous (to Sandy Bridge) Intel graphics hardware is pretty much junk. It's not very DX capable or OpenGL capable as far as 3D goes. The chips are riddled with bugs. They can't play even simple games and the performance is awful. Last time I checked the 3D portion of the hardware did not seem to be open source and documented, although that might have changed.

3. I don't believe any of the major 3D chips vendors has a fully open source driver That include Nvidia, ATI, S3, and Imagination Technologies. Where fully open source would mean that the compiler and other tools for their shader units, thread dispatchers, and so on is all open source. Or the code that reads the
EDID via DDI/I2c from the monitor.

Actually these chips will never be fully open source because of licensing constraints for things like HDCP and Blu-ray.

leb


At 1:16 PM -0500 1/27/11, erik quanstrom wrote:
> > I deleted a line from the post that said the big issue will be the
>> quality of the hardware doc, if there even is any.  Still, if it ends
>> up running linux there might be some hope.
>
>the problem is that the linux driver guy is generally getting it straight
>from the source, and not from the docs.  so (from experience)
>existence of a linux driver does not imply that there are docs,
>or if there are, a new driver could be written from them.
>
>> I wouldn't rule out getting doc on everything but the 3D unit.  Even
>> in that case, some of Nvidia's 3D GPUs instruction sets have been
>> reverse engineered.
>
>if you have time for that kind of nonsense, go ahead.  there
>are full specs on the intel video hardware, and nobody has
>done a think with that.
>
>> In case you haven't noticed all of the major 3D chips guys are closed
>> source.
>
>ati?
>
>- erik


--
leb@iridescent.org

Re: [9fans] Cute plan9/inferno client?

[Richard Miller]

>  once had a high level discussion with some 802.11 chip companies about open source drivers. Their biggest fear seems to be in revealing all of the various bugs in their chips. Which I thought was interesting.

Wouldn't it be better to get early silicon and full documentation out
to 3rd party driver writers so more bugs can be found before they get
even more expensive to fix?  Or am I being naive?

Re: [9fans] Cute plan9/inferno client?

[hiro]

> Or am I being naive?
Yes, I'm sure there's more about it, but I haven't studied politics,
so I can't give any details.

Erik, your intel fanboism really bores me.

I love that some people here put a lot of work into other low power,
small form factors. And instead of useless 3d graphic cards perhaps we
will have some kind of FPGA cpu server some day :)

I just don't want to get stuck with intel when there's so much other
neat stuff floating around.

Intel is dying now, more than ever.

Re: [9fans] Cute plan9/inferno client?

[erik quanstrom]

> Erik, your intel fanboism really bores me.

[citation needed]

by the way, i don't think there is any room for
personal attacks on this list.

> I love that some people here put a lot of work into other low power,
> small form factors. And instead of useless 3d graphic cards perhaps we
> will have some kind of FPGA cpu server some day :)
>
> I just don't want to get stuck with intel when there's so much other
> neat stuff floating around.
>
> Intel is dying now, more than ever.

[cititation needed].

there's always "neat stuff" floating around.  unfortunately there
is limited time, and many different things to consider.

i don't think it's unreasonable to not be interested in an overo
they're very slow, have little memory and have poor or no networking.
in addition, i think we still need a number of drivers to make them
useful.

i also don't see any reason to spend any time on nvidia.  arguments
alluding to how other companies have opened up in other areas
don't cut it for me.  show me the docs!  (i couldn't find any on
nvidia's web site.  might have missed them?)

the kirkwood can be interesting.  and i've written a couple of
drivers for it, but it often doesn't fit my needs (not enough
sata ports) and i'm not comfortable depending on it.
the hardware doesn't appear reliable.  it runs too hot.

now the pc platform, for all its many flaws, has answers for all
my questions.
- i can buy nearly any chassis i want and mate it with any power
supply and nearly any motherboard i want.  i can get as many nics
or sata ports as i wish.  i can plug as much memory in as i want.
i can get consumer or workstation grade components.  &c.
- i can run old os code on new architecture processors
- and, it's (mostly, except for stuff like i²c) plug-and-play.

the arm universe doesn't have answers to these questions, and
i think it's going to be an impediment until some of these
questions are answered.

if you want to be a cynic, a non-pluggable architecture is super
for hardware companies.  they can segment the heck out of
the market and get to the bad old says of charging big bucks
for little extra features.  since there's no way to add them yourselves.

- erik

Re: [9fans] Cute plan9/inferno client?

[Venkatesh Srinivas]

[-- Attachment #1: Type: text/plain, Size: 117 bytes --]

>
> The Tegra is probably one of the best ARM chips out there...
>

No it isn't; the Tegra lacks NEON.

-- vs

[-- Attachment #2: Type: text/html, Size: 319 bytes --]

Re: [9fans] Cute plan9/inferno client?

[erik quanstrom]

On Fri Jan 28 16:33:03 EST 2011, me@acm.jhu.edu wrote:

> >
> > The Tegra is probably one of the best ARM chips out there...
> >
>
> No it isn't; the Tegra lacks NEON.

why do you think simd support is key?  we have simd on x86
and ignore it.

- erik

Re: [9fans] Cute plan9/inferno client?

[hiro]

Hey, nothing against you erik, it's more or less the whole pc market
I'm bragging against.
Everyone here should know by now that I've never done anything
valuable, I'm just here to tell my dreams. And in the end everything
will become the exact opposite :D

But still, why should anybody put his time into intel graphics?!

Re: [9fans] Cute plan9/inferno client?

[Venkatesh Srinivas]

[-- Attachment #1: Type: text/plain, Size: 985 bytes --]

On Fri, Jan 28, 2011 at 4:56 PM, erik quanstrom <quanstro@quanstro.net>wrote:

> On Fri Jan 28 16:33:03 EST 2011, me@acm.jhu.edu wrote:
> > >
> > > The Tegra is probably one of the best ARM chips out there...
> > >
> >
> > No it isn't; the Tegra lacks NEON.
>
> why do you think simd support is key?  we have simd on x86
> and ignore it.
>  - erik
>

Doesn't matter for Plan 9, its true. I was thinking more generally. For
media stuff or signals stuff, NEON is totally key for these ARM widgets,
even more than x86en, which can muscle by on OOE and chutzpah.

But even disregarding SIMD... the Tegra was based on the ARM11; the ARM11
isn't a terribly competitive core compared to the Cortex A8 in use by many
of the ARMed chips; the A8 was a wider (dual-issue) design with a
lower-power sleep, bigger/faster cache, and a much better branch predictor.
So I'm not sure sure by what metric you'd say "the Tegra is one of the best
ARM chips out there...".

-- vs

[-- Attachment #2: Type: text/html, Size: 1389 bytes --]

Re: [9fans] Cute plan9/inferno client?

[Lawrence E. Bakst]

At 4:31 PM -0500 1/28/11, Venkatesh Srinivas wrote:
>The Tegra is probably one of the best ARM chips out there...
>
>
>No it isn't; the Tegra lacks NEON.
>
>-- vs

1. I should have said:

 "The Tegra 2 is probably one of the best ARM based SOCs out there ..."

2. It lacks NEON because it doesn't need that ISA. The chip has an integrated GPU which does the same thing a different way.

--
leb@iridescent.org

Re: [9fans] Cute plan9/inferno client?

[Venkatesh Srinivas]

[-- Attachment #1: Type: text/plain, Size: 772 bytes --]

On Sat, Jan 29, 2011 at 5:51 AM, Lawrence E. Bakst <ml@iridescent.org>wrote:

> At 4:31 PM -0500 1/28/11, Venkatesh Srinivas wrote:
> >The Tegra is probably one of the best ARM chips out there...
> >
> >
> >No it isn't; the Tegra lacks NEON.
> >
> >-- vs
>
> 1. I should have said:
>
>  "The Tegra 2 is probably one of the best ARM based SOCs out there ..."
>
> 2. It lacks NEON because it doesn't need that ISA. The chip has an
> integrated GPU which does the same thing a different way.
>

1. Oh, huh, the Tegra 2 is available now? I wasn't aware, sorry about that.
Looks like it is A9-based, yay.

2. If your task is amenable to the integrated GPU, sure; but its not fair to
say that NEON and GPUs (or the OMAP's DSP) do the same thing.

-- vs

[-- Attachment #2: Type: text/html, Size: 1160 bytes --]

Re: [9fans] Cute plan9/inferno client?

[erik quanstrom]

> > 2. It lacks NEON because it doesn't need that ISA. The chip has an
> > integrated GPU which does the same thing a different way.
> >
>
> 1. Oh, huh, the Tegra 2 is available now? I wasn't aware, sorry about that.
> Looks like it is A9-based, yay.
>
> 2. If your task is amenable to the integrated GPU, sure; but its not fair to
> say that NEON and GPUs (or the OMAP's DSP) do the same thing.

nor does it address the problem of programability.
you can get the neon specs.  what about the nvidia
gpu specs?

- erik

Re: [9fans] Cute plan9/inferno client?

[Noah Evans]

Can list members treat each other with greater civility? The list has
been quite nice lately.

Noah



On Sat, Jan 29, 2011 at 12:09 PM, Venkatesh Srinivas
<me@endeavour.zapto.org> wrote:
> On Sat, Jan 29, 2011 at 5:51 AM, Lawrence E. Bakst <ml@iridescent.org>
> wrote:
>>
>> At 4:31 PM -0500 1/28/11, Venkatesh Srinivas wrote:
>> >The Tegra is probably one of the best ARM chips out there...
>> >
>> >
>> >No it isn't; the Tegra lacks NEON.
>> >
>> >-- vs
>>
>> 1. I should have said:
>>
>>  "The Tegra 2 is probably one of the best ARM based SOCs out there ..."
>>
>> 2. It lacks NEON because it doesn't need that ISA. The chip has an
>> integrated GPU which does the same thing a different way.
>
> 1. Oh, huh, the Tegra 2 is available now? I wasn't aware, sorry about that.
> Looks like it is A9-based, yay.
>
> 2. If your task is amenable to the integrated GPU, sure; but its not fair to
> say that NEON and GPUs (or the OMAP's DSP) do the same thing.
>
> -- vs
>

Re: [9fans] Cute plan9/inferno client?

[Ethan Grammatikidis]

On 28 Jan 2011, at 1:36 pm, erik quanstrom wrote:
>
> if you want to be a cynic, a non-pluggable architecture is super
> for hardware companies.  they can segment the heck out of
> the market and get to the bad old says of charging big bucks
> for little extra features.  since there's no way to add them
> yourselves.

there goes my worldview. ;) i had been thinking integrated solutions
were the way to go because they can be so much simpler but i had
forgotten about this possibility. thanks for the reminder.

come to think of it, it's not just a matter of the hardware companies
being able to charge big bucks for extra features. it can also take a
lot of work to add a little extra to a non-pluggable design.

Re: [9fans] Cute plan9/inferno client?

[hiro]

> i can get as many nics or sata ports as i wish.  i can plug as much memory in as i want.

Really? You can upgrade more easily, but there are still limits.

Instead of upgrading S-ATA interfaces, CPUs and RAM, I can plug more
SOCs into my switch. I agree that this is not a very economic option
yet, but it's an old, ongoing trend as I see it.

In the end I only want Gigabit Ethernet, S-ATA, DVI and USB.
Lets see what other pc buses the future brings.

Did I miss anything?

Re: [9fans] Cute plan9/inferno client?

[erik quanstrom]

> Instead of upgrading S-ATA interfaces, CPUs and RAM, I can plug more
> SOCs into my switch. I agree that this is not a very economic option
> yet, but it's an old, ongoing trend as I see it.

yes, people have been doing this with pcs (in 1u/2u/3u boxes)
for so long that all the parts are standardized.

- erik