Sun Niagara 2 CPU Now Open Source

downix writes "Late last night Sun Microsystems announced the immediate availability of the UltraSPARC T2, also known as the Niagara 2 CPU. While we all might not have a silicon fab in the basement, the access to this source code reaffirms Sun's commitment to open source, and in addition gives us FPGA-lovers something new to play with. The source code can be downloaded (with registration) from OpenSPARC.net. Already the previously open sourced T1 has spawned spin-off projects, such as the Simple RISC S1."

Comments

george bush sucks!!!!!

[Anonymous Coward]

see subject

Re:george bush sucks!!!!!

[Anonymous Coward]

then why did he get nearly the same number of votes as Kerry and Gore in the last two elections?

Home fabbing

[QuantumG]

While we all might not have a silicon fab in the basement Does anyone? About how much would such a luxury cost?

Re:Home fabbing

[AuMatar]

Oh, a few tens of millions. Not too much.

If you were making embedded systems you could afford to pay a fab to make a batch, and likely save money over buying a new CPU from Sun/Intel/AMD. But this kind of thing isn't possile or cost effective for one off runs.

Re:Home fabbing

[irondonkey]

http://en.wikipedia.org/wiki/Fab_(semiconductors) Wiki says over 1 billion, probably close, given the relative rarity of them even amongst commercial companies.

Re:Home fabbing

[QuantumG]

Another side effect of the cost has been the challenge to make use of older fabs. For many companies these older fabs are useful for producing designs for unique markets, such as embedded processors, flash memory, and microcontrollers. However for companies with more limited product lines, it's often best to either rent out the fab, or close it entirely. This is due to the tendency of the cost of upgrading an existing fab to produce devices requiring newer technology to exceed the cost of a completely new fab. So it sounds like the second hand obsolete market for fab equipment is a real steal.. and with the industry moving to 300mm wafer sizes soon..........

Re:Home fabbing

[GwaihirBW]

See the sibling post below parent . . . this figure is way low for modern processors. There's a reason that there aren't many upstart processor manufacturers. The fabs are expensive and require significant expertise to work out all the fiddly problems that tend to crop up when dealing with a 65 nm process.

Take, for example, the recent $2.5 Billion Intel plant in China.

Re:Home fabbing

[s_p_oneil]

All you need are Shrinky Dinks, a printer, and a toaster oven. ;-)

http://science.slashdot.org/article.pl?sid=07/12/04/1940203&threshold=-1

Re:Home fabbing

[tlhIngan]

http://en.wikipedia.org/wiki/Fab_(semiconductors) Wiki says over 1 billion, probably close, given the relative rarity of them even amongst commercial companies.

A billion is low-end fabs. High end cutting edge or even near-cutting edge technology costs much more. Maybe a billion for "old-school" tech like 130nm.

No, your best bet is to just pay the few million to have someone fab it for you - there are very few companies that have their own fabs and can do it inhouse (e.g., Intel, IBM, AMD, Freescale (Motorola), Samsung, Toshiba), at least, cutting edge fabs. Low end fabs can be had for cheap (1um and larger), which is great if you don't particularly care about density (e.g., Gemplus - those smartcards have HUGE silicon for 32k memory and not much more).

Most companies are fabless. They contract out the fab work to places like TSMC (amongst others - they're all well known). These include even heavyweight giants like nVidia, Altera, Xilinx and such. The only real downside is that delays can happen if machinery breaks down, or everyone submits a fab order simultaneously that causes backups at the fab and thus delays shipments. The turnaround time (from tapeout to getting chips back) can be 3 months or more. Luckily, most people test their designs out on FPGAs first to work out their bugs before committing them to silicon. Even places like Intel use computer simulation, discrete circuits, FPGAs, and such before they fab it out to their own fabs just because of the turnaround time.

Of course, what I want to know is what's the smallest FPGA one can put this on and still have something workable. (Where things like bus timings and memory clocks still in the realm of "practical" and "in spec").

Re:Home fabbing

[Anonymous Coward]

Look here for fab costs.
http://www.mosis.com/

Packaging is crucial to making the thing work too, however.

CAD tools to convert the RTL into GDS is also very expensive.

Re:Home fabbing

[FrankSchwab]

Actually, the cost for one-offs is significantly lower than your estimate.
By using a Shuttle run, where the fab batches together a bunch of designs and runs them through using a single mask set, you can get 20 or so instances of a 130 nm design for roughly $100K. Of course, this assumes that you've already done the layout and verification steps yourself...

Re:Home fabbing

[Anonymous Coward]

You're a child beater

Re:Home fabbing

[alienw]

I'd say you can do significantly cheaper, at least for small die sizes. I have seen prices as low as $20k for 100+ chips (on a multiproject wafer). Of course, this pretty much depends on the process and on how long you can wait. And the chip layout/synthesis/verification software costs several hundred thousand bucks per year per seat.

Re:Home fabbing

[FooAtWFU]

See the sibling post below parent . . . this figure is way low for modern processors. There's a reason that there aren't many upstart processor manufacturers. The fabs are expensive and require significant expertise to work out all the fiddly problems that tend to crop up when dealing with a 65 nm process.

Take, for example, the recent $2.5 Billion Intel plant in China. Yeah, but the Intel plant is designed to create thousands, millions, of low-power / high-performance devices. A hobbyist might be content with just a few.

Re:Home fabbing

[GwaihirBW]

Er, that does *not* make it easier. Because you do get economies of scale, and in particular, the touchyness of high-precision machinery like this is going to mean the first few wafers through are likely to be useful primarily for fixing calibrations. Now, with some of the nice designs that were (to my knowledge) pioneered by TI around 1993, there are good systems for making small-batch chip facilities that can be switched to new designs rapidly . . . however, the initial setup of the equipment to enable this is still decidedly uncheap. Really, you want to be producing a stream of chips to justify your setup investment, not a small one-off batch.

Now, if you simply buy an old one, that's pretty cheap (relatively speaking) - Intel etc have more of these than they know what to do with, because they keep pushing the technology and obsoleting their relatively recent investments. However, this won't help you build any of these new designs, all of which are 65nm process.

Re: Home fabbing

[the_kanzure]

An email I just sent to QuantumG:

On Wednesday 12 December 2007, QuantumG wrote:
> Maybe one day it won't be silly :)

Absolutely. I'm doing some research in the background. Try this:

Litho-history [PDF]
Simple Lithography [blog]
News story, semirelevant

query: lenses circuits wax Intel
result: MY OWN PAGE *argh* (re: analytical instrumentation)
I also get: "Bryan's page on semiconductor manufacturing" which is also,
alas, me. But the page contains just as far as I've been able to get so
far: how to take sand and extract silicon from it, into giant crystals
that you can pull. The next step is to use diamond saws or diamond wire
to cut the silicon crystals into the wafers that you start with. The
diamond wire is very costly: it apparently degrades after a few uses.
So I haven't been able to figure out how to make my own diamond wire
yet. Another option might be to use high-powered lasers to cut the
silicon. Not sure about that one. Then, the next step is to go back
over Wikibooks book re: photolithography and the chemical etching
process, which I also have on my /notes.html file on my website. I need
to flow chart the whole operation soon.

How can we achieve fabbing@home?
- distribute the components across a city, absorb cost
- ship each incremental stage to different members
-- costly packaging, but pricing absorbed by people able to afford each individual component of the stage.

- Bryan

Re:Home fabbing

[afidel]

The software's expensive but it's not THAT expensive. Four years ago when I supported Cisco the cost per seat for the design software was in the mid five figure range and support/updates were 20% annual. Kind of made workstation pricing irrelevant, when the designer is making 6 figures and the software is that expensive a $25K workstation that would be amortized over 2-3 years was cheap.

Re: Home fabbing

[monsted]

How can we achieve fabbing@home? You go to playboy.com, get out your lotion and tissues and... oh, faBBing... Never mind me.

Re:Home fabbing

[Criton]

I'm surprised the government allowed Intel to take 65NM fab tech out of the country and to china of all places that really should not have been allowed and landed some big wigs in Leavenworth because it should have. BTW if nano assemblers tiny assembly lines for nanotech not to be confused with nanites which is pure scifi ever become possible yes home fabbing would become possible maybe even affordable. BTW the niagra chip even when built with 130nm tech still can do more flops per watt then intel's core architecture.

turds now tasty

[Anonymous Coward]

A few years ago, while browsing around the library downtown, I
had to take a piss. As I entered the john a big beautiful all-American
football hero type, about twenty-five, came out of one of the booths.
I stood at the urinal looking at him out of the corner of my eye as he
washed his hands. He didn't once look at me. He was "straight" and
married - and in any case I was sure I wouldn't have a chance with
him.

As soon as he left I darted into the booth he'd vacated,
hoping there might be a lingering smell of shit and even a seat still
warm from his sturdy young ass. I found not only the smell but the
shit itself. He'd forgotten to flush. And what a treasure he had left
behind. Three or four beautiful specimens floated in the bowl. It
apparently had been a fairly dry, constipated shit, for all were fat,
stiff, and ruggedly textured. The real prize was a great feast of turd
- a nine inch gastrointestinal triumph as thick as a man's wrist.

I knelt before the bowl, inhaling the rich brown fragrance and
wondered if I should obey the impulse building up inside me. I'd
always been a heavy rimmer and had lapped up more than one little
clump of shit, but that had been just an inevitable part of eating ass
and not an end in itself. Of course I'd had jerk-off fantasies of
devouring great loads of it (what rimmer hasn't), but I had never done
it. Now, here I was, confronted with the most beautiful five-pound
turd I'd ever feasted my eyes on, a sausage fit to star in any fantasy
and one I knew to have been hatched from the asshole of the world's
handsomest young stud.

Why not? I plucked it from the bowl, holding it with both
hands to keep it from breaking. I lifted it to my nose. It smelled
like rich, ripe limburger (horrid, but thrilling), yet had the
consistency of cheddar. What is cheese anyway but milk turning to shit
without the benefit of a digestive tract?

I gave it a lick and found that it tasted better then it
smelled. I've found since then that shit nearly almost does.

I hesitated no longer. I shoved the fucking thing as far into
my mouth as I could get it and sucked on it like a big brown cock,
beating my meat like a madman. I wanted to completely engulf it and
bit off a large chunk, flooding my mouth with the intense, bittersweet
flavor. To my delight I found that while the water in the bowl had
chilled the outside of the turd, it was still warm inside. As I chewed
I discovered that it was filled with hard little bits of something I
soon identified as peanuts. He hadn't chewed them carefully and they'd
passed through his body virtually unchanged. I ate it greedily,
sending lump after peanutty lump sliding scratchily down my throat. My
only regret was the donor of this feast wasn't there to wash it down
with his piss.

I soon reached a terrific climax. I caught my cum in the
cupped palm of my hand and drank it down. Believe me, there is no more
delightful combination of flavors than the hot sweetness of cum with
the rich bitterness of shit.

Afterwards I was sorry that I hadn't made it last longer. But
then I realized that I still had a lot of fun in store for me. There
was still a clutch of virile turds left in the bowl. I tenderly fished
them out, rolled them into my handkerchief, and stashed them in my
briefcase. In the week to come I found all kinds of ways to eat the
shit without bolting it right down. Once eaten it's gone forever
unless you want to filch it third hand out of your own asshole. Not an
unreasonable recourse in moments of desperation or simple boredom.

I stored the turds in the refrigerator when I was not using
them but within a week they were all gone. The last one I held in my
mouth without chewing, letting it slowly dissolve. I had liquid shit
trickling down my throat for nearly four hours. I must have had six
orgasms in the process.

I often think of that lovely young guy dropping solid gold out
of his sweet, pink asshole every day, never knowing what joy it could,
and at least once did, bring to a grateful shiteater.

Re:turds now tasty

[Anonymous Coward]

tucker max!

I just witnessed a minor miracle, I think

[Anonymous Coward]

Someone corrected the spelling of "Niagra" to "Niagara" - ScuttleMonkey, if it was you, I congratulate you!

Honestly :) Good job!

Re:I just witnessed a minor miracle, I think

[eulernet]

Yes, and I read Viagra. It's time to go to bed !

Re:I just witnessed a minor miracle, I think

[Enry]

Niagara Falls!

Slowly I turned....

Re:I just witnessed a minor miracle, I think

[Anonymous Coward]

Step by Step, Mile by Mile...

Openbsd

[Anonymous Coward]

I can remember when the OpenBSD crew was having issues getting sparc specs. My how times have changed.

Re:Openbsd

[Dan Ost]

I suspect that this would never have happened if it weren't for the efforts made by the OpenBSD crew to get information out of Sun.

Any insiders want to post anonymously about how we got here with Sun?

Re:Openbsd

[fm6]

I doubt if the OpenBSD thing has anything to do with this. It's more in response to the fact that it's hard to sell SPARC-based systems as long as the CPU is perceived as a Sun-proprietary technology. So Sun opens up the SPARC design, and this allows them to claim that their chips are "commodity", just like AMD and Intel's.

Re:Openbsd

[chriscappuccio]

No, OpenBSD still can't get the specs. It's very odd that Sun open-sources their CPU design, yet they still keep the rest of the chips so proprietary that you can't even get software documentation to program them. OpenBSD (and everyone else) still can't get basic specs to anything beyond the CPU itself. All support of newer Sun systems is done by reverse engineering OpenSolaris and Linux code.

Re:Openbsd

[AKAImBatman]

The Microprocessor's instruction set has been open for decades. It's all the hardware around the SPARC processor that OpenBSD had trouble getting info on. Sun used to make a huge number of hardware devices for which they provided no documentation on the internals. Of course, these days, most of it is pretty standard. But back then, hearing words like "SBUS" used to make people shudder.

Re:Openbsd

[Anonymous Coward]

No you can't. There's never been a problem getting sparc specs. It's been an openly published spec for decades. You are confused.

Re:Openbsd

[Anonymous Coward]

The OpenBSD people will always have issues. They expect the world to kneel to their greatness and kiss their asses.

Open Source friendly?

[LingNoi]

From the criticisms I have been reading over the net, Sun has been hiding information on how to get their hardware working in Linux. Hardly what I would call a committed player to open source software. Am I wrong on this?

In any case what they have just done sounds more like they just want people to peer review their work rather then release anything useful. I wonder what they'd do if someone started selling processors based on the information they just released.

Re:Open Source friendly?

[georgewilliamherbert]

I wonder what they'd do if someone started selling processors based on the information they just released.

The RTL code (Verilog) is GPLed:
http://www.opensparc.net/faqs/licensing/

Other people have built and are shipping product with the prior T1 version, the SimpleRISC folks:
http://www.srisc.com/?s1

The licensing pretty much says "Here, have it, have fun!"

Re:Open Source friendly?

[EvanED]

Sun's in sort of a weird position. They have largely embraced OSS (witness the fact that they open sourced most notably SPARC, OpenSolaris, and Java), but I think are not entirely at ease with it. From at least one point of view, they have a lot to lose from Linux taking over for Solaris. OpenSolaris is also under the CDDL, which is not GPL-compatible. I wouldn't be surprised if the incompatibility is viewed as a benefit by a lot of Sun, rather than a bug.

So they may not be perfect, but they are a heck of a lot better than most companies.

Re:Open Source friendly?

[tcampb01]

Yes, I think you are wrong on this.

The processor is open-source (OpenSPARC), Sun's own operating system port on the processor is also open-source (OpenSolaris), and of course Linux is open-source. If someone wants to go through the labor-of-love to port Linux to SPARC then there's nothing "hidden".

Sun sells x86-based machines that run Linux and I think they'll even sell you a Linux distro to go with the box (of course you don't need to get from them... any distro will do.)

I fail to see the conspiracy angle.

Re:Open Source friendly?

[davecb]

I suspect the more stringent licenses are for things that can be embraced and extended, possibly by a particular well-known competitor to Java (;-))

--dave

Re:Open Source friendly?

[Anonymous Coward]

Sun licensed SPARC to Texas Instruments, Cypress Semiconductor, and Fujitsu and used their chips in their machines.

I've head the OSS guy at Sun say that they are happy for anyone to build them and they'll be happy to buy them from whoever gives them the best deal.

Re:Open Source friendly?

[init100]

OpenSolaris is also under the CDDL, which is not GPL-compatible.

Isn't it also released under the GPLv3? Anyway, I'm pretty certain that they chose both CDDL and GPLv3 expressly to be incompatible with the Linux kernel, which could otherwise have all their nice features (like ZFS) leaving OpenSolaris with no advantage over Linux.

Re:Open Source friendly?

[QuantumG]

I'd love to know where people like you get this kind of insight from.. Sun is an open source company.. they always have been. They were doing open source before "the revolution" even happened. They're more than "at ease" with it, they have some of the oldest living open source hackers on payroll.

As for this whole "oh noes!! Linux might takeover from Solaris!!" crap, why do people continue to ascribe the Microsoft world view - everything and everyone is a competitor - to other companies? Sun and IBM and every other normal company (read: not a monopoly) has one business strategy: give the customer what they want. If the customer wants Linux, Sun will sell them Linux. If the customer wants Solaris, Sun will sell them Solaris. If the customer wants Intel or AMD or SPARC, Sun will sell them that. IBM will also sell you Linux or Aix or Intel or AMD or SPARC.. if you want to pay them for that, that's what you'll get. It's only Microsoft who seems to think they can dictate the solutions to the customers instead of the other way around.

Re:Open Source friendly?

[EvanED]

If the customer wants Linux, Sun will sell them Linux. If the customer wants Solaris, Sun will sell them Solaris. If the customer wants Intel or AMD or SPARC, Sun will sell them that. IBM will also sell you Linux or Aix or Intel or AMD or SPARC.

Yes, this is true. But I bet that if you asked them, they would prefer to be selling you Sparc and Solaris, and would prefer if Linux were not as big as it is.

Re:Open Source friendly?

[QuantumG]

Not really. Obviously they would prefer that the solutions they have expertise in be the ones wanted by the customer, so they can best serve the customer, but that's about it.

Re:Open Source friendly?

[Anonymous Coward]

Sun is definitely opening up their hardware specs / documentation to the public. All you have to do is ask. A lot of documentation is out there already. See the following for details:

http://wikis.sun.com/display/FOSSdocs/Home This is their FOSS Open Hardware Documentation information that has all the info OpenBSD and Linux developers have been asking for about the processors and supporting chipsets for things like Gigabit Ethernet, PCIe root complexes, PCI bridges, UPA bridges, and even their brand new 10 Gig Ethernet Chip, Neptune.

You can request more docs on this page. So if there's any Sun hardware that's not supported in your favorite open source *NIX, request the hardware docs there!

2.6.23 or 2.6.24 has a Sun Neptune Network driver added to it (presumably coded using these docs).

http://www.sun.com/processors/documentation.html (sun.com) This is more Hardware documentation on their chips some of which is linked from their wiki

http://blogs.sun.com/barton808/entry/let_there_be_docs (blogs.sun.com) Here is the background information on the OpenDocs Project.

Re:Open Source friendly?

[Anonymous Coward]

They may prefer BSDs from a technical aspect, though, since SunOS was originally a BSD varient and they have long had a good relationship with the FreeBSD community (FreeBSD gjails->Solaris Zones, while FreeBSD were the first to adopt DTrace and ZFS). Otherwise, they'll happily support Microsoft if you really want them to.

Re:Open Source friendly?

[JustPutt2]

Well said,,,,, The new Open System memtality has benefited all users, and developers. With the exception of MS

Re:Open Source friendly?

[FranklinDelanoBluth]

When did "the revolution" happen? I didn't know about it because it wasn't televised.

Re:Open Source friendly?

[Anne Thwacks]

I AM the customer, I want Sun to support their own video cards. I want them to support the Creator3D and the ATI rage cards they sell with Sparc64 systems so that they work properly in text mode as well as grapics mode at boot time in the OpenBoot prom. Not, I would argue, very demanding.

(Actually I'd like them supported on OpenBSD, which they now claim to support, as well)

Re:Open Source friendly?

[mha]

"to want" is defined as "willing to pay" in economy terms. Should I start a looooong list of all the things I would like? This is so silly... are you willing to spend an afternoon working my garden in exchange for the small favor you want form me - THAT'S the question, or do you want something from me without doing anything in return?

This is what I HATE about open source. Too many kids grow up with that "I get everything for free" mentality. And don't try to argue they give back to open source - what's the number of people USING open source vs. those actually contributing? Hint: Posting in discussion forums is NOT a contribution (or very very rarely, even most replies to "support questions" are useless).

I'm all for openness and liked Sun's open source Tcl/Tk *a lot* (I wrote a pretty complete and nice looking file manager GUI in ca. 200 lines of code) long before Linux was hot, but I'm against working for free and even more against trying to establish that expectation for SOME people (strange, none of those "free as in free beer" folks expects not to pay for food or their haircut). Has nothing to do with the original ./-subject we're posting under, but couldn't resist...

Re:Open Source friendly?

[Raenex]

why do people continue to ascribe the Microsoft world view - everything and everyone is a competitor - to other companies? Sun and IBM and every other normal company (read: not a monopoly) has one business strategy: give the customer what they want. I think you need to remove your anti-Microsoft colored glasses. Every company is aware of their competitors. Every company would love to lock their customers in to proprietary solutions. That's the normal position. IBM was sued for being a monopoly decades ago. Sun competes with Linux. IBM embraced Linux and Sun embraced open source only because they were losing out in their old ways, and they wanted to ride the new waves. They both still have proprietary products. Microsoft only stands out from Sun and IBM because they have achieved market dominance, whereas IBM lost it and Sun never quite attained it.

Pics?

[Anonymous Coward]

Pics or it never happened!

Re:Pics?

[riff420]

This is not the Something Awful Forums. FYAD.

Re:Pics?

[kakofb]

insightful commentary itt

you cant hide what doesnt exist

[Anonymous Coward]

You are implying that they know how to get it working in Linux. Do you really think they put resources into that .. into figuring it out and or documenting it? For what purpose??

Re:you cant hide what doesnt exist

[LingNoi]

I am saying that from what I have been reading that people are complaining about them [Sun] not releasing information they [open source programmers] need to get their hardware working in Linux.

This is...

[larpon]

Sun Viagra 2 CPU... Ok.. I need glasses

Re:This is...

[Anonymous Coward]

Sun Viagra 2 CPU... Ok.. I need glasses

Sun should market it as such, after all, you never want your server to go down.

Re:This is...

[larpon]

I woooon't let the sun go down on... argh forget it..

Re:This is...

[Anonymous Coward]

You can imagine how many times we made that joke over the last few years.

Re:This is...

[Anonymous Coward]

If your server's up for more than 4 hours, consult a physician?

Re:This is...

[geschild]

"Sun Viagra 2 CPU... Ok.. I need glasses"
I didn't know far-sightedness is a known side-effect of V1@9r@?

Blindness is, though...

Time to see a medical professional? ;D

Re:This is...

[amias]

at last stiff competition for AMD and Intel.

Re:This is...

[Anonymous Coward]

Unfortunately if they marketed it as such, you'd have to contact your administrator if your server stays up for more than 4 hours.

Too bad...

[Anonymous Coward]

SPARC kinda bites... there are other RISC processors that are more interesting.

Re:Too bad...

[downix]

I prefer MIPS, but SPARC is a close second for favor, mostly due to the open source nature of the design. This is no less than the 4th major SPARC CPU in open source.

Re:Too bad...

[Anonymous Coward]

Yes, but while the register windows were an interesting research project, my opinion is that they weren't worth the complexity and wasted resources they became. After all, you can only make so many calls before the windows run out and then it's similar to a regular CPU but still paying the complexity of having to dork with the windows. And with so many registers, it probably would have been more interesting to have all of them available to the programmer/compiler than just a small subset (and the compiler could have done windowing on its own if it wanted instead of being forced by hardware) with all the hardware complexity.

SPARCs never were about performance though (although there have been some that had reasonable performance for their time). I also liked the MIPS processors as far as load/store went. Other than that, Alpha was OK. I cut my teeth on the M68K family so I guess I'll always like them :)

I worked with Suns from the 3/50 days, 4, and then SPARCs up until the UltraIIIs (desktops and servers). I always liked the machines (they were usually very stable, there were notable exceptions but most of the time they were solid machines). Good workhorses but not necessarily racehorses.

I actually have the spindle/platters from an old IPI drive (500M or so, IIRC) sitting in my living room, pulled out of drive when we were decomissioning all of our IPI drives.

Re:Too bad...

[downix]

Well, that "too complex" windowing capability is one of the reasons why the Niagra core can run 4 threads simultaneously. It has 4 windows, and swaps them automatically, enabling a deeper pipeline without breaking the ISA. So, the compiler sees ALU's rather than 1.

In addition, I rolled my own 32-bit SPARC once with only 2 register windows, with the compiler did not control. What you had was a "program" window and an "interrupt" window. So when interrupts happened, rather than having to save the register state before dealing with it, you immediately switched windows and bam, have a completely clean slate of registers to work with.

Re:Too bad...

[wik]

At least in T1, there are 8 windows for each strand (thread), so 32 windows total. The register file design supports fast window switches within a strand (3 cycles) and can switch between strands on every cycle. The megacell guide included with the RTL has a detailed description of how this really works.

The T1 pipeline is just 6 stages (not terribly deep), while the T2 just adds a bypass stage.

Re:Too bad...

[Anonymous Coward]

Well, that "too complex" windowing capability is one of the reasons why the Niagra core can run 4 threads simultaneously.

No... it isn't. You need four sets of register files to run 4xSMT, one for each context. You can do this with non-register-windowed processors just fine. There have been many, many 2xSMT non-SPARC processors sold (many millions). Going from that to 4xSMT isn't rocket science. Trying to do all this in one register file (windowed one even) is even more complex than that. I'm not sure how they'd do it.

Yeah... windowing is fine (although there are many registers that aren't currently being used, lots of 'wasted' resources, IMO) until you run out of windows (write a recursive function or have a deep call stack). Then what does the it have to do?

Your description of your own 32-bit 'SPARC' isn't that accurrate. That's simply a set of 2 register files, one for one mode, another for system/interrupt mode. That's done in other systems already (and has been for ages). Windowing is what SPARC does with that 192 registers (or however many now) with the overlapping windows and only being able to see 32 of them in your program at any one time (although they may be a different physical registers).

Relevance to Joe Consumer

[flowerp]

I kind of wonder what the relevance of the availability of the
blueprint of a modern multithreaded special-purpose server
CPU means to the average Joe.

Probably not much, unless Joe has got an degree with a specialization
in computer science or electrical engineering.

Re:Relevance to Joe Consumer

[forkazoo]

I kind of wonder what the relevance of the availability of the
blueprint of a modern multithreaded special-purpose server
CPU means to the average Joe.

Probably not much, unless Joe has got an degree with a specialization
in computer science or electrical engineering.

The vast majority of (bachelors level) computer science degrees don't involve anywhere near enough focus on hardware issues for the "blueprint" of their CPU to be of any real use. The low level source of a CPU is of direct use to a vanishingly small subset of people. But, so is the source of the Linux kernel. I've never submitted a patch to the kernel. I wouldn't know where to start, frankly. And, I'm moderately qualified to do so, having done a fair amount of C, and a bit of embedded programming. I'm certainly more qualified to tinker with the kernel than I am with CPU source.

But, that sort of isn't the point. The fact that you and I wouldn't know where to start with something like that doesn't change the fact that such people do exist. And, there are some people who can't do anything with it, but are really curious to know more about what it is, and this may be the spark that makes them decide to learn. You and I may get the result of one of those guys having access to this. so, even though my own project plans won't be influenced by the availablity, I do expect that you and I will be effected by it indirectly.

Re:Relevance to Joe Consumer

[nurb432]

It might end up driving costs down for alternatives to intel/amd.

It might not too, but its a least something to consider.

Re:Relevance to Joe Consumer

[init100]

You and I may get the result of one of those guys having access to this. so, even though my own project plans won't be influenced by the availablity, I do expect that you and I will be effected by it indirectly.

Which, incidentally, also applies to open source software. Many people sneer at open source software, saying that they and 99% of all people wouldn't know what to do with it anyway, so why should they care. You just wrote a nice explanation for those people.

Re:Relevance to Joe Consumer

[alain94040]

Sun is opening the source code of their processors because they are losing the battle against x86 processors. Their marketshare of workstations and servers (not to mention PCs) is crashing. They hope (it won't happen) to get some momentum, maybe some second-tier vendor will build their next chip using a T2. I believe a few did for the T1 (the previous open source processor from Sun). It made no difference in marketshare.

If (it won't happen) more chips use Sparc cores inside, then the need for Sparc software and knowledge will grow, and Sun will eventually profit. They don't want the Sparc ecosystem (whatever is left of it) to die.

So I'll thank Sun for releasing such a big piece of RTL. In the field of hardware design, I think it's the largest open chip available. It means EDA tools will have one more testcase, their quality will improve. And that, in the end, will help the entire semiconductor industry. But not Sun especially. I'd recommend to sell your SUN stock if you have any.

Re:Relevance to Joe Consumer

[Anonymous Coward]

Yes indeed, the open sourcing of another SPARC processor will indeed hurt Sunoco, Inc, won't it? (hint, Sun Microsystems stock ticker is JAVA)

Re:Relevance to Joe Consumer

[NovaX]

While software folks may not understand the hardware world, its quite sad that hardware folks rarely understand the software side as well. One of the most challenging jobs, which gets little attention, is software-hardware codesign. Those applications, like Cadence VLSI suite, are quite challenging as they require EE expertise to implement features while software mastery to develop the product. This results in very advanced, but also very cryptic, software stacks.

I don't think open source hardware is too interesting or valuable, but I really hope software developers will feel comfortable enough to begin reading through HDL code just like they do their favorite open source project. Verilog/VHDL are both fairly old languages, though capable of doing some absolutely amazing tasks (I was in awe the first time I compiled my VHDL chip into a VLSI layout). It would be a really great to see language gurus apply the same innovative spirit towards modernizing HDL languages as they currently do in trying out different techniques in software languages (Haskell, Erlang, etc).

Re:Relevance to Joe Consumer

[spir0]

While software folks may not understand the hardware world, its quite sad that hardware folks rarely understand the software side as well.

Not any more. back when coding apps/games/demos in assembly was the thing to do, you had *plenty* of guys that knew the hardware and software inside and out. And while they're only 30-40 years old now, it's just a redundant practise.

Re:Relevance to Joe Consumer

[GennarinoParsifalle]

Zero for consumer direct use, agreed.
But there are companies that use the open sourced design to produce compatible processors: Simply RISC is one and, IIRC, also the Leon2-FT from ESA is a SPARC-derivative processor. There are many others around that, thanks to the GPL, are too open source.
If engineering cost of processors can be "shared" there will be good coming for consumers too. After all, why not a SPARC T1 derivative in your next phone?

Re:Relevance to Joe Consumer

[nebosuke]

Not really. All serious competitors in demo competitions know the ISA and performance characteristics of the target architecture very well, but that is nowhere near knowing the hardware 'inside and out'. It's the difference between being very familiar with the API vs. the actual code implementing a library.

Re:Relevance to Joe Consumer

[RobBebop]

I've never submitted a patch to the kernel. I wouldn't know where to start, frankly.

I have also never submitted a kernel patch, but I want to try to answer your excuse. Start at kernel.org. Read through the Bugzilla Open Issues. I have read this book and it does an excellent job introducing the tools and techniques needed to work with the kernel.

To attack the meat of the article (Open Source Sun CPU), it is valuable because it gives the specialized community a rallying point to get behind. There might be less than 100,000 people qualified to do anything meaningful with this... but I assume you that the majority are not within the Sun umbrella and thus the release serves to benefit the majority of those who are qualified. And I would guess that a handful of the people who would care are old enough to be retired and would take an interest purely as a hobby that they wouldn't be able to do otherwise.

The other value is that as a student of computer science and engineering who graduated in '01... I haven't dealt with ANY of this type of work and that is a damned shame. A single university probably can't afford the $1B price tag, but I would bet a consortium of schools (MIT, CalTech, Stanford, CMU, Northeastern, RPI, UChicago, UPenn, et cetera) would jump on this. Could 20 schools inject $50M each? I think so. Add the value of educating 50 students per year at each of those schools (1,000 per year) would overvalue the $1B initial cost. After five years, the average cost per student would be as low as $50,000 because economies of scale would kick in (and I think you would agree that I have used extremely modest estimates).

So yeah, this is a VERY GOOD THING (TM).

Re:Relevance to Joe Consumer

[spir0]

the C64's and early Amigas were pretty static, hardware-wise. so once you knew what the CPU and custom chips did, you knew the hardware.

Commodore released excellent books in the way of the ROM Kernel Manuals for the Amiga.

what was the topic about? :)

Re:Relevance to Joe Consumer

[Anonymous Coward]

The topic was the insides of the chips, not their software interfaces. I have heard there were wiring diagrams in the C64 reference manuals, but there certainly weren't any in the Amiga reference manuals I have seen. Show me one demo coder who has reimplemented or improved the Amiga custom chips in hardware.

When is Open Source actually news?

[br1an.warner]

This does bring up a point. You can open source anything (and probably pick up some decent press off of it these days). But in a case like this, what does it really mean? It gets you points from those who either 1) don't understand you need a fab to do anything with this, 2) don't understand why you would open source a project, or 3) don't read the article. So really, if nobody's going to contribute changes, make a derivative work, or build one of these things from scratch, does this really mean anything at all? I'm inclined to say no.

Re:When is Open Source actually news?

[QuantumG]

Dude. There's people who have made derivative works mentioned in the summary. You can't even be bothered reading the summary before posting? Wow.

Re:When is Open Source actually news?

[JebusIsLord]

It means some other (large) company can take the design, improve it, release a product based on it without paying huge fees, and then contribute their changes back to the public domain. So no, it doesn't help joe user, but it does help industry players save on R&D, which is still a good thing.

Re:When is Open Source actually news?

[ChrisA90278]

don't understand you need a fab to do anything with this,

Not quite.... One can burn this into an FPGA. I don't know how fast it would run but if the goal is to study and experiment with processor design then an FPGA is the tool. The purpose ere is to allow people to study and modify the CPU

Wikipeadia of course has some info
http://en.wikipedia.org/wiki/Field-programmable_gate_array

Re:When is Open Source actually news?

[nahpets77]

Not to mention that soft-core processors are quite useful (for FPGA or ASIC). See the Xilinx Microblaze and Altera Nios II, which are commercial offerings. There is a large demand for open hardware (see http://opencores.org/ and a processor core is an important building block. Just because you may never read the source, you can still instantiate them in your design and use them. How many of you have looked at the glibc source code? Doesn't stop people from using it.

Re:When is Open Source actually news?

[Anonymous Coward]

your average fpga isn't going to have enough gates (or whatever they're called) for this. the risc s1 had to be compiled with a commercial compiler before it would fit onto an fpga and they've dropped 7 of the cores.

21st century business plan

[karmaflux]

1. Release unsalable product as open-source 2. ??? 3. Profit

Re:21st century business plan

[pimpimpim]

I do wonder how Sun is going to make money the next century. They're trying new stuff, like opening up all their sources (java, solaris), but the money doesn't seem to be flowing back from these actions yet. Don't get me wrong, I think this might end up being really good for them selling the CPU in other devices, but I also think Sun makes pretty good products, and one of the last alternatives to the X86s, (for a huge price) and it would be sad if they went down.

Re:21st century business plan

[dupup]

I do wonder how Sun is going to make money the next century.

Disclosure: I work for Sun in their software division. This is not secret information, but Sun plans on making money in the next century by selling hardware. Lots and lots of hardware. Why buy it from Sun when you can get it cheaper from elsewhere? That's the other part of Sun's super secret master plan: support contracts. Business do tend to buy from Sun if they have already done so. Maybe it's just easier, maybe it makes the original decision to buy from Sun look better, I don't know. But Sun still sells $billions in hardware each year. The software revenues are a whole lot less.

Re:21st century business plan

[jcnnghm]

Hardware. I transitioned all of my company's servers to Sun (when they started selling x86 servers) over the last few years, and I couldn't be happier. The equipment is more expensive, but it is also of a superior quality and features real enterprise management features. Opening up Solaris had a lot to do with my initial decision, and I wouldn't be surprised if we don't transition to the T2 in the future. Since all of our servers are virtualized anyway, it isn't really a bad idea to move lots of small servers to a few big servers, and this open architecture will undoubtedly give some experts the opportunity to publish very in depth analysis of the chips.

Re:21st century business plan

[Anonymous Coward]

This is quite easy

Open sourcing a CUP is a great idea financially. Geeks tweak the hell out of things for you so you don't have to pay expensive engineers. Then you print the chips and sell them to the designers (and others)

Re:21st century business plan

[Anonymous Coward]

I'd love to tweak a CUP by putting some tea in it. If I'm feeling radical I'll pour grape soda into it. But I do have one question - how do you print a chip on a cup? That doesn't make any sense. Is there some kind of new pringles printer or something?

Re:21st century business plan

[TheRaven64]

I don't work at Sun, but it's pretty clear that their business model has only changed very slightly. What they have always sold is vertical integration; you buy the hardware, software, and support from Sun and they guarantee that the hardware and software work well together, and the support includes people who can fix either if they don't.

The difference is that they are now marketing a lack of lock-in as a value-added service. Previously, migrating away from Sun was hard because you had to port your code from SPARC/Solaris to POWER/AIX, PA-RISC/HP-UX, or whatever. Now, they give away all of the individual components. If you are willing to invest in some fab time, you can make your own SPARCs (not for the first time, but I believe the T1 was the first SPARCv9 to be open sourced). You can download the source code for Solaris and modify it. You can buy the support from anyone willing to do this. The only difference is, you'll be buying support from people who have used the tools, rather than from people who built them.

Joel Spolsky criticised Sun for commoditising their own product, when smart businesses commoditise their products that complement theirs. He missed the point, however, that Sun are not selling products anymore, they are selling expertise. It is in their best interests to commoditise all of their products, because their products are the complements of their expertise.

Is the hardware any good though?

[LingNoi]

Awesome, do you happen to know how well these things work compared to other processors?

My only experience with Sun hardware is from the slow out dated machines we have running here.

Re:Is the hardware any good though?

[_merlin]

It's basically a multi-core barrel processor - it switches threads on every cycle (similar to the old Cray and CDC gear). This gives it time to fetch data while other threads are being serviced to reduce the likelihood of cache misses. This makes it ideal for highly parallel workloads, like web application servers and multi-user database servers. But the workload needs to be highly parallel for it to perform: a quad core Niagara needs at least sixteen threads to be fully utilised. It has weak floating point, too, since web application and database workloads don't usually depend heavily on this. It's pretty much useless for a typical desktop or workstation workload.

Re:Is the hardware any good though?

[downix]

In my experience, in a single threaded task, an AMD or Intel of the same spec will blow it away. Crank up the threads, and something remarkable happens. The AMD/Intel stays running the same speed, this thing speeds up. Run 4 threads simultaneously, and this thing is running circles around similar spec'd AMD/Intel, and doing it in a lot less silicon, which means theoretically a lower price. Think of it as a marathon runner vs a sprinter. In short term races, the sprinter comes out on top. In a long distance (heavy load) race, the marathon runner comes out on top.

Re:Is the hardware any good though?

[jd]

Then I guess one area that open source hardware nerds can work on is to rework the FPU to be faster. If people can load the files into an FPGA and get comparable results on maths-heavy software to a full hardware implementation of the T2 as it stands, I would imagine the scientific computing folk would go for the FPGA solution as it would be cheaper so they could build more nodes for the same amount of money.

Sun is a lot of things, some unprintable, but stupid isn't one of them. If it can be shown that a T2 with stronger maths will sell better than the T2 as-is, then you will see a T2 with stronger maths in very short order.

There are other things Sun could include in the processor. I am, to this day, a devoted fan of direct CPU-to-CPU channels for multi-processor systems. Inmos' Transputer let you build hypercubes of processors as large as you liked without scaling issues. Xyron's ZOTS also seems an interesting technology, even if nobody uses it at this time. The potential wishlist of things that could be added without wrecking the design is large. Given that the core is GPLed, it would seem to make sense to experiment with some of those ideas. See what would actually work in practice, with the possibility that some vendor (not necessarily Sun) will chase the idea and turn it into a reality.

Re:Is the hardware any good though?

[EvanED]

Then I guess one area that open source hardware nerds can work on is to rework the FPU to be faster.

It's not so much that the FPU is slow as, at least on the T1 (Niagra 1), there was only one of them for the whole chip. The applications the Niagra targets don't really need FPU power (how much FP work does /.'s webserver do?), and Sun is fairly close to production of the Rock, a processor that in some sense is similar to the Niagra but will also have much heavier FP capabilities. (In a development that is pretty exciting for the architecture people down the hall, the Rock will be the first commercial system supporting transactional memory.)

Re:Is the hardware any good though?

[Anonymous Coward]

It is fancier than a barrel processor, there are two integer pipes per core and the thread picker dynamically picks threads. In other words a single thread can go down the pipe more often than once every eight cycles. It is true though that you'll get more work done on a core with several threads running than with just one.

The floating point unit is pretty powerful as well, the benchmarks are quite good.

And it works fairly well as my desktop, though I use it via VNC. I don't run any heavy graphics on it though.

Re:Is the hardware any good though?

[dr2chase]

(disclaimer, I work for Sun, but I manage to be completely clueless about many things, including the official names of Sun's products)

Niagara 1, had one FPU per chip. Niagara 2, has more than one.

The way you pose it -- doesn't perform unless you can find the parallelism -- is not the right way. Some clever person found a market where there was parallelism, and that turns the problem around. "Given that I have all this work to do, what's the throughput per watt?" Niagara wins there. And it happened that those people, or a lot of them, didn't have a burning need for floating point.

Or to use a lame car analogy, a schoolbus is no good unless you can find 32 kids to haul, whereas a minicooper is cool and zippy with only two. But if you regularly have 32 kids to haul, and some people do, you want a schoolbus, and a minicooper is not very efficient.

Re:Is the hardware any good though?

[afidel]

On the T2 there is one FPU per core, the new FPU's are also accessed in 6 cycles vs the 40 of the T1 and the full ISA is done in hardware (some rare media instructions were emulated in the T1 FPU). The biggest win for the typical customer though is the enhanced crypto performance, now you can do SSL at the rate that the T1 could do normal HTTP.

Re:Is the hardware any good though?

[Anonymous Coward]

I work for Sun, but I manage to be completely clueless about many things

"but"?

Re:Is the hardware any good though?

[dr2chase]

Well, product names, I am especially clueless. I have few clues about other things.

Seriously, when I ran a workstealing benchmark (which is all about interprocessor synchronization) it ran just as fast at 32 threads as it did at 1. The 1-thread performance is nothing to shout about, but at 16, it's suddenly kicking the ass of every other computer I can get my hands on (because of all the synchronization) and it can do that all the way out to 32. I have heard, through the grapevine, that the Niagara 1 (and perhaps the 2) provides spectacular GUPS/watt performance (GUPS is a random-access general update benchmark).

The "interesting times" part of this much multicore is that it is now possible to run out of all sorts of stuff -- should you devote chip area to floating point, or to store buffers, or to L2 cache, or to one more core? People are studying scheduling algorithms where they actually (sometimes) shut down a core to reduce resource thrashing on chip, for loads that thrash.

Re:Is the hardware any good though?

[Anonymous Coward]

> barrel processor

Huh. Now we know why they called it "Niagara".

Re:Is the hardware any good though?

[georgewilliamherbert]

Sun's late-2008 T3 "Victoria Falls" model will include a 2-way interconnect with chips that otherwise spec-wise appear to be T2 cores.

See http://www.sun.com/processors/UltraSPARC-T2/datasheet.pdf

Not clear what the interconnect model is yet.

I personally would like to see the open source T2 re-released with the PCI-X and 10 gig ethernet, as Sun hints they will do once licensing is fixed:

Note - OpenSPARC T2 currently does not include PCI-Express and 10Gigabit
Ethernet design implementation due to current legal restrictions. Equivalent models
may be available in the subsequent releases of OpenSPARC T2.

http://opensparc-t2.sunsource.net/specs/OpenSPARCT2_Core_Micro_Arch.pdf pp 1-3

Re:Is the hardware any good though?

[jd]

Since the core is software-defined, there should be relatively little difficulty in defining a bus abstraction and then having modules for HyperTransport, PCI Express, or any other bus they care to support, where perhaps only one module can be included at "compile-time". Why go for something this complex? It makes things future-proof. If a new bus or a new revision of a bus becomes the "in-thing", you only change the module. The chip architecture remains fundamentally the same. (Both HyperTransport and PCI Express have been through multiple revisions, with next to nobody using the latest. They are likely to be revised extensively in the future. Laying out and validating a CPU is non-trivial, so the less you need to do to support such changes, the better.)

JONATHAN SCHWARTZ ..

[jvd]

ISAMOTHERFUCKINGCOCKSUCKINGPOSER. Uh... sorry, they keys are like right next to each other.

What about patents?

[LingNoi]

I noticed they released it under the GPL 2 (or is the chip design released under a different license?).

Does this mean they could attack a company that started selling their processor or one based on this information with a patent?

Re:What about patents?

[downix]

Only if said company did not pay for the license. Incidentally, the basic SPARC license is $100.

Re:What about patents?

[TheRaven64]

I noticed they released it under the GPL 2 A license which you obviously have not read. Had you read it, you would see that Clause 7 specifically answers your question in the negative.

FPGA Huggers

[Doc Ruby]

in addition gives us FPGA-lovers something new to play with

How big an FPGA would be required to run this? Can you really download the configs and run it on an FPGA at a reasonable speed? Which Xilinx model?

How about running Linux on that simulated Niagara2, like you can uCLinux on a Microblaze? The exciting part would be replacing parts of the OS, like the TCP/IP stack, with "HW" configs for really high performance, customized per app. None of your processes use some dozen instructions? Drop their microcode in favor of a faster multiplier...

Re:FPGA Huggers

[nurb432]

If you have to ask, i bet they are out of your price range. The big ones are NOT cheap.

But ya, it can be done.

Re:FPGA Huggers

[Anonymous Coward]

"The numbers above clearly show that the S1 Core will not fit into any existing Spartan device; if you really plan to use FPGA technology you might consider using a Virtex 4 or 5 device with at least 100K LUTs." that is the S1 core wich are a simpleified version of the Niagara T1.
i guess the t2 is bigger but this still make's it possible to take some good code and build a smaller version

Re:FPGA Huggers

[Darinbob]

You know, I've been tempted to get an FPGA eval kit just to play with. But it might just end up as more stuff filling up the closet.

When I first got interested in computer it was because I thought digital electronics were neat. Nands and nors and 7400 series TTL building blocks, etc. Then I ran across a book at a nearby junior college that was a bit higher level and discussed half-adders, registers, and built up to a 4-bit CPU near the end (I never found this book again). Then I was hooked, and that was the major I wanted when I went to college. The idea that hardware was just a bunch of Lego(tm) pieces that could be built up into something complex was attractive.

Except that I ended up doing software. Which isn't bad, I just wanted to learn both. The CS side of the department didn't do hardware except for VLSI routing algorithms, and the EE side didn't do digital. This was also before MOSIS was widely available as well, and we didn't have any cool projects like RISC or MIPS.

Later when I heard about FPGAs and CPLDs and such, I thought these would be great to play around with. The snag is that the design languages for them are too far removed from the gates underneath, and it seems more like writing a C program than snapping together Lego blocks. So I play around with TKGate now and then instead.

Re:FPGA Huggers

[Doc Ruby]

The best apps for FPGA are glue logic. So if you want to lego together different digital chips, FPGA is a great way to go. And then there's a lot more, which your SW skills will help with. Take the plunge and do something new. There's plenty left to do that no one has yet before.

Re:FPGA Huggers

[Anonymous Coward]

You can't fit a whole N2 in any FPGA, but one of eight cores will fit and demonstrates the concepts. There is a reason Sun uses a custom LSI fab to build the chip after all.

And no, the FPGA is not terribly fast.

If you just want to run Linux you can do that on a T5120 as is, it works.

Re:FPGA Huggers

[hjf]

you need to check your facts a little. I've seen FPGAs inside HDTV sets and Cisco routers (actually some Cisco WICs but that's not the point). And considering that FPGAs can work well over 500MHz and give you 600 Billion MACs/s (multiply-and-accumulate, the basic DSP operation), that's far away from your 7400 and 4500 families.

Re:FPGA Huggers

[hjf]

Come on, it's just $150 for the Spartan-3E starter kit (xilinx.com)

Also, it seems like writing a C program, but it's not. Remember that VHDL doesn't run line-by-line, most of the time is running all the lines at once. It's a description language, not a programming language.

Re:FPGA Huggers

[eigendude]

You can do quite a bit more in FPGA's than only glue logic (unless you meant CPLD's)...

Nowadays, even a relatively modest FPGA has enough logic resources to host a few 32-bit microcontroller cores such as Xilinx Microblaze or Altera NIOS. I am not talking about some 1 k$ FPGA here, but a Cyclone-II from Altera or a Spartan-III from Xilinx which can be bought for 30$ or so in quantities of 1.

Take any of the larger Virtex-II to Virtex-4 or Stratix FPGA's and you can host a whole lot of custom logic. More than anyone can design in many years of effort (unless you repeat the logic, or make 1000 bits wide buses, obviously). The only reasons that the industry still uses ASIC's a lot is because they are much cheaper in volume and you can usually get the logic to run at higher speed.

Re:FPGA Huggers

[Doc Ruby]

Well, that is why I also mentioned MicroBlaze in this thread. But I was encouraging someone to start, whose original interest in electronics was based on "pluggable logic", which is a good intro to FPGAs.

Re:FPGA Huggers

[Doc Ruby]

How does anything you just said contradict anything that I just said?

Re:FPGA Huggers

[smorken]

Is the book you are thinking of an edition of: John D. Carpinelli, Computer Systems Organization and Architecture, Addison Wesley We used the 4bit CPU specification from this text to code a simulator for a class in college for an assignment. Our group also made an assembler for it.

Re:FPGA Huggers

[Darinbob]

I don't know what the book was, but I read it in 1980. It's probably undergone a few revisions and name changes since then as the state of the art changed.

Re:FPGA Huggers

[Anonymous Coward]

Forget it. As far as I'm aware, FPGA's are several orders of magnitude slower than dedicated ASIC - which means your Core 2 duo on an FPGA might be the equivalent of running it at about 100 MHz. And that's with FPGA's that are extremely expensive. FPGAs are strictly used to make sure that the CPU works correctly for perhaps the first few hundred million cycles and perhaps some test cases in the middle of normal operation.

Also, your dream of using FPGAs for customized designs are nice, but a pipe dream - first, they're usually too big for a PCI card (at least with decent enough performance). Then it's also hard to correctly design & debug RTL code, which means that it's better to do it in software - by the time you've implemented your super-duper optimized TCP/IP stack, computers have gotten fast enough that the software version is better (and cheaper & easier to maintain).

Also, your basic premise has a much more interesting application; is it possibly to dynamically optimize the hardware via some kind of hardware JIT compiler for the software that's being executed. However, AFAIK, aside from some vapourware announcements, it's still a research topic.

Re:FPGA Huggers

[Doc Ruby]

There are already "IP core" component configs available for some things, like TCP/IP, that some SW implements in instructions. So in fact the "pipe dream" of customizing HW to accelerate SW is entirely realistic. And when FPGA reconfig speeds increase, the dream of reconfiguring as a process demands such HW will also eventually come true. But it all requires people who dream those dreams to work on it, even when other people discourage the future.

Re:FPGA Huggers

[hjf]

Well, you said it in the typical "Macs are only good for design" sort of way. I was just filling in for you.

Re:FPGA Huggers

[Doc Ruby]

No, I said that glue logic is the best FPGA app. Which is probably exactly what the FPGA you saw in a TV was doing. I was encouraging someone with an interest in FPGA, who started in electronics because they thought they could just plug digital chips together like lego, to try that natural application of FPGA.

What you did was "make a mistake" by converting "the best" into "the only". Check your facts before snidely telling me to do so.

Re:FPGA Huggers

[imgod2u]

If you're writing Verilog or VHDL like you're doing a C program, you're doing it wrong. It's actually not very complicated to figure out the translation if you stick with RTL constructs. The key thing to remember is that those HDL languages provide more than the description of gates. They provide programming behavior so that you can create models and testbenches to test your gates. *BUT* that also means that you have to actively avoid using those constructs when describing real hardware. If you limit yourself to what synthesis will be able to deal with (registers and combinational logic) it is *very* easy to "see" the gates as you code. And that's how it should be.

Way too many people get thrown off when they first pick up HDL languages because they treat it like a programming language. It isn't, you're describing hardware. A very good book (and one that's unfortunately no longer in print and is rare and good enough that it still sells for $200+) that does an excellent job of explaining it all is HDL Chip Design by Douglas J. Smith.

Re:FPGA Huggers

[imgod2u]

The embedded systems market is flooded with FPGA-based designs. One does not need to redesign a custom TCP/IP stack. IP can be bought (or downloaded) for it and you simply glue the pieces together. The performance-centric FPGA's do draw quite a bit of power for what they offer, but if your power budget allows it, there's no reason not to.

The primary reason why FPGA's aren't used in things such as network cards has nothing to do with technical feasibility but rather cost. ASICs are significantly lower-cost per-chip (for the same features) than an FPGA is. For custom-designed systems on the order of thousands, FPGA's are usually the way to go. On the hundreds of thousands, however, people usually go the ASIC route.

Re:FPGA Huggers

[Muad'Dave]

Doc Ruby, I'd love to have an offline chat with you re: FPGAs and their use as combinatorial problem solvers. Please drop me an email at dtiller (at) captechNOSPAMPLEASEventures (dot) com. Remove NOSPAMPLEASE for great justice. Thanks!

Re:FPGA Huggers

[OrangeTide]

you need a Virtex. Expect to spend around $1200-$3000 on a kit to run a single core (like the S-RISC). You can't fit the full 8-core design in any single FPGA. And there is a huge performance and complexity factor of spreading a design out onto multiple chips.

The Niagra is already pretty optimal in terms of reducing the area required for unused instructions. And there are very few instructions that aren't used in a typical kernel+apps on a Sparc. Also microcode doesn't take up much space, especially if you don't go for an optimized microcoded version. If you have one microcoded instruction, then it's not a big deal to have 100 of them. As long as your instruction decoding can be very regular (it is on all RISCs).

Re:FPGA Huggers

[Anonymous Coward]

you must be new here

FPGA lovers

[nurb432]

Ya, we all have those HUGE ( read : expensive ) FPGAs required to implement something like this.

Many of us are lucky to fit a Z80 into what we have.

Re:FPGA lovers

[nahpets77]

You may not be able to use them at home, but most university labs have expensive FPGAs lying around collecting dust. This is good news for people doing research at universities, where they often like to tinker with the hardware to try out new ideas.

Re:FPGA lovers

[marcansoft]

I can fit the entire PACMAN arcade board (including the Z80) inside my S3E. 25% use, not counting RAM blocks, which are all in use because I'm too lazy to switch the code over to use an external ROM for the code. I have the $150 Spartan 3E Starter Kit.

Z80s aren't that big.

Re:FPGA lovers

[Criton]

I seen 32bit risc cpus implemented in fpgs but Niagara is a cutting edge 64bit design that makes core look crude for server and super computer applications. Not really suited for desktop apps yet as they tend to be single threaded still and can't make use of the 32 threads.

Re:FPGA lovers

[nurb432]

I was being a bit sarcastic with the Z80 comment. You can fit an entire Sinclair Spectrum in medium sized chip.

Spice model of this chip

[Anonymous Coward]

I've translated this chip design into an LTSpice model. I now run all my software on this spice model..

    and my computer is 100 times faster!

Government Technology Embargoes

[BearRanger]

Granted, you can't build a fab in your basement. But I imagine governments don't have this problem.

What are the implications of Sun doing this? There are countries that wouldn't be allowed to buy their finished Niagara servers that could now, given time, reproduce their technology. Doesn't this make a mockery of the U.S. technology embargo against certain countries?

Perhaps I'm simply missing something, but if AMD can get into hot water over their processors showing up in Iran why does Sun get a pass for revealing how to construct similar technology? It can't just be immediacy. If anything wouldn't the "blueprints" be of more value than the finished product?

Re:Government Technology Embargoes

[Anonymous Coward]

Definitely not an expert on this, but I believe the answer is "Yes, they could." The second edition of Applied Cryptography has a note somewhere explaining that while the book with printed source code is exportable, the source code on disk isn't. I wouldn't be too surprised if this was a similar situation. Take that with a grain of salt, though, cause I'm really not sure.

Re:Government Technology Embargoes

[Waffle Iron]

That's easy to fix. All they have to do is put a comment at the top of the CPU source code:

/* Assumed main clock frequency: 25MHz */

But seriously, this design info is just about useless without a $Billion fab full of equipment that these countries also aren't supposed to get. But why would they bother with all that when they can just buy quad core consumer PCs that rival the performance of these chips, and which are available to them on the open market from most any country other than the USA?

New GPL can of worms opening...

[droopycom]

It is going to be interesting to see how the GPL is applied to RTL code.

For example, what constitutes derivatives and what can be considered mere-aggregation.

Also can I license an RTL block from another vendor and combine the two in a new chip ?

sun4m and other "they don't exist, don't ask"'s

[sethstorm]

You are right - some of their older hardware that still is quite usable except for them cutting out support at inopportune times. They've kept 8bit cg* framebuffers yet dropped 24bit ZX's from existence. They kept [very limited] sbus in Opensolaris, yet have made a conscious effort to erase Sparcstations from ever existing in the code.

Should you run into a SunPC or similar, that will bite doubly for being Solaris only (and for versions that may not be in circulation).

Now if you run into something on the order of an E10k, and dont mind powering it, Sun would rather you not.

If they were to clean up Solaris 9 and have it up to speed as best as you can expect a SS/10 (or a Ross SS/20) to run it, that codebase would probably be fine enough.

When you have to pull teeth for their own hardware, they certainly are not going to be any better (See SunPC, E10k's with their hardware license keys) with Linux.

bmc, this definitely applies to you(and those who've dropped the axe):
"The past has been erased, the erasure was forgotten, the lie became truth"

OSFPPC!!

[syousef]

While we all might not have a silicon fab in the basement

You don't? How tragic. I'm afraid you'll have to hand in your geek card. In the meantime I wonder if the OLPC guys would consider a OSFPPC (One Silicon Fab Plant Per Child) program.

Someone try to synthesize it!!

[defro]

I'd love to see the usage statistics of this design synthesized with ISE, Quartus or Synplify. How much would we have to cut-out to have it fit in an LX160 or similar????

Re:Someone try to synthesize it!!

[mako1138]

Well, I'm trying to synthesize a single core. ISE gets really sluggish when the file count gets large, so I'm doing it on the command line. So far I'm having to fix stuff like

ERROR:Xst - "/..\opensparc\design\sys\iop\spc\dec\rtl\dec_del_ctl.v" line 1102:
Module <dec_del_ctl_spare_ctl_macro__flops_0__num_6> has no port.

which points to something like

dec_del_ctl_spare_ctl_macro__flops_0__num_6 spares (
);

Perhaps there's a way to ignore these, but I don't know XST well enough.

Sun uses Synopsys for their synthesis needs, it seems.

Re:Someone try to synthesize it!!

[wik]

The statistics the T1 are available here:

http://fpga.sunsource.net/

The most recent release of the T1 code has a few options for removing functionality (dropping to 1 core and 1 thread) such that it will fit on some of the larger available FPGAs.

Re:Someone try to synthesize it!!

[mako1138]

Haha, after 45 minutes of working, XST says:

Mapping all equations...
ERROR:Portability:3 - This Xilinx application has run out of memory or has encou
ntered a memory conflict. Current memory usage is 2090592 kb. Memory problems
may require a simple increase in available system memory, or possibly a fix to t
he software or a special workaround. To troubleshoot or remedy the problem, fir
st: Try increasing your system's RAM. Alternatively, you may try increasing yo
ur system's virtual memory or swap space. If this does not fix the problem, ple
ase try the following: Search the Answers Database at support.xilinx.com to loc
ate information on this error message. If neither of the above resources produc
es an available solution, please use Web Support to open a case with Xilinx Tech
nical Support off of support.xilinx.com. As it is likely that this may be an un
foreseen problem, please be prepared to submit relevant design files if necessar
y.

Looks like you need a serious amount of memory to synthesize even a single core. I was targeting a Virtex-5 LX50.

I don't think there's any more DDR2 here at work, so it's time to go home.

Is it really released? I can't find a link

[whoever57]

I searched for a link to an actual download (yes, I have a use for the code). Opensparc.net just refers to the "Sun download Center" (no link). Searching on Sun's site, I can only turn up OpenSPARC T1 and not T2.

Re:Is it really released? I can't find a link

[mako1138]

Yeah, it's there.

http://www.opensparc.net/opensparc-t2/downloads.html

Be warned, the 233MB file decompresses to about 1.5 GB.

All right...

[sootman]

who's the considerate jerk who tagged this story 'thanks'? We don't work that way here at Slashdot, buddy. When a company does something like this, you're supposed to tag it 'whocares' or 'toolittletoolate' or something equally dismissive. Damn noobs...

Re:All right...

[Anonymous Coward]

it's newbs BRO

we all might not have silicon fab in the basement

[unity100]

not yet pal, just not yet. you'll see.

Sun is the new Bell Labs

[teknopurge]

I've had this position for about 2 years now.

IMO, Sun is one of the only companies left innovating.

-Google is just rehashing old ideas.(Gmail? come on....I had webmail 10 years ago.)
-Oracle(eh... RDBMS v45.2 anyone?)
-IBM(If I see one more pointless black-and-white commercial about "ideas" I'm going to scream. IBM should listen to their marketing department and instead of telling us to "Stop thinking, start doing" they should create something that isn't AIX)

And, I will be the lone voice and dare to say that Microsoft, yes them, has a few teams that are starting to 'get it'. Apple is doing a great job with human-computer interaction.

Show me new, for I am tired of your old.

Bootstrapping issue

[iabervon]

So I was all set to download the source and build some chips in my basement fab, but then I looked at the system requirements. It's only for Solaris, and worse, it's only for SPARC. How the heck am I supposed to run this software so I can build a CPU when I need that same CPU to run the software? Obviously, Sun is going to have a lot of SPARCs sitting around from earlier development, so they wouldn't have this problem themselves, but they should have thought of it, at least, and provided binaries for Cell or ARM or something.

Too bad there isn't a cheap way to get a Niagra

[OrangeTide]

MIPS is a terrible design. It's no fun to program for it.

I think DLX and Microblaze is a more elegant low-end implementation of a MIPS-like. And I think SPARC and PowerPC are much more practical and useful high-end RISCs.

I would like to see Sun put together a cheap developer's kit, they could market a Mac Mini type form factor as a J2EE developer workstation or something. But the rest of us could pick one up too to see what sorts of cool projects can be done with it. I'm fine with 4-core or 8-core version on such a workstation if it could be sold for under $800. (ideally under $500)

I really don't find it practical to pick up a low end SunFire to use the chip. I think Java developers could benefit from testing and developing their code on highly threaded core. (each core does 2x or 4x SMT i believe)

They wouldn't have done it if N2 was competitive

[Glasswire]

Sorry, I mean I love that they're giving silicon designs out the world, but if there was any really important innovation in the intellectual property behind the Niagara II that would give Sun an advantage in the marketplace they would not be exposing it to all their competitors.

The only business reason I can imagine that Sun would do this is the hope that lot's of Niagara foundries would bloom and thereby cut their costs for sourcing the part.