Slashdot Mirror
Sun Opening Microprocessor Technology
bjb writes "The Wall Street Journal reported this morning that Sun plans to announce later today that they are going to distribute designs of their microprocessors to outside developers for free. Similar to the Java license that recently came out, you can modify it, but if you sell you have to pay. The schedule appears that they will release the PicoJava first, the 32-bit SPARC technology by the summer, and the 64-bit UltraSPARC technology by the end of the year." This
article requires a paid login to read- what a crock. Anyway,
someone please submit a free link.
Update: 03/02 12:09 by S :
News.com is now carrying the story. It's interesting to
see how this move fits into Forbes' analysis
Sun's strategy of getting attention with buzz around
Java, Jini, and now Sparc Processors, in order to attack
the very high-end more effectively. In the world of
Starfire and
Serengeti (supercomputers), Sun is probably telling
the truth when they say that Linux does not compete with them
(long term).
This article requires a paid login to read- what a crock
Let's see here, they paid the reporter, payed to publish the information, have bandwidth costs, but they shouldn't charge for it? What happened to 'if you write [the code, but it applies to anything] you determine the license? Oh, that only applies to people who choose to give things away.
Sunworld is NOT run by sun.. Its run by idg..
They can jump past PIII, if they can have a look at this new chip from russians and keep a big gap margin between dektop and server areas. See in microprocessor report.
I'm glad to see Sun supporting open source and open hardware this way. Moves by such a big player take the spotlight off unproductive bickering between people like Perens, Raymond, Stallman, etc. over the "purity" of the open source model. The time is right to realize that there can be multiple models of open source development that are valid for different contexts even if it takes a large commercial organization to drive the point home.
Does anyone actually have the resources to make use of a microprocessor design and NOT have to at least sell something? Or did I miss that sale on the Personal Chip Fabricator(TM) ?
This looks like more of Sun's "pretend to give something away, but not really" philosophy.
Um, Rob wrote the remark you're deriding as "crap." I, too, wish he would start moderating crappy comments.
What you missed is the fact that smaller companies often cannot affort the up-front money usually required to license designs. This was pointed out in the article. Of course they will have to sell something to make money. The point is that they will have to pay Sun for their technology only when they actually ship a product, which reduces their initial risk.
Not yet at this capacity... but just you wait.
They made their OS free for personal use.
They keep opening up Java more and more as time goes by (I'm still not sure whether it's open enough for my tastes, but I definitely like their direction). The source for Jini was made available. They are making the source for Solaris available. Now they are makin the specs and design for their chips available...
Sun may not be perfect, but I wish more companies were as open as this! Kudos to Sun!
Can one sell wallpaper or artwork incorporating
the new Sun chip designs without royalties?
This is short-sighted. "Open-ness" is not limited just to things that are strictly free. "Open" and "free" don't necessarily mean the same thing. Hardware can not normally be given away for free. In the case of hardware especially, but in some software contexts as well, there needs to be a model of open development that allows the developers to ultimately get paid for their work, especially if commercial organizations like Sun are to be brought on board. Sun's approach may or may not be the best (time will tell), but it's a start. Bringing companies like Sun into the open development camp will be important for Linux, even if it means that it isn't always free.
It could possibly help out with designing compiler
backends for the processors.
It also helps in the case where chip bugs need to
be worked around in the compiler or the os, since
more people that could do with knowing these sort
of things will know about them.
The position is good for hardware, since hardware
COSTS money to build -- though the Sun stance on
Java still needs to be opened up (i.e. the license
changed sufficiently to allow projects like
java.blackdown to release pre-beta code)
It would be cool if Compaq did this with the Alpha architecture.
Alpha might even be able to compete with x86
if that happened...
I'm looking forward to the Sparc Ultra64 clones!
Who would buy it?
Remember, think "free speech", not "free beer".
Given what it costs to make hardware (let alone what it cost to make a cip fab), this is about as close to Free Hardware/Open Hardware as it's likely to ever get.
As a former Sun marketdroid myself, I have to hand it to whomever had the guts to "light up the sky" with this piece of PR. It is really only a slight change, but will generate a lot of positive press for Sun.
In reality, SPARC Int'l has offered anybody who cared to shell out $99 the rights to this same info for years (for SPARC only, not picoJava.)
That's right, for years you've been able to spend less than $100.00 and build your own SPARC chips. Of course, there's that darn billion-dollar fab you'll need to actually do it...
Anyhow, a good PR move if no real change in Sun's policy - maybe this shows they're finally learning how to market their stuff! This may impact their revenue by up to $500/year.
Dub Dublin
dub@psw.com
People still pay millions of dollars for mainframes. The CIO at Citibank was given an award from Sun for purchasing something like the thousandth E10k. But, he still said he's still buying mainframes.
When a platform hits a plateau, you just can't kill it. This has happened with mainframes and will happen with both Sun hardware and Solaris.
Sort of. I've heard that argument a lot, that the extremists like RMS are good because they move the middle ground. I'm not sure if I agree- first of all, I most definitely do disagree with the extremists. Given that, it makes just as much sense to say, "Microsoft (or Sun, or whomever) are good, because they move the middle ground away from the everything-must-be-free crowd". I bet you (or most of the "support the extremists even if you don't agree with them" crowd) wouldn't agree with that, would you? Are you thankful that MS is around to balance out folks like RMS? Do you ever tell anyone, "don't bash MS- even if you don't like them, they help move the middle ground away from RMS". Probably not, huh? :)
I am a big fan of Linux and I've been using it since the early days. It's great to see it increase in scalability, stability and popularity.
What's obvious though, is that people making the remarks about Linux strength and improved performance on an Enterprise system haven't been there in "real life".
I admit I haven't looked very deep into Linux' high availabilty possibilities. But after reading the documentation about striping, mirroring and logged devices it turns out to be quite a disaster compared to for example Solstice DiskSuite (software RAID isn't common in >E450 systems but anyways). Then there are other solutions like veritas volume mangager. What I'm saying is that I would never take responsability for a, let's say E450 system running Linux today. What if another disc is to be put in and they want it to extend a current (meta)device? With Solaris you just pop in the disc in the internal array type drvconfig and discs to regenerate the device tree (which really needs cleaning up in Linux). Then there are utilities for both VVM and SDS (graphical and console) to extend a running striped device. I don't even dare to think how Linux would respond to a hot swap of a mobo on a E4500. Other areas like monitoring are also highly integrated into Sun software products which simply can't be replaced by some perl hacking. SyMon monitors everything from power supply to hard drive and CPU temperature. The point being that an oddly performing disc or power supply (there are several) will be replaced under warranty by Sun before it brakes. The more you work with high end systems the more you realize how solid it has to be. Yes, Linux is far better than NT but this doesn't mean it's ready for main frame computing (and no, NT isn't either yet they're making money from it).
I'd like Linux to go where Sun is heading but be careful with your remarks. If you haven't been there, don't comment on it, you just don't know what it takes.
Where did the name PicoJava come from? Maybe a year ago I posted something on the Compuserve Digitalk Smalltalk forum about some ideas I had for a class of languages that all had Pico prefixed to the names. There was PicoSmalltalk first of all. The Pico originally comes from the idea started by Little Smalltalk the language that was started by a professor in Oregon. If you want his name you can look it up on Amazon.com by searching for A Little Smalltalk as a book title. He is the author.
Also there was a slogan:
"PicoSmalltalk : the meanest badasses on the Pecos."
Where "mean" is as in "it was a mean life" as opposed to the "sadistic, hateful" meaning of mean. "it was a mean life", as in wirey and well worn old geezers.
So going one step further beyond little there was milli, micro, pico, nano, femto, unless I am wrong. So then there came PicoSmalltalk which was to be the next step in the evolution of Little Smalltalk. Next came PicoLARC and PicoProLARC. PicoLARC is to be a combination of Smalltalk and Lisp. L for Lisp and ARC for PARC as in Xerox. The simple flexible parenthesis of Lisp with the Class framework and
( aCollectionObject at: key put: value ) self documenting syntax of Smalltalk. Also it's mostly purely Object Oriented except when it's not. As in embedded C and assembler. A completely documented internal RAM Object Structure. And wide and highly automated development pathways to every other language: C, C++, Java, Pearl, Python, Smalltalk, etc. Then PicoProLARC is to be a hybrid of Smalltalk, Lisp, and Prolog, heavy on the Prolog. These three languages are called the PicoLanguages and will all have the exact same RAM Object structure so that Objects can be shared between them using shared memory. Between them they will piggy back on top of all the other operating systems and generate a plethora of PicoOperatingSystems which would be good for people who are of a Smalltalk, Lisp, and Prolog mind and not of a C, C++, Java and Basic mind. We of the flexible mind can all go and live there instead , thank you very much, you can keep your C, you can keep your Basic and you can keep your Java. Thank you, and again I thank you.
Each PicoOperatingSystem or PicOS will be a single Object pointing out to other resource Objects or other OperatingSystems. All the physical pointers in the system can either be local or interprocess or longdistance ( URLs ). Each PicoLanguage would be able to compile chunks of itself into binary or ascii that could be sent across the network and executed on another instance of that PicoLanguage.
The first main thing about the PicoLanguages is that they will piggyback. Piggyback will be their middle name. They will have wide highly automated developer pathways to all the other languages.
The second main thing about the PicoLanguages is that they will be highly changable from the ground up. You can have built versions of a PicoLanguage that has no user interface and is only 30K big ( Like Little Smalltalk. ). You can have other builds that are 300Megabytes big.
The main thing about a PicoLanguage is that it starts with a single line of ascii text which is it's starting point. There is also a highly automated system of SourceCode Objects with tracked dependencies that exist in a cloud around this one single line of ascii text. Build pathways can be tracked outwards from this start line through the SourceCode Objects following the lines of dependency. And thus a new version of the PicoLanguage can be constructed from the ground up in a highly automated Object Oriented and not Text File Oriented way. Thus in the case of PicoSmalltalk for instance the initial build sequence includes all the source code for making the virtual machine. Once that is done the rest of the build sequence would build up the Smalltalk side of Smalltalk. By tracing various pathways through the Source Code cloud you could build up various different PicoSmalltalks. Like for instance if you wanted to change memory managers.
Can you say Modularity?
Can you say Protocal?
Can you say Patterns?
Can you say Parts?
Can you say Plug and Play?
Can you say Standard?
Can you say Open Source?
Can you say Slide the Old one Out and Slide the New one IN while the power is still on?
This development system will be highly automated and recursive in nature. We will not use text files we will use ObjectOriented Databases ( and boot text files ).
( boot text files that will also be in the OODBs and automatically managed. None of this manual stuff. No scratching the head looking at miles and miles of archane uninstrumented text. )( but you can still output the entire mess as a mess of text files and a tiny boot compiler which can recompile the larger compilers from scratch and so rebuild the whole system from ascii text. )( The goal is to make it so that an alien coming from another universe could come and use the development environment without much trouble. Without having to scratch his head too much. )( The goal is to make it so that kids can take over complete control of the entire system if they have got a yen for it. This was the original reason for the self documenting nature of Smalltalk, so that Kids could program in it. But the funny thing is that once you make it so that kids can program in it, COBOL and C programmers can't anymore. Oh well, you can please some of the people all of the time and all of the people some of the time but You can't teach an old dog new tricks. )
The snake will be eating its tail Oroborus style. The system will contain and maintain the plan for itself. It will be heavily documented from top to bottom in a highly automated way. It will generate an e-mail confederation of open source developers. It will implement the BAZAR development system. It's documentation will contain stories. And manual speak will be encouraged against. It will contain whatever anyone wants to jack in there. A world wide frictionless highly automated development system. The first of the Frictionless InCorporations to actually call itself a frictionless Corp. Where Frictionless means that people and entities can come and go, use or not use, add or subtract, compete or collaborate at will. With no gatekeepers imposed or legislated but by popular demand and always with the possibility that the little guy can mount a competing version.
If such a frictionless corporation could actually make money in a frictionless way, like the stock and option markets make money in a frictionless way with standardized contracts then that might be the death knell of the old Dilbertian corporations. In a frictionless corporation / collaborations? / would be formed and broken at will just like buying and selling shares. But how would it work? I think that Linux is a frictionless incorporation. And Microsoft should be scared. Perhaps we should all be scared. Perhaps the web will become self aware and we will all be like bits and pieces in a whirl wind. We will be Borg.
Not me man.
Hey, keep the implants to yourself. I may be as obsolete as punched cards but nobody's messin with my brain except in the usual way by commercial break or banner ad.
Yeah right, you old fogey. You're such an Anac.
So I was just wondering where they got the name PicoJava from? Did it have anything to do with Little Smalltalk?
Also you should all know that the reason you can click on the right mouse button and get up the Cut Copy Paste Delete popup menu is because Smalltalk did it first on the Xerox desktop computer at PARC in 1977. Took Gates long enough.
What then would be the point in spending all the time it takes to design something like this?
You're right, I'm not and I don't. The problem with your argument, though, is that it assumes a sort of Heaven-and-Hell premise, i.e., that free software extremists and proprietary extremists are diametrical opposites engaged in some sort of eternal struggle.
The reality is that the proprietary extreme has been the status quo. The free software extremists have acted as agents of change and through their action (and extremism) established a middle ground that the more moderate parties on both sides can accept.
have you actually done any of this? or are you just ranting?
If you remember well, all GNU/Linux users were once called extremists. Why? Because we had a different view of how things should be done. Latter, Open Source was recalled as extremist idealism. Now, every company is releasing its Open Source.
And now youre calling free software advocatees extremist? Because they have different ideas, or want more of what we have? Really, if extremists like Richard Stallman and Linus Torvald didnt exist, we would now see some extremists who actually think that Microsoft software is bad!
"Sun will charge a royalty only if customers ship the processors for revenue"
"The maneuver is not unlike the open-source movement"
Oh, really? Obviously, it's completely unlike the open-source movement. I think that whoever is running OSI these days needs to run another publicity/identity blitz so that people stop confusing these fakes with the real thing.
i'd have to look up the url... here's the link
Its also on CNN, but there isn't much detail: http://cnnfn.com/digitaljam/9903/02/sun/
Three words: SGI Visual Workstation. They have a 3.2GBps system bus, unless I am mistaken. They're not comparable to an E-anything, though.
No, I don't think you can get them transparent. ;)
I think this is the closest that Sun will get to OSH (Open Source Hardware). In my office (we're a hardware manufacturer) I can throw a stick and hit 3 different software people. I can also hit 2 or 3 hardware designers. No chip designers.
OSS proliferates because everyone has easy access to the means to manufacture (compile) the code. OSH will have a harder time because it has to be physically produced, and there is no way that you'll ever get a hardware company to produce hardware for free. It won't happen.
All this being said, I think this (releasing the designs of chips) is the closest you'll get to actual OSH. However, this may make products like UltraPenguin better, since the coders can take a look inside the chip...
Bravo to Sun for doing this. I hope more hardware manufacturers follow this model.
So....t his means I can make one at home? Cool.
"shop smart:shop s-mart" ash
This is a first, small baby step for a large corporation to wholeheartedly embrace the concept of open source. Sun has spoken with *action* here, not just "endorsement"... So I'm glad.
No, it's not "pure" open source, but it *does* increase one's freedom WHILE insuring Sun get's their due if you sell it. That's the important thing.
Will it work? Time will tell. I'm hopeful.
-Stu
The release hasn't appeared on Sun's web site yet - their site normally gets updated late in the day.
Personal gripe - I submitted this story 5 hours ago with some extra info - not some 'paid subscription' junk, but did it get posted? No.
Btw, at the register, they also have this article titled Intel, Sun schmoozing for a chip cruising. The info is rather basic. Apparantly Microsoft is pissed off...
Anyway, ARM is the current champion of chip IP. They don't make, or get made, any chips themselves. Instead anyone can get a liscence to make them, or to make designs with ARMs. About 50 million ARM 'chips' were made last year, up from 10m the year before. ARM are expected to capture 70% of the digital mobile phone market this year... 3Com are going to start using them in network cards, and they're already appearing in hard-discs and stuff.
It's not particularly sexy, but ARM is worth nearly $2bn, a 300% increase since they floated a year ago. They're doing pretty damn well ^-^
However, ARM do charge an upfront fee. This cost for the ARM 9 is rumoured to be around $5-10million. The cost per chip is about $0.10 to $0.20 though. Incidentally, liscencing Sun's Jini costs nothing up front, and has an individual cost of about $0.20, with a max of $250,000 I think.
With regards to the 3.2GBps system bus/interconnect, what were you referring to? The main memory bus on a processor board is somewhere around that level. However, the main data bus is nearly 13GByte/s and about 15GB on the newer, slightly revised ones. It also has a 6GByte/s IO bus.
A university could make a PicoCore derivative or maybe even 32 bit SPARC implementation. It depends on the size of the program. Along the way those busily working graduate students would innovate and contribute to the universities I.P., which in turn contributes to the universities prestige and bottom line (assuming they have a non-brain dead industrial partnership/seeding system)
The larger designs provide valuable insight into the state of the art in various techniques and technologies at the time of the chips design.
This is valuable stuff. The fact that Joe Blow can't make use of it is not the point anymore than the fact that Joe Blow can't actually do anything useful with the source code provided by OpenSource software projects. Most people type download binary code. Some people download source code and type make. Relatively few download the source code and make non-trivial modifications. It takes more than the availibility of a compiler and source code to do so.
Surely this should be short term? Yes, at the moment, Solaris scales better than Linux, but for how long? I've personally run Linux an an Ultra Enterprise 4000, and UltraPenguin runs on the E10K ("StarFire"). OK, so Linux only currently scales to 16 CPUs, but given access to appropriate hardware that'll improve. Apart from anything else, it's very rare to have a whole StarFire configured as a single domain anyway. We have 4 here, and the largest single domain has (I think) 22 CPUs.
The point is, yes, Solaris is currently a much better choice at the high end than Linux, but Linux has consistently pushed upwards for the last few years, and it'll catch up with the commercial Unix vendors (including Sun) in a few years time. Yes, Sun and the others will have moved on by then, but my guess is Linux will push into the high end quicker than the high end will advance.
"The invisible and the non-existent look very much alike." -- Delos B. McKown
Sorry, I call 'em like I see 'em. There are trade-offs involved in making a kernel scale to big numbers of cpus, and Linus isn't willing to make them (rightly so, I might add). Let me put it to you this way: would you accept a serious slowdown on your 1-4 processor home machine just so you could brag that linux scales well on 64-processor machines (that very, very few people own)?
well gee, maybe hitachi, intel, amd, motorola, ibm, compaq... get the point, dimwad?
"The lie, Mr. Mulder, is most convincingly hidden between two truths."
--
And Justice for None
i havent seen an update to their page in ages anyway but i'm wondering if this effectively kills it.
"The lie, Mr. Mulder, is most convincingly hidden between two truths."
--
And Justice for None
Can we now start the 'Free Hardware Foundation', Stan Kelly-Bootle's idea of the logical extension of the FSF? I think this is a great idea :-)
Eric
> Let me put it to you this way: would you accept a serious slowdown on your 1-4 processor home machine just
> so you could brag that linux scales well on 64-processor machines (that very, very few people own)?
It seems to me that you could select the 'CPU Count' your kernel supports as a kernel compile option. I'm not much of a kernel hacker, and I can see how it could make a mess of some code, but it could be done. I can just see the options: Uniprocessor, 1-4 CPUs, 1-16 CPUs, or 1-64 CPUs. The only performance drawback would be when a kernel from a system with lots of CPUs is used on a system with only one or two. Oh, and when you set up a system you would need to put the right kernel on it to take proper advantage of the number of CPUs you have.
Eric
open source hardware. Now that's something you won't fine from transmeta-eat that Linus! :)
---
they're open sourcing sparc too. Everyone uses sparc in high-end servers
---
thank RMS and Perens for bickering. yeah right-thank you very much.
Thank ESR, BSD and the Mozilla project for proving open source is compatible with commercial products. Not free but not proprietary either
---
Hi All!
I made a new Samovar award for this Sun's move:
March awards
Andrew
Free speech vs. free beer. Sun Microsystems embraces the business model of the fastest growing computer
d ex.html
technology - Linux. It opens everything: Java, Jini, Solaris, Sparc and PicoJava. With only very small exception:
you may not redistribute Sun technologies without passing compatibility tests and paying royalty. That is what is
called freedom in Linux world. The early but strong sound that Linux should fire back was made few weeks ago by
Stallman. Maybe most of us don't want to shed blood in the crusade for the freedom, but it seems that this small
article defines most of what will happens in industry the next decade. In history pragmatism always wins over
ideology, but we still don't know which approach is more pragmatic. We can't predict how Linux and Sun will
divide the market but it is certain that everybody who can't open the source code will be out of business very
soon.
http://www.ecsl.cs.sunysb.edu/~andrew/awards/in
Andrew
I don't know what you're thinking of, but the open source in this case seems to be Register Transfer Level designs, which just specifies where data goes and when. I don't think Sun is going to release the masks used...
Giving this stuff away free is the only way anyone is going to really adopt it - Java computers, Java Operating Systems, Java ICs...all of these concepts have had little or no adoption.
I see this as very usefull in a way,
but OTOH:
- Anybody saw any license?
- Should anybody look at their data to find out that he/she can't ever do anything without getting in trouble with Sun?
- Can Sun claim any chip designed by someone that saw their desings as derivative work?
- How many designers/companies will be sued by Sun in the next years?
so does this mean that sombody might make a cheaper system?
i just put in
Speaking as an electronic CAD research person, things like this are very useful. One of the toughest problems, especially for university types, is getting realistic designs for testing purposes. There's a big difference between algorithms that work for small benchmark circuits and those that can handle 100K+ gates, and unless you have good examples, it's easy to fool yourself.
David
On Sun's site you could look at http://www.sunworld.com/swol-03-1999/swol-03-csl.h tml for Jivi.
Fortuna favet fatuis (Fortuna favors fools, and most of them run windows)
All of you are right about SunWorld. So her's the Sun url : http://www.sun.com/smi/Press/sunflash/9903/sunflas h.990302.4.html;$sessionid$KAP5OSYAACSU1 AMUVFZE4GUBSSUXEUDO