Sun Microsystems is releasing the specifications of its new UltraSPARC T2 processor, formally code named Niagara 2, to the open-source community Dec. 12, as part of the company's ongoing effort to build more of a community around its signature chip. The goal of releasing Niagara 2 into the open-source community through the General Public License is to create a larger community around the chip and increase the number of operating systems and applications that can use the processor, said Shrenik Mehta, senior director for Fronted Technologies and the OpenSPARC Program at Sun. In 2005, the company released the specification for the UltraSPARC T1 processor and the designs have been downloaded 6500 times since then, Mehta said.
Post a Comment
Member since:
2006-08-03
I'm not a developer, but understand the benefits of open source software. Is this immediately only beneficial to developers? I can see in the long run that software could be much better designed to run on the processors, but I don't really see "people" trying to make their own processor/motherboard logic.
How is this really different as compared to when Intel releases developer documentation about their platforms?
-m
Member since:2007-08-21
But people *are* making their own processor and logic. Look up the SimpleRISC S1, which is a stripped down T1 for embedded applications. In addition, the docs Intel and AMD provide is always limited to what they want you to know. By having the T1's sources one could better optimize a compiler to the processor. In addition, there is the comfort that should a bug develop, you could analyze the sources, find the issue, and notify Sun, thereby saving everyone time and money.
Plus, you could take the T2, modify it for a specific purpose, and approach a silicon vendor to have it produced to your specifications, legally. And Sun benefits because your improvements they can then fold into future versions they themselves make. It is a win-win situation.
So you can really play on a fully open sourced hardware platform.
Member since:2007-01-27
@downix
You were right a few weeks ago and I believe it's very important for free operating systems to support free hardware designs, particularly considering the huge potential for optimisation.
Also please check your messages. I sent you one a few weeks ago, in case you haven't noticed yet. Now I'm only waiting for T1/T2 and derivatives to reach the bulk processor market.
Member since:2007-08-21
Member since:2007-01-27
Member since:2007-08-21
Member since:2007-01-27
Yes, please look at the right side for OSNews v4, reachable by following http://www4.osnews.com. You would probably not see all the options when you're reading this web site in v3 mode.
Member since:2007-08-21
Member since:2006-01-03
This is simply amazing
Any student/professional in u-electronics design will certainly appreciate this. We're talking about a brand new processor, flagship of an important company.
Most of people certainly does not have the installation to build a full Niagara 2 chip, but the community can always get inspired by a particular module of a last-gen processor.
In counterpart, Sun can have new ideas for the design of the future generation of their chips, even if in these moments maybe the design of Niagara 3 is totally finished, which is very common in chip design.
By opensourcing the Niagara 2 design, Sun is offering to the community an impressive amount of working-time. I think we can just welcome this step from Sun in becoming a more opensource company.
Member since:
2006-03-13
Member since:2007-08-21
Member since:2006-03-13
Member since:
2005-07-06
Software developers (even those who develop an OS) couldn't care less about the CPU's implementation specifications (*), but they care about the chipset API that Sun doesn't give in general.
But Hardware developers do care about the CPU implementation specifications.
*: it was suggested above that this could be helpful for debuggers: I doubt this, the view of a CPU that a debugger have is very, very simplified compared to an implementation specification, sure you could use the implementation specification to generate the debugger view but this would be useless work as Sun has most probably already ported gdb to their CPU..
Member since:2007-08-21
I was talking about debugging on the hardware level eg: t northbridge chip the company I worked for used a few years back shipped with a serious data corruption bug that occured whenever a bus mastering over PCI occured. It took years to track down, and was never truely gotten rid of, only patched. Had the source for the chip been available, then possibly a lot of the workaround and reverse engineering necessary for the patches could have been avoided, possibly.
Edited 2007-12-12 14:30