A Rage II could do some pretty nice rasterization at 500x300 too, faster than 7.5 fps usually.
I would guess that the performance of Unreal on the Rage II would be a fair comparison to UT2003 on the RPU. The Rage II did about 8-9 FPS at 640x480 with a slower CPU handling the geometry updates for Unreal, and the RPU did 7.5FPS at 500x300 with UT2003. So I'd say it's comparable performance for comparable hardware.
OpenRT, a highly optimised software renderer on a 2.66 GHz Pentium 4, got 8FPS for the same game. A software OpenGL renderer running UT2003 rasterized on similar hardware (2.8 GHz Pentium 4) got about 60 FPS at the same resolution.
Now that the hardware has gotten to that point the quality and other advantages of raytracing are going to make it take off.
Based on the figures in the Siggraph '05 paper and the benchmarked performance of equivalent rasterizing hardware running the same engine, if ATI or nVidia had Dr Slusallek's hardware design in '96 when the Rage II came out, they could have a decent dedicated raytracing engine, in hardware, by the turn of the millenium. Or if nVidia had started cranking up a raytracing card in 2005 after Siggraph they'd probably have something that did 60 FPS at HD resolution for $100 by now.
According to your third point DNS is not an open standard.
So long as the properties I mentioned exist then there is nothing stopping an alternative implementation from being created when required.
And when that happens then you can start talking about it being open enough to make its openness really matter.
Openness is not a binary quantity, but there are certain tipping points where openness really becomes meaningful.
* Encumbrance. An interface or protocol can be publicly documented, defined, and implemented, but be patent-encumbered.
* Control. An interface where the *effective* definition belongs to an open standards body. This is one place both OOXML and Flash fail. It's also been an area where Java has historically fallen short of openness.
* Implementation. "It's not open until it's forked" is one way of putting it. If you can take the reference implementation and replace it without losing essential functionality, then it's open. You can replace it by forking an pen source code base, or by making a new implementation from scratch.
* Platform independence. Flash fails this test because the reference implementation isn't available in a portable form. This is a common way for a single vendor to fall down.
The classic open systems platform is, of course, UNIX. UNIX was explicitly not encumbered: AT&T donated the only patent essential to its implementation to the public domain. The API was small, tight, and easily implemented... by 1983, there were multiple hosted implementations like Eunice and Phoenix, native implementations like Regulus, workalikes like Idris and OS/9, and hybrids like Lanetix and Cromix... not to mention the Software Tools project to provide the majority of the UNIX API purely as a library. AT&T owned the trademark and the reference implementation, but they had no control over the standard... as witness the complete failure of "Streams" in the networking market. As for platform independence... it pretty much defined the term.
Flash, on the other hand, has a single reference implementation, under the control of that implementation's owner, and it's complex enough that an independent implementation will never catch up to it any more than WINE can catch up to Windows.
Without diving into the journal papers I couldn't find any mention of the resolutions his chip can run in realtime.
At 66 Mhz, about 6 million gates, 512x384, 7.5FPS. I don't have the performance for the later 90 MHz prototype on hand, but even the 66 MHz prototype was faster than any desktop CPU in 2005. It also required less memory bandwidth than rendering the same scene using a rasterizer... with about the same hardware resources as a Rage II.
Using an FPGA significantly limits the performance. Using current process technology you'd be looking at 400 MHz and 600 million gates at least. Raytracing is "embarrassingly parallelizable", and the RPU design had to be trimmed down to fit into an FPGA implementation, so I don't see any reason you couldn't get comparable performance to a rasterizer for similar scenes if you gave them a level playing field, otherwise it's like you're comparing a Rage II to a geForce 8800.
I have a lovely 256kbit core memory plane mounted in a contrasting brass-colored frame... which reminds me, I need to pull that out of storage and put it up in my cubicle again.
Additionally the exemptions sound like they apply solely to this bill.
These things have a habit of spreading... consider the effect of the "CAN-SPAM" act, which has effectively raised barriers to fighting spam by forcing people to prove over and over again that compliance with the CAN-SPAM act is not de-facto legalization.
If this bill replaces current laws restricting unauthorized access to computers, then we have a real problem.
That is the effect it will have in a large number of cases.
To plagiarize another saying about theory and practice, the difference between de-facto and de-jure is less de-facto than de-jure.
The archive's sheer sprawl reveals both the possibilities and the limits of Otlet's original vision. Otlet envisioned a team of professional catalogers analyzing every piece of incoming information, a philosophy that runs counter to the bottom-up ethos of the Web.
This seems more like a real-time encyclopedia than the entirety of the web, like the next step beyond the Encyclopedia Britannica with its professional editing of contributed articles. The Britannica would have been in the process of switching to updating through supplements when he conceived this, and he could hardly have missed the controversy surrounding that move. This was a grand vision, and the fact that he was able to implement it even in abbreviated form is remarkable.
Writing a spec doesn't make a platform open. Using open components doesn't make a platform open. An open platform is one that isn't tied to a single vendor, which may involve the reference implementation being open source (eg, sockets), it may involve having multiple implementations (eg, HTML), but when you need to buy Adobe's development environment to edit and author arbitrary SWF then it's not "open" in any meaningful sense.
So tell me, mister bones, how does one develop general purpose flash without using Adobe's proprietary development environment?
Open systems with a single proprietary key component are open in name only.
Sometimes you can't just power cycle it...
on
Bone-Headed IT Mistakes
·
· Score: 2, Insightful
Turn it off, turn it on. Nothing was written to running-config.
In this case that's probably going to work.
When the router is at a power station in Guangzhou and you'd have to wait until 3AM and call in one of the people in the company who speaks the language so you can call the local office during their business hours and get them to power-cycle it for you... you're a lot happier that you remembered "reload in 15".
Well, you could look at the diffs and track the percentage of each file that's got code from each committer... making sure to look for reverted code and giving credit to the person who originally committed the reverted code...
Indeed, sorry for being less than clear, that's what I meant.
Either way, this would be pretty unpleasant.
Another thing I thought of, though: if the bacteria is generating hydrocarbons instead of fats that it's using for energy reserves, then it's inherently not competitive with wild strains and thus unlikely to establish itself, but I'd love to see an explicit reference to that.
As mentioned above, there were two driver interfaces for the original USB standard, and the only people who knew were driver writers and nerds compiling their own custom kernel.
And people who use drivers and custom kernels. I didn't do either, but I still had to deal with users whose USB controller cards didn't work with NT4 because of driver conflicts, and then there were devices that had custom drivers that only worked with one or the other chipset. It wasn't until Windows 2000 that USB was really reliable on Windows.
If they have developed a microbe that basically can eat through any organic material
They've got a microbe that can eat any sugar. It already exists, it's called E. Coli. Your gut is full of it.
They didn't change the input side, they changed the output side, so it produces hydrocarbon fuels instead of fats.
The thing that worries me is... what happens if you get this modified E. Coli in your gut... I suspect you'd get pretty sick, like you'd taken a swig of gasoline.
But, man, this one has the best user interface I've seen, by far. And the other ones are either specific motion tracking games or are otherwise specific to some small part of the problem... plugins for particular applications, and the like.
I suspect there will be an open source clone out within six months though... possibly based on Intel's open source computer vision library.
I suspect Slusallek's custom ASIC cost a great deal more than a quad core system and certainly more than a modern GPU.
Sure, making anything in small quantity costs a lot more per unit.
But now what would it cost, per unit, if ATI or nVidia was making RPUs the way they make GPUs? Then it's largely a matter of process and gate count, and the RPU had a fraction of the gates in a modern GPU.
Phillip Slusallek was demonstrating full screen real time raytracing using a custom RPU (raytrace processing unit) in 2005, and that unit was running at less than 100 MHz. For a fraction of the hardware cost of a quad quad core system, you could do real time raytracing with less hardware investment than a modern GPU.
There is one caveat that you have to watch out for, if you run Linux hosted on another kernel. The FSF has taken the position that interfaces supported by a hosted implementation of an operating system are not native operating system interfaces... at least when the hosted implementation of the OS is "thin" enough: at least one UNIX-on-Windows implementation has had to avoid running GCC under their software and instead use a DOS/Windows port of GCC alongside their UNIX implementation. Running Linux in a VM, even an enhanced VM with specific APIs that Linux can call, would seem to be safe, but you would need to be doubly careful in a shallower kernel-on-kernel implementation.
The most straightforward "bright line" would still be a dual-CPU approach. This has apparently been used by other smartphone manufacturers to allow unrestricted application development alongside the cellular network. This approach would also be useful in personal computers. If the operating system, for example, is only used to deliver encrypted media to an audiovisual card with its own DRM firmware and codecs running on its own processor, then it doesn't matter whether the OS is Windows Vista, Linux, or OS X... and as an added bonus it would make all that encryption overhead Microsoft added to the Vista kernel look even more wasteful and foolish.
Wait until you've got mac fanboys flaming you for hating Apple and microserfs flaming you for being a Mac fanboy... in comments on the same article. The only constant in/. is that if someone disagrees with you you're a troll.
Slashdot Mirror
I think apple will have to face defecations.
A bit of Metamucil will clear that up.
A Rage II could do some pretty nice rasterization at 500x300 too, faster than 7.5 fps usually.
I would guess that the performance of Unreal on the Rage II would be a fair comparison to UT2003 on the RPU. The Rage II did about 8-9 FPS at 640x480 with a slower CPU handling the geometry updates for Unreal, and the RPU did 7.5FPS at 500x300 with UT2003. So I'd say it's comparable performance for comparable hardware.
OpenRT, a highly optimised software renderer on a 2.66 GHz Pentium 4, got 8FPS for the same game. A software OpenGL renderer running UT2003 rasterized on similar hardware (2.8 GHz Pentium 4) got about 60 FPS at the same resolution.
Now that the hardware has gotten to that point the quality and other advantages of raytracing are going to make it take off.
Based on the figures in the Siggraph '05 paper and the benchmarked performance of equivalent rasterizing hardware running the same engine, if ATI or nVidia had Dr Slusallek's hardware design in '96 when the Rage II came out, they could have a decent dedicated raytracing engine, in hardware, by the turn of the millenium. Or if nVidia had started cranking up a raytracing card in 2005 after Siggraph they'd probably have something that did 60 FPS at HD resolution for $100 by now.
According to your third point DNS is not an open standard.
So long as the properties I mentioned exist then there is nothing stopping an alternative implementation from being created when required.
And when that happens then you can start talking about it being open enough to make its openness really matter.
Openness is not a binary quantity, but there are certain tipping points where openness really becomes meaningful.
* Encumbrance. An interface or protocol can be publicly documented, defined, and implemented, but be patent-encumbered.
* Control. An interface where the *effective* definition belongs to an open standards body. This is one place both OOXML and Flash fail. It's also been an area where Java has historically fallen short of openness.
* Implementation. "It's not open until it's forked" is one way of putting it. If you can take the reference implementation and replace it without losing essential functionality, then it's open. You can replace it by forking an pen source code base, or by making a new implementation from scratch.
* Platform independence. Flash fails this test because the reference implementation isn't available in a portable form. This is a common way for a single vendor to fall down.
The classic open systems platform is, of course, UNIX. UNIX was explicitly not encumbered: AT&T donated the only patent essential to its implementation to the public domain. The API was small, tight, and easily implemented... by 1983, there were multiple hosted implementations like Eunice and Phoenix, native implementations like Regulus, workalikes like Idris and OS/9, and hybrids like Lanetix and Cromix... not to mention the Software Tools project to provide the majority of the UNIX API purely as a library. AT&T owned the trademark and the reference implementation, but they had no control over the standard... as witness the complete failure of "Streams" in the networking market. As for platform independence... it pretty much defined the term.
Flash, on the other hand, has a single reference implementation, under the control of that implementation's owner, and it's complex enough that an independent implementation will never catch up to it any more than WINE can catch up to Windows.
Without diving into the journal papers I couldn't find any mention of the resolutions his chip can run in realtime.
At 66 Mhz, about 6 million gates, 512x384, 7.5FPS. I don't have the performance for the later 90 MHz prototype on hand, but even the 66 MHz prototype was faster than any desktop CPU in 2005. It also required less memory bandwidth than rendering the same scene using a rasterizer... with about the same hardware resources as a Rage II.
Using an FPGA significantly limits the performance. Using current process technology you'd be looking at 400 MHz and 600 million gates at least. Raytracing is "embarrassingly parallelizable", and the RPU design had to be trimmed down to fit into an FPGA implementation, so I don't see any reason you couldn't get comparable performance to a rasterizer for similar scenes if you gave them a level playing field, otherwise it's like you're comparing a Rage II to a geForce 8800.
I have a lovely 256kbit core memory plane mounted in a contrasting brass-colored frame... which reminds me, I need to pull that out of storage and put it up in my cubicle again.
Additionally the exemptions sound like they apply solely to this bill.
These things have a habit of spreading... consider the effect of the "CAN-SPAM" act, which has effectively raised barriers to fighting spam by forcing people to prove over and over again that compliance with the CAN-SPAM act is not de-facto legalization.
If this bill replaces current laws restricting unauthorized access to computers, then we have a real problem.
That is the effect it will have in a large number of cases.
To plagiarize another saying about theory and practice, the difference between de-facto and de-jure is less de-facto than de-jure.
The final quote, "I find the whole experience totally extraterrestrial", wins the Internet.
Careful, those 5 words are probably found in an AP story!
The archive's sheer sprawl reveals both the possibilities and the limits of Otlet's original vision. Otlet envisioned a team of professional catalogers analyzing every piece of incoming information, a philosophy that runs counter to the bottom-up ethos of the Web.
This seems more like a real-time encyclopedia than the entirety of the web, like the next step beyond the Encyclopedia Britannica with its professional editing of contributed articles. The Britannica would have been in the process of switching to updating through supplements when he conceived this, and he could hardly have missed the controversy surrounding that move. This was a grand vision, and the fact that he was able to implement it even in abbreviated form is remarkable.
That would likely be a solid white screen. :)
Writing a spec doesn't make a platform open. Using open components doesn't make a platform open. An open platform is one that isn't tied to a single vendor, which may involve the reference implementation being open source (eg, sockets), it may involve having multiple implementations (eg, HTML), but when you need to buy Adobe's development environment to edit and author arbitrary SWF then it's not "open" in any meaningful sense.
So tell me, mister bones, how does one develop general purpose flash without using Adobe's proprietary development environment?
Open systems with a single proprietary key component are open in name only.
Turn it off, turn it on. Nothing was written to running-config.
In this case that's probably going to work.
When the router is at a power station in Guangzhou and you'd have to wait until 3AM and call in one of the people in the company who speaks the language so you can call the local office during their business hours and get them to power-cycle it for you... you're a lot happier that you remembered "reload in 15".
What's the IQ of a project tracking FOSS projects that's not a FOSS project?
Well, you could look at the diffs and track the percentage of each file that's got code from each committer... making sure to look for reverted code and giving credit to the person who originally committed the reverted code...
Linux kernel, Free/Net/OpenBSD, gcc, ... the core infrastructure
Maybe you just missed the news that Microsoft bought Yahoo, and that Yahoo! search is now "MSN Livesearch"...
Not quite yet, Bill, keep your pants on.
Indeed, sorry for being less than clear, that's what I meant.
Either way, this would be pretty unpleasant.
Another thing I thought of, though: if the bacteria is generating hydrocarbons instead of fats that it's using for energy reserves, then it's inherently not competitive with wild strains and thus unlikely to establish itself, but I'd love to see an explicit reference to that.
Bruteforcing an interrogation subject can be very quick indeed.
There should be a great Princess Bride joke here, but I'm too tired to think one up.
As mentioned above, there were two driver interfaces for the original USB standard, and the only people who knew were driver writers and nerds compiling their own custom kernel.
And people who use drivers and custom kernels. I didn't do either, but I still had to deal with users whose USB controller cards didn't work with NT4 because of driver conflicts, and then there were devices that had custom drivers that only worked with one or the other chipset. It wasn't until Windows 2000 that USB was really reliable on Windows.
If they have developed a microbe that basically can eat through any organic material
They've got a microbe that can eat any sugar. It already exists, it's called E. Coli. Your gut is full of it.
They didn't change the input side, they changed the output side, so it produces hydrocarbon fuels instead of fats.
The thing that worries me is... what happens if you get this modified E. Coli in your gut... I suspect you'd get pretty sick, like you'd taken a swig of gasoline.
But, man, this one has the best user interface I've seen, by far. And the other ones are either specific motion tracking games or are otherwise specific to some small part of the problem... plugins for particular applications, and the like.
I suspect there will be an open source clone out within six months though... possibly based on Intel's open source computer vision library.
I suspect Slusallek's custom ASIC cost a great deal more than a quad core system and certainly more than a modern GPU.
Sure, making anything in small quantity costs a lot more per unit.
But now what would it cost, per unit, if ATI or nVidia was making RPUs the way they make GPUs? Then it's largely a matter of process and gate count, and the RPU had a fraction of the gates in a modern GPU.
Phillip Slusallek was demonstrating full screen real time raytracing using a custom RPU (raytrace processing unit) in 2005, and that unit was running at less than 100 MHz. For a fraction of the hardware cost of a quad quad core system, you could do real time raytracing with less hardware investment than a modern GPU.
There is one caveat that you have to watch out for, if you run Linux hosted on another kernel. The FSF has taken the position that interfaces supported by a hosted implementation of an operating system are not native operating system interfaces... at least when the hosted implementation of the OS is "thin" enough: at least one UNIX-on-Windows implementation has had to avoid running GCC under their software and instead use a DOS/Windows port of GCC alongside their UNIX implementation. Running Linux in a VM, even an enhanced VM with specific APIs that Linux can call, would seem to be safe, but you would need to be doubly careful in a shallower kernel-on-kernel implementation.
The most straightforward "bright line" would still be a dual-CPU approach. This has apparently been used by other smartphone manufacturers to allow unrestricted application development alongside the cellular network. This approach would also be useful in personal computers. If the operating system, for example, is only used to deliver encrypted media to an audiovisual card with its own DRM firmware and codecs running on its own processor, then it doesn't matter whether the OS is Windows Vista, Linux, or OS X... and as an added bonus it would make all that encryption overhead Microsoft added to the Vista kernel look even more wasteful and foolish.
Wait until you've got mac fanboys flaming you for hating Apple and microserfs flaming you for being a Mac fanboy... in comments on the same article. The only constant in /. is that if someone disagrees with you you're a troll.