Origins are not SMP, they're NUMA. I find it hard to believe they can get decent performance/scalability on a NUMA machine without an OS designed for memory locality.
I searched long and hard for pictures of the Darth Vader Gargoyle on the
Washington National Cathedral, but there just don't seem to be any.
--
Patrick Doyle
When on earth, do as the humans do.
And if we visit them, the same applies in reverse.
I don't think it's quite that simple. What if the bouncy rubber ball creatures show approval by dropping each other off tall buildings? I would like to think they would be considerate enough not to do that to me when I visit them. --
Patrick Doyle
No, nothing short of an infinitely-parallel machine can do O(n) for sorting. Remember, the big-O notation refers to asymptotic complexity, which means the problem size increases without bound. If your computer is not infinitely parallel, then there will exist some n which is too large to fit. --
Patrick Doyle
Re:The emperor looks great in those new clothes
on
FPGA Supercomputers
·
· Score: 2
FPGAs can't outperform custom hardware. They can outperform CPUs because CPUs are general-purpose hardware that run programs in serial, while FPGAs are general-purpose hardware that run programs in parallel. But special-purpose hardware will always win. (Whatever technology you make the FPGA out of, you could just make the custom hardware on the same process and get an order of magnitude improvement.) --
Patrick Doyle
You can think of an FPGA as a digital circuit simulator. You can design any digital circuit, and an FPGA can simulate it roughly one or two orders of magnitude slower than the circuit would run if you made a real IC out of it.
Logic operations can be described with truth tables. FPGAs contain programmable truth tables (called lookup tables, or LUTs), so you can implement whatever logic operation you want. They also contain programmable interconnects that allow you to join your LUTs in any way you want.
Usually, they also contain some memory, because it takes a lot of LUTs and interconnects to build memory, and the resulting memory would be very slow and wasteful.
How is this faster than a CPU? Well, the win comes when you design a custom circuit to perform a certain task, rather than using a general-purpose CPU. For instance, if you could make a citcuit to do something at 100MHz when it would take, say, 100 Pentium instructions, then your FPGA would outperform a 10GHz Pentium!
Used in this way, FPGAs are the ultimate parallel computer. They have many thousands of very small processing units (LUTs). --
Patrick Doyle
Is it really true that the last mile of copper wire is the bandwidth bottleneck? I always thought that was a myth, and the problem was the 4kHz filter the phone guys put on voice lines so they can multiplex several calls on one wire. --
Patrick Doyle
NASA, in the USia, spends literally millions of dollars designing the space shuttle's computing system from the ground up. Their in-house coders pour over each microchip and line of code hundereds of times looking for even the smallest bug. This superior attention to detail is not possible using the limited resources of the open-source method.
That's exactly where open-source shines. Anyone can look at the code. Look at OpenBSD for instance. LEON could end up being carefully inspected by more knowledgable people than any other processor design ever used by any space agency. Heck, if the agencies are worried about it, they can pour over the design just as they would for any other chip.
--
Patrick Doyle
I can only access the abstract of the article, but isn't the size of the problem they can solve limited to the size of their "microfluid" computer? If so, what's the breakthrough?
If you are allowed to limit the problem size, then you can solve it on a normal computer. All you need to do is make the computer big enough. For instance, you could make a travelling-salesman computer out of a circuit what builds all possible paths in parallel and then use pairwise comparison to find the lowest-cost path.
This assumes the circuit is large enough to hold the whole search space, which is the fatal flaw. Classification of a problem as NP-complete is based on its asymptotic complexity: how the computation time grows as the problem size grows without bound. Put a bound on the problem size, and you haven't solved anything.
It seems to me that the microfluid computer has the same flaw.
Does anyone know more about this? Please straighten me out. --
Patrick Doyle
You have it wrong too. If you can find a problem in NP which can't be solved in polynomial time, you'll be famous.
A problem X is in the set NP-hard if all problems in NP can be transformed into X in polynomial time. Thus, every problem in NP-hard is at least as hard as every problem in NP.
A problem is in NP-complete if it is in NP and NP-hard. Thus, NP-complete is a "representative" set of NP problems, such that solving one of these in polynomial time would also mean solving all other NP problems.
Here is the FOLDOC entry for NP-hard. It explains all this. --
Patrick Doyle
I got the strong impression that NIMA just re-examined images from the Mars Global Surveyor. Examining images at the limit of their resolution is no easy task, so it's not surprising that an agency what specializes at that would do better than NASA on the same images. --
Patrick Doyle
The "pure math" thing is not why software patents are harmful. Patents on actual devices are just patents on "pure matter", so if you construct a collection of matter that someone else has patented, you're breaking the law. Does that make patents on matter wrong?
Software patents are wrong (at least in their current incarnation) because they do not fulfill patents' stated purpose of encouraging innovation. --
Patrick Doyle
No, diving from a baloon is totally different from diving from orbit. In orbit, you're travelling at 18,000mph, so you burn up on re-entry unless you have some serious heat shielding. A baloon is essentially stationary with respect to the ground, so you're starting from a standstill. --
Patrick Doyle
Slashdot Mirror
Origins are not SMP, they're NUMA. I find it hard to believe they can get decent performance/scalability on a NUMA machine without an OS designed for memory locality.
See the k42 project.
--
Patrick Doyle
I searched long and hard for pictures of the Darth Vader Gargoyle on the Washington National Cathedral, but there just don't seem to be any.
--
Patrick Doyle
--
Patrick Doyle
Duh, from the movie First Contact. :-)
It's a good one, even if you're not a Trekkie. Go see it.
--
Patrick Doyle
--
Patrick Doyle
A guy named Erbo said commented on
this article
saying that the Prime Directive didn't exist when the Federation was formed.
--
Patrick Doyle
What if there's something they consider a basic right, which we consider immoral? Or vice-versa?
--
Patrick Doyle
Heheh, no problem. I'm glad you took my flame so lightheartedly. :-)
--
Patrick Doyle
Do I look stupid to you?
My point (right or wrong) was that the bandwidth of copper wire is not the limiting factor.
--
Patrick Doyle
No, nothing short of an infinitely-parallel machine can do O(n) for sorting. Remember, the big-O notation refers to asymptotic complexity, which means the problem size increases without bound. If your computer is not infinitely parallel, then there will exist some n which is too large to fit.
--
Patrick Doyle
FPGAs can't outperform custom hardware. They can outperform CPUs because CPUs are general-purpose hardware that run programs in serial, while FPGAs are general-purpose hardware that run programs in parallel. But special-purpose hardware will always win. (Whatever technology you make the FPGA out of, you could just make the custom hardware on the same process and get an order of magnitude improvement.)
--
Patrick Doyle
You can think of an FPGA as a digital circuit simulator. You can design any digital circuit, and an FPGA can simulate it roughly one or two orders of magnitude slower than the circuit would run if you made a real IC out of it.
Logic operations can be described with truth tables. FPGAs contain programmable truth tables (called lookup tables, or LUTs), so you can implement whatever logic operation you want. They also contain programmable interconnects that allow you to join your LUTs in any way you want.
Usually, they also contain some memory, because it takes a lot of LUTs and interconnects to build memory, and the resulting memory would be very slow and wasteful.
How is this faster than a CPU? Well, the win comes when you design a custom circuit to perform a certain task, rather than using a general-purpose CPU. For instance, if you could make a citcuit to do something at 100MHz when it would take, say, 100 Pentium instructions, then your FPGA would outperform a 10GHz Pentium!
Used in this way, FPGAs are the ultimate parallel computer. They have many thousands of very small processing units (LUTs).
--
Patrick Doyle
Is it really true that the last mile of copper wire is the bandwidth bottleneck? I always thought that was a myth, and the problem was the 4kHz filter the phone guys put on voice lines so they can multiplex several calls on one wire.
--
Patrick Doyle
--
Patrick Doyle
I can only access the abstract of the article, but isn't the size of the problem they can solve limited to the size of their "microfluid" computer? If so, what's the breakthrough?
If you are allowed to limit the problem size, then you can solve it on a normal computer. All you need to do is make the computer big enough. For instance, you could make a travelling-salesman computer out of a circuit what builds all possible paths in parallel and then use pairwise comparison to find the lowest-cost path.
This assumes the circuit is large enough to hold the whole search space, which is the fatal flaw. Classification of a problem as NP-complete is based on its asymptotic complexity: how the computation time grows as the problem size grows without bound. Put a bound on the problem size, and you haven't solved anything.
It seems to me that the microfluid computer has the same flaw.
Does anyone know more about this? Please straighten me out.
--
Patrick Doyle
A problem X is in the set NP-hard if all problems in NP can be transformed into X in polynomial time. Thus, every problem in NP-hard is at least as hard as every problem in NP.
A problem is in NP-complete if it is in NP and NP-hard. Thus, NP-complete is a "representative" set of NP problems, such that solving one of these in polynomial time would also mean solving all other NP problems.
Here is the FOLDOC entry for NP-hard. It explains all this.
--
Patrick Doyle
Mutant microbes don't need Mir to get down to earth. If they could survive reentry on Mir, cound't they make it here on their own?
I don't think Mir has added significantly to the risk of invasion by mutant microbes.
--
Patrick Doyle
I got the strong impression that NIMA just re-examined images from the Mars Global Surveyor. Examining images at the limit of their resolution is no easy task, so it's not surprising that an agency what specializes at that would do better than NASA on the same images.
--
Patrick Doyle
What is there to learn from the Napster case? Napster lost.
--
Patrick Doyle
Having trouble finding online dictionaries?
--
Patrick Doyle
"Primary bottleneck" and "all the other bottlenecks"? How does that work?
--
Patrick Doyle
The "pure math" thing is not why software patents are harmful. Patents on actual devices are just patents on "pure matter", so if you construct a collection of matter that someone else has patented, you're breaking the law. Does that make patents on matter wrong?
Software patents are wrong (at least in their current incarnation) because they do not fulfill patents' stated purpose of encouraging innovation.
--
Patrick Doyle
Actually, that's 44100 * 2 * 16 bits/sec. Somehow you came to the right total though. :-)
--
Patrick Doyle
No, diving from a baloon is totally different from diving from orbit. In orbit, you're travelling at 18,000mph, so you burn up on re-entry unless you have some serious heat shielding. A baloon is essentially stationary with respect to the ground, so you're starting from a standstill.
--
Patrick Doyle
Encyclopoedias are full of "statements of fact". Does that make them non-copyrightable?
You may be right that lists of mp3s are not copyrightable, but I don't buy the "statement of fact" criterion.
--
Patrick Doyle