Legally I think you are wrong, in that the UK could not extradite for something policital like wikileaks. Sweden could "lend" him to the US to face charges as I understand it rather than extradite.

You are for sure wrong on a practical level though, as the UK has fundementally failed to extradite the guy to sweden let alone the US. In Sweden he would not have been out on on bail, but in jail so the claiming assilum in the embassy is not possible.

True, but "porting" something to the PHI so that it runs at roughly the same speed as the host GPU is hardly progress. The phis have lower clockspeed than the host cpus and in-order execution.

I fully accept that there is a lot of stuff that needs a lot of CPU that will run badly on a GPU. What I am not convinced about is that there is much that does not map to a lot of SIMD threads that WILL run well on a Phi. You need 240 threads, you need to fill the 512 bit vector registers with SIMD operations, to get peak performance you need huge amounts of parallelism. The same as you need on a GPU.
You have the same problem with the PCIe bus being horrendously slow as well, you can offload the whole thing (which is coming to GPUs as well, with ARM procoessors for the more general purpose stuff), but you are still limited to one card, and getting from one card to another is a PITA, particularly if they are on different nodes.
For general purpose stuff needing bucket loads of GPUs you use normal CPUs with over twice the clock speed, out of order execution and a shorter vector length,much more memory per core. Potentially together with MPI and fast inter-node communication (e.g. infiniband, Cray Aries etc).
For some things GPUs are great, for others they are horrible. My gut feeling is pretty much the same things will be great or horrible for Phis as well. I am not really familiar with the algorythms for the stuff you are talking about, but sure if they do not come back to a lot of SIMD threads they will work badly on a GPU. What remains to be shown is if they can work well on a Phi.

Huh?

Are you seriously suggesting that some stuff will work well on Xeon Phi but badly on a Kepler? To get good performance out of a GPU you need thousands of threads. To get good performance out of a Xeon Phi you need thousands of SIMD instructions. A Phi has 60 cores, to get maximum performance you need 120 threads minimum because it needs to alternate between at least two threads to get peak performance, IIRC it can't schedule the same thread consecutively, so if you don't have at least 120 threads it is not even possible to fill the machine theoretically. It is recommended to have at least 240 threads to hide the memory latency (like a GPU it has very low overhead thread swapping so you can just move to another thread while you are waiting for your data from memory). Then it has 512 bit vector instructions (which is enough for 16 floats or 8 double precision numbers), so you are looking at around 4 thousand SIMD floats at LEAST in flight just to even get close to filling the machine.

Getting stuff to work on a Phi is easier than getting it to work on a GPU, but anything that works close to peak on a Phi will also work well on a GPU.

Yeah was wondering when someone would mention OpenACC, for sure it is the most painless way to start programming on GPUs, as it is compiler directives which means that the compiler deals with most of the heavy lifting for you. For general purpose machines you need to use the PGI or CAPS compilers, which are not free.

Cuda is not too bad either as it assumes a GPGPU, and is a relatively straightforward extension of C. OpenCL is a mess in my opinion as it does not assume anything so you have to spend 10 complicated function calls explaining you have a GPU before you can actually do anything.

OpenACC also has the advantage that it is compatible with CUDA so you can write important functions in CUDA for improved performance, and leave the less performance critical parts in OpenACC.

The first step though really is to understand the hardware, writing GPU code is very different from writing normal code if you want to get good performance, althoguh assuming it vector code with a SIMD size/vector length of 32 will get you a long way as well.

Not sure where you got 1,000,000 from, it is 1,000 petaflops in an exaflop

http://physics.nist.gov/cuu/Units/prefixes.html

For sure there are significant challanges, in programming methods as much as anything else. Cores are not getting much faster, what you get are more and more of them, which makes huge demands on the amount of parallelism you have. And means IO becomes comparatively very slow indeed.

Well scaling is always a problem, but not at the 50 CPU level. In the basement at work we have a machine wtih 30,000 cores and running on that many is definately a scaling problem.

Cores are not getting any faster though (due to power footprint mainly), so scaling is a problem that is going to come to everyone sooner or later.

Well a quick google and we get this

https://asc.llnl.gov/computing_resources/sequoia/configuration.html

16 cores per CPU/chip. (or according to wikipeda 18, but one is used for the OS and one is saved as a spare).

Note also that each dimension of the torus does not have to be the same, so the constraint is

No of cores = 16*A*B*C*D*E

That is assuming each node on the torus has 16 cores (potentially could be a multiple of 16).

Anyway according to
http://en.wikipedia.org/wiki/Blue_Gene

The system is 96 racks
Each rack is 32 node cards, each node card has 32 modules, each module has 16 cores.=1,572864 total cores.

I would guess there are two dimensions between the racks (12 x 8) then the remaining dimensions are inside the racks, potentially 12x8x16x2x32 but probably split a bit more than that so the longest dimension is not 32, 12x8x16x4x16 maybe, where there are two seperate links inside the node cards splitting them into two groups of 16 modules.

There is a couple of different ways they could wire up the 3 dimensions in the cabinets so that part it is hard to be sure.

Well if you had read the document that he pulled from the wayback machine it would make sense to you. This was an International Broadcasting Trust meeting

"The International Broadcasting Trust(IBT) has been lobbying the BBC, on behalf of all the major UK aid and development agencies to improve its coverage of the developing world. One of the aims is to take this coverage out of the box of news and current affairs so that the lives of people in the rest of the world, and the issues that affect them become a regular feature of a much wider range of BBC programs, for example drama and features."

Could not find an evidence that this was the meeting where the BBC decided global warming was real, but that will never stop the FUD.

You are getting two events completely confused

http://en.wikipedia.org/wiki/Iranian_Embassy_siege

http://en.wikipedia.org/wiki/Murder_of_Yvonne_Fletcher

Especially as the storming happened in 1980 and the shooting in 1984 and one was the Iranian embassy and the other one the Lybian one.

I guess they did not get anyone that technical to write that article or the summary.

For I/O I guess they mean memory bandwidth. GPUs have a LOT of memory bandwidth from their cache memory, the problem is that they sit at the end of a PCIe bus from the CPU and the CPU has to handle most of the book keeping (and the actual IO, i.e. taking data from an external source).

So what is important is the compute density i.e. how much computation you do for each piece of data. Getting stuff into the GPU is slow, getting stuff out is slow, but doing stuff on the data is very very fast (because you have so many compute units and so much memory bandwidth).

That is also the way they are programmed, with the main code running on the CPU, and then the kernals getting launched on the GPU with explicit or implict transfer of data from the CPU memory to the GPU memory and back again.

I do have high hopes for stuff like Fusion ( http://en.wikipedia.org/wiki/AMD_Fusion ) which gets rid of the PCIe bus, and make it a lot easier to get data to the GPU cores and back again.

And if you are going to mention GPU machines, why not mention titan ? ( http://www.olcf.ornl.gov/computing-resources/titan/ )

USians really seem to think that WW1 was very much the same as WW2. There are a lot of similarities (The US arriving rather late for one), but there are important differences.

The first point was that in WW1 the war ended in 1918, it started in 1914 and the US arrived in 1917. Well they declared war on 6 April 1917, but did not provide much assistence until 1918. According to the wikipedia page, 10,000 soldiers a month were arriving in the summer of 1918, but that was mere months before the armastice. The US intervetion just changed the situation from a war the germans were losing, to a war the germans had no chance at all of winning. Their population was slowly starving from the Naval blockade. Note that at the end of the war the Germans still had gained territory (which is how the "we were winning, jews sold us out" stuff that lead to WW2 was plausable to the ignorant).

Clearly WW2 was rather different and the US involvement was decisive. I do find it ironic how USians can criticise Chamberlin for the Munich agreement while ignoring the fact that the US only got involved over two years from the start of the war and after they were themselves attacked.

The fact remains that the while the US was a major power before WW2, it was only WW2 that showed how powerful the US was to the world. Before WW1 the british empire was 20% of the worlds surface and 25% of the worlds population and by far the most powerful world force. Between the wars the british empire declined and the US grew, but it was only WW2 that showed how the balance of power had shifted.

The US is declining though, the US economy is still in big trouble and the government becoming more authoritarian (in the name of fighting terrorism). George W Bush with is "we dont need the rest of the world" attitude started the decline. There culture of cooperation between the western allies was pretty much destroyed by the Bush administration. Clinton was hugely popular at least in the UK (he made a huge difference in making peace in Northern Ireland for example). Obama repaired some of the damage, but the US political climiate is horrible with noone willing to work towards the centre.

There is one scary thing though for sure, if you thought a US dominated world was bad, a chinese one will almost certainally be worse.

The problem is that most Leftists and many Europeans can no longer distinguish true friend from foe, good from evil. It doesn't matter how good the US is, they can't recognize it. It is kind of similar to the way that anti-Semitism is rapidly growing in Europe.

Unless trends reverse, I expect there will be major wars in Europe in our lifetime.

The problem is that most lefts and Europeans can no longer accept that the ends justify the means, and actually understand how the world works, that doing evil in the name of good (eg torturing suspected terrorists), cutailing civil liberties and cenralising power. Freedom has a huge price, many revolutions bloody civil wars in Europe are testemony to that. If we give away freedom for security, then getting it back again is going to have a very high price indeed. I guess you are from the US, you should go and read what the founding fathers wrote, they were a smart bunch.

"They who can give up essential liberty to obtain a little temporary safety, deserve neither liberty nor safety" Ben Franlin

"The price of freedom is eternal vigilance" Thomas Jefferson.

The only people who can take away the Freedom in the US is the US government, so you have to watch what they do like a hawk, unfortunately the threat of terrorism, which we have lived with in Europe for decades (ETA, the IRA etc - note also the IRA were funded by weathly americans, in a similar way that AlQuida has been funded by wealthy Saudis)

Anti-semetism is growing in Europe because the Israeli right wing (and part of the current government) wants a greater Israel from the river Jorden to the sea. Unfortunately quite a few Palestinians live there and don't want to leave. So they think they can treat them like crap, kill thousands and eventually that will happen. In Europe we remember what it is like being a colonial power and taking over land that does not belong to you. America was built on taking someone elses land, which might explain the difference. We see the Israelis as the evil british empire, and the palestinians as the natives. You see settlers and indians.

Sorting out the Israel-Palestinain problem i probably the most effective thing that could be done to stop the terrorist threat against the US, that requires the US facing down the Israeli right wing, which is unfortuntely not going to happen.

Slightly pedantic point. I said "from my bank I can get either a Mastercard or a Visa card and that is about it", Discover is not very big outside the US, American express issue only the cards themsleves as I understand it, and not accepted everywhere but anyway.

Please though can we get away from the missuse of the the word terrorism. Terrorism involves killing people and threatening to kill people to get what you want. This is civil disobeidence. The best analogy is a something like a picket. It is annoying, it gets in everyones way and costs money. It has nothing to do with terrorism, and trying to connect them in any way cheapens the whole thing.

The whole "we need these draconian rules to deal wiht terrorists" to using those rules in situations that have nothing to do with terrorism is dangerous. The next time you try to change the rules to "deal with terrorists" you will get a fight.

I do wonder though what you think a reasonable response would be to MC and Visa? Democracy requires the right to protest, so what do you think would be ok? An actual picket of the MC and Visa head offices? Honestly I think the economic damage you quote is overstated, but that is getting very offtopic.

In terms of Mastercard and Visa, I would say it was a bit more more complicated, as they have a near duopoly of card processing methods. I have no problem for example with BOA, or Amazon etc. as there are plenty of alternatives but from my bank I can get either a Mastercard or a Visa card and that is about it.

Also you do not have an absolute right to turn down service. In a lot of countries I would suggest if you put up a "No Blacks" etc. sign outside your shop etc then you would quickly get in legal trouble. Discrimination on grounds of skin colour, sexual orientation etc is in most places illigal, and rightly so. I would suggest that discrimination on the grounds that the government does not like you is something much worse.

The biggest problem I have is that it does look (to me at least) like government pressure was behind the decision, and that makes it very different to then deciding for themselves that they don't want to deal with wikileaks. Well that and the "we need these draconian rules to deal wiht terrorists" being subverted and the powers that be trying to use them on everyone (i.e. the attempt to put wikileaks on the list of terror orginisations.)

Ok as a first point I have to agree with some of that, there is some common ground there. First of all though what I disagree with. Your "Military Solution" is something I find hard to provide a polite description for.

First of all Sanctions on Saudi Arabia. The trade between the US and SA consists of the US buying Saudi oil, and the Saudi's buying US military equipment (combined with generous kickbacks) and assorted Luxuary items. Now if you stop that trade you get a replay of 1973 (http://en.wikipedia.org/wiki/1973_oil_crisis ). And there are plenty of people queuing up to sell SA arms (which is why large kickbacks can be and are demanded). So sactions is basically the US shooting itself in the foot.

Now bombing, wow. I would hope you would accept that the current Saudi government are broadly pro american, and certain rich individuals are giving money to islamic charities, that funnel funds to terrorists. Now true the government could be doing more to clamp down on this, but man, you bomb SA and either you make the next Iraq or the next Iran, which is not exactly going to help your cause.

So what can be done? As you say moving away from Oil, that would be great start. Not buying their oil because you don't need it is a great way to stop the flow of cash. The fact remains while you depend on the oil, the US needs SA a lot more than SA needs the US (all the US does that SA wants is provide regional stability/security).

The other thing you missed is finally pressurize Israel to the negotiating table. There is a deal to be made there along the lines make the final land borders the pre 1967 borders more or less then have some land swaps. Something like that would work probably for the palistinian side, but is completely unacceptable to the Israeli right wing (who want greater/biblical Isreal that is also a jewish state). For once there is a lull, and the US needs to push for a deal.

A lot of donations from rich americans to republican charities linked to the IRA dried up after they stopped seeing the UK as an enemy, and realised that terrorism was not the answer. A just solution to the Isreal-palistinan problem would go some way for doing the same thing in SA.

The solution is actually soft power, dipolomacy and intelligence. Unfortunately those seem to be in disrepute in the US.