Um... while I agree with your OP (ie. hell will freeze over ere the GOP as a whole will put itself firmly behind the results of this study committee), don't you think it sounds a bit hypocrite to first give Socrates credit for proposing the concept of the freedom to share ideas, then in the next sentence decry the "stealing of ideas from others"? I'm even honestly wondering whether your post is a parody. Anyone campaigning against overreaching copyright laws that stifle progress will be happy if a study committee (in particular a republican one) picks up some of their ideas. Its kinda the whole point of activism.
Try this: I give this study committee credit for having the balls to put truth and integrity over years of party policy. I know in an ideal world, I'm giving them credit for just doing their job, but in the actual world I live in, this is a rare enough occurrence to merit commendation.
It's not "directly derived from"; it's "loosely inspired by" http://en.wikipedia.org/wiki/DOS#Origins
Which makes your point rather far-fetched. CP/M was an entirely different product, sharing no code base with any product marketed as "DOS". So, no, DOS was never 8-bit software. Otherwise you could also state that "windows at one point ran on the (Xerox Alto/Apple Lisa)" and "Linux at one point ran on the PDP-11/20".
Thank you for the clarification. I'm not a native English speaker so I didn't understand the slang in the parenthetical part. In fact, I read it several times wondering what it was about. Even now, with your hint and after a stop at urbandictionary.com , I'm not entirely sure how the sentence should be parsed./. has a brutal learning curve for non-Americans; only after years of lurking from my home country, then spending a few years in the US, I got brave enough to create an account and start posting stuff. And, as you see, I still miss a beat now and then.
That said, I did think it was probably satire; that's why I started with "you're probably trolling". I replied anyway because I know for a fact that there are a lot of people believing crap like that - if not the poster, then maybe a gullible reader. It's not even all that ridiculous to someone who doesn't have any DSP background.
Oh c'mon, can you possibly get more obtuse? This is just colorful speech, and the "not that I advocate that" was specifically added to drive home that fact. If GP would have winked, you'd just have written 'see, they winked to indicate they are not serious about the "not that I advocate that"'. Either way, no-one in their right mind will see this as a "call for the death of their opponents". And yes, I am implying with this that you're not in your right mind. Whether you're on crack or it is a permanent condition is a different question.
You're probably trolling, but just in case you aren't. Talking about math, are you familiar with the concept "standard error"? Because the standard error on the "level" on a real-life vinyl record is far greater than the difference between two of your 65536 audio levels. Hell, you'll probably lose a level here and there every time you play back the record. And that's just one problem with vinyl. There are inherent issues with low frequencies, and many more. There's a nice long list: http://en.wikipedia.org/wiki/Vinyl_record#Shortcomings
Anyhow, to put the original point differently, CDs have a greater dynamic range than the maximum real-life signal-to-noise ratio of vinyl records. If there's a problem with a CD's dynamic range, it's because it's mastered so that only a tiny fraction of the available dynamic range is used; see "loudness wars".
Nope, these beats are a physical phenomenon that occurs whenever two waves travel through the same medium. Its mathematical description is very simple and it's commonly taught in Physics 101 classes. http://en.wikipedia.org/wiki/Beat_(acoustics)
I think you're confused with binaural beats, where the brain essentially simulates the interference between two tones that are not traveling through the same medium (assuming headphones are being used). Now, if you can demonstrate audible binaural beats between two ultrasonic frequencies under controlled, scientifically sound, double-blind circumstances, then you have a point. And a paper in Nature/Science. http://en.wikipedia.org/wiki/Binaural_beats
I hate the fact that my taxes go to support this organisation that I have no control of, and can write laws that I have no influence over.
Assuming that you're an US voter, I'll would like you to know that your elected government has a vote in most if not all UN organs. Yes, you're not directly voting for the proposals that are on the table. Yes, your vote is not weighted by the population of your country. Sounds a bit like, say, electoral colleges, don't you think?
I'm glad that I don't live in Europe though - the EU seems a much more nasty (and expensive) equivalent. And they can enforce laws.
Have you tried talking to a few Europeans? I for one enjoyed living under my EU overlords. They're much more pragmatic in their decision making and there's much less over-the-top ideological drama/gridlock. Comparing it with the US federal government, I'm always amazed that this country hasn't collapsed under its own inertia yet.
Admittedly, the EU still has far less power over its member states than the federal government.
Sheesh, take a cold shower or something. Opening a debate about something doesn't equal lining up laws. It seems that in the country of freedom, some things (like putting limits on said freedom) cannot even be talked about.
And moderators, what's the thing with this being modded +5 interesting? Apparently, the US hive mind still hasn't gotten over the UN making poor Dubya go to Iraq without its stamp of legitimacy. Which is weird considering that most do admit the war itself was, in fact, not really legitimate.
- A lot of Windows users are anything but Microsoft devotees; they're just using Windows because of a perceived lack of other viable options - and not liking it. I was like that once.
- I use Linux just because it's rock-stable, responsive, secure, light on system resources, there's nothing I don't want going on behind my back (or at least much less than on some other operating systems), I can literally customize every last bit of it, I paid 0$ on software yet I'm running 0 pirated programs, and all my (open-source) software gets updated through one unified "package manager" in a fraction of the time it takes to run updates on Mac and Linux. I install closed-source software on my Linux boxen whenever I feel it's opportune - now what is this "slavish devotion" you're talking about? Also, it appears you're calling Android and Red Hat, inc market failures. Care to expound on that?
- Oh wait, now I remember the "slavish devotion" thing - that brings us to Mac. I give you OSX is not a bad operating system, though I personally could never live with the vendor lock-in. But since you're clearly not into arguing technical merits, well, it seems to me that the Mac userbase has the highest percentage of slavish devotees of all operating systems (though they're usually better than you at giving reasons to use Mac; your attitude and ignorance makes me scared to even come close to anything branded "Apple"). Some of them even realize there would never exist a Mac OSX without its open-source base. With the resources Apple (and NeXT) had at the time, there's just no way they could have coded up a modern competitive operating system without using BSD and Mach (among others).
Mostly agreed, except that they could be a bit more careful to either have no identifiable account information in the URL or not have trackers on pages that contain account information; preferably both. What would be a better term than "leaking" for this blatant disregard of basic security practices? (That's not a rhetorical question - more an attempt to start a lame semantic discussion.)
BTW, this, boys and girls, is why you should never browse without a decent script blocker. The revenue and sustainability of ad-supported sites is secondary to your privacy - or so I believe.
First, let me say... I was quite an asshole there, wasn't I? My only excuse was having my threshold for "probably an idiot fanboy" set to a hair trigger after slogging through a typical slashdot CPU flamewar (i.e. one full of fanboys)
No problem, I can relate. I tend to visit/. predominantly when I'm too tired to do anything useful (like some people plop in front of the TV) so I tend to be a bit short on patience and judgement myself. My original post is a good example of this; I got ticked off by the person starting the thread bluntly stating that (paraphrased) the shared FPU kills the Bulldozer for HPC applications. To me, this is simply false; the shared FPU was twice as wide as the (then-competing) Westmere, so all else being equal (which obviously isn't the case), the architecture of the bulldozer module with a 256-bit FPU supporting 2 "virtual cores" would be expected to give similar FP throughput than 2 Westmere cores (supporting 4 "virtual cores") with two 128-bit FPUs. In the case of poorly vectorizable code with non-optimal parallel scaling (which happens to be our case), the shared 256-bit FPU even has a distinct advantage because you need half as many parallel threads to benefit from, as you put it, branch mispredict bubbles being filled in with instructions from the other thread. In my opinion, the shared FPU by itself is fundamentally not a bad design and is totally not the reason for Bulldozer's competitive failure against Westmere; the real reasons here are many and slightly more subtle. Admittedly, that's a bit of a historical discussion since Sandy Bridge came out not so long afterwards touting one 256-bit FPU per core. But it still it infuriates me that people keep on blaming the shared FPU, which explains why my original post was quite rambling. I can easily understand how it set off your BS detector. Anyhow, I'm glad that we succeeded to turn a potential flamewar into a civilized discussion and I appreciate your insights.
I think you'll find most serious HPC users are willing to do benchmarking instead of always taking someone else's word for it. If you're going to fill a few racks full of compute nodes, it's pretty easy to justify buying one small node of each flavor to do some competitive evaluation, or perhaps rent time on someone else's systems.
There are large numbers of small to medium-sized HPC shops out there that don't even come close to the top 500 but probably contribute significantly to chipmakers' HPC sales. This is our situation; my boss was somewhat annoyed at the loss of time associated with our elaborate benchmarks. In the end, I think we did right, but it did delay our deployment by a couple of weeks. Time is money; unless there are clear red flags, bosses in this situation typically prefer speedy deployment over a small chance (remember our benchmark results were completely against expectations) of getting a financially slightly better deal. Therefore, they tend to default with what they know and trust, and that's often "Intel Inside". Again, we almost did just that. It's difficult to quantify the impact of this effect on overall sales, but I wouldn't dismiss it, especially given how overwhelmingly negative the Bulldozer's press coverage is.
Sandy Bridge has a smaller branch mispredict penalty (because its execution pipeline isn't as deep as Bulldozer's -- BD targets higher frequencies and that's the tradeoff for it), and significantly better branch predictors too. Since you say your software is branchy, I suspect Bulldozer needs to fill branch mispredict bubbles with instructions from the other thread.
The conditions for branching in our software are mostly based on simulation data that is inherently chaotic; it's hard for me to imagine a branch predictor doing much better than 50%, no matter how good it is. You're right that Bulldozer will have bubbles, but so will Sandy bridge. That (and the fact that I didn't think our code is that bra
That's pretty sad for AMD considering Westmere doesn't even implement AVX.
It's not sad at all considering that our code only gets a modest boost from AVX (as I said, a lot of it is not easily vectorizable) and that that similarly-priced AMD parts are clocked way higher.
Did you try Sandy Bridge, which does support AVX?
FYI, Romley (which I talked about earlier) was Intel's code name for Sandy Bridge-EP, mister Anonymous Expert. To give a little bit more information, on a per-clock basis, it ran our code about 35% faster than than Westmere. But on a per-price basis, AMD still won comfortably.
Uh, you really think marketing copy is all that meaningful?
No, but it supports our benchmarks, which are (at least for us).
Judging by sales figures available to the general public, the percentage of HPC customers for whom AMD is the best fit is fairly small.
Gee, I wonder why. Can't be because all that bad press and people bashing the "shared FPU"... We almost bought it too; we were ready to skip Bulldozer in favour of Nehalem, if it wouldn't have been for some of our people who insisted on elaborate benchmarks.
You have no idea what the fuck you're talking about here. AMD "clustered multithreading" is a different take on hyperthreading intended to provide a large speedup for integer code by providing dedicated integer execution resources to each thread. Because the FPU is shared, for FP code AMD CMT is effectively equivalent to Intel HT. Whatever benefit you see from saturating one FPU with two threads instead of one? It'll be there on Intel too.
I apologize for going too fast, mister Anonymous Expert. Allow me to rectify my mistake by explaining it a bit more elaborately. So, as I said, our code is only modestly vectorizable. It also has a significant amount of branching. Therefore, we think that on AMD, a single thread will not saturate the 256-bit FPU (which can only be accomplished through vectorization). Starting a second thread on the same module leads to the FPU being used more efficiently. This manifests itself in a nice performance boost between using half the logical cores on the AMD chip and using all of them. By comparison, doing the same with the Intel part gave us no performance boost at all which we speculate is because the FPUs (or its data pathway) is already saturated with a single thread. For Westmere, you could say that this has nothing to do with CMT and everything with 2 threads sharing a 256-bit FPU on AMD vs a 128-bit FPU on Intel, but we're seeing the same thing on Sandy Bridge, so we speculate that SMT vs CMT may have something to do with it. I'm looking forward to hearing your opinion on this matter, mister Anonymous Expert. But then again, you're clearly not one of the low level optimization wizards either, are you?
Whatever, your original post is a fallacious anti-Obama troll and you didn't answer my question as to why exactly you're so afraid of invoking Godwin's law (other than the obvious explanation that you were knowingly trolling and hence expecting emotional responses).
At least, I'm answering your question in my other post: all these people endorse Obama because that's simply the default option from an international vantage point; the Republican option is not a serious one.
This just in: http://www.bbc.co.uk/news/world-us-canada-20008687
Basically, an overwhelming majority of people in the following countries prefer Obama over Romney (in order of decreasing percentage of Obama supporters):
France
Australia
Kenya
Nigeria
Canada
Panama
UK
Brazil
Germany
Indonesia
South Korea
Spain
Mexico
Peru
India
Poland
Turkey
Japan
Malasya
China
Or basically, every country polled by the BBC (most of them being good allies of the US) except... Pakistan. My point being that from an international perspective, the Republicans simply don't have a serious agenda and everyone sees Romney as a potential troublemaker. Whether you're asking "The Chinese" or Putin or a teacher in the Germany or an engineer in France or a nurse in South Korea doesn't matter, mostly everybody prefers Obama (notable exceptions being 15% of the Pakistanis and a small half of the US). The only thing your post proves is that "mostly everybody" includes people *not* in their right mind. What fallacy is that again? Timothy McVeigh was a white American Christian so all white American Christians are terrorists?
I can't believe people keep on complaining about the bulldozer's FP performance. Does nobody realize the shared floating point unit is a 256-bit one, TWICE AS WIDE as the usual 128-bit? Also, assuming you're an HPC user, did you run actual benchmarks? Because we did, and for our (only modestly parallelizable) HPC workload, compiled with a bleeding-edge compiler (not Intel) that supports AVX and running on a bleeding-edge Linux kernel, Bulldozer was remarkably competitive with Intel's offerings at the time, with Interlagos and Westmere getting about the same amount of useful work done per clock cycle. There are some HPC benchmarks on AMDs website that seem very unlikely in light of the mainstream press. However, in light of our own benchmark results, they seem quite reasonable (although we never quite could make it look that good for AMD; probably because AMD didn't go to the same lengths to squeeze the maximum performance out of the Intel systems). Either way, AMD simply blew Intel away on a per-node-price basis, even when compared to Romley. All the way, I was the one arguing that "we should try Intels" based on reviews I saw online, but once we got all the benchmark results in, I simply couldn't argue anymore.
Also, if AMD's FP performance is truly that abysmal, please explain this? AMD bribed Dell more than Intel so that they now market Bulldozer-based Opterons as "excellent for oil and gas exploration, scientific and medical research, video rendering and other challenging HPC projects"???
Of course, this is all for a very specific workload and may not hold for all HPC workloads, but I have a strong feeling that even generally spoken, the Bulldozer's FP performance for HPC applications is just fine. It's just that most FP-intensive applications used in most of the benchmarks we're seeing in "end-user" space are not compiled to take full advantage of it and/or not running on an Operating System that takes full advantage of it and/or not very relevant test cases for the Bulldozer's parallel HPC potential. For example, one of the things we found out is that Intel's frequency scaling is more aggressive than AMD's, so Intel suffers badly if you put all the cores on a die to work at once. Also, Intel's improved HyperThreading still ain't worth shit if you saturate the FP units, while AMD's "clustered multithreading" succeeds to squeeze out a significant advantage owing to the fact that not all of our FP code is easily vectorizable so that sharing the 256-bit FP unit between 2 execution threads works better than trying to keep it busy with 1 thread's vectorized instructions.
The word he or she is looking for is probably "populist". While fascists are almost by definition also populists, and populism might arguably be seen as a first step on the slippery slope towards fascism, I do agree that our populists are not quite fascists yet by a long shot. Getting there, though.
Yep. You have to figure the author of the system wanted it to be deflationary, and he likely has at least a 100K himself if not a million. Apparently most coins are hoarded, not traded. At even $1/coin, the original guys have to have huge balls not to sell all their imaginary money for real money.
If they were to do that, the price would plummet so badly that it would hardly be worth it. I'd figure they'll release them bit by bit (sorry lame pun), using the same roundabout mechanisms as described in TFA so that they won't raise suspicion. The BitCoin economy is still a fragile thing that needs to be carefully nurtured and not abused too much, so that it can produce maximal financial gain for its creators. I suspect that big initial transaction TFA talks about was done in the same spirit; sacrifice a bit to bootstrap a profitable system (or in other words, priming the pump).
I see the biggest threat to the price of a BTC as a new, better system. The world does need an electronic currency which is easily traded, but bitcoins are simply to hard to get hold of.
Agreed on this account too, except that I feel BTC is doomed from the onset. If it really takes off, governments all over the world will scramble to somehow outlaw it and/or launch their own government-backed alternatives (like MintChip). These alternatives will have a psychological edge over BTC because there's a big government gambling its economy on it retaining a reasonable value. This is bound to happen; few people realize how strongly our economy, our system of taxes, law enforcement, option trading,... relies on money being traceable. A currency that is difficult to trace by design is a huge threat to the system (not to mention a money launderer's heaven).
most of the mainstream news media were far to the left (over 70%), and Fox News was only slightly to the left (IIRC 52%, within experimental error).
This makes you sound just like those climate change deniers who are always pointing to a handful of deeply flawed studies (often published by shills) and ignore a massive body of scientific work that goes against their agenda.
News outlets including CNN cited a study of several major media outlets by a UCLA political scientist and a University of Missouri-Columbia economist purporting to "show a strong liberal bias." But the study employed a measure of "bias" so problematic that its findings are next to useless, and the authors -- both former fellows at conservative think tanks cited in the study to illustrate liberal bias -- seem unaware of the substantial scholarly work that exists on the topic.
-- http://mediamatters.org/research/2005/12/21/former-fellows-at-conservative-think-tanks-issu/134514
Oh yeah, and even though the numbers coming out of the study were outrageous, you succeeded to push them into the realm of the ludicrous. The UCLA study (with all its flaws) still classified Fox News as "right of center"... Then again, that may just be your memory, as your "last year" turns out to be 7 years ago.
I'm not even going to start arguing against all the other BS in your post. Except this: yes, academia is somewhat left-leaning (although far less than you'd have people believe). What does that prove? Smart people see through the populist lies being perpetuated by the American right?
Simple: teenagers will be teenagers. Some of them eventually grow up. Speaking from first-hand experience here (at least concerning the first sentence).
You should be ashamed of yourself, sir! If everyone would act like you, there would be so few BTC in circulation that their value would be artific... oh wait...
Your post seems to be based on a truckload of assumptions. Yes, there are people like you describe. I don't see any indications that the OP is one of them. Did you even read the story?
Slashdot Mirror
Um... while I agree with your OP (ie. hell will freeze over ere the GOP as a whole will put itself firmly behind the results of this study committee), don't you think it sounds a bit hypocrite to first give Socrates credit for proposing the concept of the freedom to share ideas, then in the next sentence decry the "stealing of ideas from others"? I'm even honestly wondering whether your post is a parody. Anyone campaigning against overreaching copyright laws that stifle progress will be happy if a study committee (in particular a republican one) picks up some of their ideas. Its kinda the whole point of activism.
Try this: I give this study committee credit for having the balls to put truth and integrity over years of party policy. I know in an ideal world, I'm giving them credit for just doing their job, but in the actual world I live in, this is a rare enough occurrence to merit commendation.
It's not "directly derived from"; it's "loosely inspired by"
http://en.wikipedia.org/wiki/DOS#Origins
Which makes your point rather far-fetched. CP/M was an entirely different product, sharing no code base with any product marketed as "DOS". So, no, DOS was never 8-bit software. Otherwise you could also state that "windows at one point ran on the (Xerox Alto/Apple Lisa)" and "Linux at one point ran on the PDP-11/20".
Thank you for the clarification. I'm not a native English speaker so I didn't understand the slang in the parenthetical part. In fact, I read it several times wondering what it was about. Even now, with your hint and after a stop at urbandictionary.com , I'm not entirely sure how the sentence should be parsed. /. has a brutal learning curve for non-Americans; only after years of lurking from my home country, then spending a few years in the US, I got brave enough to create an account and start posting stuff. And, as you see, I still miss a beat now and then.
That said, I did think it was probably satire; that's why I started with "you're probably trolling". I replied anyway because I know for a fact that there are a lot of people believing crap like that - if not the poster, then maybe a gullible reader. It's not even all that ridiculous to someone who doesn't have any DSP background.
Oh c'mon, can you possibly get more obtuse? This is just colorful speech, and the "not that I advocate that" was specifically added to drive home that fact. If GP would have winked, you'd just have written 'see, they winked to indicate they are not serious about the "not that I advocate that"'. Either way, no-one in their right mind will see this as a "call for the death of their opponents". And yes, I am implying with this that you're not in your right mind. Whether you're on crack or it is a permanent condition is a different question.
/a stands for "per annum" - aka. per year.
You're probably trolling, but just in case you aren't. Talking about math, are you familiar with the concept "standard error"? Because the standard error on the "level" on a real-life vinyl record is far greater than the difference between two of your 65536 audio levels. Hell, you'll probably lose a level here and there every time you play back the record. And that's just one problem with vinyl. There are inherent issues with low frequencies, and many more. There's a nice long list:
http://en.wikipedia.org/wiki/Vinyl_record#Shortcomings
Anyhow, to put the original point differently, CDs have a greater dynamic range than the maximum real-life signal-to-noise ratio of vinyl records. If there's a problem with a CD's dynamic range, it's because it's mastered so that only a tiny fraction of the available dynamic range is used; see "loudness wars".
Nope, these beats are a physical phenomenon that occurs whenever two waves travel through the same medium. Its mathematical description is very simple and it's commonly taught in Physics 101 classes.
http://en.wikipedia.org/wiki/Beat_(acoustics)
I think you're confused with binaural beats, where the brain essentially simulates the interference between two tones that are not traveling through the same medium (assuming headphones are being used). Now, if you can demonstrate audible binaural beats between two ultrasonic frequencies under controlled, scientifically sound, double-blind circumstances, then you have a point. And a paper in Nature/Science.
http://en.wikipedia.org/wiki/Binaural_beats
I hate the fact that my taxes go to support this organisation that I have no control of, and can write laws that I have no influence over.
Assuming that you're an US voter, I'll would like you to know that your elected government has a vote in most if not all UN organs. Yes, you're not directly voting for the proposals that are on the table. Yes, your vote is not weighted by the population of your country. Sounds a bit like, say, electoral colleges, don't you think?
I'm glad that I don't live in Europe though - the EU seems a much more nasty (and expensive) equivalent. And they can enforce laws.
Have you tried talking to a few Europeans? I for one enjoyed living under my EU overlords. They're much more pragmatic in their decision making and there's much less over-the-top ideological drama/gridlock. Comparing it with the US federal government, I'm always amazed that this country hasn't collapsed under its own inertia yet.
Admittedly, the EU still has far less power over its member states than the federal government.
Sheesh, take a cold shower or something. Opening a debate about something doesn't equal lining up laws. It seems that in the country of freedom, some things (like putting limits on said freedom) cannot even be talked about.
And moderators, what's the thing with this being modded +5 interesting? Apparently, the US hive mind still hasn't gotten over the UN making poor Dubya go to Iraq without its stamp of legitimacy. Which is weird considering that most do admit the war itself was, in fact, not really legitimate.
Stupid troll is stupid.
- A lot of Windows users are anything but Microsoft devotees; they're just using Windows because of a perceived lack of other viable options - and not liking it. I was like that once.
- I use Linux just because it's rock-stable, responsive, secure, light on system resources, there's nothing I don't want going on behind my back (or at least much less than on some other operating systems), I can literally customize every last bit of it, I paid 0$ on software yet I'm running 0 pirated programs, and all my (open-source) software gets updated through one unified "package manager" in a fraction of the time it takes to run updates on Mac and Linux. I install closed-source software on my Linux boxen whenever I feel it's opportune - now what is this "slavish devotion" you're talking about? Also, it appears you're calling Android and Red Hat, inc market failures. Care to expound on that?
- Oh wait, now I remember the "slavish devotion" thing - that brings us to Mac. I give you OSX is not a bad operating system, though I personally could never live with the vendor lock-in. But since you're clearly not into arguing technical merits, well, it seems to me that the Mac userbase has the highest percentage of slavish devotees of all operating systems (though they're usually better than you at giving reasons to use Mac; your attitude and ignorance makes me scared to even come close to anything branded "Apple"). Some of them even realize there would never exist a Mac OSX without its open-source base. With the resources Apple (and NeXT) had at the time, there's just no way they could have coded up a modern competitive operating system without using BSD and Mach (among others).
Mostly agreed, except that they could be a bit more careful to either have no identifiable account information in the URL or not have trackers on pages that contain account information; preferably both. What would be a better term than "leaking" for this blatant disregard of basic security practices? (That's not a rhetorical question - more an attempt to start a lame semantic discussion.)
BTW, this, boys and girls, is why you should never browse without a decent script blocker. The revenue and sustainability of ad-supported sites is secondary to your privacy - or so I believe.
First, let me say... I was quite an asshole there, wasn't I? My only excuse was having my threshold for "probably an idiot fanboy" set to a hair trigger after slogging through a typical slashdot CPU flamewar (i.e. one full of fanboys)
No problem, I can relate. I tend to visit /. predominantly when I'm too tired to do anything useful (like some people plop in front of the TV) so I tend to be a bit short on patience and judgement myself. My original post is a good example of this; I got ticked off by the person starting the thread bluntly stating that (paraphrased) the shared FPU kills the Bulldozer for HPC applications. To me, this is simply false; the shared FPU was twice as wide as the (then-competing) Westmere, so all else being equal (which obviously isn't the case), the architecture of the bulldozer module with a 256-bit FPU supporting 2 "virtual cores" would be expected to give similar FP throughput than 2 Westmere cores (supporting 4 "virtual cores") with two 128-bit FPUs. In the case of poorly vectorizable code with non-optimal parallel scaling (which happens to be our case), the shared 256-bit FPU even has a distinct advantage because you need half as many parallel threads to benefit from, as you put it, branch mispredict bubbles being filled in with instructions from the other thread. In my opinion, the shared FPU by itself is fundamentally not a bad design and is totally not the reason for Bulldozer's competitive failure against Westmere; the real reasons here are many and slightly more subtle. Admittedly, that's a bit of a historical discussion since Sandy Bridge came out not so long afterwards touting one 256-bit FPU per core. But it still it infuriates me that people keep on blaming the shared FPU, which explains why my original post was quite rambling. I can easily understand how it set off your BS detector. Anyhow, I'm glad that we succeeded to turn a potential flamewar into a civilized discussion and I appreciate your insights.
I think you'll find most serious HPC users are willing to do benchmarking instead of always taking someone else's word for it. If you're going to fill a few racks full of compute nodes, it's pretty easy to justify buying one small node of each flavor to do some competitive evaluation, or perhaps rent time on someone else's systems.
There are large numbers of small to medium-sized HPC shops out there that don't even come close to the top 500 but probably contribute significantly to chipmakers' HPC sales. This is our situation; my boss was somewhat annoyed at the loss of time associated with our elaborate benchmarks. In the end, I think we did right, but it did delay our deployment by a couple of weeks. Time is money; unless there are clear red flags, bosses in this situation typically prefer speedy deployment over a small chance (remember our benchmark results were completely against expectations) of getting a financially slightly better deal. Therefore, they tend to default with what they know and trust, and that's often "Intel Inside". Again, we almost did just that. It's difficult to quantify the impact of this effect on overall sales, but I wouldn't dismiss it, especially given how overwhelmingly negative the Bulldozer's press coverage is.
Sandy Bridge has a smaller branch mispredict penalty (because its execution pipeline isn't as deep as Bulldozer's -- BD targets higher frequencies and that's the tradeoff for it), and significantly better branch predictors too. Since you say your software is branchy, I suspect Bulldozer needs to fill branch mispredict bubbles with instructions from the other thread.
The conditions for branching in our software are mostly based on simulation data that is inherently chaotic; it's hard for me to imagine a branch predictor doing much better than 50%, no matter how good it is. You're right that Bulldozer will have bubbles, but so will Sandy bridge. That (and the fact that I didn't think our code is that bra
That's pretty sad for AMD considering Westmere doesn't even implement AVX.
It's not sad at all considering that our code only gets a modest boost from AVX (as I said, a lot of it is not easily vectorizable) and that that similarly-priced AMD parts are clocked way higher.
Did you try Sandy Bridge, which does support AVX?
FYI, Romley (which I talked about earlier) was Intel's code name for Sandy Bridge-EP, mister Anonymous Expert. To give a little bit more information, on a per-clock basis, it ran our code about 35% faster than than Westmere. But on a per-price basis, AMD still won comfortably.
Uh, you really think marketing copy is all that meaningful?
No, but it supports our benchmarks, which are (at least for us).
Judging by sales figures available to the general public, the percentage of HPC customers for whom AMD is the best fit is fairly small.
Gee, I wonder why. Can't be because all that bad press and people bashing the "shared FPU"... We almost bought it too; we were ready to skip Bulldozer in favour of Nehalem, if it wouldn't have been for some of our people who insisted on elaborate benchmarks.
You have no idea what the fuck you're talking about here. AMD "clustered multithreading" is a different take on hyperthreading intended to provide a large speedup for integer code by providing dedicated integer execution resources to each thread. Because the FPU is shared, for FP code AMD CMT is effectively equivalent to Intel HT. Whatever benefit you see from saturating one FPU with two threads instead of one? It'll be there on Intel too.
I apologize for going too fast, mister Anonymous Expert. Allow me to rectify my mistake by explaining it a bit more elaborately. So, as I said, our code is only modestly vectorizable. It also has a significant amount of branching. Therefore, we think that on AMD, a single thread will not saturate the 256-bit FPU (which can only be accomplished through vectorization). Starting a second thread on the same module leads to the FPU being used more efficiently. This manifests itself in a nice performance boost between using half the logical cores on the AMD chip and using all of them. By comparison, doing the same with the Intel part gave us no performance boost at all which we speculate is because the FPUs (or its data pathway) is already saturated with a single thread. For Westmere, you could say that this has nothing to do with CMT and everything with 2 threads sharing a 256-bit FPU on AMD vs a 128-bit FPU on Intel, but we're seeing the same thing on Sandy Bridge, so we speculate that SMT vs CMT may have something to do with it. I'm looking forward to hearing your opinion on this matter, mister Anonymous Expert. But then again, you're clearly not one of the low level optimization wizards either, are you?
Whatever, your original post is a fallacious anti-Obama troll and you didn't answer my question as to why exactly you're so afraid of invoking Godwin's law (other than the obvious explanation that you were knowingly trolling and hence expecting emotional responses).
At least, I'm answering your question in my other post: all these people endorse Obama because that's simply the default option from an international vantage point; the Republican option is not a serious one.
This just in:
http://www.bbc.co.uk/news/world-us-canada-20008687
Basically, an overwhelming majority of people in the following countries prefer Obama over Romney (in order of decreasing percentage of Obama supporters):
France
Australia
Kenya
Nigeria
Canada
Panama
UK
Brazil
Germany
Indonesia
South Korea
Spain
Mexico
Peru
India
Poland
Turkey
Japan
Malasya
China
Or basically, every country polled by the BBC (most of them being good allies of the US) except... Pakistan. My point being that from an international perspective, the Republicans simply don't have a serious agenda and everyone sees Romney as a potential troublemaker. Whether you're asking "The Chinese" or Putin or a teacher in the Germany or an engineer in France or a nurse in South Korea doesn't matter, mostly everybody prefers Obama (notable exceptions being 15% of the Pakistanis and a small half of the US). The only thing your post proves is that "mostly everybody" includes people *not* in their right mind. What fallacy is that again? Timothy McVeigh was a white American Christian so all white American Christians are terrorists?
I can't believe people keep on complaining about the bulldozer's FP performance. Does nobody realize the shared floating point unit is a 256-bit one, TWICE AS WIDE as the usual 128-bit? Also, assuming you're an HPC user, did you run actual benchmarks? Because we did, and for our (only modestly parallelizable) HPC workload, compiled with a bleeding-edge compiler (not Intel) that supports AVX and running on a bleeding-edge Linux kernel, Bulldozer was remarkably competitive with Intel's offerings at the time, with Interlagos and Westmere getting about the same amount of useful work done per clock cycle. There are some HPC benchmarks on AMDs website that seem very unlikely in light of the mainstream press. However, in light of our own benchmark results, they seem quite reasonable (although we never quite could make it look that good for AMD; probably because AMD didn't go to the same lengths to squeeze the maximum performance out of the Intel systems). Either way, AMD simply blew Intel away on a per-node-price basis, even when compared to Romley. All the way, I was the one arguing that "we should try Intels" based on reviews I saw online, but once we got all the benchmark results in, I simply couldn't argue anymore.
Also, if AMD's FP performance is truly that abysmal, please explain this? AMD bribed Dell more than Intel so that they now market Bulldozer-based Opterons as "excellent for oil and gas exploration, scientific and medical research, video rendering and other challenging HPC projects"???
Of course, this is all for a very specific workload and may not hold for all HPC workloads, but I have a strong feeling that even generally spoken, the Bulldozer's FP performance for HPC applications is just fine. It's just that most FP-intensive applications used in most of the benchmarks we're seeing in "end-user" space are not compiled to take full advantage of it and/or not running on an Operating System that takes full advantage of it and/or not very relevant test cases for the Bulldozer's parallel HPC potential. For example, one of the things we found out is that Intel's frequency scaling is more aggressive than AMD's, so Intel suffers badly if you put all the cores on a die to work at once. Also, Intel's improved HyperThreading still ain't worth shit if you saturate the FP units, while AMD's "clustered multithreading" succeeds to squeeze out a significant advantage owing to the fact that not all of our FP code is easily vectorizable so that sharing the 256-bit FP unit between 2 execution threads works better than trying to keep it busy with 1 thread's vectorized instructions.
/rambling rant
The word he or she is looking for is probably "populist". While fascists are almost by definition also populists, and populism might arguably be seen as a first step on the slippery slope towards fascism, I do agree that our populists are not quite fascists yet by a long shot. Getting there, though.
Just the facts please (ps remember Godwin's law please)
Interrresting... for what reason would you, a Romney supporter, be afraid that the replies to your post^H^H^H^Htroll would invoke Godwin's law?
Even better, with a buzzer that goes off whenever 4 out of 6 rate a statement at false. Not that's entertainment!
It's probably a lot healthier to stick with proper OO languages.
That statement makes you not crazy enough to be a code monkey.
Yep. You have to figure the author of the system wanted it to be deflationary, and he likely has at least a 100K himself if not a million. Apparently most coins are hoarded, not traded. At even $1/coin, the original guys have to have huge balls not to sell all their imaginary money for real money.
If they were to do that, the price would plummet so badly that it would hardly be worth it. I'd figure they'll release them bit by bit (sorry lame pun), using the same roundabout mechanisms as described in TFA so that they won't raise suspicion. The BitCoin economy is still a fragile thing that needs to be carefully nurtured and not abused too much, so that it can produce maximal financial gain for its creators. I suspect that big initial transaction TFA talks about was done in the same spirit; sacrifice a bit to bootstrap a profitable system (or in other words, priming the pump).
I see the biggest threat to the price of a BTC as a new, better system. The world does need an electronic currency which is easily traded, but bitcoins are simply to hard to get hold of.
Agreed on this account too, except that I feel BTC is doomed from the onset. If it really takes off, governments all over the world will scramble to somehow outlaw it and/or launch their own government-backed alternatives (like MintChip). These alternatives will have a psychological edge over BTC because there's a big government gambling its economy on it retaining a reasonable value. This is bound to happen; few people realize how strongly our economy, our system of taxes, law enforcement, option trading,... relies on money being traceable. A currency that is difficult to trace by design is a huge threat to the system (not to mention a money launderer's heaven).
most of the mainstream news media were far to the left (over 70%), and Fox News was only slightly to the left (IIRC 52%, within experimental error).
This makes you sound just like those climate change deniers who are always pointing to a handful of deeply flawed studies (often published by shills) and ignore a massive body of scientific work that goes against their agenda.
News outlets including CNN cited a study of several major media outlets by a UCLA political scientist and a University of Missouri-Columbia economist purporting to "show a strong liberal bias." But the study employed a measure of "bias" so problematic that its findings are next to useless, and the authors -- both former fellows at conservative think tanks cited in the study to illustrate liberal bias -- seem unaware of the substantial scholarly work that exists on the topic.
-- http://mediamatters.org/research/2005/12/21/former-fellows-at-conservative-think-tanks-issu/134514
Oh yeah, and even though the numbers coming out of the study were outrageous, you succeeded to push them into the realm of the ludicrous. The UCLA study (with all its flaws) still classified Fox News as "right of center"... Then again, that may just be your memory, as your "last year" turns out to be 7 years ago.
I'm not even going to start arguing against all the other BS in your post. Except this: yes, academia is somewhat left-leaning (although far less than you'd have people believe). What does that prove? Smart people see through the populist lies being perpetuated by the American right?
Simple: teenagers will be teenagers. Some of them eventually grow up. Speaking from first-hand experience here (at least concerning the first sentence).
You should be ashamed of yourself, sir! If everyone would act like you, there would be so few BTC in circulation that their value would be artific... oh wait...
Your post seems to be based on a truckload of assumptions. Yes, there are people like you describe. I don't see any indications that the OP is one of them. Did you even read the story?