..the predecessor to DirectCompute was a little.NET library that came out of Microsoft Research called Accelerator which was initially available to the public in 2006.
..thats several years before CUDA (2008) and OpenCL (also 2008)
Microsoft has actually been the innovator on this one.
so the guy that brought AMD to a position where they're successfully launching 3 products in one year (which they've never done before) is not someone you want to keep around? Are you kidding?
Are you honestly asking this question? If you are going to pretend to know anything about the business world, then you should at least pretend to also know that some CEO's are specialists at bringing companies out of financial trouble and even bankruptcy.
For example (from my industry) there is Scott Butera, a CEO that has brought more than one casino out of financial trouble, who has just been picked up by the Foxwoods Resort Casino in Connecticut because of its very serious financial troubles (billions in debt, defaulting on loans..)
Often what these specialists bring to the table is their trusted contacts in the financial industries. The primary goal is often to maintain a credit line while the problems are resolved (because no large business can run without credit, regardless of how much cash they have.)
No amount of bribes in the world can account for the fact that Intel's latest processors have been significantly outperforming AMD's for the last few years now.
Yet AMDs significantly outperformed Intels for quite a few years.. but only managed 50% market share at its height because of Intels illegal (no "questionable" about it) practices.
They're not cheaper because they want to be, they're cheaper because they have to be. They can't compete on performance
Intel is more expensive because they have to be. They can't compete on value.
See what I did there?
The only metric worth anything is performance per dollar. You have not used that metric but tried to draw a conclusion as if you did.
Before you reply in fanboy rage, lets try it with cars:
Ford is not cheaper (than Ferari) because they want to be, they're cheaper because they have to be. They can't compete on performance, so they try to do it on price.
With cars your bullshit logic has no teeth. Now why the fuck would it apply to processors but not cars?
Worse is for those who DO want a fast CPU but also want expandability. At least those who want a low end CPU and an expandable platform can go AMD. Those who want a fast processor don't really have that option.
This isnt entirely true. AMD has options that arent considered "consumer grade" but are as cheap as Intels $1000 high end "consumer grade" and I'll speak more of this is a moment.
The blunt fact is that the high end sandy bridge processors beat every previous quad core CPU from BOTH AMD and INTEL by quite a significant margin. They don't quite keep up with the 980x in highly multithreaded benchmarks
If you want to build a high end system for multi-threaded performance without breaking the bank with $1000 parts, a pair of Opteron 6128's score only slightly worse than a single i7 980X.
Thats a true 16-core server system you will be building, only benchmarks slightly worse than Intels flagship i7 980X, and its notably cheaper (2 x 6128 + 1 x motherboard for $952, less than that i7-980X chip for $1000 that still needs a motherboard.)
..and the 8086 wasnt designed to be compatible with the 8080 either.. well, sort of..
It was designed to be source-code compatible (assembler source code) so that an assembler for the 8086/8088 when given 8080 code, could still produce 1 opcode per 8080 instruction.
RISC was better 20 years ago because CISC chips were using close to 50% of the die area for complex decoders. RISC chips could use 5%, giving them vastly more space to cram on ALUs and so on. Then the transistor budget increased, but the decoder complexity stayed pretty constant. 50% became 25%, then 10%, and now the increased space on RISC chips is pretty much irrelevant and the space-saving in instruction cache offsets it.
Just to be clear, with Intel's first version of HyperThreading (P4 Northwood in 2002), Intel reported that the extra decoder and other HT needs cost 5% of chip space.
The P4 had ~42 Million Transistors (MT) so about 2.1 MT for the decoder.
The i7-980 Gulftown has ~1170 MT, so its down to about 0.2%
I bet that in the future we will see chips with simpler (read RISC) architectures with more on-chip memory and special compilers designed to optimize tasks to minimize random memory access.
I bet that in the future, you still wont know how much chip space is used by its various components, leading you to continue believing that risc is some sort of space-saving advantage.
With any instruction set, execution can only be as fast as instruction decoding. Both RISC and CISC machines now have similar execution units, so CISC architectures can feed more execution units per instruction decoded..
To get the same sort of raw performance on RISC, the decoder needs to be faster than on CISC. When the decoders are already operating at the clock rate, the only way that this can be accomplished is with more decoders.
Sure, CISC decoders are larger.. but you need less of them to get the same benefits.. so its just about moot. But back to the original point, in neither case are these decoders anywhere near the size of the vast space devoted to cache memory, or execution units.
We are living in a time when two of the Big Three major chip makers differ significantly over how to organize the many extra decoders they trivially pack on.
Intel is feeding two banks of decoders into one scheduler and one backing set of execution units while AMD is about to release its new strategy of feeding two banks of decoders into eight schedulers with two banks of integer execution units and one shared floating point unit.
...and they do this with the majority of the chip being devoted to cache memory and their controllers.
When Intel first introduced HyperThreading in the Pentium 4, which is essentially accomplished by tacking on an extra set of CISC decoders to a core, it only took 5% more space than not doing it. A few doublings of the transistor count later, the space those CISC decoders now use is even more meaningless.
AMD is now introducing its own form of "pack on extra decoders" with its Bulldozer architecture.
If ARM (a true RISC design) wants to compete on performance, they too would have to pack on extra decoders to feed banks of execution units. ARM is going the other route, tho.. and is focusing on energy efficiency and as such, even cache memory is light on that architecture.
Just to be clear: Of the Big Three cpu manufacturers (ARM, Intel, and AMD,) the one manufacturer focusing on RISC also offers the *least* amount of cache memory.. the exact opposite of what you think will happen.
GDI was most certainly hardware accelerated in XP, if the driver so chose to do it (nVidia and ATI certainly did.) Since just about everything XP itself renders is GDI (that includes explorer) then thats accelerated too.
Only a few GDI functions were done in software with a modern GPU, and Alpha blending was one of them but that was because behind the scenes that function was actually GDI+ (it was introduced into GDI at the same time as GDI+ was released)
GDI+ was not hardware accelerated *at all* on XP (no driver hooks!), which is why many applications these days render so damn slowly on XP.
In windows 7 (and maybe Vista?), as long as your GPU has DX9 drivers that do full hardware acceleration, GDI and GDI+ are completely hardware accelerated (manufacturers dont have to do GDI-specific drivers now.. just DirectX drivers)
When a scientist has some hypothesis he wants to test and goes and looks into the data.. thats great. The problem is that there are far fewer reasonable hypothesis than there are ways to examine the data...
..so once all the reasonable hypothesis have been tested, whats left in the data is only the large number of false-positive "significance" you can find if you look hard enough! For data that can only be generated slowly (for whatever reason) the only thing that you CAN do with it is "look harder."
So there they are.. with government grants.. looking harder at the same data lots of others have already looked at years and years ago.
Then comes the "adjustments" to the data, adjustments that may only have been "justified" by someone elses false positive, creating a (perhaps slightly) different set of data to start dredging.
There is a lot of science where new data is not generated at a rate where true reproducibility is an option.
For example, anything to do with the general health of a person can only really be measured over long time scales (decades), as well as measurements of the climate and things like that.
In those cases, 'reproduction' means taking the same data, sifting it in possibly the same way (but maybe not), and getting the same or similar result.
Now take this fact in the context of data dredging.
Data dredging does not have to be intentional (ie: an intent to defraud, although it certainly can be.)
If you take 1000 scientists and give them all the same data, they will probably look at that data in several thousand ways. If you are dealing with 95% intervals, and the data is looked at in 2000 ways, then about 100 of those ways will present something 'significant' by simple random chance.
The same phenomenon exists in that whole bullshit "Equidistant Letter Spacing" Bible-Code crap, but is much easier to dismiss because you have to believe something extremely unlikely (God exists, and orchestrated the translation of the bible into English so that it would have hidden codes.)
When you really get into dismissing Bible Code in a mathematical manner, you end up realizing that in any data set there exists many things of statistically significance and yet also completely bullshit.
But on the other hand, if the impersonation is done with intent to harm, intimidate, threaten, or defraud, why can't we just prosecute people for fraud, criminal intimidation, or whatnot?
Because we rarely prosecute for that, and thus there is an epidemic, and epidemics need legislation (..or enforcement... but legislators arent enforcers, so they legislate)
Define that key word of yours, 'credibly', and then struggle with the ramification of that definition in the context of this law.
For instance, even if you are very very bad at your comic impersonations, someone may still believe that you are the person you are impersonating for any number of reasons.
Isn't their belief, possibly based on the fact that they have never even heard of the person you are so poorly impersonating prior to the event, evidence that your terribly bad impersonation was still credible none-the-less?
Sure, the law apparently also talks about Intent.. but combining Credibility with Intent doesnt actually seem to add anything of value for the above reason.
His tool only found a few bugs ("several") in Internet Explorer, found about two dozen in Webkit ("some" problems still unfixed), about 60 bugs in Mozilla ("several" still unfixed), and that for Opera some of the bugs arent fixed ("several".)
So what we see here is that of the browsers, Internet Explorer didnt have nearly as many problems identifiable by his tool as the others to begin with, and that it still doesnt have more than the other browsers now even after all parties had 6 months.
Could it be that all of the remaining bugs for all of the browsers require good reproducibility to address reasonably? Could it be that the person you replied to is correct, rather than that your "but not mozilla, webkit team and opera?" bullshit is just that, bullshit?
You will note that the author never states that the 6 month old tool reliably reproduces the bugs in question.
That would be something that, if true, he would have stated. This is so because the complaint he is facing is that only the newest tool reliably reproduces them, that further that this has been an ongoing complaint about his tool even by other parties besides Microsoft.
Ergo, its probably false. The tool did not reliably reproduce the bugs in question 6 months ago.
Pudding laced with brandy most likely won't up your BAC or intoxicate you
Were you under the impression that breathalyzers measure Blood Alcohol Content?
Yes. It looks like you were. Nobody needs to read what you said any farther than that, because conclusions drawn from faulty assumptions are meaningless.
After that, I frankly can't blame 'em for not letting the Republicans offer amendments
The health care bill was, in fact, one of the bills rammed through without a chance for public scrutiny. The call for cloture was friday, the vote for cloture was monday. The public had exactly 2 full days to read TWO THOUSAND FOUR HUNDRED AND NINE pages.
You are claiming that the health care bill was the reason that the Democrats later rammed things through.. YET WE ACTUALLY FIND THAT THIS VERY BILL WAS RAMMED THROUGH.
Your eyes are closed and you are sleeping.
The Democrats have been ramming EVERYTHING through since the beginning of 2009. EVERYTHING.
Now, you were saying about the Republics being 'at fault' for this ramming? From day one eh? Right from the start eh? What was that first thing then, eh?
Slashdot Mirror
Just to add to this..
..the predecessor to DirectCompute was a little .NET library that came out of Microsoft Research called Accelerator which was initially available to the public in 2006.
..thats several years before CUDA (2008) and OpenCL (also 2008)
Microsoft has actually been the innovator on this one.
so the guy that brought AMD to a position where they're successfully launching 3 products in one year (which they've never done before) is not someone you want to keep around? Are you kidding?
Are you honestly asking this question? If you are going to pretend to know anything about the business world, then you should at least pretend to also know that some CEO's are specialists at bringing companies out of financial trouble and even bankruptcy.
For example (from my industry) there is Scott Butera, a CEO that has brought more than one casino out of financial trouble, who has just been picked up by the Foxwoods Resort Casino in Connecticut because of its very serious financial troubles (billions in debt, defaulting on loans..)
Often what these specialists bring to the table is their trusted contacts in the financial industries. The primary goal is often to maintain a credit line while the problems are resolved (because no large business can run without credit, regardless of how much cash they have.)
No amount of bribes in the world can account for the fact that Intel's latest processors have been significantly outperforming AMD's for the last few years now.
Yet AMDs significantly outperformed Intels for quite a few years.. but only managed 50% market share at its height because of Intels illegal (no "questionable" about it) practices.
They're not cheaper because they want to be, they're cheaper because they have to be. They can't compete on performance
Intel is more expensive because they have to be. They can't compete on value.
See what I did there?
The only metric worth anything is performance per dollar. You have not used that metric but tried to draw a conclusion as if you did.
Before you reply in fanboy rage, lets try it with cars:
Ford is not cheaper (than Ferari) because they want to be, they're cheaper because they have to be. They can't compete on performance, so they try to do it on price.
With cars your bullshit logic has no teeth. Now why the fuck would it apply to processors but not cars?
Intel has a very large range of processor models at a very large range of price points. Yet they don't top the charts with whats on the market today.
Intel charges more because it has brand recognition which was propped up with the illegal activities of this convicted monopolist.
Walmart?
They told me that if I voted for McCain, that shit like this would happen.
They were right.
You can play that protected video, but your CPU can be disabled remotely.
..and thats over 3G, and even works if the system isnt even turned on.
Worse is for those who DO want a fast CPU but also want expandability. At least those who want a low end CPU and an expandable platform can go AMD. Those who want a fast processor don't really have that option.
This isnt entirely true. AMD has options that arent considered "consumer grade" but are as cheap as Intels $1000 high end "consumer grade" and I'll speak more of this is a moment.
The blunt fact is that the high end sandy bridge processors beat every previous quad core CPU from BOTH AMD and INTEL by quite a significant margin. They don't quite keep up with the 980x in highly multithreaded benchmarks
If you want to build a high end system for multi-threaded performance without breaking the bank with $1000 parts, a pair of Opteron 6128's score only slightly worse than a single i7 980X.
Thats a true 16-core server system you will be building, only benchmarks slightly worse than Intels flagship i7 980X, and its notably cheaper (2 x 6128 + 1 x motherboard for $952, less than that i7-980X chip for $1000 that still needs a motherboard.)
..and the 8086 wasnt designed to be compatible with the 8080 either.. well, sort of..
It was designed to be source-code compatible (assembler source code) so that an assembler for the 8086/8088 when given 8080 code, could still produce 1 opcode per 8080 instruction.
Exactly.
In the case of the civil war, Lincoln famously said that he would gladly give up his opposition to slavery if it would end the war.
But the war would not end that way, because it was about a lot more than slavery.
RISC was better 20 years ago because CISC chips were using close to 50% of the die area for complex decoders. RISC chips could use 5%, giving them vastly more space to cram on ALUs and so on. Then the transistor budget increased, but the decoder complexity stayed pretty constant. 50% became 25%, then 10%, and now the increased space on RISC chips is pretty much irrelevant and the space-saving in instruction cache offsets it.
Just to be clear, with Intel's first version of HyperThreading (P4 Northwood in 2002), Intel reported that the extra decoder and other HT needs cost 5% of chip space.
The P4 had ~42 Million Transistors (MT) so about 2.1 MT for the decoder.
The i7-980 Gulftown has ~1170 MT, so its down to about 0.2%
I bet that in the future we will see chips with simpler (read RISC) architectures with more on-chip memory and special compilers designed to optimize tasks to minimize random memory access.
I bet that in the future, you still wont know how much chip space is used by its various components, leading you to continue believing that risc is some sort of space-saving advantage.
...and they do this with the majority of the chip being devoted to cache memory and their controllers.
With any instruction set, execution can only be as fast as instruction decoding. Both RISC and CISC machines now have similar execution units, so CISC architectures can feed more execution units per instruction decoded..
To get the same sort of raw performance on RISC, the decoder needs to be faster than on CISC. When the decoders are already operating at the clock rate, the only way that this can be accomplished is with more decoders.
Sure, CISC decoders are larger.. but you need less of them to get the same benefits.. so its just about moot. But back to the original point, in neither case are these decoders anywhere near the size of the vast space devoted to cache memory, or execution units.
We are living in a time when two of the Big Three major chip makers differ significantly over how to organize the many extra decoders they trivially pack on.
Intel is feeding two banks of decoders into one scheduler and one backing set of execution units while AMD is about to release its new strategy of feeding two banks of decoders into eight schedulers with two banks of integer execution units and one shared floating point unit.
When Intel first introduced HyperThreading in the Pentium 4, which is essentially accomplished by tacking on an extra set of CISC decoders to a core, it only took 5% more space than not doing it. A few doublings of the transistor count later, the space those CISC decoders now use is even more meaningless.
AMD is now introducing its own form of "pack on extra decoders" with its Bulldozer architecture.
If ARM (a true RISC design) wants to compete on performance, they too would have to pack on extra decoders to feed banks of execution units. ARM is going the other route, tho.. and is focusing on energy efficiency and as such, even cache memory is light on that architecture.
Just to be clear: Of the Big Three cpu manufacturers (ARM, Intel, and AMD,) the one manufacturer focusing on RISC also offers the *least* amount of cache memory.. the exact opposite of what you think will happen.
And its very solid.
The latest Steam Survey has 64-bit Windows 7 as the #1 installed OS at 35.09%, with 32-bit XP second at 24.64%
All told, 64-bit Windows OS's on the survey are now totaling 46.41%, with another ~4.5% going to 64-bit OS/X's.
So nearly half of gamers are now using 64-bit OS's, this is in spite of the fact that very very few games actually come with 64-bit binaries.
GDI was most certainly hardware accelerated in XP, if the driver so chose to do it (nVidia and ATI certainly did.) Since just about everything XP itself renders is GDI (that includes explorer) then thats accelerated too.
Only a few GDI functions were done in software with a modern GPU, and Alpha blending was one of them but that was because behind the scenes that function was actually GDI+ (it was introduced into GDI at the same time as GDI+ was released)
GDI+ was not hardware accelerated *at all* on XP (no driver hooks!), which is why many applications these days render so damn slowly on XP.
In windows 7 (and maybe Vista?), as long as your GPU has DX9 drivers that do full hardware acceleration, GDI and GDI+ are completely hardware accelerated (manufacturers dont have to do GDI-specific drivers now.. just DirectX drivers)
I just put those shortcuts to less frequently used programs into a folder on the desktop.
Its in a category all its own, because the author has written the best program on earth.
All kidding aside, the categories in ubuntu drive me nuts too. Its a rare program that would fit into a single category.
Take something like R.. its a statistics/math package.. its a graphing package.. its a programming language.. etc..
Something further to wet your noodle.
..so once all the reasonable hypothesis have been tested, whats left in the data is only the large number of false-positive "significance" you can find if you look hard enough! For data that can only be generated slowly (for whatever reason) the only thing that you CAN do with it is "look harder."
When a scientist has some hypothesis he wants to test and goes and looks into the data.. thats great. The problem is that there are far fewer reasonable hypothesis than there are ways to examine the data...
So there they are.. with government grants.. looking harder at the same data lots of others have already looked at years and years ago.
Then comes the "adjustments" to the data, adjustments that may only have been "justified" by someone elses false positive, creating a (perhaps slightly) different set of data to start dredging.
There is a lot of science where new data is not generated at a rate where true reproducibility is an option.
For example, anything to do with the general health of a person can only really be measured over long time scales (decades), as well as measurements of the climate and things like that.
In those cases, 'reproduction' means taking the same data, sifting it in possibly the same way (but maybe not), and getting the same or similar result.
Now take this fact in the context of data dredging.
Data dredging does not have to be intentional (ie: an intent to defraud, although it certainly can be.)
If you take 1000 scientists and give them all the same data, they will probably look at that data in several thousand ways. If you are dealing with 95% intervals, and the data is looked at in 2000 ways, then about 100 of those ways will present something 'significant' by simple random chance.
The same phenomenon exists in that whole bullshit "Equidistant Letter Spacing" Bible-Code crap, but is much easier to dismiss because you have to believe something extremely unlikely (God exists, and orchestrated the translation of the bible into English so that it would have hidden codes.)
When you really get into dismissing Bible Code in a mathematical manner, you end up realizing that in any data set there exists many things of statistically significance and yet also completely bullshit.
But on the other hand, if the impersonation is done with intent to harm, intimidate, threaten, or defraud, why can't we just prosecute people for fraud, criminal intimidation, or whatnot?
Because we rarely prosecute for that, and thus there is an epidemic, and epidemics need legislation (..or enforcement... but legislators arent enforcers, so they legislate)
Define that key word of yours, 'credibly', and then struggle with the ramification of that definition in the context of this law.
For instance, even if you are very very bad at your comic impersonations, someone may still believe that you are the person you are impersonating for any number of reasons.
Isn't their belief, possibly based on the fact that they have never even heard of the person you are so poorly impersonating prior to the event, evidence that your terribly bad impersonation was still credible none-the-less?
Sure, the law apparently also talks about Intent.. but combining Credibility with Intent doesnt actually seem to add anything of value for the above reason.
Just to be fucking honest...
His tool only found a few bugs ("several") in Internet Explorer, found about two dozen in Webkit ("some" problems still unfixed), about 60 bugs in Mozilla ("several" still unfixed), and that for Opera some of the bugs arent fixed ("several".)
So what we see here is that of the browsers, Internet Explorer didnt have nearly as many problems identifiable by his tool as the others to begin with, and that it still doesnt have more than the other browsers now even after all parties had 6 months.
Could it be that all of the remaining bugs for all of the browsers require good reproducibility to address reasonably? Could it be that the person you replied to is correct, rather than that your "but not mozilla, webkit team and opera?" bullshit is just that, bullshit?
You will note that the author never states that the 6 month old tool reliably reproduces the bugs in question.
That would be something that, if true, he would have stated. This is so because the complaint he is facing is that only the newest tool reliably reproduces them, that further that this has been an ongoing complaint about his tool even by other parties besides Microsoft.
Ergo, its probably false. The tool did not reliably reproduce the bugs in question 6 months ago.
Pudding laced with brandy most likely won't up your BAC or intoxicate you
Were you under the impression that breathalyzers measure Blood Alcohol Content?
Yes. It looks like you were. Nobody needs to read what you said any farther than that, because conclusions drawn from faulty assumptions are meaningless.
After that, I frankly can't blame 'em for not letting the Republicans offer amendments
The health care bill was, in fact, one of the bills rammed through without a chance for public scrutiny. The call for cloture was friday, the vote for cloture was monday. The public had exactly 2 full days to read TWO THOUSAND FOUR HUNDRED AND NINE pages.
You are claiming that the health care bill was the reason that the Democrats later rammed things through.. YET WE ACTUALLY FIND THAT THIS VERY BILL WAS RAMMED THROUGH.
Your eyes are closed and you are sleeping.
The Democrats have been ramming EVERYTHING through since the beginning of 2009. EVERYTHING.
Now, you were saying about the Republics being 'at fault' for this ramming? From day one eh? Right from the start eh? What was that first thing then, eh?
He probably means all the people swayed by formula science news crap
..which is pretty much everyone you will ever encounter who would like to discuss this subject with you, be they for or against.