为什么摩尔定律需要并行计算?

时间:2022-05-19 13:51:53

This was a question in one of my CS textbooks. I am at a loss. I don't see why it necessarily would lead to parallel computing. Anyone wanna point me in the right direction?

这是我的一本CS教科书中的一个问题。我很茫然。我不明白为什么它必然会导致并行计算。有人想指出我正确的方向吗?

16 个解决方案

#1


Moore's law just says that the number of transistors on a reasonably priced integrated circuit tends to double every 2 years.

摩尔定律刚刚说,价格合理的集成电路上的晶体管数量每两年就会增加一倍。

Observations about speed or transistor density or die size are all somewhat orthogonal to the original observation.

关于速度或晶体管密度或管芯尺寸的观察结果与原始观察结果有些正交。

Here's why I think Moore's law leads inevitably to parallel computing:

这就是为什么我认为摩尔定律不可避免地导致并行计算:

If you keep doubling the number of transistors, what are you going to do with them all?

如果你把晶体管的数量增加一倍,那你打算怎么做呢?

  • More instructions!
  • Wider data types!
  • 更广泛的数据类型!

  • Floating Point Math!
  • 浮点数学!

  • More caches (L1, L2, L3)!
  • 更多缓存(L1,L2,L3)!

  • Micro Ops!
  • More pipeline stages!
  • 更多管道阶段!

  • Branch prediction!
  • Speculative execution!
  • Data Pre-Fetch!
  • Single Instruction Multiple Data!
  • 单指令多数据!

Eventually, when you've implemented all the tricks you can think of to use all those extra transistors, you eventually think to yourself: why don't we just do all those cool tricks TWICE on the came chip?

最后,当你实现了所有可以想到使用所有这些额外晶体管的技巧时,你最终会想到:为什么我们不在芯片上做那些很酷的技巧TWICE?

Bada bing. Bada boom. Multicore is inevitable.

巴达冰。巴达热潮。多核是不可避免的。


Incidentally, I think the current trend of CPUs with multiple identical CPU cores will eventually subside as well, and the real processors of the future will have a single master core, a collection of general purpose cores, and a collection of special purpose coprocessors (like a graphics card, but on-die with the CPU and caches).

顺便提一下,我认为目前具有多个相同CPU核心的CPU的趋势也将最终平息,未来真正的处理器将拥有一个主核心,一系列通用核心和一系列专用协处理器(如一个显卡,但在CPU和缓存的模具上)。

The IBM Cell processor (in the PS3) is already somewhat like this. It has one master core and seven "synergistic processing units".

IBM Cell处理器(在PS3中)已经有点像这样了。它有一个主核心和七个“协同处理单元”。

#2


One word - Heat.

一个字 - 热。

Due to an inability to dissipate heat at current transistor levels, engineers are using their every growing transistor budgets to create more cores instead of creating more complex (and hot) pipelines and faster processors.

由于无法在当前晶体管级散热,工程师正在利用其每个不断增长的晶体管预算来创建更多内核,而不是创建更复杂(和热)的流水线和更快的处理器。

Moore's law is not at all dead - moore's law is about transistor density at a given cost. It just so happens that for various reasons (like marketing) engineers decided to use their transistor budget to increase clock cycle. Now they decided (because of the heat issue) to start using the transistors for parallelism, plus 64bit computing and reducing power consumption.

摩尔定律根本没有死 - 摩尔定律是关于给定成本的晶体管密度。事实上,由于各种原因(如市场营销),工程师决定使用他们的晶体管预算来增加时钟周期。现在他们决定(因为热量问题)开始使用晶体管进行并行,加上64位计算并降低功耗。

#3


Moore's law describes the trend that performance of chips effectively doubles due to the addition of more transistors to a circuit board.

摩尔定律描述了由于在电路板上增加了更多晶体管,芯片性能有效翻倍的趋势。

Since devices are not increasing in size (if anything the reverse is true) then clearly the space for these additional transistors only becomes available due to chip technology becoming smaller and manufacturing becoming ever better.

由于器件尺寸没有增加(如果反之亦然),那么由于芯片技术变得越来越小并且制造变得越来越好,这些额外晶体管的空间显然也变得可用。

At somepoint however you reach the stage where transistors cannot be minimized any further. It also becomes impossible to increase the size of chips beyond a certain point due to the amount of heat generated and the manufacturing costs involved.

然而,在某些时候,您将达到晶体管无法进一步最小化的阶段。由于产生的热量和所涉及的制造成本,也不可能将芯片的尺寸增加到超过某一点。

These limits necessitate a means of increasing performance beyond simply producing more complex chips.

这些限制需要一种提高性能的方法,而不仅仅是生产更复杂的芯片。

One such method is to employ cheaper and less complex chips in parallel architectures, another is to move away from the traditional integrated chip to something like quantum computing - which by it's very definition is parallel processing.

一种这样的方法是在并行架构中采用更便宜和更简单的芯片,另一种方法是从传统的集成芯片转向类似量子计算的东西 - 其定义是并行处理。

It's worth noting that the title of this question relates more to the observed results of the law (performance increase) rather than the actual law itself which was largely an observation about transistor count.

值得注意的是,这个问题的标题更多地与观察到的法律结果(性能提高)有关,而不是实际的法律本身,这主要是对晶体管数量的观察。

#4


I think it is a reference to the free lunch is over article

我认为这是免费午餐的参考文章

basically, the original version of Moore's law, about transistor density, still holds. But one important derived law, about processing speed doubling every xx months, has hit a wall.

基本上,关于晶体管密度的摩尔定律的原始版本仍然存在。但是,关于处理速度每xx个月加倍的一个重要的衍生法则已经触底了。

So we are facing a future where processor speeds will go up only slightly but we will have more core's and cache to play with.

因此,我们面临的未来处理器速度将略有上升,但我们将拥有更多核心和缓存。

#5


That is an odd question. Moore's law doesn't necessitate anything it is just an observation of the progression of computing power, it doesn't dictate that it must increase at a certain rate.

这是一个奇怪的问题。摩尔定律不需要任何东西,它只是观察计算能力的进展,它并没有规定它必须以一定的速度增长。

#6


Increasing the speed of processors would make the operating temperature so extremely high it would burn a hole in your desk. The makers of the chips are running up against certain limitations they can't get around... like the speed of light, for instance. Parallel computing will allow them to speed up the computers without starting a fire.

提高处理器的速度会使操作温度如此之高,以至于会在桌面上烧掉一个洞。这些芯片的制造商正在遇到一些他们无法解决的限制......比如光速。并行计算将使他们能够加速计算机而不会引发火灾。

#7


Transistors and cpus and whatnot are getting smaller and smaller and faster and faster. Alas, the heat and voltage costs for computing are going up. The heat and voltage issues are as much of a concern as the actual physical size minimums. A 100ghz chip would suck up too much voltage and get too hot but 100 1ghz chips would have less of an issue with this.

晶体管和cpu等等越来越小,越来越快。唉,计算的热量和电压成本正在上升。热量和电压问题与实际物理尺寸最小值一样令人担忧。一个100ghz的芯片会吸收过多的电压而变得太热,但100个1ghz的芯片会产生较少的问题。

#8


Interestingly, the idea proposed in the question that parallel computing is "necessitated" is thrown into question by Amdahl's Law, which basically says that having parallel processors will only get you so far unless 100% of your program is parallelizable (which is never the case in the real world).

有趣的是,在问题中提出并行计算是“必要的”的想法受到Amdahl定律的质疑,基本上说,并行处理器只会让你到目前为止,除非100%的程序是可并行化的(从来都不是这样)在现实世界)。

For example, if you have a program which takes 20 minutes on one processor and is 50% parallelizable, and you buy a large number of processors to speed things up, your minimum time to run would still be over 10 minutes. This is ignoring the cost and other issues involved.

例如,如果您的程序在一个处理器上占用20分钟且可并行化50%,并且您购买大量处理器以加快速度,那么您的最短运行时间仍将超过10分钟。这忽略了所涉及的成本和其他问题。

#9


The real answer is completely un-technical, not that the hardware explanations aren't fantastic. It's that Moore's Law has become less and less of an observation, and more of an expectation. This expectation of computers growing exponentially has become the driving force of the industry, which necessitates all the parallelism.

真正的答案完全不是技术性的,并不是说硬件解释不是很棒。这就是摩尔定律越来越少的观察,更多的是期望。这种对计算机成倍增长的预期已成为该行业的驱动力,这需要所有的并行性。

#10


Moore's law says that the number of transistors in an IC relative to cost increases exponentially year on year.

摩尔定律表明,IC中晶体管的数量相对于成本的数量同比呈指数增长。

Historically, this was partly due to a decrease in transistor size, and smaller transistors also switched faster. Because you got faster transistors in step with Moore's law, clock speed increased. So there's a confusion that say Moore's law means faster processors rather than just wider.

从历史上看,这部分是由于晶体管尺寸减小,而较小的晶体管也更快地切换。因为根据摩尔定律得到更快的晶体管,时钟速度增加了。因此,存在一种混淆,即摩尔定律意味着更快的处理器而不仅仅是更宽的处理器。

Heat dissipation caused the speed increase to top out at around 3 GHz for economically produced silicon.

对于经济生产的硅,散热导致速度增加达到3 GHz左右。

So if you want more cheap computation, it's easier to add more, slower circuits. So the current state-of-the-art commodity processors are multi-core - they are getting wider, but no faster.

因此,如果您想要更便宜的计算,则更容易添加更多,更慢的电路。因此,目前最先进的商品处理器是多核的 - 它们越来越宽,但速度越来越快。

Graphene film transistors require less power, and are performing at around 30 GHz, with theoretical limits at around 0.6 THz.

石墨烯薄膜晶体管需要较少的功率,并且在大约30GHz下执行,理论极限在大约0.6THz。

When graphene technology matures to commodity level in a few years, expect there to be another sea change and no-one will care about using parallel cores for performance, and go back to narrow, fast cores. On the other hand, concurrent computing will still matter for the problems it is a natural fit for, so you'll still have to know how to handle more than one execution unit though.

当石墨烯技术在几年内达到商品化水平时,预计会出现另一次大变化,没有人会关心使用并行核心来提高性能,并回到狭窄的快速核心。另一方面,并​​发计算对于它自然适合的问题仍然很重要,所以你仍然必须知道如何处理多个执行单元。

#11


Because orthogonal computing has failed. We should go quantum.

因为正交计算失败了。我们应该去量子。

#12


Moore's law necessitates parallel computing because Moore's law is on the verge of/is dead. So taking that into consideration, if it is becoming harder and harder to cram transistors onto an IC (due to some of the reasons noted elsewhere) then the remaining options are to add more processors ala Parallel processing or go Quantum.

摩尔定律需要并行计算,因为摩尔定律已经濒临死亡。因此,考虑到这一点,如果将晶体管塞入IC变得越来越难(由于其他地方提到的一些原因),那么剩下的选择是增加更多的处理器和并行处理或去量子。

#13


Moore's law still holds. Transistor counts are still increasing. The problem is figuring out something useful to do with all those transistors. We can't just keep increasing the instruction level parallelism by making pipelines deeper and wider because the circuitry necessary to prove independence between instructions scales terribly in the number of instructions you need to prove independence of. We can't just keep cranking up clock speeds because of heat. We could just keep increasing cache size, but we've hit a point of diminishing returns here. The only use left for the transistors seems to be putting more cores on a chip, which means that the engineer's job of figuring out what to do with the transistors is just pushed up the abstraction ladder, and now programmers have to figure out what to do with all those cores.

摩尔定律仍然存在。晶体管数量仍在增加。问题在于找出对所有这些晶体管有用的东西。我们不能通过使管道更深更宽来不断增加指令级并行性,因为证明指令之间独立性所必需的电路在你需要证明独立性的指令数量上非常大。我们不能因为热量而不断提高时钟速度。我们可以继续增加缓存大小,但我们已经达到了收益递减点。晶体管的唯一用途似乎是在芯片上放置更多内核,这意味着工程师找出如何处理晶体管的工作只是推动了抽象阶梯,现在程序员必须弄清楚要做什么与所有这些核心。

#14


I don't think Moores law necessitates parallel computing, but it does necessiate an eventual shift away from pure miniturization. Multiple solutions exist. One of them is Parallel computing, another is co-processing (which is realted, but not the same thing as parallel computing. co-processing is when you offload work to a special purpose CPU such as a GPU, DSP, etc..)

我不认为摩尔定律需要并行计算,但它确实需要最终从纯粹的小型化转向。存在多种解决方案其中一个是并行计算,另一个是协同处理(它是实现的,但与并行计算不同。协同处理就是当你将工作卸载到专用CPU,如GPU,DSP等时。)

#15


I honestly don't really know, but my guess would be that transistors at some point could get no smaller requiring processing power to be spread out in parallel.

老实说我真的不知道,但我的猜测是晶体管在某些时候可能不会变小,要求处理能力并行分散。

#16


It's because we're all addicted to increasing speed in our processors. Years of conditioning have led us to expect more processing power, year after year. But the physical constraints caused by densely packed transistors have finally put a limit on clock speeds, so increases have to come from a different perspective.

这是因为我们都沉迷于加速处理器的速度。经过多年的调节,我们预计会有更多的处理能力,年复一年。但由密集的晶体管引起的物理限制最终限制了时钟速度,因此增加必须来自不同的视角。

It doesn't have to be this way. The success of the Intel Atom processor shows that processors could just get smaller and cheaper instead. The processor companies will try to keep us on the "bigger, faster" treadmill though, to keep their profits up. And we'll be willing participants, because we'll always find a way to use more power.

它不一定是这样的。英特尔凌动处理器的成功表明,处理器可以变得更小,更便宜。处理器公司将努力让我们保持“更大,更快”的跑步机,以保持他们的利润。我们愿意参与者,因为我们总能找到一种方法来使用更多的力量。

#1


Moore's law just says that the number of transistors on a reasonably priced integrated circuit tends to double every 2 years.

摩尔定律刚刚说,价格合理的集成电路上的晶体管数量每两年就会增加一倍。

Observations about speed or transistor density or die size are all somewhat orthogonal to the original observation.

关于速度或晶体管密度或管芯尺寸的观察结果与原始观察结果有些正交。

Here's why I think Moore's law leads inevitably to parallel computing:

这就是为什么我认为摩尔定律不可避免地导致并行计算:

If you keep doubling the number of transistors, what are you going to do with them all?

如果你把晶体管的数量增加一倍,那你打算怎么做呢?

  • More instructions!
  • Wider data types!
  • 更广泛的数据类型!

  • Floating Point Math!
  • 浮点数学!

  • More caches (L1, L2, L3)!
  • 更多缓存(L1,L2,L3)!

  • Micro Ops!
  • More pipeline stages!
  • 更多管道阶段!

  • Branch prediction!
  • Speculative execution!
  • Data Pre-Fetch!
  • Single Instruction Multiple Data!
  • 单指令多数据!

Eventually, when you've implemented all the tricks you can think of to use all those extra transistors, you eventually think to yourself: why don't we just do all those cool tricks TWICE on the came chip?

最后,当你实现了所有可以想到使用所有这些额外晶体管的技巧时,你最终会想到:为什么我们不在芯片上做那些很酷的技巧TWICE?

Bada bing. Bada boom. Multicore is inevitable.

巴达冰。巴达热潮。多核是不可避免的。


Incidentally, I think the current trend of CPUs with multiple identical CPU cores will eventually subside as well, and the real processors of the future will have a single master core, a collection of general purpose cores, and a collection of special purpose coprocessors (like a graphics card, but on-die with the CPU and caches).

顺便提一下,我认为目前具有多个相同CPU核心的CPU的趋势也将最终平息,未来真正的处理器将拥有一个主核心,一系列通用核心和一系列专用协处理器(如一个显卡,但在CPU和缓存的模具上)。

The IBM Cell processor (in the PS3) is already somewhat like this. It has one master core and seven "synergistic processing units".

IBM Cell处理器(在PS3中)已经有点像这样了。它有一个主核心和七个“协同处理单元”。

#2


One word - Heat.

一个字 - 热。

Due to an inability to dissipate heat at current transistor levels, engineers are using their every growing transistor budgets to create more cores instead of creating more complex (and hot) pipelines and faster processors.

由于无法在当前晶体管级散热,工程师正在利用其每个不断增长的晶体管预算来创建更多内核,而不是创建更复杂(和热)的流水线和更快的处理器。

Moore's law is not at all dead - moore's law is about transistor density at a given cost. It just so happens that for various reasons (like marketing) engineers decided to use their transistor budget to increase clock cycle. Now they decided (because of the heat issue) to start using the transistors for parallelism, plus 64bit computing and reducing power consumption.

摩尔定律根本没有死 - 摩尔定律是关于给定成本的晶体管密度。事实上,由于各种原因(如市场营销),工程师决定使用他们的晶体管预算来增加时钟周期。现在他们决定(因为热量问题)开始使用晶体管进行并行,加上64位计算并降低功耗。

#3


Moore's law describes the trend that performance of chips effectively doubles due to the addition of more transistors to a circuit board.

摩尔定律描述了由于在电路板上增加了更多晶体管,芯片性能有效翻倍的趋势。

Since devices are not increasing in size (if anything the reverse is true) then clearly the space for these additional transistors only becomes available due to chip technology becoming smaller and manufacturing becoming ever better.

由于器件尺寸没有增加(如果反之亦然),那么由于芯片技术变得越来越小并且制造变得越来越好,这些额外晶体管的空间显然也变得可用。

At somepoint however you reach the stage where transistors cannot be minimized any further. It also becomes impossible to increase the size of chips beyond a certain point due to the amount of heat generated and the manufacturing costs involved.

然而,在某些时候,您将达到晶体管无法进一步最小化的阶段。由于产生的热量和所涉及的制造成本,也不可能将芯片的尺寸增加到超过某一点。

These limits necessitate a means of increasing performance beyond simply producing more complex chips.

这些限制需要一种提高性能的方法,而不仅仅是生产更复杂的芯片。

One such method is to employ cheaper and less complex chips in parallel architectures, another is to move away from the traditional integrated chip to something like quantum computing - which by it's very definition is parallel processing.

一种这样的方法是在并行架构中采用更便宜和更简单的芯片,另一种方法是从传统的集成芯片转向类似量子计算的东西 - 其定义是并行处理。

It's worth noting that the title of this question relates more to the observed results of the law (performance increase) rather than the actual law itself which was largely an observation about transistor count.

值得注意的是,这个问题的标题更多地与观察到的法律结果(性能提高)有关,而不是实际的法律本身,这主要是对晶体管数量的观察。

#4


I think it is a reference to the free lunch is over article

我认为这是免费午餐的参考文章

basically, the original version of Moore's law, about transistor density, still holds. But one important derived law, about processing speed doubling every xx months, has hit a wall.

基本上,关于晶体管密度的摩尔定律的原始版本仍然存在。但是,关于处理速度每xx个月加倍的一个重要的衍生法则已经触底了。

So we are facing a future where processor speeds will go up only slightly but we will have more core's and cache to play with.

因此,我们面临的未来处理器速度将略有上升,但我们将拥有更多核心和缓存。

#5


That is an odd question. Moore's law doesn't necessitate anything it is just an observation of the progression of computing power, it doesn't dictate that it must increase at a certain rate.

这是一个奇怪的问题。摩尔定律不需要任何东西,它只是观察计算能力的进展,它并没有规定它必须以一定的速度增长。

#6


Increasing the speed of processors would make the operating temperature so extremely high it would burn a hole in your desk. The makers of the chips are running up against certain limitations they can't get around... like the speed of light, for instance. Parallel computing will allow them to speed up the computers without starting a fire.

提高处理器的速度会使操作温度如此之高,以至于会在桌面上烧掉一个洞。这些芯片的制造商正在遇到一些他们无法解决的限制......比如光速。并行计算将使他们能够加速计算机而不会引发火灾。

#7


Transistors and cpus and whatnot are getting smaller and smaller and faster and faster. Alas, the heat and voltage costs for computing are going up. The heat and voltage issues are as much of a concern as the actual physical size minimums. A 100ghz chip would suck up too much voltage and get too hot but 100 1ghz chips would have less of an issue with this.

晶体管和cpu等等越来越小,越来越快。唉,计算的热量和电压成本正在上升。热量和电压问题与实际物理尺寸最小值一样令人担忧。一个100ghz的芯片会吸收过多的电压而变得太热,但100个1ghz的芯片会产生较少的问题。

#8


Interestingly, the idea proposed in the question that parallel computing is "necessitated" is thrown into question by Amdahl's Law, which basically says that having parallel processors will only get you so far unless 100% of your program is parallelizable (which is never the case in the real world).

有趣的是,在问题中提出并行计算是“必要的”的想法受到Amdahl定律的质疑,基本上说,并行处理器只会让你到目前为止,除非100%的程序是可并行化的(从来都不是这样)在现实世界)。

For example, if you have a program which takes 20 minutes on one processor and is 50% parallelizable, and you buy a large number of processors to speed things up, your minimum time to run would still be over 10 minutes. This is ignoring the cost and other issues involved.

例如,如果您的程序在一个处理器上占用20分钟且可并行化50%,并且您购买大量处理器以加快速度,那么您的最短运行时间仍将超过10分钟。这忽略了所涉及的成本和其他问题。

#9


The real answer is completely un-technical, not that the hardware explanations aren't fantastic. It's that Moore's Law has become less and less of an observation, and more of an expectation. This expectation of computers growing exponentially has become the driving force of the industry, which necessitates all the parallelism.

真正的答案完全不是技术性的,并不是说硬件解释不是很棒。这就是摩尔定律越来越少的观察,更多的是期望。这种对计算机成倍增长的预期已成为该行业的驱动力,这需要所有的并行性。

#10


Moore's law says that the number of transistors in an IC relative to cost increases exponentially year on year.

摩尔定律表明,IC中晶体管的数量相对于成本的数量同比呈指数增长。

Historically, this was partly due to a decrease in transistor size, and smaller transistors also switched faster. Because you got faster transistors in step with Moore's law, clock speed increased. So there's a confusion that say Moore's law means faster processors rather than just wider.

从历史上看,这部分是由于晶体管尺寸减小,而较小的晶体管也更快地切换。因为根据摩尔定律得到更快的晶体管,时钟速度增加了。因此,存在一种混淆,即摩尔定律意味着更快的处理器而不仅仅是更宽的处理器。

Heat dissipation caused the speed increase to top out at around 3 GHz for economically produced silicon.

对于经济生产的硅,散热导致速度增加达到3 GHz左右。

So if you want more cheap computation, it's easier to add more, slower circuits. So the current state-of-the-art commodity processors are multi-core - they are getting wider, but no faster.

因此,如果您想要更便宜的计算,则更容易添加更多,更慢的电路。因此,目前最先进的商品处理器是多核的 - 它们越来越宽,但速度越来越快。

Graphene film transistors require less power, and are performing at around 30 GHz, with theoretical limits at around 0.6 THz.

石墨烯薄膜晶体管需要较少的功率,并且在大约30GHz下执行,理论极限在大约0.6THz。

When graphene technology matures to commodity level in a few years, expect there to be another sea change and no-one will care about using parallel cores for performance, and go back to narrow, fast cores. On the other hand, concurrent computing will still matter for the problems it is a natural fit for, so you'll still have to know how to handle more than one execution unit though.

当石墨烯技术在几年内达到商品化水平时,预计会出现另一次大变化,没有人会关心使用并行核心来提高性能,并回到狭窄的快速核心。另一方面,并​​发计算对于它自然适合的问题仍然很重要,所以你仍然必须知道如何处理多个执行单元。

#11


Because orthogonal computing has failed. We should go quantum.

因为正交计算失败了。我们应该去量子。

#12


Moore's law necessitates parallel computing because Moore's law is on the verge of/is dead. So taking that into consideration, if it is becoming harder and harder to cram transistors onto an IC (due to some of the reasons noted elsewhere) then the remaining options are to add more processors ala Parallel processing or go Quantum.

摩尔定律需要并行计算,因为摩尔定律已经濒临死亡。因此,考虑到这一点,如果将晶体管塞入IC变得越来越难(由于其他地方提到的一些原因),那么剩下的选择是增加更多的处理器和并行处理或去量子。

#13


Moore's law still holds. Transistor counts are still increasing. The problem is figuring out something useful to do with all those transistors. We can't just keep increasing the instruction level parallelism by making pipelines deeper and wider because the circuitry necessary to prove independence between instructions scales terribly in the number of instructions you need to prove independence of. We can't just keep cranking up clock speeds because of heat. We could just keep increasing cache size, but we've hit a point of diminishing returns here. The only use left for the transistors seems to be putting more cores on a chip, which means that the engineer's job of figuring out what to do with the transistors is just pushed up the abstraction ladder, and now programmers have to figure out what to do with all those cores.

摩尔定律仍然存在。晶体管数量仍在增加。问题在于找出对所有这些晶体管有用的东西。我们不能通过使管道更深更宽来不断增加指令级并行性,因为证明指令之间独立性所必需的电路在你需要证明独立性的指令数量上非常大。我们不能因为热量而不断提高时钟速度。我们可以继续增加缓存大小,但我们已经达到了收益递减点。晶体管的唯一用途似乎是在芯片上放置更多内核,这意味着工程师找出如何处理晶体管的工作只是推动了抽象阶梯,现在程序员必须弄清楚要做什么与所有这些核心。

#14


I don't think Moores law necessitates parallel computing, but it does necessiate an eventual shift away from pure miniturization. Multiple solutions exist. One of them is Parallel computing, another is co-processing (which is realted, but not the same thing as parallel computing. co-processing is when you offload work to a special purpose CPU such as a GPU, DSP, etc..)

我不认为摩尔定律需要并行计算,但它确实需要最终从纯粹的小型化转向。存在多种解决方案其中一个是并行计算,另一个是协同处理(它是实现的,但与并行计算不同。协同处理就是当你将工作卸载到专用CPU,如GPU,DSP等时。)

#15


I honestly don't really know, but my guess would be that transistors at some point could get no smaller requiring processing power to be spread out in parallel.

老实说我真的不知道,但我的猜测是晶体管在某些时候可能不会变小,要求处理能力并行分散。

#16


It's because we're all addicted to increasing speed in our processors. Years of conditioning have led us to expect more processing power, year after year. But the physical constraints caused by densely packed transistors have finally put a limit on clock speeds, so increases have to come from a different perspective.

这是因为我们都沉迷于加速处理器的速度。经过多年的调节,我们预计会有更多的处理能力,年复一年。但由密集的晶体管引起的物理限制最终限制了时钟速度,因此增加必须来自不同的视角。

It doesn't have to be this way. The success of the Intel Atom processor shows that processors could just get smaller and cheaper instead. The processor companies will try to keep us on the "bigger, faster" treadmill though, to keep their profits up. And we'll be willing participants, because we'll always find a way to use more power.

它不一定是这样的。英特尔凌动处理器的成功表明,处理器可以变得更小,更便宜。处理器公司将努力让我们保持“更大,更快”的跑步机,以保持他们的利润。我们愿意参与者,因为我们总能找到一种方法来使用更多的力量。