Single-Instruction Stream Multiple-Data Stream Architecture

Single-Instruction Stream Multiple-Data Stream Architecture Introduction to SIMD Architectures SIMD (Single-Instruction Stream Multiple-Data Stream) architectures are essential in the parallel world of computers. Their ability to manipulate large vectors and matrices in minimal time has created a phenomenal demand in such areas as weather data and cancer radiation research. The power behind this type of architecture can be seen when the number of processor elements is equivalent to the size of your vector. In this situation, componentwise addition and multiplication of vector elements can be done simultaneously. Even when the size of the vector is larger than the number of processors elements available, the speedup, compared to a sequential algorithm, is immense. There are two types of SIMD architectures we will be discussing. The first is the True SIMD followed by the Pipelined SIMD. Each has its own advantages and disadvantages but their common attribute is superior ability to manipulate vectors. True SIMD: Distributed Memory The True SIMD architecture contains a single contol unit(CU) with multiple processor elements(PE) acting as arithmetic units(AU). In this situation, the arithmetic units are slaves to the control unit. The AUs cannot fetch or interpret any instructions. They are merely a unit which has capabilities of addition, subtraction, multiplication, and division. Each AU has access only to its own memory. In this sense, if a AU needs the information contained in a different AU, it must put in a request to the CU and the CU must manage the transferring of information. The advantage of this type of architecture is in the ease of adding more memory and AUs to the computer. The disadvantage can be found in the time wasted by the CU managing all memory exchanges. True SIMD: Shared Memory Another True SIMD architecture, is designed with a configurable association between the PEs and the memory modules(M). In this architecture, the local memories that were attached to each AU as above are replaced by memory modules. These Ms are shared by all the PEs through an alignment network or switching unit. This allows for the individual PEs to share their memory without accessing the control unit. This type of architecture is certainly superior to the above, but a disadvantage is inherited in the difficulty of adding memory. Pipelined SIMD Pipelined SIMD architecture is composed of a pipeline of arithmetic units with shared memory. The pipeline takes different streams of instructions and performs all the operations of an arithmetic unit. The pipeline is a first in first out type of procedure. The size of the pipelines are relative. To take advantage of the pipeline, the data to be evaluated must be stored in different memory modules so the pipeline can be fed with this information as fast as possible. The advantages to this architecture can be found in the speed and efficiency of data processing assuming the above stipulation is met. SIMD BASICS Early microprocessors didnt actually have any floating-point capabilities; they were strictly integer crunchers.? Floating-point calculations were done on separate, dedicated hardware, usually in the form of a math coprocessor.? Before long though, transistor sizes shrunk to the point where it became feasible to put a floating-point unit directly onto the main CPU die, and the modern integer/floating-point microprocessor was born.? Of course, the addition of floating-point hardware meant the addition of floating-point instructions.? For the x86 world, this meant the introduction of the x87 floating-point architecture and its (now hopelessly archaic) stack-based register model. Actually, the addition of SIMD instructions and hardware to a modern, superscalar CPU is a bit more drastic than the addition of floating-point capability.? A microprocessor is a SISD device (Single Instruction stream, Single Data stream), and it has been since its inception. As you can see from the above picture, a SIMD machine exploits a property of the data stream called data parallelism.? You get data parallelism when you have a large mass of data of a uniform type that needs the same instruction performed on it.? A classic example of data parallelism is inverting an RGB picture to produce its negative.? You have to iterate through an array of uniform integer values (pixels), and perform the same operation (inversion) on each one multiple data points, a single operation.? Modern, superscalar SISD machines exploit a property of the instruction stream called instruction-level parallelism (ILP).? In a nutshell, this means that you execute multiple instructions at once on the same data stream.? (See my other articles for more detailed discussions of ILP).? So a SIMD machine is a different class of machine than a normal microprocessor.? SIMD is about exploiting parallelism in the data stream, while superscalar SISD is about exploiting parallelism in the i nstruction stream. There were some early, ill-fated attempts at making a purely SIMD machine (i.e., a SIMD-only machine).? The problem with these attempts is that the SIMD model is simply not flexible enough to accoodate general purpose code.? The only form in which SIMD is really feasible is as a part of a SISD host machine that can execute conditional instructions and other types of code that SIMD doesnt handle well.? This is, in fact, the situation with SIMD in todays market.? Programs are written for a SISD machine, and include in their code SIMD instructions. SIMD Machines The three SIMD machines covered in this paper are the Connection Machine by Danny Hillis, the Abacus Project at the MIT AI Lab, and the CAM-8 machine by Norman Margolus. These three machines give a pretty accurate sampling of the type of SIMD machines that were constructed as well as an idea of the motivations for creating the machines in the first place. The Connection Machine was composed of 65,536 bit processors. Each die consisted of 16 processors with each processor capable of communicating with each other via a switch. These 4,096 dies formed the nodes of a 12th dimension hypercube network. Thus, a processor was guaranteed to be within 12 hops of any other processor in the machine. The hypercube network also facilitated communication by providing alternative routes from source processor to destination. Each node was given a 12-bit node ID, and different paths between two nodes in the network could be traversed based on how the node ID was read. The network allowed for both packet and circuit-based communication for flexibility. The second machine discussed is the Abacus machine created at the MIT AI Lab. This machine was constructed primarily for vision processing. The machine consisted of 1024 bit processing elements set in a 2D mesh. The primary concept of interest from the design was that the processing elements were configurable, and used reconfigurable bit parallel RBP algorithms instead of traditional bit serial computation. This means that each PE emulated logic for part of an arithmetic circuit (be it an adder, shifter, multiplier,etc) based on a RBP algorithm. The motivation for having these configurable processingelements was to save on the silicon area needed to implement arithmetic. However,because there was a necessary overhead for reconfiguration and the implementation did not easily allow for pipelining due to data dependencies, it was not clear that having configurable processing elements was a definite win. SIMD versus Loop Pipelining We can consider two different models for mapping loops onto coarse-grained reconfigurable architecture SIMD and loop pipelining. SIMD computation model is efficient for computation intensive,data-parallel applications requiring less context words to configure reconfigurable processing elements. Since data load and computation are temporarily separated in this model, array elements are not efficiently utilized. In the case of loop pipelining, different operations in a loop can be executed simultaneously in a pipeline. With this flexibility, data load and computation can be simultaneously executed and all reconfigurable array elements can be efficiently used. In some loops, the performance of pipelining is roughly the same as the performance of SIMD. However, if a loop has frequent memory operations, the pipelining will render much higher performance. Reconfigurable Architecture The reconfigurable architecture that we propose consists of an ARM 926EJ-S processor, an SDRAM, a DMA controller, and a coarse-grained Reconfigurable Core Module (RCM) template, which is similar to Morphosys and specified in the DSE flow. The communication bus is AMBA AHB ,which couples the ARM 926EJ-S processor and the DMA controller as master devices and the RCM as a slave device. The ARM 926EJ-S processor executes control intensive, irregular code segments and the RCM executes data-intensive, kernel code segments. Design Space Exploration The design space exploration (DSE) flow of coarse-grained reconfigurable architecture. A design starts from profiling and partitioning of target application and defining an architecture from the tem plate. Data intensive, regular loops are selected from the profiling result and the rest of the application is modified to take care of synchronization. The selected loops are analyzed to determine the RCM structure from the template and the configuration words are generated. Design space exploration flow From the architecture specification, we can generate a SystemC description for fast architecture evaluation . Then the loop pipelining model is applied to the SystemC description. Binary configuration data are included in the executable code and overall performance of the application is evaluated on the transaction level platform. The transaction level modeling enables fast design space exploration at early stage . Finally, the architecture is designed at the RT level from the SystemC mo del and the performance is evaluated on the RTL platform. The RTL architecture is verified by FPGA prototyping. RCM Template Architecture RCM specification starts from the template architecture similar to Morphosys. Whereas the memory structure (frame buffer and configuration cache) of Morphosys support only the SIMD model, we support both SIMD and pipelining by modifying the memory structure. Types of memory:- Frame Buffer Frame buffer (FB) of Morphosys does not support concurrency between the load of two operands and the store of result in a same column. It is not needed in SIMD mapping. However, in the case of loop pipelining, concurrent load and store operations can happen between mapped loop iterations. So we modified the FB and bit-width of data bus is specified in the DSE flow. We simply added a bank to each set. Therefore, a bank can be connected to the write bus while the other two banks are connected to the read buses. Any combination of one-to-one mapping between the three banks and the three buses is possible. Configuration Cache Context memory of Morphosys is designed for broadcast configuration. So RCs in the same row or column share the same context word for SIMD operation. However, in the case of loop pipelining, each RC can be configured by different context word. So we modified the context memory and designated it as Configuration Cache. Configuration cache is composed of 64 Cache Elements(CE) and Cache Control Unit(CCU) for controlling each CE. Each CE has enough layers that enable dynamic reconfiguration and the number of layers is specified in the DSE flow. CCU supports 4 configuration modes(three broadcast modes and one individual mode) for efficient data assignment. RC Array Execution Control Unit If the main processor directly controls the RC array execution through AMBA AHB, it will cause high overhead in the main processor. In addition, the latency of the control will degrade the performance of the whole system, especially when dynamic reconfiguration is used. So we implement a control unit to control the execution of the RC array every cycle. The RC Array Execution Control Unit (RCECU) receives the encoded data for controlling RC execution from the main processor. The encoded data includes execution cycles, chip select, read/write mode, and addresses of FB and CCU for guaranteeing correct operations of the RC array. RCM Specification From profiling result, we find that ME and DCT functions occupymost of the execution time ME takes about 70% and DCT takes about 7.40%. Specifically, Sum of Absolute Differences (SAD) function called by ME takes about 47.7%. Furthermore, the two functions have regular loops that fit well with the RC array. We determine the RCM structure by analyzing the DCT and ME functions. The structure is similar to Morphosys but the bit-widthof the data bus is extended to 16 and some interconnects between RCs are added for the DCT function. In the case of Morphosys, horizontal and vertical express lanes exist to guarantee connectivity between quadrants but express lanes dont support concurrent data exchange between symmetrical RCs in the same row or column. Therefore the interconnects are added for removing data arrangement cycles . We do not expect much increase in the area with this modification but need quantitative analysis to see the effect.

The Broken Spears by Miguel Leon-Portilla Essay -- Aztecs Broken Spear

"The Broken Spears" by Miguel Leon-Portilla The author argues that the Spanish were completely at fault for the total destruction of the Aztec Empire. In Broken spears, the author explains how many factors other than Spanish power contributed to the downfall of the Aztecs. Not only did the Spanish have many advantages over the Aztecs, but also they also exploited them and took advantage of the cultural difference. The main key aspects to the Spanish victory, is that the Spanish were viewed as gods at first because of their appearance, the Aztecs welcomed the Spanish with gifts and festivities, which showed the Spanish had total control of people. The Aztecs also held a ritual ceremony for the arrival of the â€Å"god† that included a human sacrifice, which was seen by the Spanish as a disgusting act, this trigged Spanish hatred for the Aztecs. When the Spanish saw the resources that the Aztecs had, they quickly started taking advantage of the Aztecs weaknesses. The Spanish had more power; they were better equipped than the Aztecs and more advanced. The Aztecs’ leader Montecuhzoma was a weak leader, he was more of a coward, more concerned with his well being and safety than the safety of his people and kingdom so he quickly surrendered under pressure. The Spanish used surrounding enemy tribes to take over the Aztec lands, then began to massacre men, women and children. When the Aztec became angry and started to fight back, they were quickly wrecked by the diseases the Spanish ...

Communication Breakdown

Communication Breakdown Deborah Tannen in her essay â€Å"Sex Lies and Conversation† points out the problems in communication between genders. She focuses on conversational styles and on how a person’s expectations of one’s reaction may cause misunderstanding. She emphasizes the importance of something she calls cross cultural communication. Tannen believes that in order to communicate properly with the opposite sex we must learn their culture and adapt to conversation with them. First, Tannen describes the differences in how children socialize.Most of what we know about communication we learned when we were children. When women are little girls, they have best friends, ones we can share our secrets with. When we get a little older and have a relationship with the opposite sex we expect the same kind of communication. Tannen believes the importance is not necessarily the topic but the feeling that you get from conversation and points out ,â€Å"What is important is not the individual subjects that are disclosed but the sense of closeness, of a life shared, that emerges when people tell their thoughts, feelings and impressions† (51).We must learn to socialize with the opposite sex the same way we do with the same sex. If we could learn to connect and have the â€Å"sense of closeness† with the opposite sex we might be able to communicate successfully. Tannen’s research has shown that men socialize differently as boys and â€Å"Since they don’t assume talk is the cement that binds the relationship, men don’t know what kind of talk women want, and they don’t miss it when it isn’t there†(51). Tannen points out that boys learn to communicate in larger groups, often struggling not to be in the subaltern position of the conversation.Men do not like to listen; they like to feel like they are the superior person in the conversation. (51) Communication starts in childhood and what we have learne d to expect from conversations. Next, Tannen observes how men and women listen to each other and how this can cause misunderstanding between genders . She feels that there is confusion about what women expect and declares, â€Å"When women talk to each other in a close, comfortable setting, they often overlap, finish each other’s sentences and anticipate what the other is about to say† (53).Tannen also talks about listener noise. Women often will say â€Å"mhm† or â€Å"uhuh† and men do not do these things when they talk to each other. Women are looking for that â€Å"listener noise† and if he is silent, she thinks he is ignoring her but on the same token Tannen notes, â€Å"Men who expect silent attention interpret a stream of listener noise as overreaction or impatience† (53). Body language and conversational habits are causing women to feel like they are being ignored and men to feel as if they are being interrupted.When a woman is having a conversation they expect the listener to â€Å"express agreement† and be supportive of the conversation. Men, on the other hand, feel it is their conversational duty to express other positions in the argument. (53) Learning the importance of building a rapport with the person you are talking to may be another way to have successful conversation with the opposite sex. Finally, Tannen believes that learning how the opposite sex communicates is the key. If we can learn what to expect when communicating with someone then we can learn how to respond.Tannen suggests â€Å" A sociolinguistic approach by which male-female conversation is seen as cross- cultural communication allows us to understand the problem and forge solutions without blaming either party†(54). Tannen stresses the importance of telling the other person how you feel about their communication with you and advises,† Women who feel abandoned and deprived when their husbands won’t listen to or re port daily news may be happy to discover their husbands trying to adapt once they understand the place of small talk in women’s relationships† (55).It is really not the male or females fault that we communicate the way we do but it is our fault if we do not speak up and tell them it bothers us. When we are in a relationship with another person we just have to adapt to the way they communicate to make the relationship work. If a woman learns to accept that her husband is not going to communicate with her like her best girlfriend did in her childhood then she can look for other people to hold those types of conversation with. (55) Learning to communicate with the opposite sex involves many different things.Communication is something that is learned in childhood but mostly with the same sex. In order to have a successful relationship with good communication we must understand that men and women have very different body language when communicating and that they do not respo nd the same. Accepting that there is a difference in the communication of genders and learning to adapt is paramount in relationships. Tannen, Deborah. â€Å"Sex, Lies and Conversation. † The Norton Mix. Judy Sieg. New York: W. W. Norton & Company, 2012. 45-55. Print.

12 Greatest Visual Artists of All Time

If you ask art historians who the greatest visual artists of all time are, there would be a multitude of different names. Of course,  there are several standards by which you can measure who the best artists of all time are. Unfortunately, the art world has been  historically dominated by men, and women artists are seldom mentioned despite their enormous contributions. It is important that women artists are also recognized as part of the canon, though, for their work is also of superior and enduring quality, and some even achieved success in their own day despite great societal obstacles. There are several factors that would determine if a particular artist makes the list. One of them is the trend and fashion of the moment that the artist lived, another is the longevity of an artists popularity. The impact made by an artist on his contemporaries is yet another factor to consider. In the long run, determining who are the greatest artists of all time may be a subjective opinion;  however, based on public opinion and what the museums say, the 10  best visual artists of all time are: 1.  Michelangelo (1475-1564) Michelangelo is regarded as the greatest sculptor and painter of all time. He was a major figure of the Renaissance in Italy, especially in Florence and Rome. Even today, some of his marble carvings have a flawless beauty. Michelangelo is popularly known for the Italian Renaissance sculpture as well as the Sistine Chapel frescoes, among other incredible works of art. 2.  Rembrandt van Rijn  (1606-1669) Rembrandt is a Dutch painting genius who created several masterpieces. Rembrandt is particularly renowned for his passion for history painting and still life portraits. He made several canvasses famous for their emotional content, which is why Rembrandt was branded as a realist painter. 3. Pablo Picasso (1881-1973) Pablo Picasso, the founder of cubism, is one of the most influential artists in history. He was a sculptor, printmaker, painter, and ceramicist. He produced several of the greatest paintings of the 20th century. Born to a painter father, Picasso received all the needed encouragement to build a successful painting career. This gave him the privilege to study art in some of the finest art schools in Spain. He was among the 250 sculptors who exhibited the 3rd Sculpture International held in America. His style and technique was exceptionally prolific throughout his lifetime, producing a total number of 50,000 artworks, including drawings, paintings, and sculptures, among others. Of all  forms of art, Picasso excelled most in painting. 4. Leonardo da Vinci (1452-1519) Leonardo was born in Florence, Italy. Despite living several centuries ago, he still remains one of the most influential artists of all time. His only training was in the field of science and he was apprenticed at an early age to a renown Florence painter of the  time.  Leonardo was regarded as an intellectual in his lifetime because of his passion for science. Leonardos contributions to the art world were small, but two of his paintings are the most popular today: Mona Lisa and The Last Supper, which is the only surviving fresco of Leonardo da Vinci. The fact that his interest ranged beyond art could have been the reason why his input was very small. In his lifetime, so engrossed was he in physics and mechanics, that he created workable artistic designs for bicycles, among other things. This is what is popularly believed to be the cause of his failure to complete several of his paintings and art projects. There are also credible reports that he spent a significant amount of time thinking and testing scientific laws, as well as writing his observations about them. 5. Claude Monet  (1840-1926) Claude Monet is popularly regarded as the founder of French Impressionist painting. Monet had an incredible passion for documenting the countryside in his numerous paintings. This was the birth of his impressionist paintings. When Monet went to Paris, he observed several artists copying the popular works of other artists. Rather than follow this trend, Monet developed the habit of sitting by an available window and painting what he saw. Gradually, Monet became popular for his impressionism. He influenced several young artists and made them commit to impressionism, and within a short period of time, Impressionism became a popular form of painting in Paris. Sometime in 1874, the first impressionism exhibition took place in Paris. In that exhibition, Monet presented a total of 12 artworks, which included five paintings and seven pastels. 6. Vincent van Gogh (1853-1890) Born in the Netherlands, van Gogh was a remarkable painter whose works of art are still sold at unbelievable prices across the world today. His contribution to art was basically painting. He created many still-life paintings, several of them were portraits of friends and acquaintances. In all, van Gogh completed about 800 paintings. One thing that distinguished him as a painter was his grasp for color relationship and unique brushwork. His work is still a great source of inspiration for  several impressionists all over the world. 7. Auguste Rodin (1840-1917) Rodin was a workaholic and a dedicated sculptor. He is a worthy successor to great minds like Michelangelo. Rodin is popularly considered the best sculptor of the modern era. His success is derived from his ability to model stone and clay in complex manners. During his lifetime, Rodin earned a living collaborating with established sculptors on several projects. 8. Jan van Eyck (1390-1441) He is one of the most remarkable Northern Renaissance artists and many of his works are greatly copied today. Aside from working for the court, van Eyck produced several paintings for private clients. One thing that was peculiar to him is that he was the only painter in his era to sign his canvasses. 9. Donatello (1386-1466) Donatello is regarded as the greatest sculptor of the 15th century. He was no doubt one of the best artists of that era. He was very skillful with wood, terracotta, and stone. 10. Peter Paul Rubens  (1571-1640) Peter Rubens was the most influential figure in counter-reformation baroque art. His contribution to art was basically paintings; however, he produced almost all possible genres of paintings. He was very versatile, producing cartoons and book illustrations as well as sketches for metal work and sculptures. 11.  Ãƒâ€°lisabeth Louise  Vigà ©e-Le Brun (1755-1842) Vigà ©e-Le Brun was one of the most renowned portraitists of 18th century France and the most famous woman artist in Europe. Through talent and perseverance, she achieved  success in one of the most turbulent periods of French and European history and became one of Queen Marie Antoinettes favorite painters. She painted more than 20 portraits of Marie Antoinette, along with portraits of other European aristocrats, actors, and writers, primarily women. She was elected to art academies in 10 different cities. She  created over 900 paintings in her lifetime, including roughly 600 portraits and 200 landscapes. 12. Georgia OKeeffe (1887-1986) Georgia OKeeffe was one of the most important and successful artists of the 20th century. She was one of the first American artists to embrace abstraction and became one of the leaders of the American Modernist  movement. Her art is personal and innovative. She is known for paintings of large abstract flowers, dramatic New York skyscrapers, and southwestern landscapes.