Misplaced Pages

#669330

77-579: The Cray EL90 series was an air-cooled vector processor supercomputer first sold by Cray Research in 1993. The EL90 series evolved from the Cray Y-MP EL minisupercomputer , and is compatible with Y-MP software, running the same UNICOS operating system . The range comprised three models: The EL90 series Input/Output Subsystem (IOS) was based on the VMEbus and a Heurikon HK68 Motorola 68000 -based processor board (or IOP). The IOP also provided

154-412: A 15-bit instruction word containing a 6-bit operation code. There are only 64 machine codes, including a no-operation code , with no fixed-point multiply or divide operations in the central processor. The 7600 has two main core memories. Small core memory holds the instructions currently being executed and the data currently being processed. It has an access time of 10 of the 27.5-ns minor cycles and

231-408: A 60-bit word length. Large core memory holds data ready to transfer to small core memory. It has an access time of 60 of the 27.5-ns minor cycles and a word length of 480 bits (512 bits with parity). Accesses are fully pipelined and buffered, so the two have the same sequential transfer rate of 60 bits every 27.5 ns. The two work in parallel, so the sequential transfer rate from one to the other

308-625: A LOAD, ADD, MULTIPLY and STORE sequence. If the SIMD width is 4, then the SIMD processor must LOAD four elements entirely before it can move on to the ADDs, must complete all the ADDs before it can move on to the MULTIPLYs, and likewise must complete all of the MULTIPLYs before it can start the STOREs. This is by definition and by design. Having to perform 4-wide simultaneous 64-bit LOADs and 64-bit STOREs

385-554: A batch of vector instructions to be pipelined into each of the ALU subunits, a technique they called vector chaining . The Cray-1 normally had a performance of about 80 MFLOPS, but with up to three chains running it could peak at 240 MFLOPS and averaged around 150 – far faster than any machine of the era. Other examples followed. Control Data Corporation tried to re-enter the high-end market again with its ETA-10 machine, but it sold poorly and they took that as an opportunity to leave

462-540: A co-processor, it is the main computer with the PC-compatible computer into which it is plugged serving support functions. Modern graphics processing units ( GPUs ) include an array of shader pipelines which may be driven by compute kernels , and can be considered vector processors (using a similar strategy for hiding memory latencies). As shown in Flynn's 1972 paper the key distinguishing factor of SIMT-based GPUs

539-468: A greater quantity of numbers in the vector register, it becomes unfeasible for the computer to have a register that large. As a result, the vector processor either gains the ability to perform loops itself, or exposes some sort of vector control (status) register to the programmer, usually known as a vector Length. The self-repeating instructions are found in early vector computers like the STAR-100, where

616-421: A high performance vector processor may have multiple functional units adding those numbers in parallel. The checking of dependencies between those numbers is not required as a vector instruction specifies multiple independent operations. This simplifies the control logic required, and can further improve performance by avoiding stalls. The math operations thus completed far faster overall, the limiting factor being

693-428: A pipeline architecture that supported both scalar and vector computations, with peak performance reaching approximately 20 MFLOPS, readily achieved when processing long vectors. Expanded ALU configurations supported "two pipes" or "four pipes" with a corresponding 2X or 4X performance gain. Memory bandwidth was sufficient to support these expanded modes. The STAR-100 was otherwise slower than CDC's own supercomputers like

770-482: A pipelined loop over 16 units for a hybrid approach. The Broadcom Videocore IV is also capable of this hybrid approach: nominally stating that its SIMD QPU Engine supports 16-long FP array operations in its instructions, it actually does them 4 at a time, as (another) form of "threads". This example starts with an algorithm ("IAXPY"), first show it in scalar instructions, then SIMD, then predicated SIMD, and finally vector instructions. This incrementally helps illustrate

847-596: A single common instruction to all of the arithmetic logic units (ALUs), one per cycle, but with a different data point for each one to work on. This allowed the Solomon machine to apply a single algorithm to a large data set , fed in the form of an array. In 1962, Westinghouse cancelled the project, but the effort was restarted by the University of Illinois at Urbana–Champaign as the ILLIAC IV . Their version of

SECTION 10

#1732787715670

924-536: A small core memory read or write. Arithmetic and logic instructions have these registers as sources and destinations. The programmer or compiler tries to fetch data in time to be used and store data before more data needs the same register, but if it is not ready, the processor goes into a wait state until it is. It also waits if one of the four floating-point arithmetic units is not ready when requested, but due to pipelining, this does not usually happen. The CDC 7600 "was designed to be machine code upward compatible with

1001-428: A special instruction, the significance compared to Videocore IV (and, crucially as will be shown below, SIMD as well) being that the repeat length does not have to be part of the instruction encoding. This way, significantly more work can be done in each batch; the instruction encoding is much more elegant and compact as well. The only drawback is that in order to take full advantage of this extra batch processing capacity,

1078-541: A vector processor. Although vector supercomputers resembling the Cray-1 are less popular these days, NEC has continued to make this type of computer up to the present day with their SX series of computers. Most recently, the SX-Aurora TSUBASA places the processor and either 24 or 48 gigabytes of memory on an HBM 2 module within a card that physically resembles a graphics coprocessor, but instead of serving as

1155-462: A very compact, but basically unrepairable module. However the same dense packing also led to the machine's biggest problem – heat. For the 7600, Cray once again turned to his refrigeration engineer, Dean Roush, formerly of the Amana company. Roush added an aluminum plate to the back of each side of the cordwood stack, which were in turn cooled by a liquid-freon system running through

1232-406: Is unable by design to cope with iteration and reduction. This is illustrated further with examples, below. Additionally, vector processors can be more resource-efficient by using slower hardware and saving power, but still achieving throughput and having less latency than SIMD, through vector chaining . Consider both a SIMD processor and a vector processor working on 4 64-bit elements, doing

1309-611: Is significantly more complex and involved than "Packed SIMD" , which is strictly limited to execution of parallel pipelined arithmetic operations only. Although the exact internal details of today's commercial GPUs are proprietary secrets, the MIAOW team was able to piece together anecdotal information sufficient to implement a subset of the AMDGPU architecture. Several modern CPU architectures are being designed as vector processors. The RISC-V vector extension follows similar principles as

1386-521: Is single-issue and uses no SIMD ALUs, only having 1-wide 64-bit LOAD, 1-wide 64-bit STORE (and, as in the Cray-1 , the ability to run MULTIPLY simultaneously with ADD), may complete the four operations faster than a SIMD processor with 1-wide LOAD, 1-wide STORE, and 2-wide SIMD. This more efficient resource utilization, due to vector chaining , is a key advantage and difference compared to SIMD. SIMD, by design and definition, cannot perform chaining except to

1463-469: Is 60 bits per 27.5 ns minor-cycle. On an operating system call, the contents of the small core memory are swapped out and replaced from the large core memory by the operating system, and restored afterward. There is a 12-word instruction pipeline, called instruction word stack in CDC documentation. All addresses in the stack are fetched, without waiting for the instruction field to be processed. Therefore,

1540-416: Is assumed that both x and y are properly aligned here (only start on a multiple of 16) and that n is a multiple of 4, as otherwise some setup code would be needed to calculate a mask or to run a scalar version. It can also be assumed, for simplicity, that the SIMD instructions have an option to automatically repeat scalar operands, like ARM NEON can. If it does not, a "splat" (broadcast) must be used, to copy

1617-505: Is comprehensive individual element-level predicate masks on every vector instruction as is now available in ARM SVE2. And AVX-512 , almost qualifies as a vector processor. Predicated SIMD uses fixed-width SIMD ALUs but allows locally controlled (predicated) activation of units to provide the appearance of variable length vectors. Examples below help explain these categorical distinctions. SIMD, because it uses fixed-width batch processing,

SECTION 20

#1732787715670

1694-489: Is instead "pointed to" by passing in an address to a memory location that holds the data. Decoding this address and getting the data out of the memory takes some time, during which the CPU traditionally would sit idle waiting for the requested data to show up. As CPU speeds have increased, this memory latency has historically become a large impediment to performance; see Random-access memory § Memory wall . In order to reduce

1771-466: Is not common to later designs, and is often referred to under a separate category, massively parallel computing. Around this time Flynn categorized this type of processing as an early form of single instruction, multiple threads (SIMT). International Computers Limited sought to avoid many of the difficulties with the ILLIAC concept with its own Distributed Array Processor (DAP) design, categorising

1848-456: Is not possible, then the operations take even longer because the LD may not be issued (started) at the same time as the first ADDs, and so on. If there are only 4-wide 64-bit SIMD ALUs, the completion time is even worse: only when all four LOADs have completed may the SIMD operations start, and only when all ALU operations have completed may the STOREs begin. A vector processor, by contrast, even if it

1925-458: Is that it has a single instruction decoder-broadcaster but that the cores receiving and executing that same instruction are otherwise reasonably normal: their own ALUs, their own register files, their own Load/Store units and their own independent L1 data caches. Thus although all cores simultaneously execute the exact same instruction in lock-step with each other they do so with completely different data from completely different memory locations. This

2002-480: Is that vector processors, inherently by definition and design, have always been variable-length since their inception. Whereas pure (fixed-width, no predication) SIMD is often mistakenly claimed to be "vector" (because SIMD processes data which happens to be vectors), through close analysis and comparison of historic and modern ISAs, actual vector ISAs may be observed to have the following features that no SIMD ISA has: Predicated SIMD (part of Flynn's taxonomy ) which

2079-496: Is these which somewhat deserve the nomenclature "vector processor" or at least deserve the claim of being capable of "vector processing". SIMD processors without per-element predication ( MMX , SSE , AltiVec ) categorically do not. Modern GPUs, which have many small compute units each with their own independent SIMD ALUs, use Single Instruction Multiple Threads (SIMT). SIMT units run from a shared single broadcast synchronised Instruction Unit. The "vector registers" are very wide and

2156-402: Is very costly in hardware (256 bit data paths to memory). Having 4x 64-bit ALUs, especially MULTIPLY, likewise. To avoid these high costs, a SIMD processor would have to have 1-wide 64-bit LOAD, 1-wide 64-bit STORE, and only 2-wide 64-bit ALUs. As shown in the diagram, which assumes a multi-issue execution model , the consequences are that the operations now take longer to complete. If multi-issue

2233-459: The CDC 7600 , but at data-related tasks they could keep up while being much smaller and less expensive. However the machine also took considerable time decoding the vector instructions and getting ready to run the process, so it required very specific data sets to work on before it actually sped anything up. The vector technique was first fully exploited in 1976 by the famous Cray-1 . Instead of leaving

2310-585: The Videocore IV ISA for a REP field, but unlike the STAR-100 which uses memory for its repeats, the Videocore IV repeats are on all operations including arithmetic vector operations. The repeat length can be a small range of power of two or sourced from one of the scalar registers. The Cray-1 introduced the idea of using processor registers to hold vector data in batches. The batch lengths (vector length, VL) could be dynamically set with

2387-541: The price-to-performance ratio of conventional microprocessor designs led to a decline in vector supercomputers during the 1990s. Vector processing development began in the early 1960s at the Westinghouse Electric Corporation in their Solomon project. Solomon's goal was to dramatically increase math performance by using a large number of simple coprocessors under the control of a single master Central processing unit (CPU). The CPU fed

Cray EL90 - Misplaced Pages Continue

2464-746: The 60-bit words, but a 30-bit instruction could not straddle two words, and control could only be transferred to the first instruction in a word. However, the instruction set itself had changed to reflect the new internal memory layout, thereby rendering it incompatible with the earlier 6600. The machines were similar enough to make porting of compilers and operating systems possible without too much trouble. The machine initially did not come with software; sites had to be willing to write their own operating system, like LTSS, NCAROS, and others; and compilers like LRLTRAN (Livermore's version of Fortran with dynamic memory management and other non-standard features). CDC also manufactured two multi-processor computers based on

2541-459: The 6600 and 7600 left mundane housekeeping tasks, printing output or reading punched cards , for instance, to a series of ten smaller 12-bit machines based on the CDC 160-A known as "Peripheral Processor Units", or PPUs. For any given cycle of the machine one of the PPUs was in control, feeding data into the memory while the main processor was crunching numbers. When the cycle completed, the next PPU

2618-403: The 6600, but to provide a substantial increase in performance". One user said: "Most users could run on either system without changes." Although the 7600 shared many features of the 6600, including hardware, instructions, and its 60-bit word size, it was not object-code compatible with the CDC 6600. In addition, it was not entirely source-code ( COMPASS ) compatible, as some instructions in

2695-427: The 6600. At the time computer memory could be arranged in blocks with independent access paths, and Cray's designs used this to their advantage. While most machines would use a single CPU to run all the functionality of the system, Cray realized that this meant each memory block spent a considerable amount of time idle while the CPU was processing instructions and accessing other blocks. In order to take advantage of this,

2772-399: The 7600 did not exist in the 6600, and vice versa. It had originally been named the CDC 6800, but was changed to 7600 when Cray decided that it could not be completely compatible. However, due to the 7600's operating system design, the 6600 and 7600 shared a "uniform software environment" despite the low-level differences. In fact, from a high-level perspective, the 7600 was quite similar to

2849-407: The 7600's notable state-of-the-art contributions, beyond extensive pipelining , was the physical C-shape, which both reduced floor space and dramatically increased performance by reducing the distance that signals needed to travel. As the 6600 neared production quality, Cray lost interest in it and turned to designing its replacement. Making a machine "somewhat" faster would not be too difficult in

2926-665: The 7600, with the model number 7700. They consisted of two 7600 machines in an asymmetric configuration: a central and an adjunct machine. They were used for missile launch and inbound tracking of USSR ICBMs . The radar simulator was a real-time simulator with a CDC 6400 for input/output front-end. These systems were to be used in the Pacific Missile Range . One computer was installed at TRW in Redondo Beach CA (later moved to Kwajalein Atoll, South Pacific), and

3003-427: The CPU, in the fashion of an assembly line , so the address decoder is constantly in use. Any particular instruction takes the same amount of time to complete, a time known as the latency , but the CPU can process an entire batch of operations, in an overlapping fashion, much faster and more efficiently than if it did so one at a time. Vector processors take this concept one step further. Instead of pipelining just

3080-664: The CPU, this would look something like this: But to a vector processor, this task looks considerably different: Note the complete lack of looping in the instructions, because it is the hardware which has performed 10 sequential operations: effectively the loop count is on an explicit per-instruction basis. Cray-style vector ISAs take this a step further and provide a global "count" register, called vector length (VL): There are several savings inherent in this approach. Additionally, in more modern vector processor ISAs, "Fail on First" or "Fault First" has been introduced (see below) which brings even more advantages. But more than that,

3157-669: The ILLIAC and DAP as cellular array processors that potentially offered substantial performance benefits over conventional vector processor designs such as the CDC STAR-100 and Cray 1. A computer for operations with functions was presented and developed by Kartsev in 1967. The first vector supercomputers are the Control Data Corporation STAR-100 and Texas Instruments Advanced Scientific Computer (ASC), which were introduced in 1974 and 1972, respectively. The basic ASC (i.e., "one pipe") ALU used

Cray EL90 - Misplaced Pages Continue

3234-407: The STAR-100's vectorisation was by design based around memory accesses, an extra slot of memory is now required to process the information. Two times the latency is also needed due to the extra requirement of memory access. A modern packed SIMD architecture, known by many names (listed in Flynn's taxonomy ), can do most of the operation in batches. The code is mostly similar to the scalar version. It

3311-504: The above action would be described in a single instruction (somewhat like vadd c, a, b, $ 10 ). They are also found in the x86 architecture as the REP prefix. However, only very simple calculations can be done effectively in hardware this way without a very large cost increase. Since all operands have to be in memory for the STAR-100 architecture, the latency caused by access became huge too. Broadcom included space in all vector operations of

3388-787: The addition of SIMD cannot, by itself, qualify a processor as an actual vector processor , because SIMD is fixed-length , and vectors are variable-length . The difference is illustrated below with examples, showing and comparing the three categories: Pure SIMD, Predicated SIMD, and Pure Vector Processing. Other CPU designs include some multiple instructions for vector processing on multiple (vectorized) data sets, typically known as MIMD (Multiple Instruction, Multiple Data) and realized with VLIW (Very Long Instruction Word) and EPIC (Explicitly Parallel Instruction Computing). The Fujitsu FR-V VLIW/vector processor combines both technologies. SIMD instruction sets lack crucial features when compared to vector instruction sets. The most important of these

3465-415: The amount of time consumed by these steps, most modern CPUs use a technique known as instruction pipelining in which the instructions pass through several sub-units in turn. The first sub-unit reads the address and decodes it, the next "fetches" the values at those addresses, and the next does the math itself. With pipelining the "trick" is to start decoding the next instruction even before the first has left

3542-436: The concept of an instruction pipeline . Each functional unit consisted of several sections that operated in turn, for instance, an addition unit might have circuitry dedicated to retrieving the operands from memory, then the actual math unit, and finally another to send the results back to memory. At any given instance only one part of the unit was active, while the rest waited their turn. A pipeline improves on this by feeding in

3619-417: The core of the machine. Since this system was mechanical, and therefore prone to failure, the 7600 was redesigned into a large "C" shape to allow access to the modules on either side of the cooling piping by walking into the inside of the "C" and opening the cabinet. The 7600 was an architectural landmark, and most of its features are still standard parts of computer design. It is a load-store computer with

3696-460: The data in memory like the STAR-100 and ASC, the Cray design had eight vector registers , which held sixty-four 64-bit words each. The vector instructions were applied between registers, which is much faster than talking to main memory. Whereas the STAR-100 would apply a single operation across a long vector in memory and then move on to the next operation, the Cray design would load a smaller section of

3773-431: The decoding of the more common instructions such as normal adding. ( This can be somewhat mitigated by keeping the entire ISA to RISC principles: RVV only adds around 190 vector instructions even with the advanced features. ) Vector processors were traditionally designed to work best only when there are large amounts of data to be worked on. For this reason, these sorts of CPUs were found primarily in supercomputers , as

3850-431: The design originally called for a 1 GFLOPS machine with 256 ALUs, but, when it was finally delivered in 1972, it had only 64 ALUs and could reach only 100 to 150 MFLOPS. Nevertheless, it showed that the basic concept was sound, and, when used on data-intensive applications, such as computational fluid dynamics , the ILLIAC was the fastest machine in the world. The ILLIAC approach of using separate ALUs for each data element

3927-482: The difference between a traditional vector processor and a modern SIMD one. The example starts with a 32-bit integer variant of the "DAXPY" function, in C : In each iteration, every element of y has an element of x multiplied by a and added to it. The program is expressed in scalar linear form for readability. The scalar version of this would load one of each of x and y, process one calculation, store one result, and loop: The STAR-like code remains concise, but because

SECTION 50

#1732787715670

4004-635: The early vector processors, and is being implemented in commercial products such as the Andes Technology AX45MPV. There are also several open source vector processor architectures being developed, including ForwardCom and Libre-SOC . As of 2016 most commodity CPUs implement architectures that feature fixed-length SIMD instructions. On first inspection these can be considered a form of vector processing because they operate on multiple (vectorized, explicit length) data sets, and borrow features from vector processors. However, by definition,

4081-418: The entire group of results. In general terms, CPUs are able to manipulate one or two pieces of data at a time. For instance, most CPUs have an instruction that essentially says "add A to B and put the result in C". The data for A, B and C could be—in theory at least—encoded directly into the instruction. However, in efficient implementation things are rarely that simple. The data is rarely sent in raw form, and

4158-417: The fetch of the target instruction of a conditional branch precedes evaluation of the branch condition. During the execution of a 10-word (up to 40 instruction) loop, all the needed instructions remain in the stack, so no instructions are fetched, leaving small core memory free for data transfers. There are eight 60-bit registers, each with an address register. Moving an address to an address register starts

4235-402: The goal of producing a machine with ten times the performance. One of the reasons the 6600 was so much faster than its contemporaries is that it had multiple functional units that could operate in parallel. For instance, the machine could perform an addition of two numbers while simultaneously multiplying two others. However, any given instruction had to complete its trip through the unit before

4312-447: The instruction itself that the instruction will operate again on another item of data, at an address one increment larger than the last. This allows for significant savings in decoding time. To illustrate what a difference this can make, consider the simple task of adding two groups of 10 numbers together. In a normal programming language one would write a "loop" that picked up each of the pairs of numbers in turn, and then added them. To

4389-416: The instructions, they also pipeline the data itself. The processor is fed instructions that say not just to add A to B, but to add all of the numbers "from here to here" to all of the numbers "from there to there". Instead of constantly having to decode instructions and then fetch the data needed to complete them, the processor reads a single instruction from memory, and it is simply implied in the definition of

4466-401: The late 1960s; the introduction of integrated circuits allowed denser packing of components and, in turn, a higher clock speed. Transistors in general were also getting somewhat faster as the production processes and quality improved. These sorts of improvements might be expected to make a machine twice as fast, perhaps as much as five times. However, as with the 6600 design, Cray set himself

4543-406: The machine as a whole. The improvement in performance generally depends on the number of steps the unit takes to complete. For instance, the 6600's multiply unit took 10 cycles to complete an instruction, so by pipelining the units it could be expected to gain about 10 times the speed. Things are never that simple, however. Pipelining requires that the unit's internals can be effectively separated to

4620-438: The machine would break down at least once a day, and often four or five times. Acceptance at installation sites took years while the bugs were worked out, and while the machine generally sold well enough given its "high end" niche, it is unlikely the machine generated any sort of real profits for CDC. The successor CDC 8600 was never completed, and Seymour Cray went on to form his own company, Cray Research . One surviving 7600

4697-467: The memory load and store speed correspondingly had to increase as well. This is sometimes claimed to be a disadvantage of Cray-style vector processors: in reality it is part of achieving high performance throughput, as seen in GPUs , which face exactly the same issue. Modern SIMD computers claim to improve on early Cray by directly using multiple ALUs, for a higher degree of parallelism compared to only using

SECTION 60

#1732787715670

4774-451: The next could be fed into it, which caused a bottleneck when the scheduler system ran out of instructions. Adding more functional units would not improve performance unless the scheduler was also greatly improved, especially in terms of allowing it to have more memory, so it could look through more instructions for ones that could be fed into the parallel units. That appeared to be a major problem. In order to solve this problem, Cray turned to

4851-420: The next instruction before the first has completed, using up that idle time. For instance, while one instruction is being added together, the operands for the next add instruction can be fetched. That way, as soon as the current instruction completes and moves to the output circuitry, the operands for the next addition are already waiting to be added. In this way each functional unit works in "parallel", as well as

4928-536: The normal scalar pipeline. Modern vector processors (such as the SX-Aurora TSUBASA ) combine both, by issuing multiple data to multiple internal pipelined SIMD ALUs, the number issued being dynamically chosen by the vector program at runtime. Masks can be used to selectively load and store data in memory locations, and use those same masks to selectively disable processing element of SIMD ALUs. Some processors with SIMD ( AVX-512 , ARM SVE2 ) are capable of this kind of selective, per-element ( "predicated" ) processing, and it

5005-514: The performance leader, continually beating the competition with a series of machines that led to the Cray-2 , Cray X-MP and Cray Y-MP . Since then, the supercomputer market has focused much more on massively parallel processing rather than better implementations of vector processors. However, recognising the benefits of vector processing, IBM developed Virtual Vector Architecture for use in supercomputers coupling several scalar processors to act as

5082-546: The pipelines tend to be long. The "threading" part of SIMT involves the way data is handled independently on each of the compute units. In addition, GPUs such as the Broadcom Videocore IV and other external vector processors like the NEC SX-Aurora TSUBASA may use fewer vector units than the width implies: instead of having 64 units for a 64-number-wide register, the hardware might instead do

5159-425: The point where each step of the operation is running on completely separate circuitry. This is rarely achievable in the real world. Nevertheless, the use of pipelining on the 7600 improved performance over the 6600 by a factor of about 3. To achieve the rest of the goal, the machine would have to run at a faster speed, now possible using new transistor designs. However, there is a physical limit to performance because of

5236-447: The scalar argument across a SIMD register: CDC 7600 The CDC 7600 was designed by Seymour Cray to be the successor to the CDC 6600 , extending Control Data 's dominance of the supercomputer field into the 1970s. The 7600 ran at 36.4 MHz (27.5 ns clock cycle) and had a 65 Kword primary memory (with a 60-bit word size) using magnetic core and variable-size (up to 512 Kword) secondary memory (depending on site). It

5313-480: The second one was installed at McDonnell Douglas in Huntington Beach, California . They were actual 7600s connected by chassis 25 to make them a 7600 MP. From about 1969 to 1975, the CDC 7600 was generally regarded as the fastest computer in the world, except for specialized units. However, even with the advanced mechanicals and cooling, the 7600 was prone to failure. Both LLNL and NCAR reported that

5390-461: The supercomputers themselves were, in general, found in places such as weather prediction centers and physics labs, where huge amounts of data are "crunched". However, as shown above and demonstrated by RISC-V RVV the efficiency of vector ISAs brings other benefits which are compelling even for Embedded use-cases. The vector pseudocode example above comes with a big assumption that the vector computer can process more than ten numbers in one batch. For

5467-436: The supercomputing field entirely. In the early and mid-1980s Japanese companies ( Fujitsu , Hitachi and Nippon Electric Corporation (NEC) introduced register-based vector machines similar to the Cray-1, typically being slightly faster and much smaller. Oregon -based Floating Point Systems (FPS) built add-on array processors for minicomputers , later building their own minisupercomputers . Throughout, Cray continued to be

5544-496: The system's serial console . All EL90 models could be powered from regular mains power. The EL90 series was superseded by the Cray J90 series. This supercomputer-related article is a stub . You can help Misplaced Pages by expanding it . Vector processor Vector machines appeared in the early 1970s and dominated supercomputer design through the 1970s into the 1990s, notably the various Cray platforms. The rapid fall in

5621-473: The time it takes signals to move between parts of the machine, which in turn is defined by its physical size. As always, Cray's design work spent considerable effort on this problem and thus allow higher operating frequencies. For the 7600, each circuit module actually consisted of up to six printed circuit boards , each one stuffed with subminiature resistors, diodes, and transistors. The six boards were stacked up and then interconnected along their edges, making

5698-414: The time required to fetch the data from memory. Not all problems can be attacked with this sort of solution. Including these types of instructions necessarily adds complexity to the core CPU. That complexity typically makes other instructions run slower—i.e., whenever it is not adding up many numbers in a row. The more complex instructions also add to the complexity of the decoders, which might slow down

5775-450: The vector into registers and then apply as many operations as it could to that data, thereby avoiding many of the much slower memory access operations. The Cray design used pipeline parallelism to implement vector instructions rather than multiple ALUs. In addition, the design had completely separate pipelines for different instructions, for example, addition/subtraction was implemented in different hardware than multiplication. This allowed

5852-498: Was generally about ten times as fast as the CDC 6600 and could deliver about 10 MFLOPS on hand-compiled code, with a peak of 36 MFLOPS. In addition, in benchmark tests in early 1970 it was shown to be slightly faster than its IBM rival, the IBM System/360 , Model 195. When the system was released in 1967, it sold for around $ 5 million in base configurations, and considerably more as options and features were added. Among

5929-529: Was given control. In this way the memory always held up-to-date information for the main processor to work on (barring delays in the external devices themselves), eliminating delays on data, as well as allowing the CPU to be built for mathematical performance and nothing else. The PPU could have been called a very smart "communications channel". Like the 6600, the 7600 used 60-bit words with instructions that were generally 15 bits in length, although there were also 30-bit instructions. The instructions were packed into

#669330