RAPPORT ARRAYPROCESSORER INNEH]LLSF\RTECKNING: INLEDNING....................................................SID OLIKA SPECIALARKITEKTURER F\R DATORER,TEKNIK, BUSSTRUKTURER MM...........................................................SID ALLM[NT ARRAYPROCESSORER.....................................SID ARRAYPRCESSOR F\R IBM PC/XT/AT...............................SID TEKNIK...............................................SID ANV[NDNINGSOMR]DEN...................................SID MARKNAD..............................................SID KONKURRENTER.........................................SID SLUTSATS.....................................................SID REFERENSER...................................................SID APPENDIX: ICKE-DETERMINISTISKA/STATISTISKA DATORER: INNEH]LLSF\RTECKNING 1.0 DISKUSSION: DETERMINISTISKA/ICKE-DETERMINISTISKA DATORER.SID 1 1.1 ALLM[NT..................................................SID 1 1.2 ANALYS AV ALGORITMKOMPLEXITET KONTRA DATORTYPER..........SID 2 1.3 MULTIPROCESSORARKITEKTURER: DC:er KONTRA NDC:er..........SID 4 1.4 NP-PROBLEMEN.............................................SID 7 1.5 VLSI-TILL[MPNINGAR F\R NDC:er............................SID 1.6 ANV[NDNINGSOMR]DEN F\R ICKE-DETERMINISTISKA DATORER......SID 1.7 SAMMARBETSPARTNER F\R NDC:er.............................SID 2.0 MARKNAD.................................................. 4.0 SLUTSATS.................................................SID 10 REFERENSER F\R NDC:er........................................SID !! KONKURRENTER .bl ____________ Olika f|retag inom segmentet arrayprocesser: ANALOGIC CORP Oms{ttning: Audubon Rd V{rddator: PDP 11, VAX, LSI 11, DG NOVA, Wakefield, MA 01880 ECLIPS, HP 1000 (617) 246-0300 Prestanda: 8.5 MFLOP/s Pris: $ 30.000-85.000 APTEC COMPUTER SYSTEMS Oms{ttning: 10180 SW Nimbus Ave, J-6 V{rddator: Portland, OR 97223 Prestanda: (503) 620-9840 Pris: COMPUTER DESIGN& Oms{ttning: APPLICATIONS Inc. V{rddator: PDP 11, VAX, LSI 11 411 Waverly Oaks Rd Prestanda: 5 MFLOP/s Waltham, MA 02154 Pris: $ 30.000- 85.000 (617) 647-1900 CSPI Oms{ttning: ca. $M20 40 Linnell Circle V{rddator: PDP 11, VAX, LSI 11 Billerica, MA 01821 Prestanda: 7 MFLOP/s (617) 272-6020 Pris: $21.000- 30.000 FLOATING POINT SYSTEMS Oms{ttning: $M125 (varav arrayp. c:a$M50) Box 23489 V{rddator: PDP 11, VAX Portland, OR 97223 Prestanda: 12 MFLOP/s (503) 641-3151 Pris: $75.000 NUMERIX CORP Oms{ttning: 320 Needham St V{rddator: PDP 11, VAX, LSI 11, MULTIBUS Newton, MA 02161 Prestanda: 30 MFLOP/s (617) 964-2500 Pris: $ 60.000 - 85.000 SKY COMPUTERS Oms{ttning: Foot of John St V{rddator: 8086, 68000, LSI 11, MULTIBUS, VERSABUS Lowell, MA 01852 Prestanda: 1 MFLOP/s (617) 454-6200 Pris: $ 6.000 - 7.000 !! STAR TECHNOLOGIES INC Oms{ttning: 1 SW Columbia, Suite 1212 V{rddator: Portland, Or 97258 Prestanda: (503) 227-2052 Pris: SYSTOLIC SYSTEM Oms{ttning: 1550-LA Pradera Drive V{rddator: IBM PC Campbell, CA 95008 Prestanda: 9,2 MFLOP/s Richard Travassos Pris: $ 20.000 408-866 5069 Totalt {r arrayprocessormarknaden v{rd c:a $M100 1984. Antalet s}lda arrayprocessorer f|r stordatorapplikationer {r c:a 8.000- 10.000 st (Floating Point Systems har s}lt c:a 5.000 p} 10 }r). Totalt ber{knas arrayprocessormarknaden oms{tta c:a $131 M 1985 (dvs. ca. 1.000 Mkr). !! REFERENSER .bl __________ A: F\RTECKNING \VER MATERIAL FR]N KONKURRENTERNA B: F\RTECKNING \VER ALLM[N TEKNISKINFORMATION C: F\RTECKNING \VER MATERIAL ANG]ENDE IBM-AT D: F\RTECKNING \VER REFERENSER ANG]ENDE MATEMATISKA ALGORITMER E: F\RTECKNING \VER B\CKER MM !! A F\RTECKNING \VER INSAMLAT MATERIAL FR]N KONKURENTER .bl _____________________________________________________ Efter brevf|rfr}gan om material ang}ende arrayprocessorer till 9 olika tillverkare, s} svarade f|ljande 7 f|retag med nedan- st}ende material. SAMLINGSH]LLARE: A01 FLOTING POINT SYSTEM .bl ____________________ 1) Brochyr: Array Processors 2) Brochyr: FPS-500 Series 3) Brochyr: FPS-164 Family 4) Brochyr: FPS-164 Scientific Computer Sprcification 5) Brochyr: FPS-164 Applications Software Brif 6) Brochyr: FPS-164/Max Scientific Computer 7) Brochyr: Software Libraries 64-bits Products 8) Brochyr: Software Libraries 38-bits Products 9) Manual: Power Spectrum analysis with the FPS 5000 Series 10) Manual: NMR-Image Reconstruction with the FPS 500 11) Manual: Multi-band Image Classification with a distributed architectur 12) Manual: FPS 5000 Served array processor specifications 13) Manual: Adistributed architectur for high througput array processor ANALOGIC .bl ________ 1) Brochyr: Series AP 400 2) Brochyr: AP 445 3) Brochyr: AP 500 Series 4) Brochyr: A new family of array processors !! NUMERIX .bl _______ 1) Brochyr: Mars 432 STAR TECHNOLOGIES INC .bl _____________________ 1) Brochyr: ST-100 2) Manual: ST-100 CSPI .bl ____ 1) Brochyr: The Min-Map Arry Processor 2) Brochyr: Min-Map software specification Fotran 3) Brochyr: The Map-6410 array processor 4) Brochyr: The Map-6410 and FEM 5) Brochyr: MAP-6420 Specifications 6) Brochyr: High-Speed array processor math 7) Brochyr: Min-Map 111-211 8) Brochyr: Min-Map MM111-PL4, MM211-PL4 9) Brochyr: Min-Map Input/Output Module SKY, INC .bl ________ 1) Brochyr: Array Processor SYTOLIC SYSTEMS .bl _______________ 1) Brochyr + Prislista SAMMLINGSH]LLARE E01: .bl ____________________ Word Slice product family: Analog Devices Ytmontering: Tre st. tema nummer i Modern Elektronik !! B F\RTECKNING \VER ALLM[N TEKNISKINFORMATION .bl _____________________________________________ P[RM B 01: .bl _________ Data flow supercomputers. Recursive machines and computering technology Data driven system for high speed parallel computering. MAUD:a dynamic single-assignment system A data flow multiprocessor. Principal components of a data flow computer. Small firm has big ideas for multiple microprocessors. Predicting thermal characteristics of ventilated electronic enclose. Applaying the hamming code to microprocessor-based system. Choose cabels with care to optimize system design. Singel error correction code miximizes memory system efficiency. Board applies codes to small-computer memory. RAM-based multiplexer systems can save big chunks of hardware. Multiport register file stremlines signal processing. Shared memory system. Asynchronous dual-port RAM simplifies multiprocessor systems. Functionally parallel architecture for array processors. Vi {r |verhopade av best{llningar (FEM-firma) Microprogrammable chips blend top performance with 32-bit structures. Bipolar arithmetic chip speeds 68000's mat throughtput. Arbiters, priority acess conflicts and "glitch" problem. Packing a signal processor onto a singel digital board. Dual-port RAM hikes throughput in input/output controler board. !! Innovative Architectual and software techniques result in low cost high performance array processor. Avancerad bildbehandling skiljer agnarna fr}n vetet. N{stan-superdator kostar bara en tiodel. Superdator marknaden m}ngdubblas p} 10 }r: Computer Sweden 850125 Handling real-time images comes naturally to systolic array chip. Associative memory calls on the talents of systolic array chip. Systolic array chip matches the pace of high-speed processing Carring out division by addition reducees hardware complexity: Computer Design june 1975 Radix conversion: Donald E. Knut Managing the flow of data is easy with programmable multipexer: Electronics/May 11, 1978 Bipolar arithmetic chip speeds 68000's math throuhput: Personal-computer engineering add-ons and add-ins. IBM n{r en hel indistri. Industriell Datateknik 1984:4 R{knar som f}. Industriell Datateknik 1984:6/7 !! P[RM B02: .bl ________ Giants in small packages: IEEE Spectrum feb 1982 Architectual and software issues in the design and application of peripheral array processors: IEEE sep 1981. Precompilation of fortran programs to faciliate array processing. Array Processors: A selected Bibliography A proposed standard for binary floating-point arithmetic Analysis of proposal for the floating-point standard Applications of the proposed IEE 754 standards for floating- point arithmetic Underflow and the denormalized numbers Programmable array processors crunch numbers effortlessly Arithmetric chips assume greater importance as uC users demand faster response C-MOS cip set stremlines floating-point processing Hardware multiplication techniques for microprocessors systems An introduction to vector processing LSI hardware implements signal processing algorithms Comparison of selected array processor architectures Array processors provides high throughput rates Intelegent memory holds the key to a limitless array of computers Multiplier-accumulator derives high performance from 1-um CMOS Dense static RAM speeds data access in memory-intesive systems Floting-point array processors evolve into tailorable systems CMOS signal processors push to highest througput of all A statistical study of the aqccuary of floating point number system On the precision attaniable with various floating-Point number system Roundings in floating-point arithmetic !! P[RM B03: .bl ________ Matrix computations on an associtiv processor. A content addressable distributed logic memory with applications to information retrieavel. Supercomputer arichitecture. Associativ and parallel processors. Vector extensions to LRLTAN. Associtve processors architecture- a survey. Solving some mathematical problems by associtve processing. TSE computers (optisk dator). Some computer organizations and their effectivness. Design motivations for multiple processor microcomputer systems. Interconection system approches for minimizing data transmission problems. Arithmetic processor chips enhance microprocessor system performans. Making minframe mathematics. Reducing system interconnection with multivaule logic. Hardware multiplication techniques for microprocessor systems. Dedicated multiplier ICs speed-up processing in fast computer systems. Medium-speed multipliers trim cost shrink bandwith in speech transmission. Multiplier/divider hardware design accelerates microprocessor throughput. Multiprocessor memory organization and memory interference. Multiprocessor interconnection. Analys of multiple-microprocessor system architectures. The three-dimensional computer: a multiplan array processor. Interprocess communication for multi-microcomputer systems. Parallel processors architectures- part: General purpose systems. Multi-microprocessors: An overview and working example. Predicting queue performance on a programmable handheld calculator. !! P[RM B04: .bl ________ Digital group selectors Synchronization N-channel asynchronous arbiter resolves resours allocation conflicts. System capabilities get a boost from a high-powered dedicated slave. Packing a signal processor onto a single digital board. Dual-port RAM hikes throughput in input/output controler board. Micro-control for parallel asynchronous computers Time-shared memory-processor interface. A large scale, homogeneous, fully didtributed parallel machine. A multiple instruction stream processor. Process coordination in multimicroprocessor systems. SPDM- a subprocessor with dynamic microprogramming. A structured multimicroprocessor system with deadlock free operation schema. Implement digital filters efficiently with a specally cofigurated computer. Hexadecimal signatures identify troubelspots in microprocessor systems. A computer capable of exchanging processors for time. A new hierachical array processor system. Beskrivning av minnessystem 128x44. Boost bit-slice microsequencer speed: Choose from eight different circuits. Signal analysers (HP). Computer interconnection structures: Taxonomy, characteristics, and examples. Motorol's advanced computer system on silicon. Microprogram med hj{lp av PAL. An implementation guide to a proposed standard for floating-point arithmetic. !! P[RM B05: .bl ________ Digital filter design. Reducing roundoff noise in digital filters. First single-chip digital output correlator compares data at a 20-MHz rate. Interconnection network for SIMD machines. Multiprocessor organization- a survey. A multiprocessor approach to numerical analysis: An application to gaming problems. An approach to modular multicomputer systems. A hierarchically structured mult-microprocessor system. User-oriented turning in a microprogrammable computer. Poly processors system PPS-1. Parallel processing with the perfect shuffle. A shuffle-exchange network with simplified control. Developments and directions in computer architecture. Asynchronous control arrays. Macro E-nets for representation of paralell systems. Develop a multiple-instruction-stream single-chip processor. The possible impact of software trends on microcomputer architecture. Superpower computers. Integrated arithmectic processing unit enhances processor execution times. Numerical interpolation for microprocessor-based systems. Multiplying with a microcomputer. Tough mathematical tasks are child's play for number crunchers. Approach to unified bus architecture sidsteps inherent drawbacks. SMS 101- A structred multiprocessor system with deadlockfree operation scheme. !! Multi-microprocessors: An overview and working example. VLSI computering: A tough nut to crack. Computer processes multiple instruction sets, multiple data streams. Speech recognition by computer. GOP: Image processing 100 000 times faster. GOP: En skr{ddarsydd bildgehandlingsdator. GOP: A fast and flexible processor for image analysis. Image processing by computer. Associative processors: A panacea or a specific? Floating point adder VLSI projected for 1981. Unit multiplies two 64-bit numbers in 284 nanoseconds. Applications of vector processing. !! !! C F\RTECKNING \VER INSAMLAT MATERIAL ANG]ENDE IBM-AT .bl ____________________________________________________ Mikrodatorn bed|mer IBM-AT Board givs PC-XT the muscle of a supermini.... IBM-ATjr och IBM-GT kan v{ntas 1988 IBM's newcommer........ Bullet -286 Mother Board Tecmar makes PC's .....better Electonic design: Teleimport ab NS 16081-6 Floating-Point Unit. BETOMA annons Mitsubishhi CMOS-statiska RAM. Annons: Cypress Semi. (snabba PROM) NS CMOS 240-serien. Fujitsu: CMOS statiska RAM NCR: Geometric arithmetic parallel processor. NCR: CMOS 16x16 multiplier. NCR: Short-form catalog-1985. Brochyr: IBM-persondatorer. ALTERA: CMOS-PAL Annons IBM: Introductiong the IBM PC Engineering/Scientific Series. Mapp CYPRESS SEMICONDUCTOR Mapp ANALOG DEVICES !! E P[RMAR: .bl ________ Compiler design and assessment. Bakplan Digital reglerteknik AP 120 Floating Point Systems array processor: Manual MEM-p{rm: Halvledarminne EMI-p{rm Couse : Digital filters and spectral analysis. The finit element method The fast fourier transform.