Abstract: A parallel array processor for massively parallel applications is formed with low power CMOS with DRAM processing while incorporating processing elements on a single chip. Eight processors on a single chip have their own associated processing element, significant memory, and I/O and are interconnected with a hypercube based, but modified, topology. These nodes are then interconnected, either by a hypercube, modified hypercube, or ring, or ring within ring network topology. Conventional microprocessor MMPs consume pins and time going to memory. The new architecture merges processor and memory with multiple PMEs (eight 16 bit processors with 32K and I/O) in DRAM and has no memory access delays and uses all the pins for networking. The chip can be a single node of a fine-grained parallel processor. Each chip will have eight 16 bit processors, each processor providing 5 MIPs performance. I/O has three internal ports and one external port shared by the plural processors on the chip.

Abstract: A parallel array processor for massively parallel applications is formed with low power CMOS with DRAM processing while incorporating processing elements on a single chip. Eight processor memory elements on a single chip have their own associated processing element, significant memory, and I/O and are interconnected with a hypercube based, but modified, topology. These nodes are then interconnected, either by a hypercube, modified hypercube, or ring, or ring within ring network topology. The architecture uses all the pins for networking. Each chip has eight 16 bit processors, and eight respective 32K memories. I/O has three internal ports and one external port shared by the plural processors on the chip. Significant software flexibility is provided to enable quick implementation of existing programs written in common languages. The scalable chip has internal and external connections for broadcast and asynchronous SIMD, MIMD and SIMIMD (SIMD/MIMD) with dynamic switching of modes.

Abstract: A parallel array processor for massively parallel applications is formed with low power CMOS with DRAM processing while incorporating processing elements on a single chip. Eight processors on a single chip have their own associated processing element, significant memory, and I/O and are interconnected with a hypercube based, but modified, topology. These nodes are then interconnected, either by a hypercube, modified hypercube, or ring, or ring within ring network topology. Conventional microprocessor MMPs consume pins and time going to memory. The new architecture merges processor and memory with multiple PMEs (eight 16 bit processors with 32K and I/O) in DRAM and has no memory access delays and uses all the pins for networking. Each chip will have eight 16 bit processors, each processor providing 5 MIPs performance. I/O has three internal ports and one external port shared by the plural processors on the chip.

Abstract: A floating point system and method according to a format that includes a sign bit, an exponent part having a plurality of bits, and a fraction part having a plurality of multi-bit blocks, wherein floating point operation is based on block shifts of the fraction part, with each shift of one block associated with an increment or decrement of the exponent part by one count. This format illustrated is implemented as a format suitable for the accuracy greater than the IEEE 32-bit floating-point format, and is intended to be implemented in machines having byte-wide (8 bit) data streams. The preferred format consists of a sign bit, 7 exponent bits and 4 fraction bytes of eight bits for a total of 40 bits. This format and implementation allows floating-point commands to be executed in a fixed small number of cycles, thus advancing the capabilities of doing floating-point arithmetic on a SIMD machine.

Abstract: A parallel array processor for massively parallel applications is formed with low power CMOS with DRAWM processing while incorporating processing elements on a single chip, with nodes connected in an n-dimensional modified non-binary hypercube. In a 4-dimensional modified non-binary hypercube embodiment, each node includes either processor memory elements on a single chip, each processor memory element having its own associated processing element, significant memory, and I/O, with each processor memory element supporting an external port. Pairs of ports are associated with each dimension, labeled X, Y, W, and Z. Eight nodes are connected in the X dimension to form a ring. Corresponding nodes from eight such rings are connected into rings in the Y dimension to form an 8.times.8 array of nodes, referred to as a cluster. Corresponding nodes of eight clusters are connected into ring (64 rings) in the Z dimension, forming an 8.times.8.times.8 array of nodes referred to as a "cluster ring".

Abstract: In arrays of processors, especially linear arrays, it is important to be able to communicate to adjacent neighbors (en masse). That is, each element of the array can communicate with its neighbor on the left simultaneously. In addition, the array processor is provided with the ability for selected elements of the array, picket processing elements, to simultaneously communicate with other elements that are further away in one dimension than the nearest neighbor in one transfer cycle. This is accomplished by causing intermediate elements to become transparent in the communication paths, thus allowing data to "slide" through intermediate nodes to the destination node. This system can be used in the implementation of fault tolerance in the array of elements.

Abstract: A parallel array processor for massively parallel applications is formed with low power CMOS with DRAM processing while incorporating processing elements on a single chip. Eight processors on a single chip have their own associated processing element, significant memory, and I/O and are interconnected with a hypercube based, but modified, topology. These nodes are then interconnected, either by a hypercube, modified hypercube, or ring, or ring within ring network topology. Conventional microprocessor MMPs consume pins and time going to memory. The new architecture merges processor and memory with multiple PMEs (eight 16 bit processors with 32K and I/O) in DRAM and has no memory access delays and uses all the pins for networking. The chip can be a single node of a fine-grained parallel processor. Each chip will have eight 16 bit processors, each processor providing 5 MIPs performance. I/O has three internal ports and one external port shared by the plural processors on the chip.

Abstract: A system implementing a methodology for determining the exponent in parallel with determining the fractional shift during normalization according to partitioning the exponent into partial exponent groups according to the fractional shift data flow, determining all possible partial exponent values for each partial exponent group according to the fractional data flow, and providing the exponent by selectively combining possible partial exponents from each partial exponent group according to the fractional data flow. There is also provided a system implementing a methodology for generating the sticky bit during normalization. Sticky bit information is precalculated and multiplexed according to the fractional dataflow. In an embodiment of the invention, group sticky signals are calculated in tree form, each group sticky having a number of possible sticky bits corresponding to the shift increment amount of the multiplexing.

Abstract: A system implementing a methodology for determining the exponent in parallel with determining the fractional shift during normalization according to partitioning the exponent into partial exponent groups according to the fractional shift data flow, determining all possible partial exponent values for each partial exponent group according to the fractional data flow, and providing the exponent by selectively combining possible partial exponents from each partial exponent group according to the fractional data flow. There is also provided a system implementing a methodology for generating the sticky bit during normalization. Sticky bit information is precalculated and multiplexed according to the fractional dataflow. In an embodiment of the invention, group sticky signals are calculated in tree form, each group sticky having a number of possible sticky bits corresponding to the shift increment amount of the multiplexing.

Abstract: A method and system which provides exactly rounded division and square root results for a designated rounding mode independently of a remainder, or equivalent calculation of the relationship between the remainder and zero, for predetermined combinations of the rounding mode and the least significant bit of an estimate that has one more bit of precision than the exactly rounded result, and has an error tolerance magnitude less than the weight of the least significant bit of the estimate. The estimate is generated in accordance with a quadratically converging division or square root algorithm. The method and system is described in connection with IEEE 754-1985 and IBM S/390 binary floating point architectures.

Abstract: A method and system which provides exactly rounded division and square root results for a designated rounding mode independently of a remainder, or equivalent calculation of the relationship between the remainder and zero, for predetermined combinations of the rounding mode and the least significant bit of an estimate that has one more bit of precision than the exactly rounded result, and has an error tolerance magnitude less than the weight of the least significant bit of the estimate. The estimate is generated in accordance with a quadratically converging division or square root algorithm. The method and system is described in connection with IEEE 754-1985 and IBM S/390 binary floating point architectures.