Abstract: A memory interface device (100) providing a fractional address interface between a data processor (104) and a memory system (102) and a method for retrieving intermediate data values from a memory system using fractional addressing. The device includes an address generator (108) for generating first and second memory addresses, the first memory address being less than or equal to a specified fractional address, the second memory address being greater than or equal to the fractional address. The device also includes a memory access unit (110) coupled to the address generator (108) for retrieving first and second data values from the memory system (102) at the first and second memory addresses, respectively. The device also includes a data access unit (112) for interpolating between the first and second data values and passing the interpolated value to the data processor (104).

Abstract: A method and apparatus for the elimination of prolog and epilog instructions in a vector processor. To eliminate the prolog, a functional unit of the vector processor has at least one input for receiving an input data value tagged with a data validity tag and an output for outputting an intermediate result tagged with a data validity tag. The data validity tags indicate the validity of the data. Before a loop is executed, the data validity tags are set to indicate that the associated data values are invalid. During execution of the loop body a functional unit checks the validity of input data. If all of the input data values are valid the functional operation is performed, the corresponding data validity tag set to indicate that the result is valid. If any of the input data values is invalid, the data validity tag of the result is set to indicate that the result is invalid. To eliminate the epilog, an iteration counter is associated with each sink unit of the vector processor.

Abstract: A system and method for calculating memory addresses in a partitioned memory in a processing system having a processing unit, input and output units, a program sequencer and an external interface. An address calculator includes a set of storage elements, such as registers, and an arithmetic unit for calculating a memory address of a vector element dependent upon values stored in the storage elements and the address of a previous vector element. The storage elements hold STRIDE, SKIP and SPAN values and optionally a TYPE value, relating to the spacing between elements in the same partition, the spacing between elements in the consecutive partitions, the number of elements in a partition and the size of a vector element, respectively.

Abstract: A method for scheduling a computation for execution on a computer with a number of interconnected functional units. The computation is representable by a data-flow graph with a number of nodes connected by edge. A loop-period of the computation is calculated (104) and the nodes are scheduled for throughput (106) by assigning an execution cycle and a functional unit to each node of the data-flow graph. The scheduling of flexible nodes is adjusted to minimize the number of interconnections required in each execution cycle (110). The edges of the data-flow graph are allocated (122) to one or more of the interconnections between functional units. The scheduling method may be used, for example, to optimize the interconnection fabric design for an ASIC or as part of a compiler for a re-configurable streaming vector processor.

Abstract: A re-configurable, streaming vector processor (100) is provided which includes a number of function units (102), each having one or more inputs for receiving data values and an output for providing a data value, a re-configurable interconnection switch (104) and a micro-sequencer (118). The re-configurable interconnection switch (104) includes one or more links, each link operable to couple an output of a function unit (102) to an input of a function unit (102) as directed by the micro-sequencer (118). The vector processor may also include one or more input-stream units (122) for retrieving data from memory. Each input-stream unit is directed by a host processor and has a defined interface (116) to the host processor. The vector processor also includes one or more output-stream units (124) for writing data to memory or to the host processor. The defined interface of the input-stream and output-stream units forms a first part of the programming model.

Abstract: A method for producing a formatted description of a computation representable by a data-flow graph and computer for performing a computation so described. A source instruction is generated for each input (502, 522) of the data-flow graph, a computational instruction is generated for each node (506, 510, 514 etc) of the data-flow graph, and a sink instruction is generated for each output (520, 540) of the data-flow graph. The computation instruction for a node includes a descriptor of the operation performed at the node and a descriptor of each instruction that produces an input to the node. The formatted description is a sequential instruction list (A, B, C, . . . , J, K, L, FIG. 2) comprising source instructions, computational instructions and sink instructions. Each instruction has an instruction identifier and the descriptor of each instruction that produces an input to the node is the instruction identifier.

Abstract: A memory interface device (100) providing a fractional address interface between a data processor (104) and a memory system (102) and a method for retrieving intermediate data values from a memory system using fractional addressing. The device includes an address generator (108) for generating first and second memory addresses, the first memory address being less than or equal to a specified fractional address, the second memory address being greater than or equal to the fractional address. The device also includes a memory access unit (110) coupled to the address generator (108) for retrieving first and second data values from the memory system (102) at the first and second memory addresses, respectively. The device also includes a data access unit (112) for interpolating between the first and second data values and passing the interpolated value to the data processor (104).

Abstract: An interconnection device (300) with a number of links (306, 308, 310, 312 and 314), each link having a number of link input ports (302), link output ports (304) and storage registers (316). An input selection switch (402) is coupled to a selected link input port to receive an input data token. The storage registers (316) may be used to store input data tokens. A storage access switch (404) is coupled to the input selection switch (402) and to the storage registers (316) and may be used to select the current input data token or a token from the storage registers as an output data token. An output selection switch (406) receives the output data token and provides it to a selected link output port. The interconnection device may, for example, be used to connect the inputs and outputs of the processing elements of a vector processor or digital signal processor.

Abstract: Fault remediation functions are embodied in a semiconductor structure in which high quality epitaxial layers of monocrystalline materials are made to overlie monocrystalline substrates such as large silicon wafers by forming a compliant substrate for growing the monocrystalline layers. An accommodating buffer layer comprises a layer of monocrystalline oxide spaced apart from a silicon wafer by an amorphous interface layer of silicon oxide. The amorphous interface layer dissipates strain and permits the growth of a high quality monocrystalline oxide accommodating buffer layer. The accommodating buffer layer is lattice matched to both the underlying silicon wafer and the overlying monocrystalline material layer. Any lattice mismatch between the accommodating buffer layer and the underlying silicon substrate is taken care of by the amorphous interface layer. Fault remediation is carried out in one instance by recognizing the presence of a fault and in another instance by providing fault correction.

Abstract: High quality epitaxial layers of monocrystalline materials can be grown overlying monocrystalline substrates such as large silicon wafers by forming a compliant substrate for growing the monocrystalline layers. An accommodating buffer layer comprises a layer of monocrystalline oxide spaced apart form a silicon wafer by an amorphous interface layer of silicon oxide. The amorphous interface layer dissipates strain and permits the growth of a high quality monocrystalline oxide accommodating buffer layer. The accommodating buffer layer is lattice matched to both the underlying silicon wafer and the overlying monocrystalline material layer. Any lattice mismatch between the accommodating buffer layer and the underlying silicon substrate is taken care of by the amorphous interface layer. In addition, formation of a compliant substrate may include utilizing surfactant enhanced epitaxy, epitaxial growth of single crystal silicon onto single crystal oxide, and epitaxial growth of Zintl phase materials.

Abstract: High quality epitaxial layers of monocrystalline materials can be grown overlying monocrystalline substrates such as large silicon wafers by forming a compliant substrate for growing the monocrystalline layers. An accommodating buffer layer comprises a layer of monocrystalline oxide spaced apart from a silicon wafer by an amorphous interface layer of silicon oxide. The amorphous interface layer dissipates strain and permits the growth of a high quality monocrystalline oxide accommodating buffer layer. The accommodating buffer layer is lattice matched to both the underlying silicon wafer and the overlying monocrystalline material layer. Any lattice mismatch between the accommodating buffer layer and the underlying silicon substrate is taken care of by the amorphous interface layer. In addition, formation of a compliant substrate may include utilizing surfactant enhanced epitaxy, epitaxial growth of single crystal silicon onto single crystal oxide, and epitaxial growth of Zintl phase materials.

Abstract: High quality epitaxial layers of monocrystalline materials can be grown overlying monocrystalline substrates such as large silicon wafers by forming a compliant substrate for growing the monocrystalline layers. An accommodating buffer layer comprises a layer of monocrystalline oxide spaced apart from a silicon wafer by an amorphous interface layer of silicon oxide. The amorphous interface layer dissipates strain and permits the growth of a high quality monocrystalline oxide accommodating buffer layer. The accommodating buffer layer is lattice matched to both the underlying silicon wafer and the overlying monocrystalline material layer. Any lattice mismatch between the accommodating buffer layer and the underlying silicon substrate is taken care of by the amorphous interface layer. In addition, formation of a compliant substrate may include utilizing surfactant enhanced epitaxy, epitaxial growth of single crystal silicon onto single crystal oxide, and epitaxial growth of Zintl phase materials.

Abstract: High quality epitaxial layers of monocrystalline materials can be grown overlying monocrystalline substrates such as large silicon wafers by forming a compliant substrate for growing the monocrystalline layers. One way to achieve the formation of a compliant substrate includes first growing an accommodating buffer layer on a silicon wafer. The accommodating buffer layer is a layer of monocrystalline oxide spaced apart from the silicon wafer by an amorphous interface layer of silicon oxide. The amorphous interface layer dissipates strain and permits the growth of a high quality monocrystalline oxide accommodating buffer layer. The accommodating buffer layer is lattice matched to both the underlying silicon wafer and the overlying monocrystalline material layer. Any lattice mismatch between the accommodating buffer layer and the underlying silicon substrate is taken care of by the amorphous interface layer.