摘要:
A vector processor having a vector register made up of elements of l.sub.2 -byte size for storing vector data made up of a plurality of elements read out from a main storage which has a plurality of storage areas and is capable of reading out data of l.sub.1 -byte size beginning from a specified address bound, and adapted to write vector data with an element size of m (l.sub.1 /m is an integer and l.sub.2 is larger or equal to m) into the vector register sequentially, read-out vector data from the vector register for computation by an arithmetic unit, and write the computational result into the vector register, wherein the processor writes elements of vector data read out from the main storage into separatte, specified locations of the vector register in an order required for subsequent operations.
摘要:
A vector processor comprises a main storage for storing scalar instruction chains and vector instruction chains for executing desired operations, and a scalar processing unit and a vector processing unit for separately fetching the scalar instruction chains and the vector instruction chains, decoding them and executing them so that the scalar processing and the vector processing are carried out in overlap.
摘要:
In an operation unit wherein a series of data is sequentially applied, a predetermined operation is performed in synchronism with the input data in a pipelined manner, and the predetermined operation is applied to an input data and the result of the predetermined operation for a preceding input data. There are provided a plurality of partial operation devices which respectively compute a plurality of different partial data of a result data to be obtained as a result of the predetermined operation, and when one of the partial data is obtained, the one partial data is immediately used for the operation for the subsequent input data. Consequently, the operation for the subsequent input data can be started before the operation for the remainder of the partial data of the preceding input data is completed.
摘要:
An approximate quotient-correcting circuit wherein an approximate quotient Q.sub.H, a divisor D, and the least significant bit of the fraction part of a dividend N are read out; the approximate quotient Q.sub.H and the divisor D are multiplied; it is decided whether the lower m digits of Q.sub.H .times.D are not all `0` and whether the m-th significant bit of Q.sub.H .times.D is coincident with the m-th significant bit of N; and when the result of the decision is positive, Q.sub.H -2.sup.-m is provided as a quotient.
摘要:
In a vector processor having vector registers, a vector buffer storage for temporarily storing vector data is arranged closer to the vector registers than to a main storage, and a vector buffer storage control including an identification storage for storing identification information of the vector data stored at storage locations of the buffer storage and a check circuit for checking if the vector data identification information is in the identificatgion storage is provided. The vector buffer storage control checks if the identification information of the vector data designated by a vector data fetch instruction for the main storage is in the indentification storage, and if it is in the identification storage, it fetches the vector data from the buffer storage and transfers it to the vector register, and if it is not in the identification storage, it instructs to fetch the vector data from the main storage, transfers the vector data fetched from the main storage to the vector register and stores it into the buffer storage.
摘要:
In a processor such as a vector processor in which a plurality of data are processed by one instruction and a plurality of instructions are parallely processed, apparatus is provided for storing, during an interruption of the program currently being executed, the instructions being executed in the conceptual order of appearance in the program of the instruction being executed, and the sequential count of the sets of data processed. The stored information is used to restart the execution of the interrupted program at the appropriate point.
摘要:
A unit operative in concurrence with a vector processing for beforehand sequentially generating page addresses containing vector data to be referred to thereafter and a unit for achieving a processing to determine whether or not a page fault occurs in a page in an address translation and responsive to an occurrence of a page fault in a page for executing processing to beforehand transfer the page to a main storage are provided. Even if a vector element existing in the page becomes necessary in the vector processing after the operation described above, another paging processing is not necessary because the page exists in the main storage.
摘要:
A data processing apparatus comprises a plurality of sub-systems each including at least one arithmetic unit, a plurality of registers, a first selector for receiving vector data and selectively outputting the input data to the registers, and a second selector for receiving the vector data from the registers and selectively outputting the input data to a plurality of output lines. The data output of the arithmetic unit in each sub-system is supplied to the first selector in the same sub-system and the first selector in another sub-system, and the arithmetic unit in each sub-system receives the output data from the second selector in the same sub-system. The data output from the second selector in at least one sub-system is supplied to a main storage unit, and the data output from the main storage unit is supplied to the first selector in at least one sub-system.
摘要:
A vector data processor includes a vector index register for consecutively and sequentially storing indirect address vectors, which may then be consecutively and sequentially read out from the vector index register to form addresses of data, thereby to execute the consecutive reading of the data from a main storage and the consecutive writing thereof into the main storage with an increased processing speed by generating addresses and storing data in overlapping operations.
摘要:
A data processing system having a plurality of FIFO memories and a plurality of ALUs and in which a FIFO memory may be selected to receive a set of data signals from an ALU and at the same time to be selected to provide a set of data signals to another ALU, with the result that the selected FIFO memory performs read and write operations concurrently and intermittently. Also, a set of data signals held by one of the FIFO memories may be transferred to a selected ALU for effecting a logical or arithmetic operation thereon, and the data signals representing the result of the logical or arithmetic operation thereon, and the data signals representing the result of the logical or arithmetic operation by the selected ALU may be transferred to another FIFO memory.