Patent EP0085435A2 successfully refreshed from Google Patents.
← Back to Patent Explorer
EP0085435A2 Expired / Lapsed EP ⏱️ 0.0 years remaining

Array processor comprised of vector processors using vector registers

Assignee: Hitachi Ltd
•
Filing: 1983-02-03
•
Granted: —
•
Expires: 2003-02-03
📄 Original PDF 🌐 Google Patents 🔄 Sync Now

Abstract

Abstract Herein disclosed is an array processor which is constructed of: a central vector processing unit (VPC) including a plurality of vector registers and a pipe-line control arithmetic and logical operation unit (ALU) and made operative to execute instruction (vector instruction) requiring processings of vectors; and a plurality of vector processing units (PE, - PE M ) including a plurality of vector registers (V1 - VK) and a pipe-line control (ALU) made operative to execute an instruction (array instruction) requiring processings of arrays. The central vector processing unit (VPC) fetches and decodes the vector instruction or the array instruction to execute the decoded instruction, when this instruction is the vector instruction but to start those vector processing units, when the decoded instruction is the array instruction. Each of the vector processing units (PE, - PE M ) executes that operation to one of plural vector data comprising array data to be an object of the operation designated by the decoded instruction, which is designated by the instruction for one of the vector data, when the decoded instruction is the array instruction. When the result of the operation designated by the array instruction is the vector data, each of the vector processing units computes and stores one element of those vector data in built-in scala registers, and sends out that element to the central vector processing unit (VPC) so that the element may be stored in vector registers in the central vector processing unit. When the result of the operation designated by the array instruction is the array data, each of the vector processing units (PE, - PE M ) computes and stores one vector data of those resultant array data in built-in vector registers (V1 - VK).

Inventors

  • 👤 Yoshikazu Tanaka
  • 👤 Shunichi Torii

Bibliographic Details

Application Number —
Jurisdiction EP
Data Source google_patents
Last Synchronized 2026-10-29 20:46
CPC Classifications
G; G06; G06F; G06F9/00; G06F9/06; G06F9/30; G06F9/38; G06F9/3885; G06F15/00; G06F15/16; G06F15/76; G06F15/80; G06F15/8007; G06F15/8015; G06F15/8053; G06F15/8092; G06F9/3887

Claims Analysis

Total Claims: 5
Claim 1
Independent Claim 314 words
An array processor characterized by: - a main storage unit (MS) for storing scalar data, vector data and array data and for separately storing scalar instruction series, vector- and array instruction series; - scalar instruction executing means (SP) for reading out said scalar instruction series and said scalar data from said main storage unit for decoding and executing said scalar instructions read out and for storing said scalar data obtained by said executions in said main storage unit; and - vector and array instruction executing means (VPC, PE) for reading out said vector instructions or array instruction series and said vector or array data of said main storage unit for decoding and executing said vector and array instructions read out and for storing said vector or array data obtained by said executions in said main storage unit, wherein said vector and array instruction executing means include: - main vector instruction executing means for executing vector instructions designating the operations of one-dimensional array data to obtain a one-dimensional array or scalar data and - a plurality of array instruction executing means (PE l - PE m ) for executing array instructions designating the operations of two-dimensional array data, each of said array instruction executive means (PE l - PE M ) being made operative to execute operations of one of a plurality of one-dimensional array data constructing said two-dimensional array data thereby to determine one element of one-dimensional array result data when said array instructions instruct that as result data said one-dimensional array result data are required and thereby to determine one of plural one-dimensional array result data constructing two-dimensional array result data, when said array instructions instruct that as result data said two-dimensional array result data are required, and - wherein said scalar instruction means include: - means for generating in response to said scalar instruction data necessary for vector and scalar instruction series.
Claim 2
Dependent Claim 314 words
An array processor characterized by: - a main storage unit (MS) for storing scalar data, vector data and array data and for separately storing scalar instruction series, vector- and array instruction series; - scalar instruction executing means (SP) for reading out said scalar instruction series and said scalar data from said main storage unit for decoding and executing said scalar instructions read out and for storing said scalar data obtained by said executions in said main storage unit; and - vector and array instruction executing means (VPC, PE) for reading out said vector instructions or array instruction series and said vector or array data of said main storage unit for decoding and executing said vector and array instructions read out and for storing said vector or array data obtained by said executions in said main storage unit, wherein said vector and array instruction executing means include: - main vector instruction executing means for executing vector instructions designating the operations of one-dimensional array data to obtain a one-dimensional array or scalar data and - a plurality of array instruction executing means (PE l - PE m ) for executing array instructions designating the operations of two-dimensional array data, each of said array instruction executive means (PE l - PE M ) being made operative to execute operations of one of a plurality of one-dimensional array data constructing said two-dimensional array data thereby to determine one element of one-dimensional array result data when said array instructions instruct that as result data said one-dimensional array result data are required and thereby to determine one of plural one-dimensional array result data constructing two-dimensional array result data, when said array instructions instruct that as result data said two-dimensional array result data are required, and - wherein said scalar instruction means include: - means for generating in response to said scalar instruction data necessary for vector and scalar instruction series.
Claim 3
Dependent Claim 160 words
An array processor according to claim 1, wherein said main vector instruction executing means include: - a plurality of first vector registers (VC1 - VCL), - a plurality of first scalar registers (SC1 - SCZ), - a first pipe-line operating means, - means for fetching said vector instructions of said array instructions, - means for decoding whether said fetched instructions are said vector instructions or said array instructions and for controlling said first vector registers, said first scalar registers and said first pipe-line operating means, when the decoded instructions are said vector instructions, thereby to execute said decoded vector instructions, - means for sending each element of said one-dimensional array data to corresponding one of said array instruction executing means when the computed results of said fetched vector instructions are said one-dimensional array data and - means for storing said element of said one-dimensional array data which has been computed by each of said plurality of array instruction executing means.
Claim 4
Dependent Claim 157 words
An array processor according to claim 1 and 2, characterized in that each of said array instruction processing means includes: - a plurality of second vector registers (V i l - V i K), - a plurality of second scalar registers (S i l - S . L), - a second pipe-line operating means, - means for controlling said second vector registers, said second scalar registers and said second pipe-line operating means, when said fetched instructions are said array instructions, thereby to execute said array instructions of the two-dimensional array data instructed by said fetched instructions which has been fetched for one of said one-dimensional array data, - means for sending one element of said one-dimensional array data to said main vector instruction executing means when the resultant data of said execution reduces to scalar data and - means for storing said one element sent out from said first means in one of said second scalar registers.
Claim 5
Dependent Claim 82 words
An array processor according to anyone of the preceeding claims, characterized in that each of said plural vector instruction executing means includes means for holding the vector length to be independently processed by each of said plural vector instruction execution means, the start address of the vector data of said main storage unit to be processed and the data increment of said vector data, all of which are to be transferred by said transfer means and for using them for controlling purposes.

References Cited (Backward Citations)

No citation records indexed.

USPTO Annuity & Maintenance Windows

Maintenance schedule is calculated for granted US utility patents.

Assigned Categories & Tags

No categories assigned to this patent yet.

Assign Category Tag