Digital computer system with cache controller coordinating both vector and scalar operations

• US 5,418,973 A
• Filed: 06/22/1992
• Issued: 05/23/1995
• Est. Priority Date: 06/22/1992
• Status: Expired due to Term

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Digital Equipment Corporation (HP Inc.)
• Inventors: Nangia, Era, Patwa, Nital, Ellis, James P., Wolrich, Gilbert M., Shah, Bhavin
• Primary Examiner(s): Lall, Parshotam S.
• Assistant Examiner(s): VU, VIET DUY
• Application Number:
  US07/902,149
• Time in Patent Office: 1,065 Days
• Field of Search: 395/375, 395/425, 395/800, 395/500
• US Class Current: 712/3
• CPC Class Codes: G06F 15/8069 using a cache
• CPC Class Codes: G06F 15/8092 Array of vector units
• CPC Class Codes: G06F 9/30036 Instructions to perform operations on packed data, e.g. vector, tile or matrix operations
• CPC Class Codes: G06F 9/30038 using a mask
• CPC Class Codes: G06F 9/3834 Maintaining memory consistency
• CPC Class Codes: G06F 9/3838 Dependency mechanisms, e.g. register scoreboarding

Abstract

A digital computer system includes a scalar CPU, a vector processor, and a shared cache memory. The scalar CPU has an execution unit, a memory management unit, and a cache controller unit. The execution unit generates load/store memory addresses for vector load/store instructions. The load/store addresses are translated by the memory management unit, and stored in a write buffer that is also used for buffering scalar write addresses and write data. The cache controller coordinates-loads and stores between the vector processor and the shared cache with scalar reads and writes to the cache. Preferably the cache controller permits scalar reads to precede scalar writes and vector load/stores by checking for conflicts with scalar writes and vector load/stores in the write queue, and also permits vector load/stores to precede vector operates by checking for conflicts with vector operate information stored in a vector register scoreboard. Preferably the cache controller includes vector logic which is responsive to vector information written in intra-processor registers by the execution unit. The vector logic keeps track of the vector length and blocks extra memory addresses generated by the execution unit for the vector elements. The vector logic also blocks the memory addresses of masked vector elements so that these addresses are not translated by the memory management unit.

Claims

1. A digital computer system comprising, in combination:

a scalar central processing unit;

a vector processor; and

a cache memory coupled to said scalar central processing unit for scalar reads and writes between said scalar central processing unit and said cache memory, and coupled to said vector processor for vector loads and stores between said vector processor and said cache memory;

wherein said scalar processing unit includes;

(a) an execution unit for executing scalar instructions and for issuing vector instructions to said vector processor,

(b) a memory management unit for translating virtual addresses of both said scalar instructions and said vector instructions to physical addresses, and

(c) a cache controller unit for coordinating data transfer between said scalar central processing unit and said cache memory; and

wherein said execution unit includes means for computing virtual memory addresses of vector elements specified by vector load and vector store instructions and sending the virtual memory addresses to said memory management unit for translation to physical addresses,

wherein said cache controller unit includes means for coordinating both;

(a) vector loads and stores between said vector processor and said cache memory, and

(b) scalar reads and writes between said scalar central processing unit and said cache memory; and

wherein said cache controller unit, said cache memory, and said vector processor are directly connected together by a common address bus and a common data bus;

wherein said cache controller unit includes a write queue for queuing both;

(a) scalar physical write addresses and scalar write data, and

(b) physical addresses of said vector elements of said vector store instructions; and

means for servicing said write queue by addressing said cache memory with said scalar physical write addresses in said write queue for scalar operations and by addressing said cache memory for vector operations with said physical addresses of said vector elements in said write queue.

2. The digital computer system as claimed in claim 1, wherein said write queue includes means for inserting physical addresses of said vector elements of vector load instructions as well as said vector store instructions in said write queue together with vector load and store commands.

3. The digital computer system as claimed in claim 1, wherein said cache controller unit includes arbitration means for granting priority to allow scalar reads to precede scalar writes and vector stores if, upon checking for conflicts with scalar writes and vector stores in said write queue, there is no conflict.

4. The digital computer system as claimed in claim 1, wherein said cache controller unit includes a vector scoreboard for storing information about vector operate instructions issued to said vector processor by said execution unit, and arbitration means for granting priority to allow vector loads and stores to precede vector operates if, upon checking for conflicts with said information about vector operate instructions, there is no conflict.

5. The digital computer system as claimed in claim 1, wherein said cache controller unit includes intra-processor registers written to by said execution unit, and vector logic means responsive to said intra-processor registers for counting vector elements having physical addresses translated by said memory management unit, and sending a termination signal to said execution unit when said count reaches a specified vector length.

6. The digital computer system as claimed in claim 1, wherein said cache controller unit includes intra-processor registers written to by said execution unit, and vector logic means responsive to said intra-processor registers and receipt of said physical addresses of vector elements from said memory management unit for generating a blocking signal for a next vector element which is masked by a specified vector mask, and said execution unit includes means for blocking a virtual address of said next vector element from being translated by said memory management unit.

7. The digital computer system as claimed in claim 1, wherein said cache controller unit includes vector logic means for receiving a physical address of a first vector element of a vector, computing physical addresses of successive vector elements, and sending a next page signal to said execution unit when the physical address of a next vector element crosses a page boundary.