Data processing apparatus and method for pre-fetching an instruction in to an instruction cache

• US 5,961,631 A
• Filed: 07/16/1997
• Issued: 10/05/1999
• Est. Priority Date: 07/16/1997
• Status: Expired due to Term

• Current Assignee: Arm Limited (SoftBank Group Corp.)
• Original Assignee: Arm Limited (SoftBank Group Corp.)
• Inventors: Salter, Nicholas Andrew, Devereux, Ian Victor
• Primary Examiner(s): Eng, David Y.
• Application Number:
  US08/893,982
• Time in Patent Office: 811 Days
• Field of Search: 711/118, 712/207, 712/225, 712/1
• US Class Current: 712/207
• CPC Class Codes: G06F 9/3802 Instruction prefetching
• CPC Class Codes: G06F 9/3877 using a slave processor, e.g. coprocessor peripheral processor G06F13/12; vector processor G06F15/8053

Abstract

The present invention provides a data processing apparatus for fetching an instruction in to an instruction cache, comprising an instruction cache for storing instructions, and a processor core for outputting an instruction address to the instruction cache on an instruction address bus, and for receiving the instruction corresponding to that instruction address on an instruction data bus The processor core is arranged to issue a predetermined control signal to the instruction cache when outputting the instruction address to cause the instruction cache to perform an instruction fetch procedure. In accordance with the present invention, a coprocessor is provided for executing a first predetermined instruction also executed by the processor core, the first predetermined instruction causing the coprocessor to issue the predetermined control signal to the instruction cache, and causing the processor core to output to the instruction cache the instruction address data for an instruction to be added to the instruction cache without issuing the predetermined control signal. Further, the instruction cache is responsive to the predetermined control signal from the coprocessor and the instruction address from the processor core to perform the instruction fetch procedure, and, if this results in a cache miss, to cause the retrieval of the instruction from memory for storage in the instruction cache, the processor core being arranged to ignore the retrieved instruction. The above approach allows an instruction to be pre-fetched and stored in the instruction cache without that instruction being returned to the processor core for execution, and without any additional logic needing to be added to the instruction cache to support pre-fetching.

Claims

1. A data processing apparatus for fetching an instruction in to an instruction cache, comprising:

an instruction cache for storing instructions;

a processor core arranged, when an instruction is required by the processor core, to output an instruction address to the instruction cache on an instruction address bus, and to issue a predetermined control signal to the instruction cache to cause the instruction cache to perform an instruction fetch procedure, the processor core arranged to receive the instruction corresponding to that instruction address on an instruction data bus only when said processor core issues said predetermined control signal;

a coprocessor arranged, when an instruction is to be added to the instruction cache without the instruction being received by the processor core, to execute a first predetermined instruction also executed by the processor core, the first predetermined instruction causing the coprocessor to also issue a predetermined control signal to the instruction cache, and causing the processor core to output to the instruction cache the instruction address data for the instruction to be added to the instruction cache without issuing the processor core originated predetermined control signal;

the instruction cache being responsive to the coprocessor originated predetermined control signal and the instruction address from the processor core to perform the instruction fetch procedure, and, if this results in a cache miss, to cause the retrieval of the instruction from memory for storage in the instruction cache, the processor core ignoring the retrieved instruction.

2. A data processing apparatus as claimed in claim 1, wherein the first predetermined instruction causes the processor core to output the instruction address data on the instruction address bus.

3. A data processing apparatus as claimed in claim 1, wherein the first predetermined instruction is a Move to Coprocessor Register (MCR) instruction.

4. A data processing apparatus as claimed in claim 1, wherein the coprocessor has a number of coprocessor registers whose stored data values are used to control the operation of the instruction cache, a second predetermined Move to Coprocessor Register (MCR) instruction being executable to cause the coprocessor to issue an instruction to one of said number of coprocessor registers to store a data value defining the cache line pointed to by the instruction cache pointer, and to cause the processor core to output that data value for storage in the coprocessor register.

5. A data processing apparatus as claimed in claim 4, further comprising:

a data cache for storing data values, the data cache being connected to the processor core by a data address bus and a data data bus;

the processor core being arranged to output data addresses to the data cache over the data address bus, and being arranged to receive over the data data bus data values corresponding to data addresses output by the processor core on the data address bus, or to output over the data data bus data values to the data cache;

the coprocessor registers being connected to the data data bus, or to the data address bus.

6. A data processing apparatus as claimed in claim 5, wherein the coprocessor register is connected to the data address bus, and the coprocessor is arranged to send a signal to the processor core to cause the processor core to output the data value on the data address bus.

7. A data processing apparatus as claimed in claim 6, wherein the coprocessor register is connected to the instruction address bus, and the coprocessor is arranged to send a signal to the processor core to cause the processor core to output the data value on the instruction address bus.

8. A data processing apparatus as claimed in claim 1, wherein the predetermined control signal is issued by the processor core over an instruction control bus connected to the instruction cache.

9. A data processing apparatus as claimed in claim 1, wherein, in the event of a cache miss, a whole cache line containing the instruction is retrieved from memory for storage in the cache.

10. A data processing apparatus as claimed in claim 4, wherein a third predetermined Move to Coprocessor Register (MCR) instruction is executable to cause the coprocessor to issue an instruction to one of said number of coprocessor registers to store a data value defining a number of cache lines which are to be locked down in the instruction cache to avoid subsequent overwriting of the instructions stored in those cache lines, and to cause the processor core to output that data value for storage in the coprocessor register.

11. A data processing apparatus as claimed in claim 10, further comprising:

a data cache for storing data values, the data cache being connected to the processor core by a data address bus and a data data bus;

the processor core being arranged to output data addresses to the data cache over the data address bus, and being arranged to receive over the data data bus data values corresponding to data addresses output by the processor core on the data address bus, or to output over the data data bus data values to the data cache;

the coprocessor registers being connected to the data data bus, or to the data address bus.

12. A data processing apparatus as claimed in claim 11, wherein the coprocessor register is connected to the data address bus, and the coprocessor is arranged to send a signal to the processor core to cause the processor core to output the data value on the data address bus.

13. A data processing apparatus as claimed in claim 12, wherein the coprocessor register is connected to the instruction address bus, and the coprocessor is arranged to send a signal to the processor core to cause the processor core to output the data value on the instruction address bus.

14. A method of operating a data processing apparatus to fetch an instruction in to an instruction cache, comprising the steps of:

outputting from a processor core an instruction address to the instruction cache on an instruction address bus;

outputting, when an instruction is required by the processor core, from the processor core a predetermined control signal to the instruction cache while outputting the instruction address to cause the instruction cache to perform an instruction fetch procedure;

receiving at the processor core on an instruction data bus the instruction corresponding to the instruction address only when the processor core has issued said predetermined control signal;

arranging, when an instruction is to be added to the instruction cache without the instruction being received by the processor core, a coprocessor and the processor core to execute a first predetermined instruction, the first predetermined instruction causing the coprocessor to issue a predetermined control signal to the instruction cache, and causing the processor core to output to the instruction cache the instruction address data for the instruction to be added to the instruction cache without issuing the processor core originated predetermined control signal;

arranging the instruction cache to perform the instruction fetch procedure in response to the predetermined control signal from the coprocessor and the instruction address from the processor core;

if the instruction fetch procedure results in a cache miss, retrieving the instruction from memory for storage in the instruction cache, the processor core being arranged to ignore the retrieved instruction.

15. A method as claimed in claim 14, wherein the first predetermined instruction causes the processor core to output the instruction address data on the instruction address bus.

16. A method as claimed in claim 14, wherein the first predetermined instruction is a Move to Coprocessor Register (MCR) instruction.