Single cycle data movement between general purpose and floating-point registers

Patent metadata

Field	Value
Publication number	US-9304767-B2
Application number	US-47663609-A
Country	US
Kind code	B2
Filing date	Jun 2, 2009
Priority date	Jun 2, 2009
Publication date	Apr 5, 2016
Grant date	Apr 5, 2016

Abstract

Systems and methods for providing single cycle movement of data between a floating-point register file (FRF) and a general purpose or integer register file (IRF) of a microprocessor system are provided. The system may include an integer execution unit operative to execute instructions with single cycle latency, a floating-point execution unit, a working register file (WRF), an FRF, and an IRF. To achieve the single cycle movement functionality, the integer execution unit may physically own the WRF, IRF, and FRF, and may monitor and control any dependencies between them. Thus, since the integer execution unit has direct read access to both the IRF and the FRF, data may be moved between the two register files using the single cycle operation of the integer execution unit, without the need to store and load the data from memory.

First claim

We claim:

1. A processor comprising:
an integer unit comprising:
an integer execution unit that is operative to execute an instruction with single cycle latency;
an integer register file (IRF) coupled to the integer execution unit; and
a floating-point register file (FRF) coupled to the integer execution unit;

a floating-point execution unit that is operative to execute one or more types of instructions with multiple cycle latency; and
a data bus that couples the floating-point execution unit to the integer unit;
wherein the integer unit is operative to execute respective move instructions to move data from a register of the IRF to a register of the FRF and to move data from a register of the FRF to a register of the IRF with single cycle latencies using the integer execution unit.

2. The processor of claim 1 , wherein the IRF and FRF are implemented physically near the integer execution unit such that the integer execution unit executes instructions having one or more operands located in the IRF or the FRF with single cycle latency.

3. The processor of claim 1 , further comprising:
bypass logic that is operative to identify data dependencies and to forward results data from the integer execution unit or the floating-point execution unit to an input of the integer execution unit or the floating-point execution unit to reduce the instruction latency of either the integer execution unit or the floating-point execution unit.

4. The processor of claim 1 , wherein the integer execution unit is a first integer execution unit, and wherein the processor further comprises:
a second integer execution unit inside the integer unit, the second integer execution unit being operative to execute one or more types of instructions with single cycle latency;
wherein the second integer execution unit is coupled to the IRF and the FRF, and wherein the integer unit is operative to move data between a register of the IRF and a register of the FRF with single cycle latency using the second integer execution unit.

5. The processor of claim 4 , wherein the processor is operative to fetch and execute multiple instructions in parallel, and wherein the processor is further operative to execute instructions in an order different than the order provided by program code that is executed on the processor.

6. The processor of claim 1 , further comprising:
a working register file (WRF) inside the integer unit, the WRF being operative to store results data that have not been committed to architectural state.

7. The processor of claim 6 , further comprising:
bypass logic that is operative to identify data dependencies and to forward results data from the integer execution unit, the floating-point execution unit, or the WRF to an input of the integer execution unit or the floating-point execution unit to reduce the instruction latency of either the integer execution unit or the floating-point execution unit.

8. The processor of claim 6 , wherein the WRF, IRF, and FRF are implemented physically near the integer execution unit such that the integer execution unit executes instructions having one or more operands located in the WRF, IRF, or the FRF with single cycle latency.

9. A method of fabricating a core adapted for providing functionality to move data between register files in a processor, the method comprising:
providing an integer unit including an integer execution unit thereinside that is operative to execute one or more types of instructions with single cycle latency;
coupling a floating-point execution unit to the integer unit with a data bus;
positioning an integer register file (IRF) proximate to the integer execution unit inside the integer unit; and
positioning a floating-point register file (FRF) proximate to the integer execution unit inside the integer unit, and wherein the integer execution unit is operable to move data from a register of the IRF to a register of the FRF and move data from a register of the FRF to a register of the IRF with single cycle latencies.

10. The method of claim 9 , further comprising:
coupling the IRF and FRF to the integer execution unit through data buses.

11. The method of claim 9 , further comprising:
interconnecting a bypass controller to at least one of the integer execution unit and the floating-point execution unit, wherein the bypass controller is operable to identify data dependencies and forward results data from the integer execution unit or the floating-point execution unit to an input of the integer execution unit or the floating-point execution unit.

12. The method of claim 9 , wherein the integer execution unit is a first integer execution unit, and wherein the method further comprises:
providing a second integer execution unit inside the integer unit; and
coupling the IRF and the FRF to the second integer execution unit through data buses, wherein the second integer execution unit is operable to move data between a register of the IRF and a register of the FRF with single cycle latency.

13. The method of claim 9 , further comprising:
coupling a working register file (WRF) to the integer execution unit and the floating-point execution unit, wherein the WRF is located inside the integer unit, and wherein the WRF is operable to store results data from the integer execution unit or the floating-point execution unit that have not been committed to architectural state.

14. The method of claim 13 , wherein the WRF, IRF, and FRF are implemented physically near the integer execution unit such that the integer execution unit executes instructions having one or more operands located in the WRF, IRF, or the FRF with single cycle latency.

15. A computer implemented method for moving data between register files of a processor, the processor comprising an integer unit including an integer execution unit inside the integer unit that is operative to execute one or more types of instructions with single cycle latency, an integer register file (IRF) coupled to the integer execution unit inside the integer unit, a floating-point register file (FRF) coupled to the integer execution unit inside the integer unit, and a working register file (WRF) coupled to the integer execution unit inside the integer unit; a floating-point execution unit that is operative to execute one or more types of instructions with multiple cycle latency; and a data bus that couples the floating-point execution unit to the integer unit, the method comprising:
receiving data from one of the IRF or FRF at the integer execution unit; and
executing a move instruction from the integer execution unit to move the data to the other of the IRF or FRF, the executing comprising:
writing the data to a register of the WRF; and
writing the data from the register of the WRF to a register of the other of the IRF or FRF, wherein the data is moved from the one of the IRF or FRF to the other of the IRF or FRF using the integer execution unit with single cycle latency.

16. The computer implemented method of claim 15 , further comprising:
storing floating-point data in the IRF or storing integer data in the FRF by executing program code on the processor.

17. The computer implemented method of claim 15 , further comprising:
running a program that includes floating-point instructions on the processor; and
storing floating-point data in the IRF to increase the number of registers available for executing the floating-point instructions.

18. The computer implemented method of claim 15 , further comprising:
running a program that includes integer instructions on the pr

Assignees

• Olson Christopher
• Golla Robert T
• Brooks Jeffrey S
• Oracle America Inc

Inventors

• Olson Christopher
• Golla Robert T
• Brooks Jeffrey S

Classifications

• G06F9/30014

  with variable precision · CPC title

• G06F9/3885

  using a plurality of independent parallel functional units · CPC title

• G06F9/384

  Register renaming · CPC title

• G06F9/3013

  according to data content, e.g. floating-point registers, address registers · CPC title

• G06F9/3851

  from multiple instruction streams, e.g. multistreaming · CPC title