Processor that eliminates mis-steering instruction fetch resulting from incorrect resolution of mis-speculated branch instructions

• US 7,076,640 B2
• Filed: 03/11/2002
• Issued: 07/11/2006
• Est. Priority Date: 02/05/2002
• Status: Active Grant

• Current Assignee: Oracle America, Inc. (Oracle Corporation)
• Original Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Inventors: Kadambi, Sudarshan
• Primary Examiner(s): TSAI, HENRY
• Application Number:
  US10/095,397
• Publication Number:
  US 20030149865A1
• Time in Patent Office: 1,583 Days
• Field of Search: 712/244, 712/233, 712/234, 712/237, 712/240, 712/219
• US Class Current: 712/244
• CPC Class Codes: G06F 9/30058 Conditional branch instructions
• CPC Class Codes: G06F 9/3867 using instruction pipelines

Abstract

A processor avoids or eliminates repetitive replay conditions and frequent instruction resteering through various techniques including resteering the fetch after the branch instruction retires, and delaying branch resolution. A processor resolves conditional branches and avoids repetitive resteering by delaying branch resolution. The processor has an instruction pipeline with inserted delay in branch condition and replay control pathways. For example, an instruction sequence that includes a load instruction followed by a subtract instruction then a conditional branch, delays branch resolution to allow time for analysis to determine whether the condition branch has resolved correctly. Eliminating incorrect branch resolutions prevents flushing of correctly predicted branches.

Claims

1. A pipelined processor comprising:

an instruction fetcher capable of fetching instructions from an instruction source;

a branch resolver coupled to the instruction fetcher and capable of resolving a branch instruction;

a delay element coupled between the branch resolver and the instruction fetcher, the delay element to allow for determination of whether the branch instruction has been resolved correctly;

replay logic coupled to the instruction fetcher and capable of determining a replay condition; and

logic coupled to receive input from the replay logic and input from the branch resolver, and the logic coupled to supply output to the instruction fetcher for determining a fetch operation of the instruction fetcher, wherein the logic selects either input from the branch resolver or input from the replay logic, wherein the output supplied to the instruction fetcher is the input selected by the logic, and wherein the input from the branch resolver is delayed by the delay element.

2. A pipelined processor according to claim 1 wherein

the replay condition includes a data cache hit/miss, a translation lookaside buffer miss, read-after-write (RAW) hazard, a load miss buffer (LMB) full condition, or an overeager issue condition.

3. A pipelined processor according to claim 1 wherein the instruction fetcher resteers an instruction fetch based on resolution of a branch instruction after the branch instruction retires.

4. A pipelined processor according to claim 1 wherein the instruction fetcher resteers an instruction fetch based on a delayed branch instruction resolution.

5. A pipelined processor according to claim 1 further comprising:

an instruction pipeline with the delay element inserted in branch condition and replay control pathways.

6. A pipelined processor according to claim 1 wherein:

the processor is a deeply pipelined processor, microprocessor, CPU, digital signal processor, sequencer, or computational logic.

7. The pipelined processor of claim 1 wherein the determination of whether the branch instruction resolves correctly is based, at least in part, on the replay logic determining a replay condition.

8. The pipelined processor of claim 1,

wherein the logic combines input from the replay logic with input from the branch resolver that has been delayed by the delay element, and

wherein output supplied from the logic to the instruction fetcher is the combined inputs.

9. The pipelined processor of claim 1 wherein the branch instruction includes a conditional branch instruction or an unconditional branch instruction.

10. A pipelined processor comprising:

an instruction fetcher capable of fetching instructions from an instruction source;

a branch resolver coupled to the instruction fetcher and capable of resolving a branch instruction;

a delay element coupled between the branch resolver and the instruction fetcher, the delay element to allow for determination of whether the branch instruction has been resolved correctly;

a load miss buffer operable to indicate detection of a replay condition for load instructions; and

the instruction fetcher coupled with the load miss buffer, the instruction fetcher operable to fetch instructions based at least in part on both replay condition indications from the load miss buffer and branch instruction resolution indications from the branch resolver.

11. The pipelined processor of claim 10 further comprising a logic gate coupled with the instruction fetcher, the load miss buffer, and the branch resolver, the logic gate coupled to receive as inputs the branch instruction instance resolution indications and the replay condition indications, and operable to logically combine the indications and output the logically combined indications.

12. The pipelined processor of claim 10 further comprising an instruction scheduling unit coupled with the fetch unit and the load miss buffer.

13. The pipelined processor of claim 10 further comprising a memory disambiguation buffer that includes the load miss buffer.

14. The pipelined processor of claim 10 further comprising an instruction execution unit that includes the branch resolver.

15. The pipelined processor of claim 10, wherein the replay condition includes one or more of a load miss buffer full condition, a load miss condition, a read-after-write data hazard condition, a translation lookaside buffer miss condition, and an overeager load condition.