# US 2023/0273797 A1 · Processor with adaptive pipeline length

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2023/0273797 A1
- Document ID: US20230273797A1
- Status: Published application
- Kind: A1
- Application number: 18/314,264
- Filed: 2023-05-09
- Published: 2023-08-31
- Earliest claimed: 2013-08-23
- Assignees: Texas Instruments Incorporated
- Applicants: Texas Instruments Incorporated
- Inventors: Christian Wiencke; Shrey Sudhir Bhatia; Jeroen Vliegen
- Web page: https://patentgrep.com/patent/US20230273797A1

## CPC

- G06F 9/3873: {Variable length pipelines, e.g. elastic pipeline}
- G06F 9/3838: {Dependency mechanisms, e.g. register scoreboarding}
- G06F 9/3867: {using instruction pipelines}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A system and method for reducing pipeline latency. In one embodiment, a processing system includes a processing pipeline. The processing pipeline includes a plurality of processing stages. Each stage is configured to further processing provided by a previous stage. A first of the stages is configured to perform a first function in a pipeline cycle. A second of the stages is disposed downstream of the first of the stages, and is configured to perform, in a pipeline cycle, a second function that is different from the first function. The first of the stages is further configured to selectably perform the first function and the second function in a pipeline cycle, and bypass the second of the stages.

## Claims

### Claim 1

A method, comprising:
identifying a first instruction being executed in a pipeline stage of an instruction processing pipeline;
determining whether an execute function is to be performed on the first instruction in an execution stage or whether the execute function and an decode function are to be performed on the first instruction in a first stage based on determining whether performing the execute function on the first instruction in the execution stage would cause a delay in execution of a second instruction that depends on a result of the first instruction; and
stalling the instruction processing pipeline for a single pipeline cycle in response to determining that the execute function is to be performed on the first instruction in the first stage.

### Claim 2 (depends on 1)

The method of claim 1, wherein the instruction processing pipeline comprises:
a plurality of stages including:
a fetch stage comprising circuitry configured to perform a fetch function on the first instruction;
a decode stage coupled to the fetch stage and comprising circuitry configured to perform a decode function on the first instruction; and
an execution stage coupled to the decode stage and comprising circuitry configured to perform an execute function on the first instruction.

### Claim 3 (depends on 2)

The method of claim 2, further comprising: selecting the fetch stage or the decode stage for a first stage.

### Claim 4 (depends on 2)

The method of claim 2, wherein one of the fetch stage or the decode stage comprises a latency reduction circuit, the latency reduction circuit comprising circuitry configured to perform the execute function on the first instruction in a respective one of the fetch stage or the decode stage.

### Claim 5 (depends on 1)

The method of claim 1, further comprising:
when determining that the execute function is to be performed on the first instruction:
performing, by a latency reduction circuit, the execute function on the first instruction in a decode stage; and
not performing on the first instruction by circuitry of the execution stage.

### Claim 6 (depends on 1)

The method of claim 1, further comprising: when performing, by a latency reduction circuit, the execute function on the first instruction, storing, by a latency reduction circuit, a result of the execute function in a memory.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
writing, by a latency reduction circuit, to a memory, the result of performing the execute function on the first instruction in a stage other than a writeback stage; and
determining whether the result of performing the execute function on the first instruction is written to the memory by the writeback stage or by the latency reduction circuit.

### Claim 8 (depends on 1)

The method of claim 1, further comprising:
performing, by a latency reduction circuit, the execute function on the first instruction in a decode stage; and
writing, by the latency reduction circuit, a result of the execute function on the first instruction produced in the decode stage.

### Claim 9

A device comprising:
a plurality of stages including:
a fetch stage including circuitry configured to perform a fetch function on a first instruction;
a decode stage coupled to the fetch stage and including circuitry configured to perform a decode function on the first instruction; and
an execution stage coupled to the decode stage and including circuitry configured to perform an execute function on the first instruction;
a latency reduction circuit included within the decode stage, the latency reduction circuit including circuitry configured to perform the execute function on the first instruction in the decode stage; and
a pipeline control circuit coupled to the latency reduction circuit and configured to:
identify the first instruction;
determine whether the execute function is to be performed on the first instruction by the execution stage or the decode function and the execute function are to be performed on the first instruction by the decode stage based on determining whether performing the execute function on the first instruction by the execution stage would cause a delay in execution of a second instruction that depends on a result of the first instruction; and
stall an instruction processing pipeline for a single pipeline cycle in response to the pipeline control circuit determining that the execution function is to be performed on the first instruction by the latency reduction circuit in the decode stage.

### Claim 10 (depends on 9)

The device of claim 9, wherein the latency reduction circuit further comprises circuitry configured to write, in a stage selected from the fetch stage, the decode stage, and the execute stage, the result of the execute function on the first instruction by the latency reduction circuit to a memory.

### Claim 11 (depends on 9)

The device of claim 9, wherein the latency reduction circuit is configured to perform functions of the circuitry of the execution stage and of circuitry configured to write a result of the execute function on the first instruction to a memory.

### Claim 12

A method comprising:
determining whether performing a execute operation on a first instruction using an execute stage delays a second instruction that depends on the first instruction;
determining whether to cause decode logic and execute logic of a fetch stage or a decode stage and a execute stage to perform a decode operation and the execute operation on the first instruction based on whether performing execute operation on the first instruction using the execute stage delays the second instruction; and
determining whether to stall the fetch stage, the decode stage, and the execute stage based on whether the execute operation is to be performed on the first instruction using the decode logic and the execute logic of the fetch stage.

### Claim 13 (depends on 12)

The method of claim 12, wherein:
performing, by the execute stage, a respective execute operation on each of a set of instructions; and
performing, by the execute logic, a respective execute operation on each of a subset of the set of instructions.

### Claim 14 (depends on 13)

The method of claim 13, wherein the subset of the set of instructions is a set of flow control instructions.

### Claim 15 (depends on 12)

The method of claim 12, further comprising: determining whether to cause writeback logic of the fetch stage or a writeback stage to perform a writeback operation on the first instruction, wherein the writeback stage is coupled to the execute stage that is different from the writeback logic.

### Claim 16 (depends on 15)

The method of claim 15, wherein the fetch stage includes:
fetch logic configured to perform a fetch operation;
decode logic configured to perform the decode operation; and
execute logic configured to perform the execute operation.

### Claim 17 (depends on 16)

The method of claim 16, wherein the decode stage is coupled to the fetch stage and is different from the decode logic of the fetch stage, and is configured to perform the decode operation.

### Claim 18 (depends on 16)

The method of claim 16, wherein the execute stage is coupled to the decode stage and is different from the execute logic of the fetch stage and is configured to perform the execute operation.
