# US 2013/0159666 A1 · Reducing issue-to-issue latency by reversing processing order in half-pumped simd execution units

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

- Number: US 2013/0159666 A1
- Document ID: US20130159666A1
- Status: Published application
- Kind: A1
- Application number: 13/326,249
- Filed: 2011-12-14
- Published: 2013-06-20
- Earliest claimed: 2011-12-14
- Assignees: International Business Machines Corporation
- Inventors: Maarten J. Boersma; Markus Kaltenbach; Christophe J. Layer; Jens Leenstra; Silvia M. Mueller
- Web page: https://patentgrep.com/patent/US20130159666A1

## CPC

- G06F 15/76: Architectures of general purpose stored program computers (with program plugboard G06F15/08; multicomputers G06F15/16)
- G06F 9/30036: {Instructions to perform operations on packed data, e.g. vector, tile or matrix operations}
- G06F 9/345: of multiple operands or results {(addressing multiple banks G06F12/06)}
- G06F 9/38: Concurrent instruction execution, e.g. pipeline or look ahead
- G06F 9/3826: {Bypassing or forwarding of data results, e.g. locally between pipeline stages or within a pipeline stage}
- G06F 9/3867: {using instruction pipelines}
- G06F 9/3869: {Implementation aspects, e.g. pipeline latches; pipeline synchronisation and clocking}
- G06F 15/8053: {Vector processors}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Techniques for reducing issue-to-issue latency by reversing processing order in half-pumped single instruction multiple data (SIMD) execution units are described. In one embodiment a processor functional unit is provided comprising a frontend unit, and execution core unit, a backend unit, an execution order control signal unit, a first interconnect coupled between and output and an input of the execution core unit and a second interconnect coupled between an output of the backend unit and an input of the frontend unit. In operation, the execution order control signal unit generates a forwarding order control signal based on the parity of an applied clock signal on reception of a first vector instruction. This control signal is in turn used to selectively forward first and second portions of an execution result of the first vector instruction via the interconnects for use in the execution of a dependent second vector instruction.

## Claims

### Claim 1

A processor functional unit comprising:
a frontend unit coupled to a register file, said register file comprising a plurality of registers to store one or more operands associated with a first vector instruction and a second vector instruction, wherein execution of said second vector instruction depends on an execution result of said first vector instruction;
an execution core unit coupled to said frontend unit, said execution core unit comprising a pipelined execution unit to execute instructions including said first vector instruction and said second vector instruction;
a backend unit coupled to said execution core unit, said backend unit comprising a result register to store said execution result;
an execution order control signal unit to generate a forwarding order control signal in dependence on a parity of a clock signal applied to said processor functional unit;
a first interconnect coupled between an output of said execution core unit and an input of said execution core unit to selectively forward at least one of a first portion of said execution result and a second portion of said execution result to said input of said execution core unit based on said forwarding order control signal; and
a second interconnect coupled between an output of said backend unit and an input of said frontend unit to selectively forward at least one of said first portion of said execution result and said second portion of said execution result to said input of said frontend unit based on said forwarding order control signal.

### Claim 2 (depends on 1)

The processor functional unit of claim 1, wherein
said instructions including said first vector instruction and said second vector instruction further comprise a scalar instruction,
said processor functional unit has a minimum issue-to-issue latency of K−1 cycles, and
said first vector instruction and said second vector instruction have an issue-to-issue latency of K cycles.

### Claim 3 (depends on 1)

The processor functional unit of claim 1, wherein
said execution order control signal unit is coupled to a signal selection unit to
forward said first portion of said execution result to said execution core unit prior to said second portion of said execution result in response to a determination that said parity of said clock signal, upon a receipt of said first vector instruction at said execution core unit, is even, and
forward said second portion of said execution result to said execution core unit prior to said first portion of said execution result in response to a determination that said parity of said clock signal, upon receipt of said first vector instruction at said execution core unit, is odd.

### Claim 4 (depends on 1)

The processor functional unit of claim 1, wherein said second interconnect is further coupled to at least one other execution core unit.

### Claim 5 (depends on 1)

The processor functional unit of claim 1, wherein said execution order control signal unit comprises a single-bit binary counter.

### Claim 6 (depends on 1)

The processor functional unit of claim 1, wherein
said first vector instruction comprises a single instruction multiple data(SIMD) instruction, and
said pipelined execution unit comprises a SIMD instruction mode to execute said SIMD instruction utilizing a half-pumped instruction execution technique.

### Claim 7

A method comprising:
receiving a plurality of instructions, wherein
said plurality of instructions comprises a first vector instruction and a second vector instruction, and
execution of said second vector instruction depends on an execution result of said first vector instruction,
executing said first vector instruction utilizing a processor functional unit, wherein said processor functional unit comprises a pipelined execution unit;
determining a forwarding order of a first portion of said execution result and a second portion of said execution result in dependence on a parity of a clock signal applied to said processor functional unit upon receipt of said first vector instruction; and
forwarding said first portion of said execution result and said second portion of said execution result from an output coupled to said pipelined execution unit to an input coupled to said pipelined execution unit according to said forwarding order.

### Claim 8 (depends on 7)

The method of claim 7, wherein
said receiving comprises receiving a set of scalar instructions and a set of vector instructions,
said set of vector instructions comprises said first vector instruction and said second vector instruction, and
said processor functional unit has a minimum issue-to-issue latency of K−1 cycles, and
said method further comprises,
issuing said second vector instruction K cycles after issuing said first vector instruction.

### Claim 9 (depends on 7)

The method of claim 7, wherein
said executing said first vector instruction further comprises:
executing a first portion of said first vector instruction to generate said first portion of said execution result, and
executing a second portion of said first vector instruction to generate said second portion of said execution result in response to an execution of said first portion of said first vector instruction; and
said forwarding said first portion of said execution result and said second portion of said execution result comprises,
forwarding said second portion of said execution result utilizing a first interconnect coupled between an output of said pipelined execution unit and an input of said pipelined execution unit, and
forwarding said first portion of said execution result utilizing a second interconnect coupled between an output of said processor functional unit and an input of said processor functional unit in response to a forward of said second portion of said execution result.

### Claim 10 (depends on 9)

The method of claim 9, wherein said second interconnect is coupled to one or more other pipelined execution units.

### Claim 11 (depends on 7)

The method of claim 7, wherein
said determining comprises determining said forwarding order utilizing a single-bit binary counter to identify said parity of said clock signal upon receipt of said first vector instruction, and
said forwarding comprises,
forwarding said first portion of said execution result prior to said second portion of said execution result in response to a determination that said parity of said clock signal upon said receipt of said first vector instruction is even, and
forwarding said second portion of said execution result prior to said first portion of said execution result in response to a determination that said parity of said clock signal upon receipt of said first vector instruction is odd.

### Claim 12 (depends on 7)

The method of claim 7, wherein
said first vector instruction comprises a single instruction multiple data (SIMD) instruction, and
said executing said first vector instruction further comprises executing said SIMD instruction utilizing a half-pumped instruction execution technique.

### Claim 13

A data handling system comprising:
a memory unit;
a bus coupled to said memory unit; and
processor functional unit coupled to said bus, said processor functional unit comprising:
a frontend unit coupled to a register file, said register file comprising a plurality of registers to store one or more operands associated with a first vector instruction and a second vector instruction, wherein execution of said second vector instruction depends on an execution result of said first vector instruction;
an execution core unit coupled to said frontend unit, said execution core unit comprising a pipelined execution unit to execute instructions including said first vector instruction and said second vector instruction;
a backend unit coupled to said execution core unit, said backend unit comprising a result register to store said execution result;
an execution order control signal unit to generate a forwarding order control signal in dependence on a parity of a clock signal applied to said processor functional unit;
a first interconnect coupled between an output of said execution core unit and an input of said execution core unit to selectively forward at least one of a first portion of said execution result and a second portion of said execution result to said input of said execution core unit based on said forwarding order control signal; and
a second interconnect coupled between an output of said backend unit and an input of said frontend unit to selectively forward at least one of said first portion of said execution result and said second portion of said execution result to said input of said frontend unit based on said forwarding order control signal.

### Claim 14 (depends on 13)

The data handling system of claim 13, wherein
said instructions including said first vector instruction and said second vector instruction further comprise a scalar instruction,
said processor functional unit has a minimum issue-to-issue latency of K−1 cycles, and
said first vector instruction and said second vector instruction have an issue-to-issue latency of K cycles.

### Claim 15 (depends on 13)

The data handling system of claim 13, wherein
said execution order control signal unit is coupled to a signal selection unit to
forward said first portion of said execution result to said execution core unit prior to said second portion of said execution result in response to a determination that said parity of said clock signal, upon a receipt of said first vector instruction at said execution core unit, is even, and
forward said second portion of said execution result to said execution core unit prior to said first portion of said execution result in response to a determination that said parity of said clock signal, upon receipt of said first vector instruction at said execution core unit, is odd.

### Claim 16 (depends on 13)

The data handling system of claim 13, wherein said second interconnect is further coupled to at least one other execution core unit.

### Claim 17 (depends on 13)

The data handling system of claim 13, wherein said execution order control signal unit comprises a single-bit binary counter.

### Claim 18 (depends on 13)

The data handling system of claim 13, wherein
said first vector instruction comprises a single instruction multiple data (SIMD) instruction, and
said pipelined execution unit comprises a SIMD instruction mode to execute said SIMD instruction utilizing a half-pumped instruction execution technique.
