# US 2013/0159669 A1 · Low latency variable transfer network for fine grained parallelism of virtual threads across multiple hardware threads

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

- Number: US 2013/0159669 A1
- Document ID: US20130159669A1
- Status: Published application
- Kind: A1
- Application number: 13/330,831
- Filed: 2011-12-20
- Published: 2013-06-20
- Earliest claimed: 2011-12-20
- Assignees: International Business Machines Corporation
- Inventors: Miguel Comparan; Russell D. Hoover; Robert A. Shearer; Alfred T. Watson, III
- Web page: https://patentgrep.com/patent/US20130159669A1

## CPC

- G06F 9/3824: {Operand accessing}
- G06F 9/38: Concurrent instruction execution, e.g. pipeline or look ahead
- G06F 9/3828: {with global bypass, e.g. between pipelines, between clusters}
- G06F 9/3851: {from multiple instruction streams, e.g. multistreaming}
- G06F 9/3891: {organised in groups of units sharing resources, e.g. clusters}
- G06F 9/544: {Buffers; Shared memory; Pipes}
- G06F 15/173: using an interconnection network, e.g. matrix, shuffle, pyramid, star, snowflake
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method and circuit arrangement utilize a low latency variable transfer network between the register files of multiple processing cores in a multi-core processor chip to support fine grained parallelism of virtual threads across multiple hardware threads. The communication of a variable over the variable transfer network may be initiated by a move from a local register in a register file of a source processing core to a variable register that is allocated to a destination hardware thread in a destination processing core, so that the destination hardware thread can then move the variable from the variable register to a local register in the destination processing core.

## Claims

### Claim 1

A circuit arrangement, comprising:
a plurality of hardware threads disposed in a plurality of processing cores, each processing core including a register file and at least one hardware thread among the plurality of hardware threads; and
a variable transfer network coupled between the plurality of processing cores, the variable transfer network configured to communicate a variable between a source hardware thread in a source processing core among the plurality of processing cores and a destination hardware thread in a destination processing core among the plurality of processing cores.

### Claim 2 (depends on 1)

The circuit arrangement of claim 1, further comprising a plurality of variable registers, each variable register allocated to an individual hardware thread among the plurality of hardware threads, wherein the variable transfer network is configured to communicate the variable to the destination processing core in response to storage of the variable in the variable register allocated to the destination hardware thread.

### Claim 3 (depends on 2)

The circuit arrangement of claim 2, wherein the plurality of variable registers are architected into the register file of each processing core, and wherein dependency logic disposed within each processing core handles dependencies for both a plurality of local registers in the register file and the plurality of variable registers.

### Claim 4 (depends on 2)

The circuit arrangement of claim 2, wherein the variable transfer network is configured to communicate the variable to the destination processing core in response to an instruction executed by the source hardware thread.

### Claim 5 (depends on 4)

The circuit arrangement of claim 4, wherein the instruction comprises a put instruction that moves the variable from a local register in the register file of the source processing core to the variable register allocated to the destination hardware thread.

### Claim 6 (depends on 5)

The circuit arrangement of claim 5, wherein the variable register allocated to the destination hardware thread includes a credit flag that indicates whether the destination hardware thread is ready to receive a variable over the variable transfer network, and wherein the source processing core is configured to stall the put instruction until the credit flag indicates that the destination hardware thread is ready to receive a variable.

### Claim 7 (depends on 2)

The circuit arrangement of claim 2, wherein the variable transfer network is configured to present the variable at the variable register associated with the source thread in the destination processing core in response to storage of the variable in the variable register allocated to the destination hardware thread in the source processing core, wherein the destination hardware thread is configured to move the variable from the variable register allocated to the source hardware thread to a local register in the register file of the destination processing core in response to an instruction executed by the destination hardware thread.

### Claim 8 (depends on 7)

The circuit arrangement of claim 7, wherein the instruction comprises a get instruction, wherein the variable register allocated to the source hardware thread in the destination processing core includes a valid flag that indicates whether valid data is stored in the variable register allocated to the source hardware thread in the destination processing core, and wherein the destination processing core is configured to stall the get instruction until the valid flag indicates that valid data is stored in the variable register allocated to the source hardware thread in the destination processing core.

### Claim 9 (depends on 7)

The circuit arrangement of claim 7, wherein the destination processing core is configured to returning a credit over the variable transfer network in response to moving the variable from the variable register allocated to the source hardware thread in the destination processing core to a local register in the register file of the destination processing core.

### Claim 10 (depends on 1)

The circuit arrangement of claim 1, wherein the variable transfer network comprises a send bus coupled between the source processing core and the destination processing core, the send bus including a plurality of data lines configured to communicate the variable, at least one thread identifier line configured to communicate an identifier for the destination hardware thread, and a valid line configured to indicate when valid data is being communicated over the plurality of data lines.

### Claim 11 (depends on 10)

The circuit arrangement of claim 10, wherein each of the data, thread identifier and valid lines are driven by the source processing core, and wherein the send bus further includes a credit line driven by the destination processing core and configured to return a credit to the source processing core in response to a successful transfer of the variable by the destination hardware thread.

### Claim 12 (depends on 10)

The circuit arrangement of claim 10, wherein the send bus comprises a dedicated bus from the source processing core to the destination processing core, and wherein the variable transfer network comprises a plurality of send buses such that each of the plurality of processing cores includes dedicated buses to each other processing core among the plurality of processing cores.

### Claim 13 (depends on 10)

The circuit arrangement of claim 10, wherein the send bus is coupled from the source processing core to each other processing core among the plurality of processing cores, and wherein the send bus further includes at least one core identifier line configured to communicate an identifier for the destination processing core.

### Claim 14 (depends on 1)

An integrated circuit device including the circuit arrangement of claim 1.

### Claim 15 (depends on 1)

A program product comprising a computer readable medium and logic definition program code stored on the computer readable medium and defining the circuit arrangement of claim 1.

### Claim 16

A method of communicating variables between a plurality of hardware threads disposed in a plurality of processing cores, each processing core including a register file and at least one hardware thread among the plurality of hardware threads, the method comprising:
with a source hardware thread in a source processing core among the plurality of processing cores, initiating a transfer of a variable to a destination hardware thread in a destination processing core among the plurality of processing cores; and
in response to initiating the transfer of the variable, communicating the variable from the source processing core to the destination processing core over a variable transfer network coupled to the source and destination processing cores.

### Claim 17 (depends on 16)

The method of claim 16, wherein initiating the transfer of the variable comprises storing the variable in a variable register allocated to the destination hardware thread, wherein the variable register is among a plurality of variable registers disposed on the multi-core processing chip, wherein each variable register is allocated to an individual hardware thread among the plurality of hardware threads.

### Claim 18 (depends on 17)

The method of claim 17, wherein the plurality of variable registers are architected into the register file of each processing core, and wherein dependency logic disposed within each processing core handles dependencies for both a plurality of local registers in the register file and the plurality of variable registers.

### Claim 19 (depends on 17)

The method of claim 17, wherein initiating the transfer of the variable includes executing a put instruction with the source hardware thread, wherein the put instruction that moves the variable from a local register in the register file of the source processing core to the variable register allocated to the destination hardware thread.

### Claim 20 (depends on 19)

The method of claim 19, wherein the variable register allocated to the destination hardware thread includes a credit flag that indicates whether the destination hardware thread is ready to receive a variable over the variable transfer network, and wherein the method further comprises stalling the put instruction until the credit flag indicates that the destination hardware thread is ready to receive a variable.

### Claim 21 (depends on 17)

The method of claim 17, wherein the variable transfer network is configured to present the variable at the variable register associated with the source thread in the destination processing core in response to storage of the variable in the variable register allocated to the destination hardware thread in the source processing core, the method further comprising moving the variable from the variable register allocated to the source thread in the destination processing core to a local register in the register file of the destination processing core in response to a get instruction executed by the destination hardware thread, wherein the variable register allocated to the source thread in the destination processing core includes a valid flag that indicates whether valid data is stored in the variable register allocated to the source thread in the destination processing core, and wherein the method further comprises stalling the get instruction until the valid flag indicates that valid data is stored in the variable register allocated to the source thread in the destination processing core.

### Claim 22 (depends on 21)

The method of claim 21, further comprising returning a credit over the variable transfer network in response to moving the variable from the variable register allocated to the source thread in the destination processing core to a local register in the register file of the destination processing core.

### Claim 23 (depends on 16)

The method of claim 16, wherein the variable transfer network comprises a send bus coupled between the source processing core and the destination processing core, the send bus including a plurality of data lines configured to communicate the variable, at least one thread identifier line configured to communicate an identifier for the destination hardware thread, and a valid line configured to indicate when valid data is being communicated over the plurality of data lines, wherein each of the data, thread identifier and valid lines are driven by the source processing core, and wherein the send bus further includes a credit line driven by the destination processing core and configured to return a credit to the source processing core in response to a successful transfer of the variable to the destination hardware thread.

### Claim 24 (depends on 23)

The method of claim 23, wherein the send bus comprises a dedicated bus from the source processing core to the destination processing core, and wherein the variable transfer network comprises a plurality of send buses such that each of the plurality of processing cores includes dedicated buses to each other processing core among the plurality of processing cores.

### Claim 25 (depends on 23)

The method of claim 23, wherein the send bus is coupled from the source processing core to each other processing core among the plurality of processing cores, and wherein the send bus further includes at least one core identifier line configured to communicate an identifier for the destination processing core.
